xref: /openbmc/linux/net/ipv4/cipso_ipv4.c (revision cd6d421e)
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 = 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  * @cipso_ptr: pointer to CIPSO IP option
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_info: NetLabel audit information
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 	list_del_rcu(&doi_def->list);
523 	spin_unlock(&cipso_v4_doi_list_lock);
524 
525 	cipso_v4_doi_putdef(doi_def);
526 	ret_val = 0;
527 
528 doi_remove_return:
529 	audit_buf = netlbl_audit_start(AUDIT_MAC_CIPSOV4_DEL, audit_info);
530 	if (audit_buf) {
531 		audit_log_format(audit_buf,
532 				 " cipso_doi=%u res=%u",
533 				 doi, ret_val == 0 ? 1 : 0);
534 		audit_log_end(audit_buf);
535 	}
536 
537 	return ret_val;
538 }
539 
540 /**
541  * cipso_v4_doi_getdef - Returns a reference to a valid DOI definition
542  * @doi: the DOI value
543  *
544  * Description:
545  * Searches for a valid DOI definition and if one is found it is returned to
546  * the caller.  Otherwise NULL is returned.  The caller must ensure that
547  * rcu_read_lock() is held while accessing the returned definition and the DOI
548  * definition reference count is decremented when the caller is done.
549  *
550  */
551 struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
552 {
553 	struct cipso_v4_doi *doi_def;
554 
555 	rcu_read_lock();
556 	doi_def = cipso_v4_doi_search(doi);
557 	if (!doi_def)
558 		goto doi_getdef_return;
559 	if (!refcount_inc_not_zero(&doi_def->refcount))
560 		doi_def = NULL;
561 
562 doi_getdef_return:
563 	rcu_read_unlock();
564 	return doi_def;
565 }
566 
567 /**
568  * cipso_v4_doi_putdef - Releases a reference for the given DOI definition
569  * @doi_def: the DOI definition
570  *
571  * Description:
572  * Releases a DOI definition reference obtained from cipso_v4_doi_getdef().
573  *
574  */
575 void cipso_v4_doi_putdef(struct cipso_v4_doi *doi_def)
576 {
577 	if (!doi_def)
578 		return;
579 
580 	if (!refcount_dec_and_test(&doi_def->refcount))
581 		return;
582 
583 	cipso_v4_cache_invalidate();
584 	call_rcu(&doi_def->rcu, cipso_v4_doi_free_rcu);
585 }
586 
587 /**
588  * cipso_v4_doi_walk - Iterate through the DOI definitions
589  * @skip_cnt: skip past this number of DOI definitions, updated
590  * @callback: callback for each DOI definition
591  * @cb_arg: argument for the callback function
592  *
593  * Description:
594  * Iterate over the DOI definition list, skipping the first @skip_cnt entries.
595  * For each entry call @callback, if @callback returns a negative value stop
596  * 'walking' through the list and return.  Updates the value in @skip_cnt upon
597  * return.  Returns zero on success, negative values on failure.
598  *
599  */
600 int cipso_v4_doi_walk(u32 *skip_cnt,
601 		     int (*callback) (struct cipso_v4_doi *doi_def, void *arg),
602 		     void *cb_arg)
603 {
604 	int ret_val = -ENOENT;
605 	u32 doi_cnt = 0;
606 	struct cipso_v4_doi *iter_doi;
607 
608 	rcu_read_lock();
609 	list_for_each_entry_rcu(iter_doi, &cipso_v4_doi_list, list)
610 		if (refcount_read(&iter_doi->refcount) > 0) {
611 			if (doi_cnt++ < *skip_cnt)
612 				continue;
613 			ret_val = callback(iter_doi, cb_arg);
614 			if (ret_val < 0) {
615 				doi_cnt--;
616 				goto doi_walk_return;
617 			}
618 		}
619 
620 doi_walk_return:
621 	rcu_read_unlock();
622 	*skip_cnt = doi_cnt;
623 	return ret_val;
624 }
625 
626 /*
627  * Label Mapping Functions
628  */
629 
630 /**
631  * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
632  * @doi_def: the DOI definition
633  * @level: the level to check
634  *
635  * Description:
636  * Checks the given level against the given DOI definition and returns a
637  * negative value if the level does not have a valid mapping and a zero value
638  * if the level is defined by the DOI.
639  *
640  */
641 static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
642 {
643 	switch (doi_def->type) {
644 	case CIPSO_V4_MAP_PASS:
645 		return 0;
646 	case CIPSO_V4_MAP_TRANS:
647 		if ((level < doi_def->map.std->lvl.cipso_size) &&
648 		    (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL))
649 			return 0;
650 		break;
651 	}
652 
653 	return -EFAULT;
654 }
655 
656 /**
657  * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
658  * @doi_def: the DOI definition
659  * @host_lvl: the host MLS level
660  * @net_lvl: the network/CIPSO MLS level
661  *
662  * Description:
663  * Perform a label mapping to translate a local MLS level to the correct
664  * CIPSO level using the given DOI definition.  Returns zero on success,
665  * negative values otherwise.
666  *
667  */
668 static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
669 				 u32 host_lvl,
670 				 u32 *net_lvl)
671 {
672 	switch (doi_def->type) {
673 	case CIPSO_V4_MAP_PASS:
674 		*net_lvl = host_lvl;
675 		return 0;
676 	case CIPSO_V4_MAP_TRANS:
677 		if (host_lvl < doi_def->map.std->lvl.local_size &&
678 		    doi_def->map.std->lvl.local[host_lvl] < CIPSO_V4_INV_LVL) {
679 			*net_lvl = doi_def->map.std->lvl.local[host_lvl];
680 			return 0;
681 		}
682 		return -EPERM;
683 	}
684 
685 	return -EINVAL;
686 }
687 
688 /**
689  * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
690  * @doi_def: the DOI definition
691  * @net_lvl: the network/CIPSO MLS level
692  * @host_lvl: the host MLS level
693  *
694  * Description:
695  * Perform a label mapping to translate a CIPSO level to the correct local MLS
696  * level using the given DOI definition.  Returns zero on success, negative
697  * values otherwise.
698  *
699  */
700 static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
701 				 u32 net_lvl,
702 				 u32 *host_lvl)
703 {
704 	struct cipso_v4_std_map_tbl *map_tbl;
705 
706 	switch (doi_def->type) {
707 	case CIPSO_V4_MAP_PASS:
708 		*host_lvl = net_lvl;
709 		return 0;
710 	case CIPSO_V4_MAP_TRANS:
711 		map_tbl = doi_def->map.std;
712 		if (net_lvl < map_tbl->lvl.cipso_size &&
713 		    map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
714 			*host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
715 			return 0;
716 		}
717 		return -EPERM;
718 	}
719 
720 	return -EINVAL;
721 }
722 
723 /**
724  * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
725  * @doi_def: the DOI definition
726  * @bitmap: category bitmap
727  * @bitmap_len: bitmap length in bytes
728  *
729  * Description:
730  * Checks the given category bitmap against the given DOI definition and
731  * returns a negative value if any of the categories in the bitmap do not have
732  * a valid mapping and a zero value if all of the categories are valid.
733  *
734  */
735 static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
736 				      const unsigned char *bitmap,
737 				      u32 bitmap_len)
738 {
739 	int cat = -1;
740 	u32 bitmap_len_bits = bitmap_len * 8;
741 	u32 cipso_cat_size;
742 	u32 *cipso_array;
743 
744 	switch (doi_def->type) {
745 	case CIPSO_V4_MAP_PASS:
746 		return 0;
747 	case CIPSO_V4_MAP_TRANS:
748 		cipso_cat_size = doi_def->map.std->cat.cipso_size;
749 		cipso_array = doi_def->map.std->cat.cipso;
750 		for (;;) {
751 			cat = netlbl_bitmap_walk(bitmap,
752 						 bitmap_len_bits,
753 						 cat + 1,
754 						 1);
755 			if (cat < 0)
756 				break;
757 			if (cat >= cipso_cat_size ||
758 			    cipso_array[cat] >= CIPSO_V4_INV_CAT)
759 				return -EFAULT;
760 		}
761 
762 		if (cat == -1)
763 			return 0;
764 		break;
765 	}
766 
767 	return -EFAULT;
768 }
769 
770 /**
771  * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
772  * @doi_def: the DOI definition
773  * @secattr: the security attributes
774  * @net_cat: the zero'd out category bitmap in network/CIPSO format
775  * @net_cat_len: the length of the CIPSO bitmap in bytes
776  *
777  * Description:
778  * Perform a label mapping to translate a local MLS category bitmap to the
779  * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
780  * size in bytes of the network bitmap on success, negative values otherwise.
781  *
782  */
783 static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
784 				     const struct netlbl_lsm_secattr *secattr,
785 				     unsigned char *net_cat,
786 				     u32 net_cat_len)
787 {
788 	int host_spot = -1;
789 	u32 net_spot = CIPSO_V4_INV_CAT;
790 	u32 net_spot_max = 0;
791 	u32 net_clen_bits = net_cat_len * 8;
792 	u32 host_cat_size = 0;
793 	u32 *host_cat_array = NULL;
794 
795 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
796 		host_cat_size = doi_def->map.std->cat.local_size;
797 		host_cat_array = doi_def->map.std->cat.local;
798 	}
799 
800 	for (;;) {
801 		host_spot = netlbl_catmap_walk(secattr->attr.mls.cat,
802 					       host_spot + 1);
803 		if (host_spot < 0)
804 			break;
805 
806 		switch (doi_def->type) {
807 		case CIPSO_V4_MAP_PASS:
808 			net_spot = host_spot;
809 			break;
810 		case CIPSO_V4_MAP_TRANS:
811 			if (host_spot >= host_cat_size)
812 				return -EPERM;
813 			net_spot = host_cat_array[host_spot];
814 			if (net_spot >= CIPSO_V4_INV_CAT)
815 				return -EPERM;
816 			break;
817 		}
818 		if (net_spot >= net_clen_bits)
819 			return -ENOSPC;
820 		netlbl_bitmap_setbit(net_cat, net_spot, 1);
821 
822 		if (net_spot > net_spot_max)
823 			net_spot_max = net_spot;
824 	}
825 
826 	if (++net_spot_max % 8)
827 		return net_spot_max / 8 + 1;
828 	return net_spot_max / 8;
829 }
830 
831 /**
832  * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
833  * @doi_def: the DOI definition
834  * @net_cat: the category bitmap in network/CIPSO format
835  * @net_cat_len: the length of the CIPSO bitmap in bytes
836  * @secattr: the security attributes
837  *
838  * Description:
839  * Perform a label mapping to translate a CIPSO bitmap to the correct local
840  * MLS category bitmap using the given DOI definition.  Returns zero on
841  * success, negative values on failure.
842  *
843  */
844 static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
845 				     const unsigned char *net_cat,
846 				     u32 net_cat_len,
847 				     struct netlbl_lsm_secattr *secattr)
848 {
849 	int ret_val;
850 	int net_spot = -1;
851 	u32 host_spot = CIPSO_V4_INV_CAT;
852 	u32 net_clen_bits = net_cat_len * 8;
853 	u32 net_cat_size = 0;
854 	u32 *net_cat_array = NULL;
855 
856 	if (doi_def->type == CIPSO_V4_MAP_TRANS) {
857 		net_cat_size = doi_def->map.std->cat.cipso_size;
858 		net_cat_array = doi_def->map.std->cat.cipso;
859 	}
860 
861 	for (;;) {
862 		net_spot = netlbl_bitmap_walk(net_cat,
863 					      net_clen_bits,
864 					      net_spot + 1,
865 					      1);
866 		if (net_spot < 0) {
867 			if (net_spot == -2)
868 				return -EFAULT;
869 			return 0;
870 		}
871 
872 		switch (doi_def->type) {
873 		case CIPSO_V4_MAP_PASS:
874 			host_spot = net_spot;
875 			break;
876 		case CIPSO_V4_MAP_TRANS:
877 			if (net_spot >= net_cat_size)
878 				return -EPERM;
879 			host_spot = net_cat_array[net_spot];
880 			if (host_spot >= CIPSO_V4_INV_CAT)
881 				return -EPERM;
882 			break;
883 		}
884 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
885 						       host_spot,
886 						       GFP_ATOMIC);
887 		if (ret_val != 0)
888 			return ret_val;
889 	}
890 
891 	return -EINVAL;
892 }
893 
894 /**
895  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
896  * @doi_def: the DOI definition
897  * @enumcat: category list
898  * @enumcat_len: length of the category list in bytes
899  *
900  * Description:
901  * Checks the given categories against the given DOI definition and returns a
902  * negative value if any of the categories do not have a valid mapping and a
903  * zero value if all of the categories are valid.
904  *
905  */
906 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
907 				       const unsigned char *enumcat,
908 				       u32 enumcat_len)
909 {
910 	u16 cat;
911 	int cat_prev = -1;
912 	u32 iter;
913 
914 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
915 		return -EFAULT;
916 
917 	for (iter = 0; iter < enumcat_len; iter += 2) {
918 		cat = get_unaligned_be16(&enumcat[iter]);
919 		if (cat <= cat_prev)
920 			return -EFAULT;
921 		cat_prev = cat;
922 	}
923 
924 	return 0;
925 }
926 
927 /**
928  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
929  * @doi_def: the DOI definition
930  * @secattr: the security attributes
931  * @net_cat: the zero'd out category list in network/CIPSO format
932  * @net_cat_len: the length of the CIPSO category list in bytes
933  *
934  * Description:
935  * Perform a label mapping to translate a local MLS category bitmap to the
936  * correct CIPSO category list using the given DOI definition.   Returns the
937  * size in bytes of the network category bitmap on success, negative values
938  * otherwise.
939  *
940  */
941 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
942 				      const struct netlbl_lsm_secattr *secattr,
943 				      unsigned char *net_cat,
944 				      u32 net_cat_len)
945 {
946 	int cat = -1;
947 	u32 cat_iter = 0;
948 
949 	for (;;) {
950 		cat = netlbl_catmap_walk(secattr->attr.mls.cat, cat + 1);
951 		if (cat < 0)
952 			break;
953 		if ((cat_iter + 2) > net_cat_len)
954 			return -ENOSPC;
955 
956 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
957 		cat_iter += 2;
958 	}
959 
960 	return cat_iter;
961 }
962 
963 /**
964  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
965  * @doi_def: the DOI definition
966  * @net_cat: the category list in network/CIPSO format
967  * @net_cat_len: the length of the CIPSO bitmap in bytes
968  * @secattr: the security attributes
969  *
970  * Description:
971  * Perform a label mapping to translate a CIPSO category list to the correct
972  * local MLS category bitmap using the given DOI definition.  Returns zero on
973  * success, negative values on failure.
974  *
975  */
976 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
977 				      const unsigned char *net_cat,
978 				      u32 net_cat_len,
979 				      struct netlbl_lsm_secattr *secattr)
980 {
981 	int ret_val;
982 	u32 iter;
983 
984 	for (iter = 0; iter < net_cat_len; iter += 2) {
985 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
986 					     get_unaligned_be16(&net_cat[iter]),
987 					     GFP_ATOMIC);
988 		if (ret_val != 0)
989 			return ret_val;
990 	}
991 
992 	return 0;
993 }
994 
995 /**
996  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
997  * @doi_def: the DOI definition
998  * @rngcat: category list
999  * @rngcat_len: length of the category list in bytes
1000  *
1001  * Description:
1002  * Checks the given categories against the given DOI definition and returns a
1003  * negative value if any of the categories do not have a valid mapping and a
1004  * zero value if all of the categories are valid.
1005  *
1006  */
1007 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1008 				      const unsigned char *rngcat,
1009 				      u32 rngcat_len)
1010 {
1011 	u16 cat_high;
1012 	u16 cat_low;
1013 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1014 	u32 iter;
1015 
1016 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1017 		return -EFAULT;
1018 
1019 	for (iter = 0; iter < rngcat_len; iter += 4) {
1020 		cat_high = get_unaligned_be16(&rngcat[iter]);
1021 		if ((iter + 4) <= rngcat_len)
1022 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1023 		else
1024 			cat_low = 0;
1025 
1026 		if (cat_high > cat_prev)
1027 			return -EFAULT;
1028 
1029 		cat_prev = cat_low;
1030 	}
1031 
1032 	return 0;
1033 }
1034 
1035 /**
1036  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1037  * @doi_def: the DOI definition
1038  * @secattr: the security attributes
1039  * @net_cat: the zero'd out category list in network/CIPSO format
1040  * @net_cat_len: the length of the CIPSO category list in bytes
1041  *
1042  * Description:
1043  * Perform a label mapping to translate a local MLS category bitmap to the
1044  * correct CIPSO category list using the given DOI definition.   Returns the
1045  * size in bytes of the network category bitmap on success, negative values
1046  * otherwise.
1047  *
1048  */
1049 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1050 				     const struct netlbl_lsm_secattr *secattr,
1051 				     unsigned char *net_cat,
1052 				     u32 net_cat_len)
1053 {
1054 	int iter = -1;
1055 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1056 	u32 array_cnt = 0;
1057 	u32 cat_size = 0;
1058 
1059 	/* make sure we don't overflow the 'array[]' variable */
1060 	if (net_cat_len >
1061 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1062 		return -ENOSPC;
1063 
1064 	for (;;) {
1065 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1066 		if (iter < 0)
1067 			break;
1068 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1069 		if (cat_size > net_cat_len)
1070 			return -ENOSPC;
1071 		array[array_cnt++] = iter;
1072 
1073 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1074 		if (iter < 0)
1075 			return -EFAULT;
1076 		cat_size += sizeof(u16);
1077 		if (cat_size > net_cat_len)
1078 			return -ENOSPC;
1079 		array[array_cnt++] = iter;
1080 	}
1081 
1082 	for (iter = 0; array_cnt > 0;) {
1083 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1084 		iter += 2;
1085 		array_cnt--;
1086 		if (array[array_cnt] != 0) {
1087 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1088 			iter += 2;
1089 		}
1090 	}
1091 
1092 	return cat_size;
1093 }
1094 
1095 /**
1096  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1097  * @doi_def: the DOI definition
1098  * @net_cat: the category list in network/CIPSO format
1099  * @net_cat_len: the length of the CIPSO bitmap in bytes
1100  * @secattr: the security attributes
1101  *
1102  * Description:
1103  * Perform a label mapping to translate a CIPSO category list to the correct
1104  * local MLS category bitmap using the given DOI definition.  Returns zero on
1105  * success, negative values on failure.
1106  *
1107  */
1108 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1109 				     const unsigned char *net_cat,
1110 				     u32 net_cat_len,
1111 				     struct netlbl_lsm_secattr *secattr)
1112 {
1113 	int ret_val;
1114 	u32 net_iter;
1115 	u16 cat_low;
1116 	u16 cat_high;
1117 
1118 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1119 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1120 		if ((net_iter + 4) <= net_cat_len)
1121 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1122 		else
1123 			cat_low = 0;
1124 
1125 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1126 					       cat_low,
1127 					       cat_high,
1128 					       GFP_ATOMIC);
1129 		if (ret_val != 0)
1130 			return ret_val;
1131 	}
1132 
1133 	return 0;
1134 }
1135 
1136 /*
1137  * Protocol Handling Functions
1138  */
1139 
1140 /**
1141  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1142  * @doi_def: the DOI definition
1143  * @len: the total tag length in bytes, not including this header
1144  * @buf: the CIPSO option buffer
1145  *
1146  * Description:
1147  * Write a CIPSO header into the beginning of @buffer.
1148  *
1149  */
1150 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1151 				unsigned char *buf,
1152 				u32 len)
1153 {
1154 	buf[0] = IPOPT_CIPSO;
1155 	buf[1] = CIPSO_V4_HDR_LEN + len;
1156 	put_unaligned_be32(doi_def->doi, &buf[2]);
1157 }
1158 
1159 /**
1160  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1161  * @doi_def: the DOI definition
1162  * @secattr: the security attributes
1163  * @buffer: the option buffer
1164  * @buffer_len: length of buffer in bytes
1165  *
1166  * Description:
1167  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1168  * actual buffer length may be larger than the indicated size due to
1169  * translation between host and network category bitmaps.  Returns the size of
1170  * the tag on success, negative values on failure.
1171  *
1172  */
1173 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1174 			       const struct netlbl_lsm_secattr *secattr,
1175 			       unsigned char *buffer,
1176 			       u32 buffer_len)
1177 {
1178 	int ret_val;
1179 	u32 tag_len;
1180 	u32 level;
1181 
1182 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1183 		return -EPERM;
1184 
1185 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1186 					secattr->attr.mls.lvl,
1187 					&level);
1188 	if (ret_val != 0)
1189 		return ret_val;
1190 
1191 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1192 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1193 						    secattr,
1194 						    &buffer[4],
1195 						    buffer_len - 4);
1196 		if (ret_val < 0)
1197 			return ret_val;
1198 
1199 		/* This will send packets using the "optimized" format when
1200 		 * possible as specified in  section 3.4.2.6 of the
1201 		 * CIPSO draft. */
1202 		if (cipso_v4_rbm_optfmt && ret_val > 0 && ret_val <= 10)
1203 			tag_len = 14;
1204 		else
1205 			tag_len = 4 + ret_val;
1206 	} else
1207 		tag_len = 4;
1208 
1209 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1210 	buffer[1] = tag_len;
1211 	buffer[3] = level;
1212 
1213 	return tag_len;
1214 }
1215 
1216 /**
1217  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1218  * @doi_def: the DOI definition
1219  * @tag: the CIPSO tag
1220  * @secattr: the security attributes
1221  *
1222  * Description:
1223  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1224  * attributes in @secattr.  Return zero on success, negatives values on
1225  * failure.
1226  *
1227  */
1228 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1229 				 const unsigned char *tag,
1230 				 struct netlbl_lsm_secattr *secattr)
1231 {
1232 	int ret_val;
1233 	u8 tag_len = tag[1];
1234 	u32 level;
1235 
1236 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1237 	if (ret_val != 0)
1238 		return ret_val;
1239 	secattr->attr.mls.lvl = level;
1240 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1241 
1242 	if (tag_len > 4) {
1243 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1244 						    &tag[4],
1245 						    tag_len - 4,
1246 						    secattr);
1247 		if (ret_val != 0) {
1248 			netlbl_catmap_free(secattr->attr.mls.cat);
1249 			return ret_val;
1250 		}
1251 
1252 		if (secattr->attr.mls.cat)
1253 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1254 	}
1255 
1256 	return 0;
1257 }
1258 
1259 /**
1260  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1261  * @doi_def: the DOI definition
1262  * @secattr: the security attributes
1263  * @buffer: the option buffer
1264  * @buffer_len: length of buffer in bytes
1265  *
1266  * Description:
1267  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1268  * size of the tag on success, negative values on failure.
1269  *
1270  */
1271 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1272 				const struct netlbl_lsm_secattr *secattr,
1273 				unsigned char *buffer,
1274 				u32 buffer_len)
1275 {
1276 	int ret_val;
1277 	u32 tag_len;
1278 	u32 level;
1279 
1280 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1281 		return -EPERM;
1282 
1283 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1284 					secattr->attr.mls.lvl,
1285 					&level);
1286 	if (ret_val != 0)
1287 		return ret_val;
1288 
1289 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1290 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1291 						     secattr,
1292 						     &buffer[4],
1293 						     buffer_len - 4);
1294 		if (ret_val < 0)
1295 			return ret_val;
1296 
1297 		tag_len = 4 + ret_val;
1298 	} else
1299 		tag_len = 4;
1300 
1301 	buffer[0] = CIPSO_V4_TAG_ENUM;
1302 	buffer[1] = tag_len;
1303 	buffer[3] = level;
1304 
1305 	return tag_len;
1306 }
1307 
1308 /**
1309  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1310  * @doi_def: the DOI definition
1311  * @tag: the CIPSO tag
1312  * @secattr: the security attributes
1313  *
1314  * Description:
1315  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1316  * attributes in @secattr.  Return zero on success, negatives values on
1317  * failure.
1318  *
1319  */
1320 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1321 				  const unsigned char *tag,
1322 				  struct netlbl_lsm_secattr *secattr)
1323 {
1324 	int ret_val;
1325 	u8 tag_len = tag[1];
1326 	u32 level;
1327 
1328 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1329 	if (ret_val != 0)
1330 		return ret_val;
1331 	secattr->attr.mls.lvl = level;
1332 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1333 
1334 	if (tag_len > 4) {
1335 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1336 						     &tag[4],
1337 						     tag_len - 4,
1338 						     secattr);
1339 		if (ret_val != 0) {
1340 			netlbl_catmap_free(secattr->attr.mls.cat);
1341 			return ret_val;
1342 		}
1343 
1344 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1345 	}
1346 
1347 	return 0;
1348 }
1349 
1350 /**
1351  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1352  * @doi_def: the DOI definition
1353  * @secattr: the security attributes
1354  * @buffer: the option buffer
1355  * @buffer_len: length of buffer in bytes
1356  *
1357  * Description:
1358  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1359  * size of the tag on success, negative values on failure.
1360  *
1361  */
1362 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1363 			       const struct netlbl_lsm_secattr *secattr,
1364 			       unsigned char *buffer,
1365 			       u32 buffer_len)
1366 {
1367 	int ret_val;
1368 	u32 tag_len;
1369 	u32 level;
1370 
1371 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1372 		return -EPERM;
1373 
1374 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1375 					secattr->attr.mls.lvl,
1376 					&level);
1377 	if (ret_val != 0)
1378 		return ret_val;
1379 
1380 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1381 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1382 						    secattr,
1383 						    &buffer[4],
1384 						    buffer_len - 4);
1385 		if (ret_val < 0)
1386 			return ret_val;
1387 
1388 		tag_len = 4 + ret_val;
1389 	} else
1390 		tag_len = 4;
1391 
1392 	buffer[0] = CIPSO_V4_TAG_RANGE;
1393 	buffer[1] = tag_len;
1394 	buffer[3] = level;
1395 
1396 	return tag_len;
1397 }
1398 
1399 /**
1400  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1401  * @doi_def: the DOI definition
1402  * @tag: the CIPSO tag
1403  * @secattr: the security attributes
1404  *
1405  * Description:
1406  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1407  * in @secattr.  Return zero on success, negatives values on failure.
1408  *
1409  */
1410 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1411 				 const unsigned char *tag,
1412 				 struct netlbl_lsm_secattr *secattr)
1413 {
1414 	int ret_val;
1415 	u8 tag_len = tag[1];
1416 	u32 level;
1417 
1418 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1419 	if (ret_val != 0)
1420 		return ret_val;
1421 	secattr->attr.mls.lvl = level;
1422 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1423 
1424 	if (tag_len > 4) {
1425 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1426 						    &tag[4],
1427 						    tag_len - 4,
1428 						    secattr);
1429 		if (ret_val != 0) {
1430 			netlbl_catmap_free(secattr->attr.mls.cat);
1431 			return ret_val;
1432 		}
1433 
1434 		if (secattr->attr.mls.cat)
1435 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1436 	}
1437 
1438 	return 0;
1439 }
1440 
1441 /**
1442  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1443  * @doi_def: the DOI definition
1444  * @secattr: the security attributes
1445  * @buffer: the option buffer
1446  * @buffer_len: length of buffer in bytes
1447  *
1448  * Description:
1449  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1450  * on success, negative values on failure.
1451  *
1452  */
1453 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1454 			       const struct netlbl_lsm_secattr *secattr,
1455 			       unsigned char *buffer,
1456 			       u32 buffer_len)
1457 {
1458 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1459 		return -EPERM;
1460 
1461 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1462 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1463 	*(u32 *)&buffer[2] = secattr->attr.secid;
1464 
1465 	return CIPSO_V4_TAG_LOC_BLEN;
1466 }
1467 
1468 /**
1469  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1470  * @doi_def: the DOI definition
1471  * @tag: the CIPSO tag
1472  * @secattr: the security attributes
1473  *
1474  * Description:
1475  * Parse a CIPSO local tag and return the security attributes in @secattr.
1476  * Return zero on success, negatives values on failure.
1477  *
1478  */
1479 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1480 				 const unsigned char *tag,
1481 				 struct netlbl_lsm_secattr *secattr)
1482 {
1483 	secattr->attr.secid = *(u32 *)&tag[2];
1484 	secattr->flags |= NETLBL_SECATTR_SECID;
1485 
1486 	return 0;
1487 }
1488 
1489 /**
1490  * cipso_v4_optptr - Find the CIPSO option in the packet
1491  * @skb: the packet
1492  *
1493  * Description:
1494  * Parse the packet's IP header looking for a CIPSO option.  Returns a pointer
1495  * to the start of the CIPSO option on success, NULL if one is not found.
1496  *
1497  */
1498 unsigned char *cipso_v4_optptr(const struct sk_buff *skb)
1499 {
1500 	const struct iphdr *iph = ip_hdr(skb);
1501 	unsigned char *optptr = (unsigned char *)&(ip_hdr(skb)[1]);
1502 	int optlen;
1503 	int taglen;
1504 
1505 	for (optlen = iph->ihl*4 - sizeof(struct iphdr); optlen > 1; ) {
1506 		switch (optptr[0]) {
1507 		case IPOPT_END:
1508 			return NULL;
1509 		case IPOPT_NOOP:
1510 			taglen = 1;
1511 			break;
1512 		default:
1513 			taglen = optptr[1];
1514 		}
1515 		if (!taglen || taglen > optlen)
1516 			return NULL;
1517 		if (optptr[0] == IPOPT_CIPSO)
1518 			return optptr;
1519 
1520 		optlen -= taglen;
1521 		optptr += taglen;
1522 	}
1523 
1524 	return NULL;
1525 }
1526 
1527 /**
1528  * cipso_v4_validate - Validate a CIPSO option
1529  * @skb: the packet
1530  * @option: the start of the option, on error it is set to point to the error
1531  *
1532  * Description:
1533  * This routine is called to validate a CIPSO option, it checks all of the
1534  * fields to ensure that they are at least valid, see the draft snippet below
1535  * for details.  If the option is valid then a zero value is returned and
1536  * the value of @option is unchanged.  If the option is invalid then a
1537  * non-zero value is returned and @option is adjusted to point to the
1538  * offending portion of the option.  From the IETF draft ...
1539  *
1540  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1541  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1542  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1543  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1544  *   that is unrecognized."
1545  *
1546  */
1547 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1548 {
1549 	unsigned char *opt = *option;
1550 	unsigned char *tag;
1551 	unsigned char opt_iter;
1552 	unsigned char err_offset = 0;
1553 	u8 opt_len;
1554 	u8 tag_len;
1555 	struct cipso_v4_doi *doi_def = NULL;
1556 	u32 tag_iter;
1557 
1558 	/* caller already checks for length values that are too large */
1559 	opt_len = opt[1];
1560 	if (opt_len < 8) {
1561 		err_offset = 1;
1562 		goto validate_return;
1563 	}
1564 
1565 	rcu_read_lock();
1566 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1567 	if (!doi_def) {
1568 		err_offset = 2;
1569 		goto validate_return_locked;
1570 	}
1571 
1572 	opt_iter = CIPSO_V4_HDR_LEN;
1573 	tag = opt + opt_iter;
1574 	while (opt_iter < opt_len) {
1575 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1576 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1577 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1578 				err_offset = opt_iter;
1579 				goto validate_return_locked;
1580 			}
1581 
1582 		if (opt_iter + 1 == opt_len) {
1583 			err_offset = opt_iter;
1584 			goto validate_return_locked;
1585 		}
1586 		tag_len = tag[1];
1587 		if (tag_len > (opt_len - opt_iter)) {
1588 			err_offset = opt_iter + 1;
1589 			goto validate_return_locked;
1590 		}
1591 
1592 		switch (tag[0]) {
1593 		case CIPSO_V4_TAG_RBITMAP:
1594 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1595 				err_offset = opt_iter + 1;
1596 				goto validate_return_locked;
1597 			}
1598 
1599 			/* We are already going to do all the verification
1600 			 * necessary at the socket layer so from our point of
1601 			 * view it is safe to turn these checks off (and less
1602 			 * work), however, the CIPSO draft says we should do
1603 			 * all the CIPSO validations here but it doesn't
1604 			 * really specify _exactly_ what we need to validate
1605 			 * ... so, just make it a sysctl tunable. */
1606 			if (cipso_v4_rbm_strictvalid) {
1607 				if (cipso_v4_map_lvl_valid(doi_def,
1608 							   tag[3]) < 0) {
1609 					err_offset = opt_iter + 3;
1610 					goto validate_return_locked;
1611 				}
1612 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1613 				    cipso_v4_map_cat_rbm_valid(doi_def,
1614 							    &tag[4],
1615 							    tag_len - 4) < 0) {
1616 					err_offset = opt_iter + 4;
1617 					goto validate_return_locked;
1618 				}
1619 			}
1620 			break;
1621 		case CIPSO_V4_TAG_ENUM:
1622 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1623 				err_offset = opt_iter + 1;
1624 				goto validate_return_locked;
1625 			}
1626 
1627 			if (cipso_v4_map_lvl_valid(doi_def,
1628 						   tag[3]) < 0) {
1629 				err_offset = opt_iter + 3;
1630 				goto validate_return_locked;
1631 			}
1632 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1633 			    cipso_v4_map_cat_enum_valid(doi_def,
1634 							&tag[4],
1635 							tag_len - 4) < 0) {
1636 				err_offset = opt_iter + 4;
1637 				goto validate_return_locked;
1638 			}
1639 			break;
1640 		case CIPSO_V4_TAG_RANGE:
1641 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1642 				err_offset = opt_iter + 1;
1643 				goto validate_return_locked;
1644 			}
1645 
1646 			if (cipso_v4_map_lvl_valid(doi_def,
1647 						   tag[3]) < 0) {
1648 				err_offset = opt_iter + 3;
1649 				goto validate_return_locked;
1650 			}
1651 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1652 			    cipso_v4_map_cat_rng_valid(doi_def,
1653 						       &tag[4],
1654 						       tag_len - 4) < 0) {
1655 				err_offset = opt_iter + 4;
1656 				goto validate_return_locked;
1657 			}
1658 			break;
1659 		case CIPSO_V4_TAG_LOCAL:
1660 			/* This is a non-standard tag that we only allow for
1661 			 * local connections, so if the incoming interface is
1662 			 * not the loopback device drop the packet. Further,
1663 			 * there is no legitimate reason for setting this from
1664 			 * userspace so reject it if skb is NULL. */
1665 			if (!skb || !(skb->dev->flags & IFF_LOOPBACK)) {
1666 				err_offset = opt_iter;
1667 				goto validate_return_locked;
1668 			}
1669 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1670 				err_offset = opt_iter + 1;
1671 				goto validate_return_locked;
1672 			}
1673 			break;
1674 		default:
1675 			err_offset = opt_iter;
1676 			goto validate_return_locked;
1677 		}
1678 
1679 		tag += tag_len;
1680 		opt_iter += tag_len;
1681 	}
1682 
1683 validate_return_locked:
1684 	rcu_read_unlock();
1685 validate_return:
1686 	*option = opt + err_offset;
1687 	return err_offset;
1688 }
1689 
1690 /**
1691  * cipso_v4_error - Send the correct response for a bad packet
1692  * @skb: the packet
1693  * @error: the error code
1694  * @gateway: CIPSO gateway flag
1695  *
1696  * Description:
1697  * Based on the error code given in @error, send an ICMP error message back to
1698  * the originating host.  From the IETF draft ...
1699  *
1700  *  "If the contents of the CIPSO [option] are valid but the security label is
1701  *   outside of the configured host or port label range, the datagram is
1702  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1703  *   returned.  The code field of the ICMP is set to 'communication with
1704  *   destination network administratively prohibited' (code 9) or to
1705  *   'communication with destination host administratively prohibited'
1706  *   (code 10).  The value of the code is dependent on whether the originator
1707  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1708  *   recipient of the ICMP message MUST be able to handle either value.  The
1709  *   same procedure is performed if a CIPSO [option] can not be added to an
1710  *   IP packet because it is too large to fit in the IP options area."
1711  *
1712  *  "If the error is triggered by receipt of an ICMP message, the message is
1713  *   discarded and no response is permitted (consistent with general ICMP
1714  *   processing rules)."
1715  *
1716  */
1717 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1718 {
1719 	unsigned char optbuf[sizeof(struct ip_options) + 40];
1720 	struct ip_options *opt = (struct ip_options *)optbuf;
1721 	int res;
1722 
1723 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1724 		return;
1725 
1726 	/*
1727 	 * We might be called above the IP layer,
1728 	 * so we can not use icmp_send and IPCB here.
1729 	 */
1730 
1731 	memset(opt, 0, sizeof(struct ip_options));
1732 	opt->optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr);
1733 	rcu_read_lock();
1734 	res = __ip_options_compile(dev_net(skb->dev), opt, skb, NULL);
1735 	rcu_read_unlock();
1736 
1737 	if (res)
1738 		return;
1739 
1740 	if (gateway)
1741 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0, opt);
1742 	else
1743 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0, opt);
1744 }
1745 
1746 /**
1747  * cipso_v4_genopt - Generate a CIPSO option
1748  * @buf: the option buffer
1749  * @buf_len: the size of opt_buf
1750  * @doi_def: the CIPSO DOI to use
1751  * @secattr: the security attributes
1752  *
1753  * Description:
1754  * Generate a CIPSO option using the DOI definition and security attributes
1755  * passed to the function.  Returns the length of the option on success and
1756  * negative values on failure.
1757  *
1758  */
1759 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1760 			   const struct cipso_v4_doi *doi_def,
1761 			   const struct netlbl_lsm_secattr *secattr)
1762 {
1763 	int ret_val;
1764 	u32 iter;
1765 
1766 	if (buf_len <= CIPSO_V4_HDR_LEN)
1767 		return -ENOSPC;
1768 
1769 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1770 	 * really a good assumption to make but since we only support the MAC
1771 	 * tags right now it is a safe assumption. */
1772 	iter = 0;
1773 	do {
1774 		memset(buf, 0, buf_len);
1775 		switch (doi_def->tags[iter]) {
1776 		case CIPSO_V4_TAG_RBITMAP:
1777 			ret_val = cipso_v4_gentag_rbm(doi_def,
1778 						   secattr,
1779 						   &buf[CIPSO_V4_HDR_LEN],
1780 						   buf_len - CIPSO_V4_HDR_LEN);
1781 			break;
1782 		case CIPSO_V4_TAG_ENUM:
1783 			ret_val = cipso_v4_gentag_enum(doi_def,
1784 						   secattr,
1785 						   &buf[CIPSO_V4_HDR_LEN],
1786 						   buf_len - CIPSO_V4_HDR_LEN);
1787 			break;
1788 		case CIPSO_V4_TAG_RANGE:
1789 			ret_val = cipso_v4_gentag_rng(doi_def,
1790 						   secattr,
1791 						   &buf[CIPSO_V4_HDR_LEN],
1792 						   buf_len - CIPSO_V4_HDR_LEN);
1793 			break;
1794 		case CIPSO_V4_TAG_LOCAL:
1795 			ret_val = cipso_v4_gentag_loc(doi_def,
1796 						   secattr,
1797 						   &buf[CIPSO_V4_HDR_LEN],
1798 						   buf_len - CIPSO_V4_HDR_LEN);
1799 			break;
1800 		default:
1801 			return -EPERM;
1802 		}
1803 
1804 		iter++;
1805 	} while (ret_val < 0 &&
1806 		 iter < CIPSO_V4_TAG_MAXCNT &&
1807 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1808 	if (ret_val < 0)
1809 		return ret_val;
1810 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1811 	return CIPSO_V4_HDR_LEN + ret_val;
1812 }
1813 
1814 /**
1815  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1816  * @sk: the socket
1817  * @doi_def: the CIPSO DOI to use
1818  * @secattr: the specific security attributes of the socket
1819  *
1820  * Description:
1821  * Set the CIPSO option on the given socket using the DOI definition and
1822  * security attributes passed to the function.  This function requires
1823  * exclusive access to @sk, which means it either needs to be in the
1824  * process of being created or locked.  Returns zero on success and negative
1825  * values on failure.
1826  *
1827  */
1828 int cipso_v4_sock_setattr(struct sock *sk,
1829 			  const struct cipso_v4_doi *doi_def,
1830 			  const struct netlbl_lsm_secattr *secattr)
1831 {
1832 	int ret_val = -EPERM;
1833 	unsigned char *buf = NULL;
1834 	u32 buf_len;
1835 	u32 opt_len;
1836 	struct ip_options_rcu *old, *opt = NULL;
1837 	struct inet_sock *sk_inet;
1838 	struct inet_connection_sock *sk_conn;
1839 
1840 	/* In the case of sock_create_lite(), the sock->sk field is not
1841 	 * defined yet but it is not a problem as the only users of these
1842 	 * "lite" PF_INET sockets are functions which do an accept() call
1843 	 * afterwards so we will label the socket as part of the accept(). */
1844 	if (!sk)
1845 		return 0;
1846 
1847 	/* We allocate the maximum CIPSO option size here so we are probably
1848 	 * being a little wasteful, but it makes our life _much_ easier later
1849 	 * on and after all we are only talking about 40 bytes. */
1850 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1851 	buf = kmalloc(buf_len, GFP_ATOMIC);
1852 	if (!buf) {
1853 		ret_val = -ENOMEM;
1854 		goto socket_setattr_failure;
1855 	}
1856 
1857 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1858 	if (ret_val < 0)
1859 		goto socket_setattr_failure;
1860 	buf_len = ret_val;
1861 
1862 	/* We can't use ip_options_get() directly because it makes a call to
1863 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1864 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1865 	 * set the IPOPT_CIPSO option. */
1866 	opt_len = (buf_len + 3) & ~3;
1867 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1868 	if (!opt) {
1869 		ret_val = -ENOMEM;
1870 		goto socket_setattr_failure;
1871 	}
1872 	memcpy(opt->opt.__data, buf, buf_len);
1873 	opt->opt.optlen = opt_len;
1874 	opt->opt.cipso = sizeof(struct iphdr);
1875 	kfree(buf);
1876 	buf = NULL;
1877 
1878 	sk_inet = inet_sk(sk);
1879 
1880 	old = rcu_dereference_protected(sk_inet->inet_opt,
1881 					lockdep_sock_is_held(sk));
1882 	if (sk_inet->is_icsk) {
1883 		sk_conn = inet_csk(sk);
1884 		if (old)
1885 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1886 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1887 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1888 	}
1889 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1890 	if (old)
1891 		kfree_rcu(old, rcu);
1892 
1893 	return 0;
1894 
1895 socket_setattr_failure:
1896 	kfree(buf);
1897 	kfree(opt);
1898 	return ret_val;
1899 }
1900 
1901 /**
1902  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1903  * @req: the connection request socket
1904  * @doi_def: the CIPSO DOI to use
1905  * @secattr: the specific security attributes of the socket
1906  *
1907  * Description:
1908  * Set the CIPSO option on the given socket using the DOI definition and
1909  * security attributes passed to the function.  Returns zero on success and
1910  * negative values on failure.
1911  *
1912  */
1913 int cipso_v4_req_setattr(struct request_sock *req,
1914 			 const struct cipso_v4_doi *doi_def,
1915 			 const struct netlbl_lsm_secattr *secattr)
1916 {
1917 	int ret_val = -EPERM;
1918 	unsigned char *buf = NULL;
1919 	u32 buf_len;
1920 	u32 opt_len;
1921 	struct ip_options_rcu *opt = NULL;
1922 	struct inet_request_sock *req_inet;
1923 
1924 	/* We allocate the maximum CIPSO option size here so we are probably
1925 	 * being a little wasteful, but it makes our life _much_ easier later
1926 	 * on and after all we are only talking about 40 bytes. */
1927 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1928 	buf = kmalloc(buf_len, GFP_ATOMIC);
1929 	if (!buf) {
1930 		ret_val = -ENOMEM;
1931 		goto req_setattr_failure;
1932 	}
1933 
1934 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1935 	if (ret_val < 0)
1936 		goto req_setattr_failure;
1937 	buf_len = ret_val;
1938 
1939 	/* We can't use ip_options_get() directly because it makes a call to
1940 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1941 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1942 	 * set the IPOPT_CIPSO option. */
1943 	opt_len = (buf_len + 3) & ~3;
1944 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1945 	if (!opt) {
1946 		ret_val = -ENOMEM;
1947 		goto req_setattr_failure;
1948 	}
1949 	memcpy(opt->opt.__data, buf, buf_len);
1950 	opt->opt.optlen = opt_len;
1951 	opt->opt.cipso = sizeof(struct iphdr);
1952 	kfree(buf);
1953 	buf = NULL;
1954 
1955 	req_inet = inet_rsk(req);
1956 	opt = xchg((__force struct ip_options_rcu **)&req_inet->ireq_opt, opt);
1957 	if (opt)
1958 		kfree_rcu(opt, rcu);
1959 
1960 	return 0;
1961 
1962 req_setattr_failure:
1963 	kfree(buf);
1964 	kfree(opt);
1965 	return ret_val;
1966 }
1967 
1968 /**
1969  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
1970  * @opt_ptr: IP option pointer
1971  *
1972  * Description:
1973  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
1974  * adjustments to the IP option structure.  Returns zero on success, negative
1975  * values on failure.
1976  *
1977  */
1978 static int cipso_v4_delopt(struct ip_options_rcu __rcu **opt_ptr)
1979 {
1980 	struct ip_options_rcu *opt = rcu_dereference_protected(*opt_ptr, 1);
1981 	int hdr_delta = 0;
1982 
1983 	if (!opt || opt->opt.cipso == 0)
1984 		return 0;
1985 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
1986 		u8 cipso_len;
1987 		u8 cipso_off;
1988 		unsigned char *cipso_ptr;
1989 		int iter;
1990 		int optlen_new;
1991 
1992 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
1993 		cipso_ptr = &opt->opt.__data[cipso_off];
1994 		cipso_len = cipso_ptr[1];
1995 
1996 		if (opt->opt.srr > opt->opt.cipso)
1997 			opt->opt.srr -= cipso_len;
1998 		if (opt->opt.rr > opt->opt.cipso)
1999 			opt->opt.rr -= cipso_len;
2000 		if (opt->opt.ts > opt->opt.cipso)
2001 			opt->opt.ts -= cipso_len;
2002 		if (opt->opt.router_alert > opt->opt.cipso)
2003 			opt->opt.router_alert -= cipso_len;
2004 		opt->opt.cipso = 0;
2005 
2006 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2007 			opt->opt.optlen - cipso_off - cipso_len);
2008 
2009 		/* determining the new total option length is tricky because of
2010 		 * the padding necessary, the only thing i can think to do at
2011 		 * this point is walk the options one-by-one, skipping the
2012 		 * padding at the end to determine the actual option size and
2013 		 * from there we can determine the new total option length */
2014 		iter = 0;
2015 		optlen_new = 0;
2016 		while (iter < opt->opt.optlen)
2017 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2018 				iter += opt->opt.__data[iter + 1];
2019 				optlen_new = iter;
2020 			} else
2021 				iter++;
2022 		hdr_delta = opt->opt.optlen;
2023 		opt->opt.optlen = (optlen_new + 3) & ~3;
2024 		hdr_delta -= opt->opt.optlen;
2025 	} else {
2026 		/* only the cipso option was present on the socket so we can
2027 		 * remove the entire option struct */
2028 		*opt_ptr = NULL;
2029 		hdr_delta = opt->opt.optlen;
2030 		kfree_rcu(opt, rcu);
2031 	}
2032 
2033 	return hdr_delta;
2034 }
2035 
2036 /**
2037  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2038  * @sk: the socket
2039  *
2040  * Description:
2041  * Removes the CIPSO option from a socket, if present.
2042  *
2043  */
2044 void cipso_v4_sock_delattr(struct sock *sk)
2045 {
2046 	struct inet_sock *sk_inet;
2047 	int hdr_delta;
2048 
2049 	sk_inet = inet_sk(sk);
2050 
2051 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2052 	if (sk_inet->is_icsk && hdr_delta > 0) {
2053 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2054 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2055 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2056 	}
2057 }
2058 
2059 /**
2060  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2061  * @req: the request socket
2062  *
2063  * Description:
2064  * Removes the CIPSO option from a request socket, if present.
2065  *
2066  */
2067 void cipso_v4_req_delattr(struct request_sock *req)
2068 {
2069 	cipso_v4_delopt(&inet_rsk(req)->ireq_opt);
2070 }
2071 
2072 /**
2073  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2074  * @cipso: the CIPSO v4 option
2075  * @secattr: the security attributes
2076  *
2077  * Description:
2078  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2079  * on success and negative values on failure.
2080  *
2081  */
2082 int cipso_v4_getattr(const unsigned char *cipso,
2083 		     struct netlbl_lsm_secattr *secattr)
2084 {
2085 	int ret_val = -ENOMSG;
2086 	u32 doi;
2087 	struct cipso_v4_doi *doi_def;
2088 
2089 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2090 		return 0;
2091 
2092 	doi = get_unaligned_be32(&cipso[2]);
2093 	rcu_read_lock();
2094 	doi_def = cipso_v4_doi_search(doi);
2095 	if (!doi_def)
2096 		goto getattr_return;
2097 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2098 	 * really a good assumption to make but since we only support the MAC
2099 	 * tags right now it is a safe assumption. */
2100 	switch (cipso[6]) {
2101 	case CIPSO_V4_TAG_RBITMAP:
2102 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2103 		break;
2104 	case CIPSO_V4_TAG_ENUM:
2105 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2106 		break;
2107 	case CIPSO_V4_TAG_RANGE:
2108 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2109 		break;
2110 	case CIPSO_V4_TAG_LOCAL:
2111 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2112 		break;
2113 	}
2114 	if (ret_val == 0)
2115 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2116 
2117 getattr_return:
2118 	rcu_read_unlock();
2119 	return ret_val;
2120 }
2121 
2122 /**
2123  * cipso_v4_sock_getattr - Get the security attributes from a sock
2124  * @sk: the sock
2125  * @secattr: the security attributes
2126  *
2127  * Description:
2128  * Query @sk to see if there is a CIPSO option attached to the sock and if
2129  * there is return the CIPSO security attributes in @secattr.  This function
2130  * requires that @sk be locked, or privately held, but it does not do any
2131  * locking itself.  Returns zero on success and negative values on failure.
2132  *
2133  */
2134 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2135 {
2136 	struct ip_options_rcu *opt;
2137 	int res = -ENOMSG;
2138 
2139 	rcu_read_lock();
2140 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2141 	if (opt && opt->opt.cipso)
2142 		res = cipso_v4_getattr(opt->opt.__data +
2143 						opt->opt.cipso -
2144 						sizeof(struct iphdr),
2145 				       secattr);
2146 	rcu_read_unlock();
2147 	return res;
2148 }
2149 
2150 /**
2151  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2152  * @skb: the packet
2153  * @doi_def: the DOI structure
2154  * @secattr: the security attributes
2155  *
2156  * Description:
2157  * Set the CIPSO option on the given packet based on the security attributes.
2158  * Returns a pointer to the IP header on success and NULL on failure.
2159  *
2160  */
2161 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2162 			    const struct cipso_v4_doi *doi_def,
2163 			    const struct netlbl_lsm_secattr *secattr)
2164 {
2165 	int ret_val;
2166 	struct iphdr *iph;
2167 	struct ip_options *opt = &IPCB(skb)->opt;
2168 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2169 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2170 	u32 opt_len;
2171 	int len_delta;
2172 
2173 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2174 	if (ret_val < 0)
2175 		return ret_val;
2176 	buf_len = ret_val;
2177 	opt_len = (buf_len + 3) & ~3;
2178 
2179 	/* we overwrite any existing options to ensure that we have enough
2180 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2181 	 * that the security label is applied to the packet - we do the same
2182 	 * thing when using the socket options and it hasn't caused a problem,
2183 	 * if we need to we can always revisit this choice later */
2184 
2185 	len_delta = opt_len - opt->optlen;
2186 	/* if we don't ensure enough headroom we could panic on the skb_push()
2187 	 * call below so make sure we have enough, we are also "mangling" the
2188 	 * packet so we should probably do a copy-on-write call anyway */
2189 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2190 	if (ret_val < 0)
2191 		return ret_val;
2192 
2193 	if (len_delta > 0) {
2194 		/* we assume that the header + opt->optlen have already been
2195 		 * "pushed" in ip_options_build() or similar */
2196 		iph = ip_hdr(skb);
2197 		skb_push(skb, len_delta);
2198 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2199 		skb_reset_network_header(skb);
2200 		iph = ip_hdr(skb);
2201 	} else if (len_delta < 0) {
2202 		iph = ip_hdr(skb);
2203 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2204 	} else
2205 		iph = ip_hdr(skb);
2206 
2207 	if (opt->optlen > 0)
2208 		memset(opt, 0, sizeof(*opt));
2209 	opt->optlen = opt_len;
2210 	opt->cipso = sizeof(struct iphdr);
2211 	opt->is_changed = 1;
2212 
2213 	/* we have to do the following because we are being called from a
2214 	 * netfilter hook which means the packet already has had the header
2215 	 * fields populated and the checksum calculated - yes this means we
2216 	 * are doing more work than needed but we do it to keep the core
2217 	 * stack clean and tidy */
2218 	memcpy(iph + 1, buf, buf_len);
2219 	if (opt_len > buf_len)
2220 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2221 	if (len_delta != 0) {
2222 		iph->ihl = 5 + (opt_len >> 2);
2223 		iph->tot_len = htons(skb->len);
2224 	}
2225 	ip_send_check(iph);
2226 
2227 	return 0;
2228 }
2229 
2230 /**
2231  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2232  * @skb: the packet
2233  *
2234  * Description:
2235  * Removes any and all CIPSO options from the given packet.  Returns zero on
2236  * success, negative values on failure.
2237  *
2238  */
2239 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2240 {
2241 	int ret_val;
2242 	struct iphdr *iph;
2243 	struct ip_options *opt = &IPCB(skb)->opt;
2244 	unsigned char *cipso_ptr;
2245 
2246 	if (opt->cipso == 0)
2247 		return 0;
2248 
2249 	/* since we are changing the packet we should make a copy */
2250 	ret_val = skb_cow(skb, skb_headroom(skb));
2251 	if (ret_val < 0)
2252 		return ret_val;
2253 
2254 	/* the easiest thing to do is just replace the cipso option with noop
2255 	 * options since we don't change the size of the packet, although we
2256 	 * still need to recalculate the checksum */
2257 
2258 	iph = ip_hdr(skb);
2259 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2260 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2261 	opt->cipso = 0;
2262 	opt->is_changed = 1;
2263 
2264 	ip_send_check(iph);
2265 
2266 	return 0;
2267 }
2268 
2269 /*
2270  * Setup Functions
2271  */
2272 
2273 /**
2274  * cipso_v4_init - Initialize the CIPSO module
2275  *
2276  * Description:
2277  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2278  * and negative values on failure.
2279  *
2280  */
2281 static int __init cipso_v4_init(void)
2282 {
2283 	int ret_val;
2284 
2285 	ret_val = cipso_v4_cache_init();
2286 	if (ret_val != 0)
2287 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2288 		      ret_val);
2289 
2290 	return 0;
2291 }
2292 
2293 subsys_initcall(cipso_v4_init);
2294