xref: /openbmc/linux/net/ipv4/cipso_ipv4.c (revision d47a97bd)
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 			if (net_spot == -2)
869 				return -EFAULT;
870 			return 0;
871 		}
872 
873 		switch (doi_def->type) {
874 		case CIPSO_V4_MAP_PASS:
875 			host_spot = net_spot;
876 			break;
877 		case CIPSO_V4_MAP_TRANS:
878 			if (net_spot >= net_cat_size)
879 				return -EPERM;
880 			host_spot = net_cat_array[net_spot];
881 			if (host_spot >= CIPSO_V4_INV_CAT)
882 				return -EPERM;
883 			break;
884 		}
885 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
886 						       host_spot,
887 						       GFP_ATOMIC);
888 		if (ret_val != 0)
889 			return ret_val;
890 	}
891 
892 	return -EINVAL;
893 }
894 
895 /**
896  * cipso_v4_map_cat_enum_valid - Checks to see if the categories are valid
897  * @doi_def: the DOI definition
898  * @enumcat: category list
899  * @enumcat_len: length of the category list in bytes
900  *
901  * Description:
902  * Checks the given categories against the given DOI definition and returns a
903  * negative value if any of the categories do not have a valid mapping and a
904  * zero value if all of the categories are valid.
905  *
906  */
907 static int cipso_v4_map_cat_enum_valid(const struct cipso_v4_doi *doi_def,
908 				       const unsigned char *enumcat,
909 				       u32 enumcat_len)
910 {
911 	u16 cat;
912 	int cat_prev = -1;
913 	u32 iter;
914 
915 	if (doi_def->type != CIPSO_V4_MAP_PASS || enumcat_len & 0x01)
916 		return -EFAULT;
917 
918 	for (iter = 0; iter < enumcat_len; iter += 2) {
919 		cat = get_unaligned_be16(&enumcat[iter]);
920 		if (cat <= cat_prev)
921 			return -EFAULT;
922 		cat_prev = cat;
923 	}
924 
925 	return 0;
926 }
927 
928 /**
929  * cipso_v4_map_cat_enum_hton - Perform a category mapping from host to network
930  * @doi_def: the DOI definition
931  * @secattr: the security attributes
932  * @net_cat: the zero'd out category list in network/CIPSO format
933  * @net_cat_len: the length of the CIPSO category list in bytes
934  *
935  * Description:
936  * Perform a label mapping to translate a local MLS category bitmap to the
937  * correct CIPSO category list using the given DOI definition.   Returns the
938  * size in bytes of the network category bitmap on success, negative values
939  * otherwise.
940  *
941  */
942 static int cipso_v4_map_cat_enum_hton(const struct cipso_v4_doi *doi_def,
943 				      const struct netlbl_lsm_secattr *secattr,
944 				      unsigned char *net_cat,
945 				      u32 net_cat_len)
946 {
947 	int cat = -1;
948 	u32 cat_iter = 0;
949 
950 	for (;;) {
951 		cat = netlbl_catmap_walk(secattr->attr.mls.cat, cat + 1);
952 		if (cat < 0)
953 			break;
954 		if ((cat_iter + 2) > net_cat_len)
955 			return -ENOSPC;
956 
957 		*((__be16 *)&net_cat[cat_iter]) = htons(cat);
958 		cat_iter += 2;
959 	}
960 
961 	return cat_iter;
962 }
963 
964 /**
965  * cipso_v4_map_cat_enum_ntoh - Perform a category mapping from network to host
966  * @doi_def: the DOI definition
967  * @net_cat: the category list in network/CIPSO format
968  * @net_cat_len: the length of the CIPSO bitmap in bytes
969  * @secattr: the security attributes
970  *
971  * Description:
972  * Perform a label mapping to translate a CIPSO category list to the correct
973  * local MLS category bitmap using the given DOI definition.  Returns zero on
974  * success, negative values on failure.
975  *
976  */
977 static int cipso_v4_map_cat_enum_ntoh(const struct cipso_v4_doi *doi_def,
978 				      const unsigned char *net_cat,
979 				      u32 net_cat_len,
980 				      struct netlbl_lsm_secattr *secattr)
981 {
982 	int ret_val;
983 	u32 iter;
984 
985 	for (iter = 0; iter < net_cat_len; iter += 2) {
986 		ret_val = netlbl_catmap_setbit(&secattr->attr.mls.cat,
987 					     get_unaligned_be16(&net_cat[iter]),
988 					     GFP_ATOMIC);
989 		if (ret_val != 0)
990 			return ret_val;
991 	}
992 
993 	return 0;
994 }
995 
996 /**
997  * cipso_v4_map_cat_rng_valid - Checks to see if the categories are valid
998  * @doi_def: the DOI definition
999  * @rngcat: category list
1000  * @rngcat_len: length of the category list in bytes
1001  *
1002  * Description:
1003  * Checks the given categories against the given DOI definition and returns a
1004  * negative value if any of the categories do not have a valid mapping and a
1005  * zero value if all of the categories are valid.
1006  *
1007  */
1008 static int cipso_v4_map_cat_rng_valid(const struct cipso_v4_doi *doi_def,
1009 				      const unsigned char *rngcat,
1010 				      u32 rngcat_len)
1011 {
1012 	u16 cat_high;
1013 	u16 cat_low;
1014 	u32 cat_prev = CIPSO_V4_MAX_REM_CATS + 1;
1015 	u32 iter;
1016 
1017 	if (doi_def->type != CIPSO_V4_MAP_PASS || rngcat_len & 0x01)
1018 		return -EFAULT;
1019 
1020 	for (iter = 0; iter < rngcat_len; iter += 4) {
1021 		cat_high = get_unaligned_be16(&rngcat[iter]);
1022 		if ((iter + 4) <= rngcat_len)
1023 			cat_low = get_unaligned_be16(&rngcat[iter + 2]);
1024 		else
1025 			cat_low = 0;
1026 
1027 		if (cat_high > cat_prev)
1028 			return -EFAULT;
1029 
1030 		cat_prev = cat_low;
1031 	}
1032 
1033 	return 0;
1034 }
1035 
1036 /**
1037  * cipso_v4_map_cat_rng_hton - Perform a category mapping from host to network
1038  * @doi_def: the DOI definition
1039  * @secattr: the security attributes
1040  * @net_cat: the zero'd out category list in network/CIPSO format
1041  * @net_cat_len: the length of the CIPSO category list in bytes
1042  *
1043  * Description:
1044  * Perform a label mapping to translate a local MLS category bitmap to the
1045  * correct CIPSO category list using the given DOI definition.   Returns the
1046  * size in bytes of the network category bitmap on success, negative values
1047  * otherwise.
1048  *
1049  */
1050 static int cipso_v4_map_cat_rng_hton(const struct cipso_v4_doi *doi_def,
1051 				     const struct netlbl_lsm_secattr *secattr,
1052 				     unsigned char *net_cat,
1053 				     u32 net_cat_len)
1054 {
1055 	int iter = -1;
1056 	u16 array[CIPSO_V4_TAG_RNG_CAT_MAX * 2];
1057 	u32 array_cnt = 0;
1058 	u32 cat_size = 0;
1059 
1060 	/* make sure we don't overflow the 'array[]' variable */
1061 	if (net_cat_len >
1062 	    (CIPSO_V4_OPT_LEN_MAX - CIPSO_V4_HDR_LEN - CIPSO_V4_TAG_RNG_BLEN))
1063 		return -ENOSPC;
1064 
1065 	for (;;) {
1066 		iter = netlbl_catmap_walk(secattr->attr.mls.cat, iter + 1);
1067 		if (iter < 0)
1068 			break;
1069 		cat_size += (iter == 0 ? 0 : sizeof(u16));
1070 		if (cat_size > net_cat_len)
1071 			return -ENOSPC;
1072 		array[array_cnt++] = iter;
1073 
1074 		iter = netlbl_catmap_walkrng(secattr->attr.mls.cat, iter);
1075 		if (iter < 0)
1076 			return -EFAULT;
1077 		cat_size += sizeof(u16);
1078 		if (cat_size > net_cat_len)
1079 			return -ENOSPC;
1080 		array[array_cnt++] = iter;
1081 	}
1082 
1083 	for (iter = 0; array_cnt > 0;) {
1084 		*((__be16 *)&net_cat[iter]) = htons(array[--array_cnt]);
1085 		iter += 2;
1086 		array_cnt--;
1087 		if (array[array_cnt] != 0) {
1088 			*((__be16 *)&net_cat[iter]) = htons(array[array_cnt]);
1089 			iter += 2;
1090 		}
1091 	}
1092 
1093 	return cat_size;
1094 }
1095 
1096 /**
1097  * cipso_v4_map_cat_rng_ntoh - Perform a category mapping from network to host
1098  * @doi_def: the DOI definition
1099  * @net_cat: the category list in network/CIPSO format
1100  * @net_cat_len: the length of the CIPSO bitmap in bytes
1101  * @secattr: the security attributes
1102  *
1103  * Description:
1104  * Perform a label mapping to translate a CIPSO category list to the correct
1105  * local MLS category bitmap using the given DOI definition.  Returns zero on
1106  * success, negative values on failure.
1107  *
1108  */
1109 static int cipso_v4_map_cat_rng_ntoh(const struct cipso_v4_doi *doi_def,
1110 				     const unsigned char *net_cat,
1111 				     u32 net_cat_len,
1112 				     struct netlbl_lsm_secattr *secattr)
1113 {
1114 	int ret_val;
1115 	u32 net_iter;
1116 	u16 cat_low;
1117 	u16 cat_high;
1118 
1119 	for (net_iter = 0; net_iter < net_cat_len; net_iter += 4) {
1120 		cat_high = get_unaligned_be16(&net_cat[net_iter]);
1121 		if ((net_iter + 4) <= net_cat_len)
1122 			cat_low = get_unaligned_be16(&net_cat[net_iter + 2]);
1123 		else
1124 			cat_low = 0;
1125 
1126 		ret_val = netlbl_catmap_setrng(&secattr->attr.mls.cat,
1127 					       cat_low,
1128 					       cat_high,
1129 					       GFP_ATOMIC);
1130 		if (ret_val != 0)
1131 			return ret_val;
1132 	}
1133 
1134 	return 0;
1135 }
1136 
1137 /*
1138  * Protocol Handling Functions
1139  */
1140 
1141 /**
1142  * cipso_v4_gentag_hdr - Generate a CIPSO option header
1143  * @doi_def: the DOI definition
1144  * @len: the total tag length in bytes, not including this header
1145  * @buf: the CIPSO option buffer
1146  *
1147  * Description:
1148  * Write a CIPSO header into the beginning of @buffer.
1149  *
1150  */
1151 static void cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
1152 				unsigned char *buf,
1153 				u32 len)
1154 {
1155 	buf[0] = IPOPT_CIPSO;
1156 	buf[1] = CIPSO_V4_HDR_LEN + len;
1157 	put_unaligned_be32(doi_def->doi, &buf[2]);
1158 }
1159 
1160 /**
1161  * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
1162  * @doi_def: the DOI definition
1163  * @secattr: the security attributes
1164  * @buffer: the option buffer
1165  * @buffer_len: length of buffer in bytes
1166  *
1167  * Description:
1168  * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
1169  * actual buffer length may be larger than the indicated size due to
1170  * translation between host and network category bitmaps.  Returns the size of
1171  * the tag on success, negative values on failure.
1172  *
1173  */
1174 static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
1175 			       const struct netlbl_lsm_secattr *secattr,
1176 			       unsigned char *buffer,
1177 			       u32 buffer_len)
1178 {
1179 	int ret_val;
1180 	u32 tag_len;
1181 	u32 level;
1182 
1183 	if ((secattr->flags & NETLBL_SECATTR_MLS_LVL) == 0)
1184 		return -EPERM;
1185 
1186 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1187 					secattr->attr.mls.lvl,
1188 					&level);
1189 	if (ret_val != 0)
1190 		return ret_val;
1191 
1192 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1193 		ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
1194 						    secattr,
1195 						    &buffer[4],
1196 						    buffer_len - 4);
1197 		if (ret_val < 0)
1198 			return ret_val;
1199 
1200 		/* This will send packets using the "optimized" format when
1201 		 * possible as specified in  section 3.4.2.6 of the
1202 		 * CIPSO draft. */
1203 		if (READ_ONCE(cipso_v4_rbm_optfmt) && ret_val > 0 &&
1204 		    ret_val <= 10)
1205 			tag_len = 14;
1206 		else
1207 			tag_len = 4 + ret_val;
1208 	} else
1209 		tag_len = 4;
1210 
1211 	buffer[0] = CIPSO_V4_TAG_RBITMAP;
1212 	buffer[1] = tag_len;
1213 	buffer[3] = level;
1214 
1215 	return tag_len;
1216 }
1217 
1218 /**
1219  * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
1220  * @doi_def: the DOI definition
1221  * @tag: the CIPSO tag
1222  * @secattr: the security attributes
1223  *
1224  * Description:
1225  * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
1226  * attributes in @secattr.  Return zero on success, negatives values on
1227  * failure.
1228  *
1229  */
1230 static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
1231 				 const unsigned char *tag,
1232 				 struct netlbl_lsm_secattr *secattr)
1233 {
1234 	int ret_val;
1235 	u8 tag_len = tag[1];
1236 	u32 level;
1237 
1238 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1239 	if (ret_val != 0)
1240 		return ret_val;
1241 	secattr->attr.mls.lvl = level;
1242 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1243 
1244 	if (tag_len > 4) {
1245 		ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
1246 						    &tag[4],
1247 						    tag_len - 4,
1248 						    secattr);
1249 		if (ret_val != 0) {
1250 			netlbl_catmap_free(secattr->attr.mls.cat);
1251 			return ret_val;
1252 		}
1253 
1254 		if (secattr->attr.mls.cat)
1255 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1256 	}
1257 
1258 	return 0;
1259 }
1260 
1261 /**
1262  * cipso_v4_gentag_enum - Generate a CIPSO enumerated tag (type #2)
1263  * @doi_def: the DOI definition
1264  * @secattr: the security attributes
1265  * @buffer: the option buffer
1266  * @buffer_len: length of buffer in bytes
1267  *
1268  * Description:
1269  * Generate a CIPSO option using the enumerated tag, tag type #2.  Returns the
1270  * size of the tag on success, negative values on failure.
1271  *
1272  */
1273 static int cipso_v4_gentag_enum(const struct cipso_v4_doi *doi_def,
1274 				const struct netlbl_lsm_secattr *secattr,
1275 				unsigned char *buffer,
1276 				u32 buffer_len)
1277 {
1278 	int ret_val;
1279 	u32 tag_len;
1280 	u32 level;
1281 
1282 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1283 		return -EPERM;
1284 
1285 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1286 					secattr->attr.mls.lvl,
1287 					&level);
1288 	if (ret_val != 0)
1289 		return ret_val;
1290 
1291 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1292 		ret_val = cipso_v4_map_cat_enum_hton(doi_def,
1293 						     secattr,
1294 						     &buffer[4],
1295 						     buffer_len - 4);
1296 		if (ret_val < 0)
1297 			return ret_val;
1298 
1299 		tag_len = 4 + ret_val;
1300 	} else
1301 		tag_len = 4;
1302 
1303 	buffer[0] = CIPSO_V4_TAG_ENUM;
1304 	buffer[1] = tag_len;
1305 	buffer[3] = level;
1306 
1307 	return tag_len;
1308 }
1309 
1310 /**
1311  * cipso_v4_parsetag_enum - Parse a CIPSO enumerated tag
1312  * @doi_def: the DOI definition
1313  * @tag: the CIPSO tag
1314  * @secattr: the security attributes
1315  *
1316  * Description:
1317  * Parse a CIPSO enumerated tag (tag type #2) and return the security
1318  * attributes in @secattr.  Return zero on success, negatives values on
1319  * failure.
1320  *
1321  */
1322 static int cipso_v4_parsetag_enum(const struct cipso_v4_doi *doi_def,
1323 				  const unsigned char *tag,
1324 				  struct netlbl_lsm_secattr *secattr)
1325 {
1326 	int ret_val;
1327 	u8 tag_len = tag[1];
1328 	u32 level;
1329 
1330 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1331 	if (ret_val != 0)
1332 		return ret_val;
1333 	secattr->attr.mls.lvl = level;
1334 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1335 
1336 	if (tag_len > 4) {
1337 		ret_val = cipso_v4_map_cat_enum_ntoh(doi_def,
1338 						     &tag[4],
1339 						     tag_len - 4,
1340 						     secattr);
1341 		if (ret_val != 0) {
1342 			netlbl_catmap_free(secattr->attr.mls.cat);
1343 			return ret_val;
1344 		}
1345 
1346 		secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1347 	}
1348 
1349 	return 0;
1350 }
1351 
1352 /**
1353  * cipso_v4_gentag_rng - Generate a CIPSO ranged tag (type #5)
1354  * @doi_def: the DOI definition
1355  * @secattr: the security attributes
1356  * @buffer: the option buffer
1357  * @buffer_len: length of buffer in bytes
1358  *
1359  * Description:
1360  * Generate a CIPSO option using the ranged tag, tag type #5.  Returns the
1361  * size of the tag on success, negative values on failure.
1362  *
1363  */
1364 static int cipso_v4_gentag_rng(const struct cipso_v4_doi *doi_def,
1365 			       const struct netlbl_lsm_secattr *secattr,
1366 			       unsigned char *buffer,
1367 			       u32 buffer_len)
1368 {
1369 	int ret_val;
1370 	u32 tag_len;
1371 	u32 level;
1372 
1373 	if (!(secattr->flags & NETLBL_SECATTR_MLS_LVL))
1374 		return -EPERM;
1375 
1376 	ret_val = cipso_v4_map_lvl_hton(doi_def,
1377 					secattr->attr.mls.lvl,
1378 					&level);
1379 	if (ret_val != 0)
1380 		return ret_val;
1381 
1382 	if (secattr->flags & NETLBL_SECATTR_MLS_CAT) {
1383 		ret_val = cipso_v4_map_cat_rng_hton(doi_def,
1384 						    secattr,
1385 						    &buffer[4],
1386 						    buffer_len - 4);
1387 		if (ret_val < 0)
1388 			return ret_val;
1389 
1390 		tag_len = 4 + ret_val;
1391 	} else
1392 		tag_len = 4;
1393 
1394 	buffer[0] = CIPSO_V4_TAG_RANGE;
1395 	buffer[1] = tag_len;
1396 	buffer[3] = level;
1397 
1398 	return tag_len;
1399 }
1400 
1401 /**
1402  * cipso_v4_parsetag_rng - Parse a CIPSO ranged tag
1403  * @doi_def: the DOI definition
1404  * @tag: the CIPSO tag
1405  * @secattr: the security attributes
1406  *
1407  * Description:
1408  * Parse a CIPSO ranged tag (tag type #5) and return the security attributes
1409  * in @secattr.  Return zero on success, negatives values on failure.
1410  *
1411  */
1412 static int cipso_v4_parsetag_rng(const struct cipso_v4_doi *doi_def,
1413 				 const unsigned char *tag,
1414 				 struct netlbl_lsm_secattr *secattr)
1415 {
1416 	int ret_val;
1417 	u8 tag_len = tag[1];
1418 	u32 level;
1419 
1420 	ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
1421 	if (ret_val != 0)
1422 		return ret_val;
1423 	secattr->attr.mls.lvl = level;
1424 	secattr->flags |= NETLBL_SECATTR_MLS_LVL;
1425 
1426 	if (tag_len > 4) {
1427 		ret_val = cipso_v4_map_cat_rng_ntoh(doi_def,
1428 						    &tag[4],
1429 						    tag_len - 4,
1430 						    secattr);
1431 		if (ret_val != 0) {
1432 			netlbl_catmap_free(secattr->attr.mls.cat);
1433 			return ret_val;
1434 		}
1435 
1436 		if (secattr->attr.mls.cat)
1437 			secattr->flags |= NETLBL_SECATTR_MLS_CAT;
1438 	}
1439 
1440 	return 0;
1441 }
1442 
1443 /**
1444  * cipso_v4_gentag_loc - Generate a CIPSO local tag (non-standard)
1445  * @doi_def: the DOI definition
1446  * @secattr: the security attributes
1447  * @buffer: the option buffer
1448  * @buffer_len: length of buffer in bytes
1449  *
1450  * Description:
1451  * Generate a CIPSO option using the local tag.  Returns the size of the tag
1452  * on success, negative values on failure.
1453  *
1454  */
1455 static int cipso_v4_gentag_loc(const struct cipso_v4_doi *doi_def,
1456 			       const struct netlbl_lsm_secattr *secattr,
1457 			       unsigned char *buffer,
1458 			       u32 buffer_len)
1459 {
1460 	if (!(secattr->flags & NETLBL_SECATTR_SECID))
1461 		return -EPERM;
1462 
1463 	buffer[0] = CIPSO_V4_TAG_LOCAL;
1464 	buffer[1] = CIPSO_V4_TAG_LOC_BLEN;
1465 	*(u32 *)&buffer[2] = secattr->attr.secid;
1466 
1467 	return CIPSO_V4_TAG_LOC_BLEN;
1468 }
1469 
1470 /**
1471  * cipso_v4_parsetag_loc - Parse a CIPSO local tag
1472  * @doi_def: the DOI definition
1473  * @tag: the CIPSO tag
1474  * @secattr: the security attributes
1475  *
1476  * Description:
1477  * Parse a CIPSO local tag and return the security attributes in @secattr.
1478  * Return zero on success, negatives values on failure.
1479  *
1480  */
1481 static int cipso_v4_parsetag_loc(const struct cipso_v4_doi *doi_def,
1482 				 const unsigned char *tag,
1483 				 struct netlbl_lsm_secattr *secattr)
1484 {
1485 	secattr->attr.secid = *(u32 *)&tag[2];
1486 	secattr->flags |= NETLBL_SECATTR_SECID;
1487 
1488 	return 0;
1489 }
1490 
1491 /**
1492  * cipso_v4_optptr - Find the CIPSO option in the packet
1493  * @skb: the packet
1494  *
1495  * Description:
1496  * Parse the packet's IP header looking for a CIPSO option.  Returns a pointer
1497  * to the start of the CIPSO option on success, NULL if one is not found.
1498  *
1499  */
1500 unsigned char *cipso_v4_optptr(const struct sk_buff *skb)
1501 {
1502 	const struct iphdr *iph = ip_hdr(skb);
1503 	unsigned char *optptr = (unsigned char *)&(ip_hdr(skb)[1]);
1504 	int optlen;
1505 	int taglen;
1506 
1507 	for (optlen = iph->ihl*4 - sizeof(struct iphdr); optlen > 1; ) {
1508 		switch (optptr[0]) {
1509 		case IPOPT_END:
1510 			return NULL;
1511 		case IPOPT_NOOP:
1512 			taglen = 1;
1513 			break;
1514 		default:
1515 			taglen = optptr[1];
1516 		}
1517 		if (!taglen || taglen > optlen)
1518 			return NULL;
1519 		if (optptr[0] == IPOPT_CIPSO)
1520 			return optptr;
1521 
1522 		optlen -= taglen;
1523 		optptr += taglen;
1524 	}
1525 
1526 	return NULL;
1527 }
1528 
1529 /**
1530  * cipso_v4_validate - Validate a CIPSO option
1531  * @skb: the packet
1532  * @option: the start of the option, on error it is set to point to the error
1533  *
1534  * Description:
1535  * This routine is called to validate a CIPSO option, it checks all of the
1536  * fields to ensure that they are at least valid, see the draft snippet below
1537  * for details.  If the option is valid then a zero value is returned and
1538  * the value of @option is unchanged.  If the option is invalid then a
1539  * non-zero value is returned and @option is adjusted to point to the
1540  * offending portion of the option.  From the IETF draft ...
1541  *
1542  *  "If any field within the CIPSO options, such as the DOI identifier, is not
1543  *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
1544  *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
1545  *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
1546  *   that is unrecognized."
1547  *
1548  */
1549 int cipso_v4_validate(const struct sk_buff *skb, unsigned char **option)
1550 {
1551 	unsigned char *opt = *option;
1552 	unsigned char *tag;
1553 	unsigned char opt_iter;
1554 	unsigned char err_offset = 0;
1555 	u8 opt_len;
1556 	u8 tag_len;
1557 	struct cipso_v4_doi *doi_def = NULL;
1558 	u32 tag_iter;
1559 
1560 	/* caller already checks for length values that are too large */
1561 	opt_len = opt[1];
1562 	if (opt_len < 8) {
1563 		err_offset = 1;
1564 		goto validate_return;
1565 	}
1566 
1567 	rcu_read_lock();
1568 	doi_def = cipso_v4_doi_search(get_unaligned_be32(&opt[2]));
1569 	if (!doi_def) {
1570 		err_offset = 2;
1571 		goto validate_return_locked;
1572 	}
1573 
1574 	opt_iter = CIPSO_V4_HDR_LEN;
1575 	tag = opt + opt_iter;
1576 	while (opt_iter < opt_len) {
1577 		for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
1578 			if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
1579 			    ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
1580 				err_offset = opt_iter;
1581 				goto validate_return_locked;
1582 			}
1583 
1584 		if (opt_iter + 1 == opt_len) {
1585 			err_offset = opt_iter;
1586 			goto validate_return_locked;
1587 		}
1588 		tag_len = tag[1];
1589 		if (tag_len > (opt_len - opt_iter)) {
1590 			err_offset = opt_iter + 1;
1591 			goto validate_return_locked;
1592 		}
1593 
1594 		switch (tag[0]) {
1595 		case CIPSO_V4_TAG_RBITMAP:
1596 			if (tag_len < CIPSO_V4_TAG_RBM_BLEN) {
1597 				err_offset = opt_iter + 1;
1598 				goto validate_return_locked;
1599 			}
1600 
1601 			/* We are already going to do all the verification
1602 			 * necessary at the socket layer so from our point of
1603 			 * view it is safe to turn these checks off (and less
1604 			 * work), however, the CIPSO draft says we should do
1605 			 * all the CIPSO validations here but it doesn't
1606 			 * really specify _exactly_ what we need to validate
1607 			 * ... so, just make it a sysctl tunable. */
1608 			if (READ_ONCE(cipso_v4_rbm_strictvalid)) {
1609 				if (cipso_v4_map_lvl_valid(doi_def,
1610 							   tag[3]) < 0) {
1611 					err_offset = opt_iter + 3;
1612 					goto validate_return_locked;
1613 				}
1614 				if (tag_len > CIPSO_V4_TAG_RBM_BLEN &&
1615 				    cipso_v4_map_cat_rbm_valid(doi_def,
1616 							    &tag[4],
1617 							    tag_len - 4) < 0) {
1618 					err_offset = opt_iter + 4;
1619 					goto validate_return_locked;
1620 				}
1621 			}
1622 			break;
1623 		case CIPSO_V4_TAG_ENUM:
1624 			if (tag_len < CIPSO_V4_TAG_ENUM_BLEN) {
1625 				err_offset = opt_iter + 1;
1626 				goto validate_return_locked;
1627 			}
1628 
1629 			if (cipso_v4_map_lvl_valid(doi_def,
1630 						   tag[3]) < 0) {
1631 				err_offset = opt_iter + 3;
1632 				goto validate_return_locked;
1633 			}
1634 			if (tag_len > CIPSO_V4_TAG_ENUM_BLEN &&
1635 			    cipso_v4_map_cat_enum_valid(doi_def,
1636 							&tag[4],
1637 							tag_len - 4) < 0) {
1638 				err_offset = opt_iter + 4;
1639 				goto validate_return_locked;
1640 			}
1641 			break;
1642 		case CIPSO_V4_TAG_RANGE:
1643 			if (tag_len < CIPSO_V4_TAG_RNG_BLEN) {
1644 				err_offset = opt_iter + 1;
1645 				goto validate_return_locked;
1646 			}
1647 
1648 			if (cipso_v4_map_lvl_valid(doi_def,
1649 						   tag[3]) < 0) {
1650 				err_offset = opt_iter + 3;
1651 				goto validate_return_locked;
1652 			}
1653 			if (tag_len > CIPSO_V4_TAG_RNG_BLEN &&
1654 			    cipso_v4_map_cat_rng_valid(doi_def,
1655 						       &tag[4],
1656 						       tag_len - 4) < 0) {
1657 				err_offset = opt_iter + 4;
1658 				goto validate_return_locked;
1659 			}
1660 			break;
1661 		case CIPSO_V4_TAG_LOCAL:
1662 			/* This is a non-standard tag that we only allow for
1663 			 * local connections, so if the incoming interface is
1664 			 * not the loopback device drop the packet. Further,
1665 			 * there is no legitimate reason for setting this from
1666 			 * userspace so reject it if skb is NULL. */
1667 			if (!skb || !(skb->dev->flags & IFF_LOOPBACK)) {
1668 				err_offset = opt_iter;
1669 				goto validate_return_locked;
1670 			}
1671 			if (tag_len != CIPSO_V4_TAG_LOC_BLEN) {
1672 				err_offset = opt_iter + 1;
1673 				goto validate_return_locked;
1674 			}
1675 			break;
1676 		default:
1677 			err_offset = opt_iter;
1678 			goto validate_return_locked;
1679 		}
1680 
1681 		tag += tag_len;
1682 		opt_iter += tag_len;
1683 	}
1684 
1685 validate_return_locked:
1686 	rcu_read_unlock();
1687 validate_return:
1688 	*option = opt + err_offset;
1689 	return err_offset;
1690 }
1691 
1692 /**
1693  * cipso_v4_error - Send the correct response for a bad packet
1694  * @skb: the packet
1695  * @error: the error code
1696  * @gateway: CIPSO gateway flag
1697  *
1698  * Description:
1699  * Based on the error code given in @error, send an ICMP error message back to
1700  * the originating host.  From the IETF draft ...
1701  *
1702  *  "If the contents of the CIPSO [option] are valid but the security label is
1703  *   outside of the configured host or port label range, the datagram is
1704  *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
1705  *   returned.  The code field of the ICMP is set to 'communication with
1706  *   destination network administratively prohibited' (code 9) or to
1707  *   'communication with destination host administratively prohibited'
1708  *   (code 10).  The value of the code is dependent on whether the originator
1709  *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
1710  *   recipient of the ICMP message MUST be able to handle either value.  The
1711  *   same procedure is performed if a CIPSO [option] can not be added to an
1712  *   IP packet because it is too large to fit in the IP options area."
1713  *
1714  *  "If the error is triggered by receipt of an ICMP message, the message is
1715  *   discarded and no response is permitted (consistent with general ICMP
1716  *   processing rules)."
1717  *
1718  */
1719 void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
1720 {
1721 	unsigned char optbuf[sizeof(struct ip_options) + 40];
1722 	struct ip_options *opt = (struct ip_options *)optbuf;
1723 	int res;
1724 
1725 	if (ip_hdr(skb)->protocol == IPPROTO_ICMP || error != -EACCES)
1726 		return;
1727 
1728 	/*
1729 	 * We might be called above the IP layer,
1730 	 * so we can not use icmp_send and IPCB here.
1731 	 */
1732 
1733 	memset(opt, 0, sizeof(struct ip_options));
1734 	opt->optlen = ip_hdr(skb)->ihl*4 - sizeof(struct iphdr);
1735 	rcu_read_lock();
1736 	res = __ip_options_compile(dev_net(skb->dev), opt, skb, NULL);
1737 	rcu_read_unlock();
1738 
1739 	if (res)
1740 		return;
1741 
1742 	if (gateway)
1743 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0, opt);
1744 	else
1745 		__icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0, opt);
1746 }
1747 
1748 /**
1749  * cipso_v4_genopt - Generate a CIPSO option
1750  * @buf: the option buffer
1751  * @buf_len: the size of opt_buf
1752  * @doi_def: the CIPSO DOI to use
1753  * @secattr: the security attributes
1754  *
1755  * Description:
1756  * Generate a CIPSO option using the DOI definition and security attributes
1757  * passed to the function.  Returns the length of the option on success and
1758  * negative values on failure.
1759  *
1760  */
1761 static int cipso_v4_genopt(unsigned char *buf, u32 buf_len,
1762 			   const struct cipso_v4_doi *doi_def,
1763 			   const struct netlbl_lsm_secattr *secattr)
1764 {
1765 	int ret_val;
1766 	u32 iter;
1767 
1768 	if (buf_len <= CIPSO_V4_HDR_LEN)
1769 		return -ENOSPC;
1770 
1771 	/* XXX - This code assumes only one tag per CIPSO option which isn't
1772 	 * really a good assumption to make but since we only support the MAC
1773 	 * tags right now it is a safe assumption. */
1774 	iter = 0;
1775 	do {
1776 		memset(buf, 0, buf_len);
1777 		switch (doi_def->tags[iter]) {
1778 		case CIPSO_V4_TAG_RBITMAP:
1779 			ret_val = cipso_v4_gentag_rbm(doi_def,
1780 						   secattr,
1781 						   &buf[CIPSO_V4_HDR_LEN],
1782 						   buf_len - CIPSO_V4_HDR_LEN);
1783 			break;
1784 		case CIPSO_V4_TAG_ENUM:
1785 			ret_val = cipso_v4_gentag_enum(doi_def,
1786 						   secattr,
1787 						   &buf[CIPSO_V4_HDR_LEN],
1788 						   buf_len - CIPSO_V4_HDR_LEN);
1789 			break;
1790 		case CIPSO_V4_TAG_RANGE:
1791 			ret_val = cipso_v4_gentag_rng(doi_def,
1792 						   secattr,
1793 						   &buf[CIPSO_V4_HDR_LEN],
1794 						   buf_len - CIPSO_V4_HDR_LEN);
1795 			break;
1796 		case CIPSO_V4_TAG_LOCAL:
1797 			ret_val = cipso_v4_gentag_loc(doi_def,
1798 						   secattr,
1799 						   &buf[CIPSO_V4_HDR_LEN],
1800 						   buf_len - CIPSO_V4_HDR_LEN);
1801 			break;
1802 		default:
1803 			return -EPERM;
1804 		}
1805 
1806 		iter++;
1807 	} while (ret_val < 0 &&
1808 		 iter < CIPSO_V4_TAG_MAXCNT &&
1809 		 doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
1810 	if (ret_val < 0)
1811 		return ret_val;
1812 	cipso_v4_gentag_hdr(doi_def, buf, ret_val);
1813 	return CIPSO_V4_HDR_LEN + ret_val;
1814 }
1815 
1816 /**
1817  * cipso_v4_sock_setattr - Add a CIPSO option to a socket
1818  * @sk: the socket
1819  * @doi_def: the CIPSO DOI to use
1820  * @secattr: the specific security attributes of the socket
1821  *
1822  * Description:
1823  * Set the CIPSO option on the given socket using the DOI definition and
1824  * security attributes passed to the function.  This function requires
1825  * exclusive access to @sk, which means it either needs to be in the
1826  * process of being created or locked.  Returns zero on success and negative
1827  * values on failure.
1828  *
1829  */
1830 int cipso_v4_sock_setattr(struct sock *sk,
1831 			  const struct cipso_v4_doi *doi_def,
1832 			  const struct netlbl_lsm_secattr *secattr)
1833 {
1834 	int ret_val = -EPERM;
1835 	unsigned char *buf = NULL;
1836 	u32 buf_len;
1837 	u32 opt_len;
1838 	struct ip_options_rcu *old, *opt = NULL;
1839 	struct inet_sock *sk_inet;
1840 	struct inet_connection_sock *sk_conn;
1841 
1842 	/* In the case of sock_create_lite(), the sock->sk field is not
1843 	 * defined yet but it is not a problem as the only users of these
1844 	 * "lite" PF_INET sockets are functions which do an accept() call
1845 	 * afterwards so we will label the socket as part of the accept(). */
1846 	if (!sk)
1847 		return 0;
1848 
1849 	/* We allocate the maximum CIPSO option size here so we are probably
1850 	 * being a little wasteful, but it makes our life _much_ easier later
1851 	 * on and after all we are only talking about 40 bytes. */
1852 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1853 	buf = kmalloc(buf_len, GFP_ATOMIC);
1854 	if (!buf) {
1855 		ret_val = -ENOMEM;
1856 		goto socket_setattr_failure;
1857 	}
1858 
1859 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1860 	if (ret_val < 0)
1861 		goto socket_setattr_failure;
1862 	buf_len = ret_val;
1863 
1864 	/* We can't use ip_options_get() directly because it makes a call to
1865 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1866 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1867 	 * set the IPOPT_CIPSO option. */
1868 	opt_len = (buf_len + 3) & ~3;
1869 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1870 	if (!opt) {
1871 		ret_val = -ENOMEM;
1872 		goto socket_setattr_failure;
1873 	}
1874 	memcpy(opt->opt.__data, buf, buf_len);
1875 	opt->opt.optlen = opt_len;
1876 	opt->opt.cipso = sizeof(struct iphdr);
1877 	kfree(buf);
1878 	buf = NULL;
1879 
1880 	sk_inet = inet_sk(sk);
1881 
1882 	old = rcu_dereference_protected(sk_inet->inet_opt,
1883 					lockdep_sock_is_held(sk));
1884 	if (sk_inet->is_icsk) {
1885 		sk_conn = inet_csk(sk);
1886 		if (old)
1887 			sk_conn->icsk_ext_hdr_len -= old->opt.optlen;
1888 		sk_conn->icsk_ext_hdr_len += opt->opt.optlen;
1889 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
1890 	}
1891 	rcu_assign_pointer(sk_inet->inet_opt, opt);
1892 	if (old)
1893 		kfree_rcu(old, rcu);
1894 
1895 	return 0;
1896 
1897 socket_setattr_failure:
1898 	kfree(buf);
1899 	kfree(opt);
1900 	return ret_val;
1901 }
1902 
1903 /**
1904  * cipso_v4_req_setattr - Add a CIPSO option to a connection request socket
1905  * @req: the connection request socket
1906  * @doi_def: the CIPSO DOI to use
1907  * @secattr: the specific security attributes of the socket
1908  *
1909  * Description:
1910  * Set the CIPSO option on the given socket using the DOI definition and
1911  * security attributes passed to the function.  Returns zero on success and
1912  * negative values on failure.
1913  *
1914  */
1915 int cipso_v4_req_setattr(struct request_sock *req,
1916 			 const struct cipso_v4_doi *doi_def,
1917 			 const struct netlbl_lsm_secattr *secattr)
1918 {
1919 	int ret_val = -EPERM;
1920 	unsigned char *buf = NULL;
1921 	u32 buf_len;
1922 	u32 opt_len;
1923 	struct ip_options_rcu *opt = NULL;
1924 	struct inet_request_sock *req_inet;
1925 
1926 	/* We allocate the maximum CIPSO option size here so we are probably
1927 	 * being a little wasteful, but it makes our life _much_ easier later
1928 	 * on and after all we are only talking about 40 bytes. */
1929 	buf_len = CIPSO_V4_OPT_LEN_MAX;
1930 	buf = kmalloc(buf_len, GFP_ATOMIC);
1931 	if (!buf) {
1932 		ret_val = -ENOMEM;
1933 		goto req_setattr_failure;
1934 	}
1935 
1936 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
1937 	if (ret_val < 0)
1938 		goto req_setattr_failure;
1939 	buf_len = ret_val;
1940 
1941 	/* We can't use ip_options_get() directly because it makes a call to
1942 	 * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
1943 	 * we won't always have CAP_NET_RAW even though we _always_ want to
1944 	 * set the IPOPT_CIPSO option. */
1945 	opt_len = (buf_len + 3) & ~3;
1946 	opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
1947 	if (!opt) {
1948 		ret_val = -ENOMEM;
1949 		goto req_setattr_failure;
1950 	}
1951 	memcpy(opt->opt.__data, buf, buf_len);
1952 	opt->opt.optlen = opt_len;
1953 	opt->opt.cipso = sizeof(struct iphdr);
1954 	kfree(buf);
1955 	buf = NULL;
1956 
1957 	req_inet = inet_rsk(req);
1958 	opt = xchg((__force struct ip_options_rcu **)&req_inet->ireq_opt, opt);
1959 	if (opt)
1960 		kfree_rcu(opt, rcu);
1961 
1962 	return 0;
1963 
1964 req_setattr_failure:
1965 	kfree(buf);
1966 	kfree(opt);
1967 	return ret_val;
1968 }
1969 
1970 /**
1971  * cipso_v4_delopt - Delete the CIPSO option from a set of IP options
1972  * @opt_ptr: IP option pointer
1973  *
1974  * Description:
1975  * Deletes the CIPSO IP option from a set of IP options and makes the necessary
1976  * adjustments to the IP option structure.  Returns zero on success, negative
1977  * values on failure.
1978  *
1979  */
1980 static int cipso_v4_delopt(struct ip_options_rcu __rcu **opt_ptr)
1981 {
1982 	struct ip_options_rcu *opt = rcu_dereference_protected(*opt_ptr, 1);
1983 	int hdr_delta = 0;
1984 
1985 	if (!opt || opt->opt.cipso == 0)
1986 		return 0;
1987 	if (opt->opt.srr || opt->opt.rr || opt->opt.ts || opt->opt.router_alert) {
1988 		u8 cipso_len;
1989 		u8 cipso_off;
1990 		unsigned char *cipso_ptr;
1991 		int iter;
1992 		int optlen_new;
1993 
1994 		cipso_off = opt->opt.cipso - sizeof(struct iphdr);
1995 		cipso_ptr = &opt->opt.__data[cipso_off];
1996 		cipso_len = cipso_ptr[1];
1997 
1998 		if (opt->opt.srr > opt->opt.cipso)
1999 			opt->opt.srr -= cipso_len;
2000 		if (opt->opt.rr > opt->opt.cipso)
2001 			opt->opt.rr -= cipso_len;
2002 		if (opt->opt.ts > opt->opt.cipso)
2003 			opt->opt.ts -= cipso_len;
2004 		if (opt->opt.router_alert > opt->opt.cipso)
2005 			opt->opt.router_alert -= cipso_len;
2006 		opt->opt.cipso = 0;
2007 
2008 		memmove(cipso_ptr, cipso_ptr + cipso_len,
2009 			opt->opt.optlen - cipso_off - cipso_len);
2010 
2011 		/* determining the new total option length is tricky because of
2012 		 * the padding necessary, the only thing i can think to do at
2013 		 * this point is walk the options one-by-one, skipping the
2014 		 * padding at the end to determine the actual option size and
2015 		 * from there we can determine the new total option length */
2016 		iter = 0;
2017 		optlen_new = 0;
2018 		while (iter < opt->opt.optlen)
2019 			if (opt->opt.__data[iter] != IPOPT_NOP) {
2020 				iter += opt->opt.__data[iter + 1];
2021 				optlen_new = iter;
2022 			} else
2023 				iter++;
2024 		hdr_delta = opt->opt.optlen;
2025 		opt->opt.optlen = (optlen_new + 3) & ~3;
2026 		hdr_delta -= opt->opt.optlen;
2027 	} else {
2028 		/* only the cipso option was present on the socket so we can
2029 		 * remove the entire option struct */
2030 		*opt_ptr = NULL;
2031 		hdr_delta = opt->opt.optlen;
2032 		kfree_rcu(opt, rcu);
2033 	}
2034 
2035 	return hdr_delta;
2036 }
2037 
2038 /**
2039  * cipso_v4_sock_delattr - Delete the CIPSO option from a socket
2040  * @sk: the socket
2041  *
2042  * Description:
2043  * Removes the CIPSO option from a socket, if present.
2044  *
2045  */
2046 void cipso_v4_sock_delattr(struct sock *sk)
2047 {
2048 	struct inet_sock *sk_inet;
2049 	int hdr_delta;
2050 
2051 	sk_inet = inet_sk(sk);
2052 
2053 	hdr_delta = cipso_v4_delopt(&sk_inet->inet_opt);
2054 	if (sk_inet->is_icsk && hdr_delta > 0) {
2055 		struct inet_connection_sock *sk_conn = inet_csk(sk);
2056 		sk_conn->icsk_ext_hdr_len -= hdr_delta;
2057 		sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
2058 	}
2059 }
2060 
2061 /**
2062  * cipso_v4_req_delattr - Delete the CIPSO option from a request socket
2063  * @req: the request socket
2064  *
2065  * Description:
2066  * Removes the CIPSO option from a request socket, if present.
2067  *
2068  */
2069 void cipso_v4_req_delattr(struct request_sock *req)
2070 {
2071 	cipso_v4_delopt(&inet_rsk(req)->ireq_opt);
2072 }
2073 
2074 /**
2075  * cipso_v4_getattr - Helper function for the cipso_v4_*_getattr functions
2076  * @cipso: the CIPSO v4 option
2077  * @secattr: the security attributes
2078  *
2079  * Description:
2080  * Inspect @cipso and return the security attributes in @secattr.  Returns zero
2081  * on success and negative values on failure.
2082  *
2083  */
2084 int cipso_v4_getattr(const unsigned char *cipso,
2085 		     struct netlbl_lsm_secattr *secattr)
2086 {
2087 	int ret_val = -ENOMSG;
2088 	u32 doi;
2089 	struct cipso_v4_doi *doi_def;
2090 
2091 	if (cipso_v4_cache_check(cipso, cipso[1], secattr) == 0)
2092 		return 0;
2093 
2094 	doi = get_unaligned_be32(&cipso[2]);
2095 	rcu_read_lock();
2096 	doi_def = cipso_v4_doi_search(doi);
2097 	if (!doi_def)
2098 		goto getattr_return;
2099 	/* XXX - This code assumes only one tag per CIPSO option which isn't
2100 	 * really a good assumption to make but since we only support the MAC
2101 	 * tags right now it is a safe assumption. */
2102 	switch (cipso[6]) {
2103 	case CIPSO_V4_TAG_RBITMAP:
2104 		ret_val = cipso_v4_parsetag_rbm(doi_def, &cipso[6], secattr);
2105 		break;
2106 	case CIPSO_V4_TAG_ENUM:
2107 		ret_val = cipso_v4_parsetag_enum(doi_def, &cipso[6], secattr);
2108 		break;
2109 	case CIPSO_V4_TAG_RANGE:
2110 		ret_val = cipso_v4_parsetag_rng(doi_def, &cipso[6], secattr);
2111 		break;
2112 	case CIPSO_V4_TAG_LOCAL:
2113 		ret_val = cipso_v4_parsetag_loc(doi_def, &cipso[6], secattr);
2114 		break;
2115 	}
2116 	if (ret_val == 0)
2117 		secattr->type = NETLBL_NLTYPE_CIPSOV4;
2118 
2119 getattr_return:
2120 	rcu_read_unlock();
2121 	return ret_val;
2122 }
2123 
2124 /**
2125  * cipso_v4_sock_getattr - Get the security attributes from a sock
2126  * @sk: the sock
2127  * @secattr: the security attributes
2128  *
2129  * Description:
2130  * Query @sk to see if there is a CIPSO option attached to the sock and if
2131  * there is return the CIPSO security attributes in @secattr.  This function
2132  * requires that @sk be locked, or privately held, but it does not do any
2133  * locking itself.  Returns zero on success and negative values on failure.
2134  *
2135  */
2136 int cipso_v4_sock_getattr(struct sock *sk, struct netlbl_lsm_secattr *secattr)
2137 {
2138 	struct ip_options_rcu *opt;
2139 	int res = -ENOMSG;
2140 
2141 	rcu_read_lock();
2142 	opt = rcu_dereference(inet_sk(sk)->inet_opt);
2143 	if (opt && opt->opt.cipso)
2144 		res = cipso_v4_getattr(opt->opt.__data +
2145 						opt->opt.cipso -
2146 						sizeof(struct iphdr),
2147 				       secattr);
2148 	rcu_read_unlock();
2149 	return res;
2150 }
2151 
2152 /**
2153  * cipso_v4_skbuff_setattr - Set the CIPSO option on a packet
2154  * @skb: the packet
2155  * @doi_def: the DOI structure
2156  * @secattr: the security attributes
2157  *
2158  * Description:
2159  * Set the CIPSO option on the given packet based on the security attributes.
2160  * Returns a pointer to the IP header on success and NULL on failure.
2161  *
2162  */
2163 int cipso_v4_skbuff_setattr(struct sk_buff *skb,
2164 			    const struct cipso_v4_doi *doi_def,
2165 			    const struct netlbl_lsm_secattr *secattr)
2166 {
2167 	int ret_val;
2168 	struct iphdr *iph;
2169 	struct ip_options *opt = &IPCB(skb)->opt;
2170 	unsigned char buf[CIPSO_V4_OPT_LEN_MAX];
2171 	u32 buf_len = CIPSO_V4_OPT_LEN_MAX;
2172 	u32 opt_len;
2173 	int len_delta;
2174 
2175 	ret_val = cipso_v4_genopt(buf, buf_len, doi_def, secattr);
2176 	if (ret_val < 0)
2177 		return ret_val;
2178 	buf_len = ret_val;
2179 	opt_len = (buf_len + 3) & ~3;
2180 
2181 	/* we overwrite any existing options to ensure that we have enough
2182 	 * room for the CIPSO option, the reason is that we _need_ to guarantee
2183 	 * that the security label is applied to the packet - we do the same
2184 	 * thing when using the socket options and it hasn't caused a problem,
2185 	 * if we need to we can always revisit this choice later */
2186 
2187 	len_delta = opt_len - opt->optlen;
2188 	/* if we don't ensure enough headroom we could panic on the skb_push()
2189 	 * call below so make sure we have enough, we are also "mangling" the
2190 	 * packet so we should probably do a copy-on-write call anyway */
2191 	ret_val = skb_cow(skb, skb_headroom(skb) + len_delta);
2192 	if (ret_val < 0)
2193 		return ret_val;
2194 
2195 	if (len_delta > 0) {
2196 		/* we assume that the header + opt->optlen have already been
2197 		 * "pushed" in ip_options_build() or similar */
2198 		iph = ip_hdr(skb);
2199 		skb_push(skb, len_delta);
2200 		memmove((char *)iph - len_delta, iph, iph->ihl << 2);
2201 		skb_reset_network_header(skb);
2202 		iph = ip_hdr(skb);
2203 	} else if (len_delta < 0) {
2204 		iph = ip_hdr(skb);
2205 		memset(iph + 1, IPOPT_NOP, opt->optlen);
2206 	} else
2207 		iph = ip_hdr(skb);
2208 
2209 	if (opt->optlen > 0)
2210 		memset(opt, 0, sizeof(*opt));
2211 	opt->optlen = opt_len;
2212 	opt->cipso = sizeof(struct iphdr);
2213 	opt->is_changed = 1;
2214 
2215 	/* we have to do the following because we are being called from a
2216 	 * netfilter hook which means the packet already has had the header
2217 	 * fields populated and the checksum calculated - yes this means we
2218 	 * are doing more work than needed but we do it to keep the core
2219 	 * stack clean and tidy */
2220 	memcpy(iph + 1, buf, buf_len);
2221 	if (opt_len > buf_len)
2222 		memset((char *)(iph + 1) + buf_len, 0, opt_len - buf_len);
2223 	if (len_delta != 0) {
2224 		iph->ihl = 5 + (opt_len >> 2);
2225 		iph_set_totlen(iph, skb->len);
2226 	}
2227 	ip_send_check(iph);
2228 
2229 	return 0;
2230 }
2231 
2232 /**
2233  * cipso_v4_skbuff_delattr - Delete any CIPSO options from a packet
2234  * @skb: the packet
2235  *
2236  * Description:
2237  * Removes any and all CIPSO options from the given packet.  Returns zero on
2238  * success, negative values on failure.
2239  *
2240  */
2241 int cipso_v4_skbuff_delattr(struct sk_buff *skb)
2242 {
2243 	int ret_val;
2244 	struct iphdr *iph;
2245 	struct ip_options *opt = &IPCB(skb)->opt;
2246 	unsigned char *cipso_ptr;
2247 
2248 	if (opt->cipso == 0)
2249 		return 0;
2250 
2251 	/* since we are changing the packet we should make a copy */
2252 	ret_val = skb_cow(skb, skb_headroom(skb));
2253 	if (ret_val < 0)
2254 		return ret_val;
2255 
2256 	/* the easiest thing to do is just replace the cipso option with noop
2257 	 * options since we don't change the size of the packet, although we
2258 	 * still need to recalculate the checksum */
2259 
2260 	iph = ip_hdr(skb);
2261 	cipso_ptr = (unsigned char *)iph + opt->cipso;
2262 	memset(cipso_ptr, IPOPT_NOOP, cipso_ptr[1]);
2263 	opt->cipso = 0;
2264 	opt->is_changed = 1;
2265 
2266 	ip_send_check(iph);
2267 
2268 	return 0;
2269 }
2270 
2271 /*
2272  * Setup Functions
2273  */
2274 
2275 /**
2276  * cipso_v4_init - Initialize the CIPSO module
2277  *
2278  * Description:
2279  * Initialize the CIPSO module and prepare it for use.  Returns zero on success
2280  * and negative values on failure.
2281  *
2282  */
2283 static int __init cipso_v4_init(void)
2284 {
2285 	int ret_val;
2286 
2287 	ret_val = cipso_v4_cache_init();
2288 	if (ret_val != 0)
2289 		panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
2290 		      ret_val);
2291 
2292 	return 0;
2293 }
2294 
2295 subsys_initcall(cipso_v4_init);
2296