[NetLabel]: CIPSOv4 engine
authorPaul Moore <paul.moore@hp.com>
Thu, 3 Aug 2006 23:48:06 +0000 (16:48 -0700)
committerDavid S. Miller <davem@sunset.davemloft.net>
Fri, 22 Sep 2006 21:53:33 +0000 (14:53 -0700)
Add support for the Commercial IP Security Option (CIPSO) to the IPv4
network stack.  CIPSO has become a de-facto standard for
trusted/labeled networking amongst existing Trusted Operating Systems
such as Trusted Solaris, HP-UX CMW, etc.  This implementation is
designed to be used with the NetLabel subsystem to provide explicit
packet labeling to LSM developers.

The CIPSO/IPv4 packet labeling works by the LSM calling a NetLabel API
function which attaches a CIPSO label (IPv4 option) to a given socket;
this in turn attaches the CIPSO label to every packet leaving the
socket without any extra processing on the outbound side.  On the
inbound side the individual packet's sk_buff is examined through a
call to a NetLabel API function to determine if a CIPSO/IPv4 label is
present and if so the security attributes of the CIPSO label are
returned to the caller of the NetLabel API function.

Signed-off-by: Paul Moore <paul.moore@hp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
include/linux/sysctl.h
net/ipv4/Makefile
net/ipv4/cipso_ipv4.c [new file with mode: 0644]
net/ipv4/sysctl_net_ipv4.c

index e4b1a4d4dcf36968ca133b0b6fd4730a6e626b92..af61d9235409015f2bbef97b169c13c8e4381041 100644 (file)
@@ -411,6 +411,10 @@ enum
        NET_IPV4_TCP_WORKAROUND_SIGNED_WINDOWS=115,
        NET_TCP_DMA_COPYBREAK=116,
        NET_TCP_SLOW_START_AFTER_IDLE=117,
+       NET_CIPSOV4_CACHE_ENABLE=118,
+       NET_CIPSOV4_CACHE_BUCKET_SIZE=119,
+       NET_CIPSOV4_RBM_OPTFMT=120,
+       NET_CIPSOV4_RBM_STRICTVALID=121,
 };
 
 enum {
index 4878fc5be85fec93bd2463b7d7fe872139afdd17..f66049e28aebba92a946b92bb0339b35e15b5aee 100644 (file)
@@ -47,6 +47,7 @@ obj-$(CONFIG_TCP_CONG_VEGAS) += tcp_vegas.o
 obj-$(CONFIG_TCP_CONG_VENO) += tcp_veno.o
 obj-$(CONFIG_TCP_CONG_SCALABLE) += tcp_scalable.o
 obj-$(CONFIG_TCP_CONG_LP) += tcp_lp.o
+obj-$(CONFIG_NETLABEL) += cipso_ipv4.o
 
 obj-$(CONFIG_XFRM) += xfrm4_policy.o xfrm4_state.o xfrm4_input.o \
                      xfrm4_output.o
diff --git a/net/ipv4/cipso_ipv4.c b/net/ipv4/cipso_ipv4.c
new file mode 100644 (file)
index 0000000..b82a101
--- /dev/null
@@ -0,0 +1,1607 @@
+/*
+ * CIPSO - Commercial IP Security Option
+ *
+ * This is an implementation of the CIPSO 2.2 protocol as specified in
+ * draft-ietf-cipso-ipsecurity-01.txt with additional tag types as found in
+ * FIPS-188, copies of both documents can be found in the Documentation
+ * directory.  While CIPSO never became a full IETF RFC standard many vendors
+ * have chosen to adopt the protocol and over the years it has become a
+ * de-facto standard for labeled networking.
+ *
+ * Author: Paul Moore <paul.moore@hp.com>
+ *
+ */
+
+/*
+ * (c) Copyright Hewlett-Packard Development Company, L.P., 2006
+ *
+ * This program is free software;  you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY;  without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See
+ * the GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program;  if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ *
+ */
+
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/rcupdate.h>
+#include <linux/list.h>
+#include <linux/spinlock.h>
+#include <linux/string.h>
+#include <linux/jhash.h>
+#include <net/ip.h>
+#include <net/icmp.h>
+#include <net/tcp.h>
+#include <net/netlabel.h>
+#include <net/cipso_ipv4.h>
+#include <asm/bug.h>
+
+struct cipso_v4_domhsh_entry {
+       char *domain;
+       u32 valid;
+       struct list_head list;
+       struct rcu_head rcu;
+};
+
+/* List of available DOI definitions */
+/* XXX - Updates should be minimal so having a single lock for the
+ * cipso_v4_doi_list and the cipso_v4_doi_list->dom_list should be
+ * okay. */
+/* XXX - This currently assumes a minimal number of different DOIs in use,
+ * if in practice there are a lot of different DOIs this list should
+ * probably be turned into a hash table or something similar so we
+ * can do quick lookups. */
+DEFINE_SPINLOCK(cipso_v4_doi_list_lock);
+static struct list_head cipso_v4_doi_list = LIST_HEAD_INIT(cipso_v4_doi_list);
+
+/* Label mapping cache */
+int cipso_v4_cache_enabled = 1;
+int cipso_v4_cache_bucketsize = 10;
+#define CIPSO_V4_CACHE_BUCKETBITS     7
+#define CIPSO_V4_CACHE_BUCKETS        (1 << CIPSO_V4_CACHE_BUCKETBITS)
+#define CIPSO_V4_CACHE_REORDERLIMIT   10
+struct cipso_v4_map_cache_bkt {
+       spinlock_t lock;
+       u32 size;
+       struct list_head list;
+};
+struct cipso_v4_map_cache_entry {
+       u32 hash;
+       unsigned char *key;
+       size_t key_len;
+
+       struct netlbl_lsm_cache lsm_data;
+
+       u32 activity;
+       struct list_head list;
+};
+static struct cipso_v4_map_cache_bkt *cipso_v4_cache = NULL;
+
+/* Restricted bitmap (tag #1) flags */
+int cipso_v4_rbm_optfmt = 0;
+int cipso_v4_rbm_strictvalid = 1;
+
+/*
+ * Helper Functions
+ */
+
+/**
+ * cipso_v4_bitmap_walk - Walk a bitmap looking for a bit
+ * @bitmap: the bitmap
+ * @bitmap_len: length in bits
+ * @offset: starting offset
+ * @state: if non-zero, look for a set (1) bit else look for a cleared (0) bit
+ *
+ * Description:
+ * Starting at @offset, walk the bitmap from left to right until either the
+ * desired bit is found or we reach the end.  Return the bit offset, -1 if
+ * not found, or -2 if error.
+ */
+static int cipso_v4_bitmap_walk(const unsigned char *bitmap,
+                               u32 bitmap_len,
+                               u32 offset,
+                               u8 state)
+{
+       u32 bit_spot;
+       u32 byte_offset;
+       unsigned char bitmask;
+       unsigned char byte;
+
+       /* gcc always rounds to zero when doing integer division */
+       byte_offset = offset / 8;
+       byte = bitmap[byte_offset];
+       bit_spot = offset;
+       bitmask = 0x80 >> (offset % 8);
+
+       while (bit_spot < bitmap_len) {
+               if ((state && (byte & bitmask) == bitmask) ||
+                   (state == 0 && (byte & bitmask) == 0))
+                       return bit_spot;
+
+               bit_spot++;
+               bitmask >>= 1;
+               if (bitmask == 0) {
+                       byte = bitmap[++byte_offset];
+                       bitmask = 0x80;
+               }
+       }
+
+       return -1;
+}
+
+/**
+ * cipso_v4_bitmap_setbit - Sets a single bit in a bitmap
+ * @bitmap: the bitmap
+ * @bit: the bit
+ * @state: if non-zero, set the bit (1) else clear the bit (0)
+ *
+ * Description:
+ * Set a single bit in the bitmask.  Returns zero on success, negative values
+ * on error.
+ */
+static void cipso_v4_bitmap_setbit(unsigned char *bitmap,
+                                  u32 bit,
+                                  u8 state)
+{
+       u32 byte_spot;
+       u8 bitmask;
+
+       /* gcc always rounds to zero when doing integer division */
+       byte_spot = bit / 8;
+       bitmask = 0x80 >> (bit % 8);
+       if (state)
+               bitmap[byte_spot] |= bitmask;
+       else
+               bitmap[byte_spot] &= ~bitmask;
+}
+
+/**
+ * cipso_v4_doi_domhsh_free - Frees a domain list entry
+ * @entry: the entry's RCU field
+ *
+ * Description:
+ * This function is designed to be used as a callback to the call_rcu()
+ * function so that the memory allocated to a domain list entry can be released
+ * safely.
+ *
+ */
+static void cipso_v4_doi_domhsh_free(struct rcu_head *entry)
+{
+       struct cipso_v4_domhsh_entry *ptr;
+
+       ptr = container_of(entry, struct cipso_v4_domhsh_entry, rcu);
+       kfree(ptr->domain);
+       kfree(ptr);
+}
+
+/**
+ * cipso_v4_cache_entry_free - Frees a cache entry
+ * @entry: the entry to free
+ *
+ * Description:
+ * This function frees the memory associated with a cache entry.
+ *
+ */
+static void cipso_v4_cache_entry_free(struct cipso_v4_map_cache_entry *entry)
+{
+       if (entry->lsm_data.free)
+               entry->lsm_data.free(entry->lsm_data.data);
+       kfree(entry->key);
+       kfree(entry);
+}
+
+/**
+ * cipso_v4_map_cache_hash - Hashing function for the CIPSO cache
+ * @key: the hash key
+ * @key_len: the length of the key in bytes
+ *
+ * Description:
+ * The CIPSO tag hashing function.  Returns a 32-bit hash value.
+ *
+ */
+static u32 cipso_v4_map_cache_hash(const unsigned char *key, u32 key_len)
+{
+       return jhash(key, key_len, 0);
+}
+
+/*
+ * Label Mapping Cache Functions
+ */
+
+/**
+ * cipso_v4_cache_init - Initialize the CIPSO cache
+ *
+ * Description:
+ * Initializes the CIPSO label mapping cache, this function should be called
+ * before any of the other functions defined in this file.  Returns zero on
+ * success, negative values on error.
+ *
+ */
+static int cipso_v4_cache_init(void)
+{
+       u32 iter;
+
+       cipso_v4_cache = kcalloc(CIPSO_V4_CACHE_BUCKETS,
+                                sizeof(struct cipso_v4_map_cache_bkt),
+                                GFP_KERNEL);
+       if (cipso_v4_cache == NULL)
+               return -ENOMEM;
+
+       for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
+               spin_lock_init(&cipso_v4_cache[iter].lock);
+               cipso_v4_cache[iter].size = 0;
+               INIT_LIST_HEAD(&cipso_v4_cache[iter].list);
+       }
+
+       return 0;
+}
+
+/**
+ * cipso_v4_cache_invalidate - Invalidates the current CIPSO cache
+ *
+ * Description:
+ * Invalidates and frees any entries in the CIPSO cache.  Returns zero on
+ * success and negative values on failure.
+ *
+ */
+void cipso_v4_cache_invalidate(void)
+{
+       struct cipso_v4_map_cache_entry *entry, *tmp_entry;
+       u32 iter;
+
+       for (iter = 0; iter < CIPSO_V4_CACHE_BUCKETS; iter++) {
+               spin_lock(&cipso_v4_cache[iter].lock);
+               list_for_each_entry_safe(entry,
+                                        tmp_entry,
+                                        &cipso_v4_cache[iter].list, list) {
+                       list_del(&entry->list);
+                       cipso_v4_cache_entry_free(entry);
+               }
+               cipso_v4_cache[iter].size = 0;
+               spin_unlock(&cipso_v4_cache[iter].lock);
+       }
+
+       return;
+}
+
+/**
+ * cipso_v4_cache_check - Check the CIPSO cache for a label mapping
+ * @key: the buffer to check
+ * @key_len: buffer length in bytes
+ * @secattr: the security attribute struct to use
+ *
+ * Description:
+ * This function checks the cache to see if a label mapping already exists for
+ * the given key.  If there is a match then the cache is adjusted and the
+ * @secattr struct is populated with the correct LSM security attributes.  The
+ * cache is adjusted in the following manner if the entry is not already the
+ * first in the cache bucket:
+ *
+ *  1. The cache entry's activity counter is incremented
+ *  2. The previous (higher ranking) entry's activity counter is decremented
+ *  3. If the difference between the two activity counters is geater than
+ *     CIPSO_V4_CACHE_REORDERLIMIT the two entries are swapped
+ *
+ * Returns zero on success, -ENOENT for a cache miss, and other negative values
+ * on error.
+ *
+ */
+static int cipso_v4_cache_check(const unsigned char *key,
+                               u32 key_len,
+                               struct netlbl_lsm_secattr *secattr)
+{
+       u32 bkt;
+       struct cipso_v4_map_cache_entry *entry;
+       struct cipso_v4_map_cache_entry *prev_entry = NULL;
+       u32 hash;
+
+       if (!cipso_v4_cache_enabled)
+               return -ENOENT;
+
+       hash = cipso_v4_map_cache_hash(key, key_len);
+       bkt = hash & (CIPSO_V4_CACHE_BUCKETBITS - 1);
+       spin_lock(&cipso_v4_cache[bkt].lock);
+       list_for_each_entry(entry, &cipso_v4_cache[bkt].list, list) {
+               if (entry->hash == hash &&
+                   entry->key_len == key_len &&
+                   memcmp(entry->key, key, key_len) == 0) {
+                       entry->activity += 1;
+                       secattr->cache.free = entry->lsm_data.free;
+                       secattr->cache.data = entry->lsm_data.data;
+                       if (prev_entry == NULL) {
+                               spin_unlock(&cipso_v4_cache[bkt].lock);
+                               return 0;
+                       }
+
+                       if (prev_entry->activity > 0)
+                               prev_entry->activity -= 1;
+                       if (entry->activity > prev_entry->activity &&
+                           entry->activity - prev_entry->activity >
+                           CIPSO_V4_CACHE_REORDERLIMIT) {
+                               __list_del(entry->list.prev, entry->list.next);
+                               __list_add(&entry->list,
+                                          prev_entry->list.prev,
+                                          &prev_entry->list);
+                       }
+
+                       spin_unlock(&cipso_v4_cache[bkt].lock);
+                       return 0;
+               }
+               prev_entry = entry;
+       }
+       spin_unlock(&cipso_v4_cache[bkt].lock);
+
+       return -ENOENT;
+}
+
+/**
+ * cipso_v4_cache_add - Add an entry to the CIPSO cache
+ * @skb: the packet
+ * @secattr: the packet's security attributes
+ *
+ * Description:
+ * Add a new entry into the CIPSO label mapping cache.  Add the new entry to
+ * head of the cache bucket's list, if the cache bucket is out of room remove
+ * the last entry in the list first.  It is important to note that there is
+ * currently no checking for duplicate keys.  Returns zero on success,
+ * negative values on failure.
+ *
+ */
+int cipso_v4_cache_add(const struct sk_buff *skb,
+                      const struct netlbl_lsm_secattr *secattr)
+{
+       int ret_val = -EPERM;
+       u32 bkt;
+       struct cipso_v4_map_cache_entry *entry = NULL;
+       struct cipso_v4_map_cache_entry *old_entry = NULL;
+       unsigned char *cipso_ptr;
+       u32 cipso_ptr_len;
+
+       if (!cipso_v4_cache_enabled || cipso_v4_cache_bucketsize <= 0)
+               return 0;
+
+       cipso_ptr = CIPSO_V4_OPTPTR(skb);
+       cipso_ptr_len = cipso_ptr[1];
+
+       entry = kzalloc(sizeof(*entry), GFP_ATOMIC);
+       if (entry == NULL)
+               return -ENOMEM;
+       entry->key = kmalloc(cipso_ptr_len, GFP_ATOMIC);
+       if (entry->key == NULL) {
+               ret_val = -ENOMEM;
+               goto cache_add_failure;
+       }
+       memcpy(entry->key, cipso_ptr, cipso_ptr_len);
+       entry->key_len = cipso_ptr_len;
+       entry->hash = cipso_v4_map_cache_hash(cipso_ptr, cipso_ptr_len);
+       entry->lsm_data.free = secattr->cache.free;
+       entry->lsm_data.data = secattr->cache.data;
+
+       bkt = entry->hash & (CIPSO_V4_CACHE_BUCKETBITS - 1);
+       spin_lock(&cipso_v4_cache[bkt].lock);
+       if (cipso_v4_cache[bkt].size < cipso_v4_cache_bucketsize) {
+               list_add(&entry->list, &cipso_v4_cache[bkt].list);
+               cipso_v4_cache[bkt].size += 1;
+       } else {
+               old_entry = list_entry(cipso_v4_cache[bkt].list.prev,
+                                      struct cipso_v4_map_cache_entry, list);
+               list_del(&old_entry->list);
+               list_add(&entry->list, &cipso_v4_cache[bkt].list);
+               cipso_v4_cache_entry_free(old_entry);
+       }
+       spin_unlock(&cipso_v4_cache[bkt].lock);
+
+       return 0;
+
+cache_add_failure:
+       if (entry)
+               cipso_v4_cache_entry_free(entry);
+       return ret_val;
+}
+
+/*
+ * DOI List Functions
+ */
+
+/**
+ * cipso_v4_doi_search - Searches for a DOI definition
+ * @doi: the DOI to search for
+ *
+ * Description:
+ * Search the DOI definition list for a DOI definition with a DOI value that
+ * matches @doi.  The caller is responsibile for calling rcu_read_[un]lock().
+ * Returns a pointer to the DOI definition on success and NULL on failure.
+ */
+static struct cipso_v4_doi *cipso_v4_doi_search(u32 doi)
+{
+       struct cipso_v4_doi *iter;
+
+       list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
+               if (iter->doi == doi && iter->valid)
+                       return iter;
+       return NULL;
+}
+
+/**
+ * cipso_v4_doi_add - Add a new DOI to the CIPSO protocol engine
+ * @doi_def: the DOI structure
+ *
+ * Description:
+ * The caller defines a new DOI for use by the CIPSO engine and calls this
+ * function to add it to the list of acceptable domains.  The caller must
+ * ensure that the mapping table specified in @doi_def->map meets all of the
+ * requirements of the mapping type (see cipso_ipv4.h for details).  Returns
+ * zero on success and non-zero on failure.
+ *
+ */
+int cipso_v4_doi_add(struct cipso_v4_doi *doi_def)
+{
+       if (doi_def == NULL || doi_def->doi == CIPSO_V4_DOI_UNKNOWN)
+               return -EINVAL;
+
+       doi_def->valid = 1;
+       INIT_RCU_HEAD(&doi_def->rcu);
+       INIT_LIST_HEAD(&doi_def->dom_list);
+
+       rcu_read_lock();
+       if (cipso_v4_doi_search(doi_def->doi) != NULL)
+               goto doi_add_failure_rlock;
+       spin_lock(&cipso_v4_doi_list_lock);
+       if (cipso_v4_doi_search(doi_def->doi) != NULL)
+               goto doi_add_failure_slock;
+       list_add_tail_rcu(&doi_def->list, &cipso_v4_doi_list);
+       spin_unlock(&cipso_v4_doi_list_lock);
+       rcu_read_unlock();
+
+       return 0;
+
+doi_add_failure_slock:
+       spin_unlock(&cipso_v4_doi_list_lock);
+doi_add_failure_rlock:
+       rcu_read_unlock();
+       return -EEXIST;
+}
+
+/**
+ * cipso_v4_doi_remove - Remove an existing DOI from the CIPSO protocol engine
+ * @doi: the DOI value
+ * @callback: the DOI cleanup/free callback
+ *
+ * Description:
+ * Removes a DOI definition from the CIPSO engine, @callback is called to
+ * free any memory.  The NetLabel routines will be called to release their own
+ * LSM domain mappings as well as our own domain list.  Returns zero on
+ * success and negative values on failure.
+ *
+ */
+int cipso_v4_doi_remove(u32 doi, void (*callback) (struct rcu_head * head))
+{
+       struct cipso_v4_doi *doi_def;
+       struct cipso_v4_domhsh_entry *dom_iter;
+
+       rcu_read_lock();
+       if (cipso_v4_doi_search(doi) != NULL) {
+               spin_lock(&cipso_v4_doi_list_lock);
+               doi_def = cipso_v4_doi_search(doi);
+               if (doi_def == NULL) {
+                       spin_unlock(&cipso_v4_doi_list_lock);
+                       rcu_read_unlock();
+                       return -ENOENT;
+               }
+               doi_def->valid = 0;
+               list_del_rcu(&doi_def->list);
+               spin_unlock(&cipso_v4_doi_list_lock);
+               list_for_each_entry_rcu(dom_iter, &doi_def->dom_list, list)
+                       if (dom_iter->valid)
+                               netlbl_domhsh_remove(dom_iter->domain);
+               cipso_v4_cache_invalidate();
+               rcu_read_unlock();
+
+               call_rcu(&doi_def->rcu, callback);
+               return 0;
+       }
+       rcu_read_unlock();
+
+       return -ENOENT;
+}
+
+/**
+ * cipso_v4_doi_getdef - Returns a pointer to a valid DOI definition
+ * @doi: the DOI value
+ *
+ * Description:
+ * Searches for a valid DOI definition and if one is found it is returned to
+ * the caller.  Otherwise NULL is returned.  The caller must ensure that
+ * rcu_read_lock() is held while accessing the returned definition.
+ *
+ */
+struct cipso_v4_doi *cipso_v4_doi_getdef(u32 doi)
+{
+       return cipso_v4_doi_search(doi);
+}
+
+/**
+ * cipso_v4_doi_dump_all - Dump all the CIPSO DOI definitions into a sk_buff
+ * @headroom: the amount of headroom to allocate for the sk_buff
+ *
+ * Description:
+ * Dump a list of all the configured DOI values into a sk_buff.  The returned
+ * sk_buff has room at the front of the sk_buff for @headroom bytes.  See
+ * net/netlabel/netlabel_cipso_v4.h for the LISTALL message format.  This
+ * function may fail if another process is changing the DOI list at the same
+ * time.  Returns a pointer to a sk_buff on success, NULL on error.
+ *
+ */
+struct sk_buff *cipso_v4_doi_dump_all(size_t headroom)
+{
+       struct sk_buff *skb = NULL;
+       struct cipso_v4_doi *iter;
+       u32 doi_cnt = 0;
+       ssize_t buf_len;
+
+       buf_len = NETLBL_LEN_U32;
+       rcu_read_lock();
+       list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
+               if (iter->valid) {
+                       doi_cnt += 1;
+                       buf_len += 2 * NETLBL_LEN_U32;
+               }
+
+       skb = netlbl_netlink_alloc_skb(headroom, buf_len, GFP_ATOMIC);
+       if (skb == NULL)
+               goto doi_dump_all_failure;
+
+       if (nla_put_u32(skb, NLA_U32, doi_cnt) != 0)
+               goto doi_dump_all_failure;
+       buf_len -= NETLBL_LEN_U32;
+       list_for_each_entry_rcu(iter, &cipso_v4_doi_list, list)
+               if (iter->valid) {
+                       if (buf_len < 2 * NETLBL_LEN_U32)
+                               goto doi_dump_all_failure;
+                       if (nla_put_u32(skb, NLA_U32, iter->doi) != 0)
+                               goto doi_dump_all_failure;
+                       if (nla_put_u32(skb, NLA_U32, iter->type) != 0)
+                               goto doi_dump_all_failure;
+                       buf_len -= 2 * NETLBL_LEN_U32;
+               }
+       rcu_read_unlock();
+
+       return skb;
+
+doi_dump_all_failure:
+       rcu_read_unlock();
+       kfree(skb);
+       return NULL;
+}
+
+/**
+ * cipso_v4_doi_dump - Dump a CIPSO DOI definition into a sk_buff
+ * @doi: the DOI value
+ * @headroom: the amount of headroom to allocate for the sk_buff
+ *
+ * Description:
+ * Lookup the DOI definition matching @doi and dump it's contents into a
+ * sk_buff.  The returned sk_buff has room at the front of the sk_buff for
+ * @headroom bytes.  See net/netlabel/netlabel_cipso_v4.h for the LIST message
+ * format.  This function may fail if another process is changing the DOI list
+ * at the same time.  Returns a pointer to a sk_buff on success, NULL on error.
+ *
+ */
+struct sk_buff *cipso_v4_doi_dump(u32 doi, size_t headroom)
+{
+       struct sk_buff *skb = NULL;
+       struct cipso_v4_doi *iter;
+       u32 tag_cnt = 0;
+       u32 lvl_cnt = 0;
+       u32 cat_cnt = 0;
+       ssize_t buf_len;
+       ssize_t tmp;
+
+       rcu_read_lock();
+       iter = cipso_v4_doi_getdef(doi);
+       if (iter == NULL)
+               goto doi_dump_failure;
+       buf_len = NETLBL_LEN_U32;
+       switch (iter->type) {
+       case CIPSO_V4_MAP_PASS:
+               buf_len += NETLBL_LEN_U32;
+               while(tag_cnt < CIPSO_V4_TAG_MAXCNT &&
+                     iter->tags[tag_cnt] != CIPSO_V4_TAG_INVALID) {
+                       tag_cnt += 1;
+                       buf_len += NETLBL_LEN_U8;
+               }
+               break;
+       case CIPSO_V4_MAP_STD:
+               buf_len += 3 * NETLBL_LEN_U32;
+               while (tag_cnt < CIPSO_V4_TAG_MAXCNT &&
+                      iter->tags[tag_cnt] != CIPSO_V4_TAG_INVALID) {
+                       tag_cnt += 1;
+                       buf_len += NETLBL_LEN_U8;
+               }
+               for (tmp = 0; tmp < iter->map.std->lvl.local_size; tmp++)
+                       if (iter->map.std->lvl.local[tmp] !=
+                           CIPSO_V4_INV_LVL) {
+                               lvl_cnt += 1;
+                               buf_len += NETLBL_LEN_U32 + NETLBL_LEN_U8;
+                       }
+               for (tmp = 0; tmp < iter->map.std->cat.local_size; tmp++)
+                       if (iter->map.std->cat.local[tmp] !=
+                           CIPSO_V4_INV_CAT) {
+                               cat_cnt += 1;
+                               buf_len += NETLBL_LEN_U32 + NETLBL_LEN_U16;
+                       }
+               break;
+       }
+
+       skb = netlbl_netlink_alloc_skb(headroom, buf_len, GFP_ATOMIC);
+       if (skb == NULL)
+               goto doi_dump_failure;
+
+       if (nla_put_u32(skb, NLA_U32, iter->type) != 0)
+               goto doi_dump_failure;
+       buf_len -= NETLBL_LEN_U32;
+       if (iter != cipso_v4_doi_getdef(doi))
+               goto doi_dump_failure;
+       switch (iter->type) {
+       case CIPSO_V4_MAP_PASS:
+               if (nla_put_u32(skb, NLA_U32, tag_cnt) != 0)
+                       goto doi_dump_failure;
+               buf_len -= NETLBL_LEN_U32;
+               for (tmp = 0;
+                    tmp < CIPSO_V4_TAG_MAXCNT &&
+                            iter->tags[tmp] != CIPSO_V4_TAG_INVALID;
+                    tmp++) {
+                       if (buf_len < NETLBL_LEN_U8)
+                               goto doi_dump_failure;
+                       if (nla_put_u8(skb, NLA_U8, iter->tags[tmp]) != 0)
+                               goto doi_dump_failure;
+                       buf_len -= NETLBL_LEN_U8;
+               }
+               break;
+       case CIPSO_V4_MAP_STD:
+               if (nla_put_u32(skb, NLA_U32, tag_cnt) != 0)
+                       goto doi_dump_failure;
+               if (nla_put_u32(skb, NLA_U32, lvl_cnt) != 0)
+                       goto doi_dump_failure;
+               if (nla_put_u32(skb, NLA_U32, cat_cnt) != 0)
+                       goto doi_dump_failure;
+               buf_len -= 3 * NETLBL_LEN_U32;
+               for (tmp = 0;
+                    tmp < CIPSO_V4_TAG_MAXCNT &&
+                            iter->tags[tmp] != CIPSO_V4_TAG_INVALID;
+                    tmp++) {
+                       if (buf_len < NETLBL_LEN_U8)
+                               goto doi_dump_failure;
+                       if (nla_put_u8(skb, NLA_U8, iter->tags[tmp]) != 0)
+                               goto doi_dump_failure;
+                       buf_len -= NETLBL_LEN_U8;
+               }
+               for (tmp = 0; tmp < iter->map.std->lvl.local_size; tmp++)
+                       if (iter->map.std->lvl.local[tmp] !=
+                           CIPSO_V4_INV_LVL) {
+                               if (buf_len < NETLBL_LEN_U32 + NETLBL_LEN_U8)
+                                       goto doi_dump_failure;
+                               if (nla_put_u32(skb, NLA_U32, tmp) != 0)
+                                       goto doi_dump_failure;
+                               if (nla_put_u8(skb,
+                                          NLA_U8,
+                                          iter->map.std->lvl.local[tmp]) != 0)
+                                       goto doi_dump_failure;
+                               buf_len -= NETLBL_LEN_U32 + NETLBL_LEN_U8;
+                       }
+               for (tmp = 0; tmp < iter->map.std->cat.local_size; tmp++)
+                       if (iter->map.std->cat.local[tmp] !=
+                           CIPSO_V4_INV_CAT) {
+                               if (buf_len < NETLBL_LEN_U32 + NETLBL_LEN_U16)
+                                       goto doi_dump_failure;
+                               if (nla_put_u32(skb, NLA_U32, tmp) != 0)
+                                       goto doi_dump_failure;
+                               if (nla_put_u16(skb,
+                                          NLA_U16,
+                                          iter->map.std->cat.local[tmp]) != 0)
+                                       goto doi_dump_failure;
+                               buf_len -= NETLBL_LEN_U32 + NETLBL_LEN_U16;
+                       }
+               break;
+       }
+       rcu_read_unlock();
+
+       return skb;
+
+doi_dump_failure:
+       rcu_read_unlock();
+       kfree(skb);
+       return NULL;
+}
+
+/**
+ * cipso_v4_doi_domhsh_add - Adds a domain entry to a DOI definition
+ * @doi_def: the DOI definition
+ * @domain: the domain to add
+ *
+ * Description:
+ * Adds the @domain to the the DOI specified by @doi_def, this function
+ * should only be called by external functions (i.e. NetLabel).  This function
+ * does allocate memory.  Returns zero on success, negative values on failure.
+ *
+ */
+int cipso_v4_doi_domhsh_add(struct cipso_v4_doi *doi_def, const char *domain)
+{
+       struct cipso_v4_domhsh_entry *iter;
+       struct cipso_v4_domhsh_entry *new_dom;
+
+       new_dom = kzalloc(sizeof(*new_dom), GFP_KERNEL);
+       if (new_dom == NULL)
+               return -ENOMEM;
+       if (domain) {
+               new_dom->domain = kstrdup(domain, GFP_KERNEL);
+               if (new_dom->domain == NULL) {
+                       kfree(new_dom);
+                       return -ENOMEM;
+               }
+       }
+       new_dom->valid = 1;
+       INIT_RCU_HEAD(&new_dom->rcu);
+
+       rcu_read_lock();
+       spin_lock(&cipso_v4_doi_list_lock);
+       list_for_each_entry_rcu(iter, &doi_def->dom_list, list)
+               if (iter->valid &&
+                   ((domain != NULL && iter->domain != NULL &&
+                     strcmp(iter->domain, domain) == 0) ||
+                    (domain == NULL && iter->domain == NULL))) {
+                       spin_unlock(&cipso_v4_doi_list_lock);
+                       rcu_read_unlock();
+                       kfree(new_dom->domain);
+                       kfree(new_dom);
+                       return -EEXIST;
+               }
+       list_add_tail_rcu(&new_dom->list, &doi_def->dom_list);
+       spin_unlock(&cipso_v4_doi_list_lock);
+       rcu_read_unlock();
+
+       return 0;
+}
+
+/**
+ * cipso_v4_doi_domhsh_remove - Removes a domain entry from a DOI definition
+ * @doi_def: the DOI definition
+ * @domain: the domain to remove
+ *
+ * Description:
+ * Removes the @domain from the DOI specified by @doi_def, this function
+ * should only be called by external functions (i.e. NetLabel).   Returns zero
+ * on success and negative values on error.
+ *
+ */
+int cipso_v4_doi_domhsh_remove(struct cipso_v4_doi *doi_def,
+                              const char *domain)
+{
+       struct cipso_v4_domhsh_entry *iter;
+
+       rcu_read_lock();
+       spin_lock(&cipso_v4_doi_list_lock);
+       list_for_each_entry_rcu(iter, &doi_def->dom_list, list)
+               if (iter->valid &&
+                   ((domain != NULL && iter->domain != NULL &&
+                     strcmp(iter->domain, domain) == 0) ||
+                    (domain == NULL && iter->domain == NULL))) {
+                       iter->valid = 0;
+                       list_del_rcu(&iter->list);
+                       spin_unlock(&cipso_v4_doi_list_lock);
+                       rcu_read_unlock();
+                       call_rcu(&iter->rcu, cipso_v4_doi_domhsh_free);
+
+                       return 0;
+               }
+       spin_unlock(&cipso_v4_doi_list_lock);
+       rcu_read_unlock();
+
+       return -ENOENT;
+}
+
+/*
+ * Label Mapping Functions
+ */
+
+/**
+ * cipso_v4_map_lvl_valid - Checks to see if the given level is understood
+ * @doi_def: the DOI definition
+ * @level: the level to check
+ *
+ * Description:
+ * Checks the given level against the given DOI definition and returns a
+ * negative value if the level does not have a valid mapping and a zero value
+ * if the level is defined by the DOI.
+ *
+ */
+static int cipso_v4_map_lvl_valid(const struct cipso_v4_doi *doi_def, u8 level)
+{
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               return 0;
+       case CIPSO_V4_MAP_STD:
+               if (doi_def->map.std->lvl.cipso[level] < CIPSO_V4_INV_LVL)
+                       return 0;
+               break;
+       }
+
+       return -EFAULT;
+}
+
+/**
+ * cipso_v4_map_lvl_hton - Perform a level mapping from the host to the network
+ * @doi_def: the DOI definition
+ * @host_lvl: the host MLS level
+ * @net_lvl: the network/CIPSO MLS level
+ *
+ * Description:
+ * Perform a label mapping to translate a local MLS level to the correct
+ * CIPSO level using the given DOI definition.  Returns zero on success,
+ * negative values otherwise.
+ *
+ */
+static int cipso_v4_map_lvl_hton(const struct cipso_v4_doi *doi_def,
+                                u32 host_lvl,
+                                u32 *net_lvl)
+{
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               *net_lvl = host_lvl;
+               return 0;
+       case CIPSO_V4_MAP_STD:
+               if (host_lvl < doi_def->map.std->lvl.local_size) {
+                       *net_lvl = doi_def->map.std->lvl.local[host_lvl];
+                       return 0;
+               }
+               break;
+       }
+
+       return -EINVAL;
+}
+
+/**
+ * cipso_v4_map_lvl_ntoh - Perform a level mapping from the network to the host
+ * @doi_def: the DOI definition
+ * @net_lvl: the network/CIPSO MLS level
+ * @host_lvl: the host MLS level
+ *
+ * Description:
+ * Perform a label mapping to translate a CIPSO level to the correct local MLS
+ * level using the given DOI definition.  Returns zero on success, negative
+ * values otherwise.
+ *
+ */
+static int cipso_v4_map_lvl_ntoh(const struct cipso_v4_doi *doi_def,
+                                u32 net_lvl,
+                                u32 *host_lvl)
+{
+       struct cipso_v4_std_map_tbl *map_tbl;
+
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               *host_lvl = net_lvl;
+               return 0;
+       case CIPSO_V4_MAP_STD:
+               map_tbl = doi_def->map.std;
+               if (net_lvl < map_tbl->lvl.cipso_size &&
+                   map_tbl->lvl.cipso[net_lvl] < CIPSO_V4_INV_LVL) {
+                       *host_lvl = doi_def->map.std->lvl.cipso[net_lvl];
+                       return 0;
+               }
+               break;
+       }
+
+       return -EINVAL;
+}
+
+/**
+ * cipso_v4_map_cat_rbm_valid - Checks to see if the category bitmap is valid
+ * @doi_def: the DOI definition
+ * @bitmap: category bitmap
+ * @bitmap_len: bitmap length in bytes
+ *
+ * Description:
+ * Checks the given category bitmap against the given DOI definition and
+ * returns a negative value if any of the categories in the bitmap do not have
+ * a valid mapping and a zero value if all of the categories are valid.
+ *
+ */
+static int cipso_v4_map_cat_rbm_valid(const struct cipso_v4_doi *doi_def,
+                                     const unsigned char *bitmap,
+                                     u32 bitmap_len)
+{
+       int cat = -1;
+       u32 bitmap_len_bits = bitmap_len * 8;
+       u32 cipso_cat_size = doi_def->map.std->cat.cipso_size;
+       u32 *cipso_array = doi_def->map.std->cat.cipso;
+
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               return 0;
+       case CIPSO_V4_MAP_STD:
+               for (;;) {
+                       cat = cipso_v4_bitmap_walk(bitmap,
+                                                  bitmap_len_bits,
+                                                  cat + 1,
+                                                  1);
+                       if (cat < 0)
+                               break;
+                       if (cat >= cipso_cat_size ||
+                           cipso_array[cat] >= CIPSO_V4_INV_CAT)
+                               return -EFAULT;
+               }
+
+               if (cat == -1)
+                       return 0;
+               break;
+       }
+
+       return -EFAULT;
+}
+
+/**
+ * cipso_v4_map_cat_rbm_hton - Perform a category mapping from host to network
+ * @doi_def: the DOI definition
+ * @host_cat: the category bitmap in host format
+ * @host_cat_len: the length of the host's category bitmap in bytes
+ * @net_cat: the zero'd out category bitmap in network/CIPSO format
+ * @net_cat_len: the length of the CIPSO bitmap in bytes
+ *
+ * Description:
+ * Perform a label mapping to translate a local MLS category bitmap to the
+ * correct CIPSO bitmap using the given DOI definition.  Returns the minimum
+ * size in bytes of the network bitmap on success, negative values otherwise.
+ *
+ */
+static int cipso_v4_map_cat_rbm_hton(const struct cipso_v4_doi *doi_def,
+                                    const unsigned char *host_cat,
+                                    u32 host_cat_len,
+                                    unsigned char *net_cat,
+                                    u32 net_cat_len)
+{
+       int host_spot = -1;
+       u32 net_spot;
+       u32 net_spot_max = 0;
+       u32 host_clen_bits = host_cat_len * 8;
+       u32 net_clen_bits = net_cat_len * 8;
+       u32 host_cat_size = doi_def->map.std->cat.local_size;
+       u32 *host_cat_array = doi_def->map.std->cat.local;
+
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               net_spot_max = host_cat_len - 1;
+               while (net_spot_max > 0 && host_cat[net_spot_max] == 0)
+                       net_spot_max--;
+               if (net_spot_max > net_cat_len)
+                       return -EINVAL;
+               memcpy(net_cat, host_cat, net_spot_max);
+               return net_spot_max;
+       case CIPSO_V4_MAP_STD:
+               for (;;) {
+                       host_spot = cipso_v4_bitmap_walk(host_cat,
+                                                        host_clen_bits,
+                                                        host_spot + 1,
+                                                        1);
+                       if (host_spot < 0)
+                               break;
+                       if (host_spot >= host_cat_size)
+                               return -EPERM;
+
+                       net_spot = host_cat_array[host_spot];
+                       if (net_spot >= net_clen_bits)
+                               return -ENOSPC;
+                       cipso_v4_bitmap_setbit(net_cat, net_spot, 1);
+
+                       if (net_spot > net_spot_max)
+                               net_spot_max = net_spot;
+               }
+
+               if (host_spot == -2)
+                       return -EFAULT;
+
+               if (++net_spot_max % 8)
+                       return net_spot_max / 8 + 1;
+               return net_spot_max / 8;
+       }
+
+       return -EINVAL;
+}
+
+/**
+ * cipso_v4_map_cat_rbm_ntoh - Perform a category mapping from network to host
+ * @doi_def: the DOI definition
+ * @net_cat: the category bitmap in network/CIPSO format
+ * @net_cat_len: the length of the CIPSO bitmap in bytes
+ * @host_cat: the zero'd out category bitmap in host format
+ * @host_cat_len: the length of the host's category bitmap in bytes
+ *
+ * Description:
+ * Perform a label mapping to translate a CIPSO bitmap to the correct local
+ * MLS category bitmap using the given DOI definition.  Returns the minimum
+ * size in bytes of the host bitmap on success, negative values otherwise.
+ *
+ */
+static int cipso_v4_map_cat_rbm_ntoh(const struct cipso_v4_doi *doi_def,
+                                    const unsigned char *net_cat,
+                                    u32 net_cat_len,
+                                    unsigned char *host_cat,
+                                    u32 host_cat_len)
+{
+       u32 host_spot;
+       u32 host_spot_max = 0;
+       int net_spot = -1;
+       u32 net_clen_bits = net_cat_len * 8;
+       u32 host_clen_bits = host_cat_len * 8;
+       u32 net_cat_size = doi_def->map.std->cat.cipso_size;
+       u32 *net_cat_array = doi_def->map.std->cat.cipso;
+
+       switch (doi_def->type) {
+       case CIPSO_V4_MAP_PASS:
+               if (net_cat_len > host_cat_len)
+                       return -EINVAL;
+               memcpy(host_cat, net_cat, net_cat_len);
+               return net_cat_len;
+       case CIPSO_V4_MAP_STD:
+               for (;;) {
+                       net_spot = cipso_v4_bitmap_walk(net_cat,
+                                                       net_clen_bits,
+                                                       net_spot + 1,
+                                                       1);
+                       if (net_spot < 0)
+                               break;
+                       if (net_spot >= net_cat_size ||
+                           net_cat_array[net_spot] >= CIPSO_V4_INV_CAT)
+                               return -EPERM;
+
+                       host_spot = net_cat_array[net_spot];
+                       if (host_spot >= host_clen_bits)
+                               return -ENOSPC;
+                       cipso_v4_bitmap_setbit(host_cat, host_spot, 1);
+
+                       if (host_spot > host_spot_max)
+                               host_spot_max = host_spot;
+               }
+
+               if (net_spot == -2)
+                       return -EFAULT;
+
+               if (++host_spot_max % 8)
+                       return host_spot_max / 8 + 1;
+               return host_spot_max / 8;
+       }
+
+       return -EINVAL;
+}
+
+/*
+ * Protocol Handling Functions
+ */
+
+#define CIPSO_V4_HDR_LEN              6
+
+/**
+ * cipso_v4_gentag_hdr - Generate a CIPSO option header
+ * @doi_def: the DOI definition
+ * @len: the total tag length in bytes
+ * @buf: the CIPSO option buffer
+ *
+ * Description:
+ * Write a CIPSO header into the beginning of @buffer.  Return zero on success,
+ * negative values on failure.
+ *
+ */
+static int cipso_v4_gentag_hdr(const struct cipso_v4_doi *doi_def,
+                              u32 len,
+                              unsigned char *buf)
+{
+       if (CIPSO_V4_HDR_LEN + len > 40)
+               return -ENOSPC;
+
+       buf[0] = IPOPT_CIPSO;
+       buf[1] = CIPSO_V4_HDR_LEN + len;
+       *(u32 *)&buf[2] = htonl(doi_def->doi);
+
+       return 0;
+}
+
+#define CIPSO_V4_TAG1_CAT_LEN         30
+
+/**
+ * cipso_v4_gentag_rbm - Generate a CIPSO restricted bitmap tag (type #1)
+ * @doi_def: the DOI definition
+ * @secattr: the security attributes
+ * @buffer: the option buffer
+ * @buffer_len: length of buffer in bytes
+ *
+ * Description:
+ * Generate a CIPSO option using the restricted bitmap tag, tag type #1.  The
+ * actual buffer length may be larger than the indicated size due to
+ * translation between host and network category bitmaps.  Returns zero on
+ * success, negative values on failure.
+ *
+ */
+static int cipso_v4_gentag_rbm(const struct cipso_v4_doi *doi_def,
+                              const struct netlbl_lsm_secattr *secattr,
+                              unsigned char **buffer,
+                              u32 *buffer_len)
+{
+       int ret_val = -EPERM;
+       unsigned char *buf = NULL;
+       u32 buf_len;
+       u32 level;
+
+       if (secattr->mls_cat) {
+               buf = kzalloc(CIPSO_V4_HDR_LEN + 4 + CIPSO_V4_TAG1_CAT_LEN,
+                             GFP_ATOMIC);
+               if (buf == NULL)
+                       return -ENOMEM;
+
+               ret_val = cipso_v4_map_cat_rbm_hton(doi_def,
+                                                   secattr->mls_cat,
+                                                   secattr->mls_cat_len,
+                                                   &buf[CIPSO_V4_HDR_LEN + 4],
+                                                   CIPSO_V4_TAG1_CAT_LEN);
+               if (ret_val < 0)
+                       goto gentag_failure;
+
+               /* This will send packets using the "optimized" format when
+                * possibile as specified in  section 3.4.2.6 of the
+                * CIPSO draft. */
+               if (cipso_v4_rbm_optfmt && (ret_val > 0 && ret_val < 10))
+                       ret_val = 10;
+
+               buf_len = 4 + ret_val;
+       } else {
+               buf = kzalloc(CIPSO_V4_HDR_LEN + 4, GFP_ATOMIC);
+               if (buf == NULL)
+                       return -ENOMEM;
+               buf_len = 4;
+       }
+
+       ret_val = cipso_v4_map_lvl_hton(doi_def, secattr->mls_lvl, &level);
+       if (ret_val != 0)
+               goto gentag_failure;
+
+       ret_val = cipso_v4_gentag_hdr(doi_def, buf_len, buf);
+       if (ret_val != 0)
+               goto gentag_failure;
+
+       buf[CIPSO_V4_HDR_LEN] = 0x01;
+       buf[CIPSO_V4_HDR_LEN + 1] = buf_len;
+       buf[CIPSO_V4_HDR_LEN + 3] = level;
+
+       *buffer = buf;
+       *buffer_len = CIPSO_V4_HDR_LEN + buf_len;
+
+       return 0;
+
+gentag_failure:
+       kfree(buf);
+       return ret_val;
+}
+
+/**
+ * cipso_v4_parsetag_rbm - Parse a CIPSO restricted bitmap tag
+ * @doi_def: the DOI definition
+ * @tag: the CIPSO tag
+ * @secattr: the security attributes
+ *
+ * Description:
+ * Parse a CIPSO restricted bitmap tag (tag type #1) and return the security
+ * attributes in @secattr.  Return zero on success, negatives values on
+ * failure.
+ *
+ */
+static int cipso_v4_parsetag_rbm(const struct cipso_v4_doi *doi_def,
+                                const unsigned char *tag,
+                                struct netlbl_lsm_secattr *secattr)
+{
+       int ret_val;
+       u8 tag_len = tag[1];
+       u32 level;
+
+       ret_val = cipso_v4_map_lvl_ntoh(doi_def, tag[3], &level);
+       if (ret_val != 0)
+               return ret_val;
+       secattr->mls_lvl = level;
+       secattr->mls_lvl_vld = 1;
+
+       if (tag_len > 4) {
+               switch (doi_def->type) {
+               case CIPSO_V4_MAP_PASS:
+                       secattr->mls_cat_len = tag_len - 4;
+                       break;
+               case CIPSO_V4_MAP_STD:
+                       secattr->mls_cat_len =
+                               doi_def->map.std->cat.local_size;
+                       break;
+               }
+               secattr->mls_cat = kzalloc(secattr->mls_cat_len, GFP_ATOMIC);
+               if (secattr->mls_cat == NULL)
+                       return -ENOMEM;
+
+               ret_val = cipso_v4_map_cat_rbm_ntoh(doi_def,
+                                                   &tag[4],
+                                                   tag_len - 4,
+                                                   secattr->mls_cat,
+                                                   secattr->mls_cat_len);
+               if (ret_val < 0) {
+                       kfree(secattr->mls_cat);
+                       return ret_val;
+               }
+               secattr->mls_cat_len = ret_val;
+       }
+
+       return 0;
+}
+
+/**
+ * cipso_v4_validate - Validate a CIPSO option
+ * @option: the start of the option, on error it is set to point to the error
+ *
+ * Description:
+ * This routine is called to validate a CIPSO option, it checks all of the
+ * fields to ensure that they are at least valid, see the draft snippet below
+ * for details.  If the option is valid then a zero value is returned and
+ * the value of @option is unchanged.  If the option is invalid then a
+ * non-zero value is returned and @option is adjusted to point to the
+ * offending portion of the option.  From the IETF draft ...
+ *
+ *  "If any field within the CIPSO options, such as the DOI identifier, is not
+ *   recognized the IP datagram is discarded and an ICMP 'parameter problem'
+ *   (type 12) is generated and returned.  The ICMP code field is set to 'bad
+ *   parameter' (code 0) and the pointer is set to the start of the CIPSO field
+ *   that is unrecognized."
+ *
+ */
+int cipso_v4_validate(unsigned char **option)
+{
+       unsigned char *opt = *option;
+       unsigned char *tag;
+       unsigned char opt_iter;
+       unsigned char err_offset = 0;
+       u8 opt_len;
+       u8 tag_len;
+       struct cipso_v4_doi *doi_def = NULL;
+       u32 tag_iter;
+
+       /* caller already checks for length values that are too large */
+       opt_len = opt[1];
+       if (opt_len < 8) {
+               err_offset = 1;
+               goto validate_return;
+       }
+
+       rcu_read_lock();
+       doi_def = cipso_v4_doi_getdef(ntohl(*((u32 *)&opt[2])));
+       if (doi_def == NULL) {
+               err_offset = 2;
+               goto validate_return_locked;
+       }
+
+       opt_iter = 6;
+       tag = opt + opt_iter;
+       while (opt_iter < opt_len) {
+               for (tag_iter = 0; doi_def->tags[tag_iter] != tag[0];)
+                       if (doi_def->tags[tag_iter] == CIPSO_V4_TAG_INVALID ||
+                           ++tag_iter == CIPSO_V4_TAG_MAXCNT) {
+                               err_offset = opt_iter;
+                               goto validate_return_locked;
+                       }
+
+               tag_len = tag[1];
+               if (tag_len > (opt_len - opt_iter)) {
+                       err_offset = opt_iter + 1;
+                       goto validate_return_locked;
+               }
+
+               switch (tag[0]) {
+               case CIPSO_V4_TAG_RBITMAP:
+                       if (tag_len < 4) {
+                               err_offset = opt_iter + 1;
+                               goto validate_return_locked;
+                       }
+
+                       /* We are already going to do all the verification
+                        * necessary at the socket layer so from our point of
+                        * view it is safe to turn these checks off (and less
+                        * work), however, the CIPSO draft says we should do
+                        * all the CIPSO validations here but it doesn't
+                        * really specify _exactly_ what we need to validate
+                        * ... so, just make it a sysctl tunable. */
+                       if (cipso_v4_rbm_strictvalid) {
+                               if (cipso_v4_map_lvl_valid(doi_def,
+                                                          tag[3]) < 0) {
+                                       err_offset = opt_iter + 3;
+                                       goto validate_return_locked;
+                               }
+                               if (tag_len > 4 &&
+                                   cipso_v4_map_cat_rbm_valid(doi_def,
+                                                           &tag[4],
+                                                           tag_len - 4) < 0) {
+                                       err_offset = opt_iter + 4;
+                                       goto validate_return_locked;
+                               }
+                       }
+                       break;
+               default:
+                       err_offset = opt_iter;
+                       goto validate_return_locked;
+               }
+
+               tag += tag_len;
+               opt_iter += tag_len;
+       }
+
+validate_return_locked:
+       rcu_read_unlock();
+validate_return:
+       *option = opt + err_offset;
+       return err_offset;
+}
+
+/**
+ * cipso_v4_error - Send the correct reponse for a bad packet
+ * @skb: the packet
+ * @error: the error code
+ * @gateway: CIPSO gateway flag
+ *
+ * Description:
+ * Based on the error code given in @error, send an ICMP error message back to
+ * the originating host.  From the IETF draft ...
+ *
+ *  "If the contents of the CIPSO [option] are valid but the security label is
+ *   outside of the configured host or port label range, the datagram is
+ *   discarded and an ICMP 'destination unreachable' (type 3) is generated and
+ *   returned.  The code field of the ICMP is set to 'communication with
+ *   destination network administratively prohibited' (code 9) or to
+ *   'communication with destination host administratively prohibited'
+ *   (code 10).  The value of the code is dependent on whether the originator
+ *   of the ICMP message is acting as a CIPSO host or a CIPSO gateway.  The
+ *   recipient of the ICMP message MUST be able to handle either value.  The
+ *   same procedure is performed if a CIPSO [option] can not be added to an
+ *   IP packet because it is too large to fit in the IP options area."
+ *
+ *  "If the error is triggered by receipt of an ICMP message, the message is
+ *   discarded and no response is permitted (consistent with general ICMP
+ *   processing rules)."
+ *
+ */
+void cipso_v4_error(struct sk_buff *skb, int error, u32 gateway)
+{
+       if (skb->nh.iph->protocol == IPPROTO_ICMP || error != -EACCES)
+               return;
+
+       if (gateway)
+               icmp_send(skb, ICMP_DEST_UNREACH, ICMP_NET_ANO, 0);
+       else
+               icmp_send(skb, ICMP_DEST_UNREACH, ICMP_HOST_ANO, 0);
+}
+
+/**
+ * cipso_v4_socket_setattr - Add a CIPSO option to a socket
+ * @sock: the socket
+ * @doi_def: the CIPSO DOI to use
+ * @secattr: the specific security attributes of the socket
+ *
+ * Description:
+ * Set the CIPSO option on the given socket using the DOI definition and
+ * security attributes passed to the function.  This function requires
+ * exclusive access to @sock->sk, which means it either needs to be in the
+ * process of being created or locked via lock_sock(sock->sk).  Returns zero on
+ * success and negative values on failure.
+ *
+ */
+int cipso_v4_socket_setattr(const struct socket *sock,
+                           const struct cipso_v4_doi *doi_def,
+                           const struct netlbl_lsm_secattr *secattr)
+{
+       int ret_val = -EPERM;
+       u32 iter;
+       unsigned char *buf = NULL;
+       u32 buf_len = 0;
+       u32 opt_len;
+       struct ip_options *opt = NULL;
+       struct sock *sk;
+       struct inet_sock *sk_inet;
+       struct inet_connection_sock *sk_conn;
+
+       /* In the case of sock_create_lite(), the sock->sk field is not
+        * defined yet but it is not a problem as the only users of these
+        * "lite" PF_INET sockets are functions which do an accept() call
+        * afterwards so we will label the socket as part of the accept(). */
+       sk = sock->sk;
+       if (sk == NULL)
+               return 0;
+
+       /* XXX - This code assumes only one tag per CIPSO option which isn't
+        * really a good assumption to make but since we only support the MAC
+        * tags right now it is a safe assumption. */
+       iter = 0;
+       do {
+               switch (doi_def->tags[iter]) {
+               case CIPSO_V4_TAG_RBITMAP:
+                       ret_val = cipso_v4_gentag_rbm(doi_def,
+                                                     secattr,
+                                                     &buf,
+                                                     &buf_len);
+                       break;
+               default:
+                       ret_val = -EPERM;
+                       goto socket_setattr_failure;
+               }
+
+               iter++;
+       } while (ret_val != 0 &&
+                iter < CIPSO_V4_TAG_MAXCNT &&
+                doi_def->tags[iter] != CIPSO_V4_TAG_INVALID);
+       if (ret_val != 0)
+               goto socket_setattr_failure;
+
+       /* We can't use ip_options_get() directly because it makes a call to
+        * ip_options_get_alloc() which allocates memory with GFP_KERNEL and
+        * we can't block here. */
+       opt_len = (buf_len + 3) & ~3;
+       opt = kzalloc(sizeof(*opt) + opt_len, GFP_ATOMIC);
+       if (opt == NULL) {
+               ret_val = -ENOMEM;
+               goto socket_setattr_failure;
+       }
+       memcpy(opt->__data, buf, buf_len);
+       opt->optlen = opt_len;
+       opt->is_data = 1;
+       kfree(buf);
+       buf = NULL;
+       ret_val = ip_options_compile(opt, NULL);
+       if (ret_val != 0)
+               goto socket_setattr_failure;
+
+       sk_inet = inet_sk(sk);
+       if (sk_inet->is_icsk) {
+               sk_conn = inet_csk(sk);
+               if (sk_inet->opt)
+                       sk_conn->icsk_ext_hdr_len -= sk_inet->opt->optlen;
+               sk_conn->icsk_ext_hdr_len += opt->optlen;
+               sk_conn->icsk_sync_mss(sk, sk_conn->icsk_pmtu_cookie);
+       }
+       opt = xchg(&sk_inet->opt, opt);
+       kfree(opt);
+
+       return 0;
+
+socket_setattr_failure:
+       kfree(buf);
+       kfree(opt);
+       return ret_val;
+}
+
+/**
+ * cipso_v4_socket_getattr - Get the security attributes from a socket
+ * @sock: the socket
+ * @secattr: the security attributes
+ *
+ * Description:
+ * Query @sock to see if there is a CIPSO option attached to the socket and if
+ * there is return the CIPSO security attributes in @secattr.  Returns zero on
+ * success and negative values on failure.
+ *
+ */
+int cipso_v4_socket_getattr(const struct socket *sock,
+                           struct netlbl_lsm_secattr *secattr)
+{
+       int ret_val = -ENOMSG;
+       struct sock *sk;
+       struct inet_sock *sk_inet;
+       unsigned char *cipso_ptr;
+       u32 doi;
+       struct cipso_v4_doi *doi_def;
+
+       sk = sock->sk;
+       lock_sock(sk);
+       sk_inet = inet_sk(sk);
+       if (sk_inet->opt == NULL || sk_inet->opt->cipso == 0)
+               goto socket_getattr_return;
+       cipso_ptr = sk_inet->opt->__data + sk_inet->opt->cipso -
+               sizeof(struct iphdr);
+       ret_val = cipso_v4_cache_check(cipso_ptr, cipso_ptr[1], secattr);
+       if (ret_val == 0)
+               goto socket_getattr_return;
+
+       doi = ntohl(*(u32 *)&cipso_ptr[2]);
+       rcu_read_lock();
+       doi_def = cipso_v4_doi_getdef(doi);
+       if (doi_def == NULL) {
+               rcu_read_unlock();
+               goto socket_getattr_return;
+       }
+       switch (cipso_ptr[6]) {
+       case CIPSO_V4_TAG_RBITMAP:
+               ret_val = cipso_v4_parsetag_rbm(doi_def,
+                                               &cipso_ptr[6],
+                                               secattr);
+               break;
+       }
+       rcu_read_unlock();
+
+socket_getattr_return:
+       release_sock(sk);
+       return ret_val;
+}
+
+/**
+ * cipso_v4_skbuff_getattr - Get the security attributes from the CIPSO option
+ * @skb: the packet
+ * @secattr: the security attributes
+ *
+ * Description:
+ * Parse the given packet's CIPSO option and return the security attributes.
+ * Returns zero on success and negative values on failure.
+ *
+ */
+int cipso_v4_skbuff_getattr(const struct sk_buff *skb,
+                           struct netlbl_lsm_secattr *secattr)
+{
+       int ret_val = -ENOMSG;
+       unsigned char *cipso_ptr;
+       u32 doi;
+       struct cipso_v4_doi *doi_def;
+
+       if (!CIPSO_V4_OPTEXIST(skb))
+               return -ENOMSG;
+       cipso_ptr = CIPSO_V4_OPTPTR(skb);
+       if (cipso_v4_cache_check(cipso_ptr, cipso_ptr[1], secattr) == 0)
+               return 0;
+
+       doi = ntohl(*(u32 *)&cipso_ptr[2]);
+       rcu_read_lock();
+       doi_def = cipso_v4_doi_getdef(doi);
+       if (doi_def == NULL)
+               goto skbuff_getattr_return;
+       switch (cipso_ptr[6]) {
+       case CIPSO_V4_TAG_RBITMAP:
+               ret_val = cipso_v4_parsetag_rbm(doi_def,
+                                               &cipso_ptr[6],
+                                               secattr);
+               break;
+       }
+
+skbuff_getattr_return:
+       rcu_read_unlock();
+       return ret_val;
+}
+
+/*
+ * Setup Functions
+ */
+
+/**
+ * cipso_v4_init - Initialize the CIPSO module
+ *
+ * Description:
+ * Initialize the CIPSO module and prepare it for use.  Returns zero on success
+ * and negative values on failure.
+ *
+ */
+static int __init cipso_v4_init(void)
+{
+       int ret_val;
+
+       ret_val = cipso_v4_cache_init();
+       if (ret_val != 0)
+               panic("Failed to initialize the CIPSO/IPv4 cache (%d)\n",
+                     ret_val);
+
+       return 0;
+}
+
+subsys_initcall(cipso_v4_init);
index 70cea9d08a38f7972b9a7ce0a7f17242de1ce41e..19b2071ff319bf256ae048f104dbd9a16999b98b 100644 (file)
@@ -17,6 +17,7 @@
 #include <net/ip.h>
 #include <net/route.h>
 #include <net/tcp.h>
+#include <net/cipso_ipv4.h>
 
 /* From af_inet.c */
 extern int sysctl_ip_nonlocal_bind;
@@ -697,6 +698,40 @@ ctl_table ipv4_table[] = {
                .mode           = 0644,
                .proc_handler   = &proc_dointvec
        },
+#ifdef CONFIG_NETLABEL
+       {
+               .ctl_name       = NET_CIPSOV4_CACHE_ENABLE,
+               .procname       = "cipso_cache_enable",
+               .data           = &cipso_v4_cache_enabled,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
+       {
+               .ctl_name       = NET_CIPSOV4_CACHE_BUCKET_SIZE,
+               .procname       = "cipso_cache_bucket_size",
+               .data           = &cipso_v4_cache_bucketsize,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
+       {
+               .ctl_name       = NET_CIPSOV4_RBM_OPTFMT,
+               .procname       = "cipso_rbm_optfmt",
+               .data           = &cipso_v4_rbm_optfmt,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
+       {
+               .ctl_name       = NET_CIPSOV4_RBM_STRICTVALID,
+               .procname       = "cipso_rbm_strictvalid",
+               .data           = &cipso_v4_rbm_strictvalid,
+               .maxlen         = sizeof(int),
+               .mode           = 0644,
+               .proc_handler   = &proc_dointvec,
+       },
+#endif /* CONFIG_NETLABEL */
        { .ctl_name = 0 }
 };