XADV-2013003 Linux Kernel eCryptfs write_tag_3_packet Heap Buffer Overflow Vulnerability Nov 04 2013 05:16AM
geinblues gmail com

| XADV-2013003 Linux Kernel eCryptfs write_tag_3_packet Heap Buffer Overflow Vulnerability |

Vulnerable versions:
- linux kernel 2.6.18

Testbed: linux kernel 2.6.18
Type: Local
Impact: kernel panic or potential local privelge escalation.
Vendor: http://www.kernel.org
Author: x90c <geinblues *nospam* gmail dot com>
Site: x90c.org


The write_tag_3_packet() in Linux Kernel eCryptfs is vulnerable to heap buffer overflow.
It lead to kernel panic and potentialy privilege escalation. The vulnerability occured
with no checks to memory copy length variable in the vulnerable function.


The ecryptfs_create() in ecryptfs/inode.c called when operation to create an directory
in the ecryptfs file system. When after ecryptfs_create() called finally to reach to
the vulnerable point of memcpy in the vulnerable function of write_tag_3_packet.

[write_tag_3_packet() in ecryptfs/keystore.c]:

} else /* no aes, no 0, 24 key size? */
auth_tok->session_key.encrypted_key_size = crypt_stat->key_size; /* (1) */

key_rec->enc_key_size = /* (2) */

/* vulnerable point (2, 3 arguments usercontrollable) */
memcpy(key_rec->enc_key, auth_tok->session_key.encrypted_key, key_rec->enc_key_size);


If see The vulnerable point, key_rec->enc_key_size is usercontrolable variable.
(1) Store the crypt_stat->key_size to authtok->session_key.encrypted_key_size and (2)
store the authtok->session_key.encrypted_key_size to key_rec->enc_key_size.
In other word, the crypt_stat->key_size to the key_rec->enc_key_size usercontrollable
variable at the vulnerable point.

The Enter to the vulnerable point, no aes and 0, 24 key size.

If can control crypt_stat->key_size variable, lead to the heap buffer overflow.
First see the call path from ecryptfs_create() in ecryptfs/inode.c to vulnerable point.

[call path to the vulnerable point]
+-> ecryptfs/inode.c::ecryptfs_initialize_file()
| -> ecryptfs/crypto.c::ecryptfs_new_file_context()
+-> ecryptfs/inode.c::ecryptfs_write_metadata()
-> ecryptfs/crypto.c::ecryptfs_write_headers_virt()
-> ecryptfs/keystore.c::ecryptfs_generate_key_packet_set()
-> ecryptfs/keystore.c::write_tag_3_packet() (vulnerable function)
-> memcpy(key_rec->enc_key, /* vulnerable point */
key_rec->enc_key_size); // XXX usercontrolable key_rec->enc_key_size!

The ecryptfs_initialize_file in ecryptfs/inode.c store the crypt_stat->key_size by
the variable passed to kernel (mount options). In ecryptfs_new_file_context
store mount_crypt_stat->global_default_cipher_key_size to crypt_stat->key_size.
the global*key_size is the variable can control at the parse mount options.

int ecryptfs_new_file_context(struct dentry *ecryptfs_dentry)

crypt_stat->key_size =



The mount_crypt_stat->global_default_cipher_key_size can be set by parse
option 'ecryptfs_opt_ecryptfs_key_bytes'. (usercontrollable variable)

static int ecryptfs_parse_options(struct super_block *sb, char *options)

case ecryptfs_opt_ecryptfs_key_bytes:
cipher_key_bytes_src = args[0].from;
cipher_key_bytes =
&cipher_key_bytes_src, 0);
mount_crypt_stat->global_default_cipher_key_size =
"The mount_crypt_stat "
"global_default_cipher_key_size "
"set to: [%d]\n", mount_crypt_stat->
cipher_key_bytes_set = 1;

The memcpy copy size at vulnerable point can be set by usercontrolable variable
via the 'ecryptfs_opt_ecryptfs_key_bytes' mount option.


/* vulnerable point */
memcpy(key_rec->enc_key, auth_tok->session_key.encrypted_key, key_rec->enc_key_size);


The second argument also can user controllable, in this advisory just commented on it.

heap buffer overflow!





2013/11/04 - The vulnerability discovered.
2013/11/04 - Advisory released on full-disclosure, bugtraq, packetstorm, exploit-db


The authors reserve the right not to be responsible for the topicality,
correctness, completeness or quality of the information provided in this
document. Liability claims regarding damage caused by the use of any information
provided, including any kind of information which is incomplete or incorrect,
will therefore be rejected.

[ reply ]


Privacy Statement
Copyright 2010, SecurityFocus