2012-09-22 02:25:40 +04:00
/* RSA asymmetric public-key algorithm [RFC3447]
*
* Copyright ( C ) 2012 Red Hat , Inc . All Rights Reserved .
* Written by David Howells ( dhowells @ redhat . com )
*
* This program is free software ; you can redistribute it and / or
* modify it under the terms of the GNU General Public Licence
* as published by the Free Software Foundation ; either version
* 2 of the Licence , or ( at your option ) any later version .
*/
# define pr_fmt(fmt) "RSA: "fmt
# include <linux/module.h>
# include <linux/kernel.h>
# include <linux/slab.h>
crypto: crypto_memneq - add equality testing of memory regions w/o timing leaks
When comparing MAC hashes, AEAD authentication tags, or other hash
values in the context of authentication or integrity checking, it
is important not to leak timing information to a potential attacker,
i.e. when communication happens over a network.
Bytewise memory comparisons (such as memcmp) are usually optimized so
that they return a nonzero value as soon as a mismatch is found. E.g,
on x86_64/i5 for 512 bytes this can be ~50 cyc for a full mismatch
and up to ~850 cyc for a full match (cold). This early-return behavior
can leak timing information as a side channel, allowing an attacker to
iteratively guess the correct result.
This patch adds a new method crypto_memneq ("memory not equal to each
other") to the crypto API that compares memory areas of the same length
in roughly "constant time" (cache misses could change the timing, but
since they don't reveal information about the content of the strings
being compared, they are effectively benign). Iow, best and worst case
behaviour take the same amount of time to complete (in contrast to
memcmp).
Note that crypto_memneq (unlike memcmp) can only be used to test for
equality or inequality, NOT for lexicographical order. This, however,
is not an issue for its use-cases within the crypto API.
We tried to locate all of the places in the crypto API where memcmp was
being used for authentication or integrity checking, and convert them
over to crypto_memneq.
crypto_memneq is declared noinline, placed in its own source file,
and compiled with optimizations that might increase code size disabled
("Os") because a smart compiler (or LTO) might notice that the return
value is always compared against zero/nonzero, and might then
reintroduce the same early-return optimization that we are trying to
avoid.
Using #pragma or __attribute__ optimization annotations of the code
for disabling optimization was avoided as it seems to be considered
broken or unmaintained for long time in GCC [1]. Therefore, we work
around that by specifying the compile flag for memneq.o directly in
the Makefile. We found that this seems to be most appropriate.
As we use ("Os"), this patch also provides a loop-free "fast-path" for
frequently used 16 byte digests. Similarly to kernel library string
functions, leave an option for future even further optimized architecture
specific assembler implementations.
This was a joint work of James Yonan and Daniel Borkmann. Also thanks
for feedback from Florian Weimer on this and earlier proposals [2].
[1] http://gcc.gnu.org/ml/gcc/2012-07/msg00211.html
[2] https://lkml.org/lkml/2013/2/10/131
Signed-off-by: James Yonan <james@openvpn.net>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Florian Weimer <fw@deneb.enyo.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2013-09-26 12:20:39 +04:00
# include <crypto/algapi.h>
2012-09-22 02:25:40 +04:00
# include "public_key.h"
MODULE_LICENSE ( " GPL " ) ;
MODULE_DESCRIPTION ( " RSA Public Key Algorithm " ) ;
# define kenter(FMT, ...) \
pr_devel ( " ==> %s( " FMT " ) \n " , __func__ , # # __VA_ARGS__ )
# define kleave(FMT, ...) \
pr_devel ( " <== %s() " FMT " \n " , __func__ , # # __VA_ARGS__ )
/*
* Hash algorithm OIDs plus ASN .1 DER wrappings [ RFC4880 sec 5.2 .2 ] .
*/
static const u8 RSA_digest_info_MD5 [ ] = {
0x30 , 0x20 , 0x30 , 0x0C , 0x06 , 0x08 ,
0x2A , 0x86 , 0x48 , 0x86 , 0xF7 , 0x0D , 0x02 , 0x05 , /* OID */
0x05 , 0x00 , 0x04 , 0x10
} ;
static const u8 RSA_digest_info_SHA1 [ ] = {
0x30 , 0x21 , 0x30 , 0x09 , 0x06 , 0x05 ,
0x2B , 0x0E , 0x03 , 0x02 , 0x1A ,
0x05 , 0x00 , 0x04 , 0x14
} ;
static const u8 RSA_digest_info_RIPE_MD_160 [ ] = {
0x30 , 0x21 , 0x30 , 0x09 , 0x06 , 0x05 ,
0x2B , 0x24 , 0x03 , 0x02 , 0x01 ,
0x05 , 0x00 , 0x04 , 0x14
} ;
static const u8 RSA_digest_info_SHA224 [ ] = {
0x30 , 0x2d , 0x30 , 0x0d , 0x06 , 0x09 ,
0x60 , 0x86 , 0x48 , 0x01 , 0x65 , 0x03 , 0x04 , 0x02 , 0x04 ,
0x05 , 0x00 , 0x04 , 0x1C
} ;
static const u8 RSA_digest_info_SHA256 [ ] = {
0x30 , 0x31 , 0x30 , 0x0d , 0x06 , 0x09 ,
0x60 , 0x86 , 0x48 , 0x01 , 0x65 , 0x03 , 0x04 , 0x02 , 0x01 ,
0x05 , 0x00 , 0x04 , 0x20
} ;
static const u8 RSA_digest_info_SHA384 [ ] = {
0x30 , 0x41 , 0x30 , 0x0d , 0x06 , 0x09 ,
0x60 , 0x86 , 0x48 , 0x01 , 0x65 , 0x03 , 0x04 , 0x02 , 0x02 ,
0x05 , 0x00 , 0x04 , 0x30
} ;
static const u8 RSA_digest_info_SHA512 [ ] = {
0x30 , 0x51 , 0x30 , 0x0d , 0x06 , 0x09 ,
0x60 , 0x86 , 0x48 , 0x01 , 0x65 , 0x03 , 0x04 , 0x02 , 0x03 ,
0x05 , 0x00 , 0x04 , 0x40
} ;
static const struct {
const u8 * data ;
size_t size ;
} RSA_ASN1_templates [ PKEY_HASH__LAST ] = {
# define _(X) { RSA_digest_info_##X, sizeof(RSA_digest_info_##X) }
2013-05-06 16:58:15 +04:00
[ HASH_ALGO_MD5 ] = _ ( MD5 ) ,
[ HASH_ALGO_SHA1 ] = _ ( SHA1 ) ,
[ HASH_ALGO_RIPE_MD_160 ] = _ ( RIPE_MD_160 ) ,
[ HASH_ALGO_SHA256 ] = _ ( SHA256 ) ,
[ HASH_ALGO_SHA384 ] = _ ( SHA384 ) ,
[ HASH_ALGO_SHA512 ] = _ ( SHA512 ) ,
[ HASH_ALGO_SHA224 ] = _ ( SHA224 ) ,
2012-09-22 02:25:40 +04:00
# undef _
} ;
/*
* RSAVP1 ( ) function [ RFC3447 sec 5.2 .2 ]
*/
static int RSAVP1 ( const struct public_key * key , MPI s , MPI * _m )
{
MPI m ;
int ret ;
/* (1) Validate 0 <= s < n */
if ( mpi_cmp_ui ( s , 0 ) < 0 ) {
kleave ( " = -EBADMSG [s < 0] " ) ;
return - EBADMSG ;
}
if ( mpi_cmp ( s , key - > rsa . n ) > = 0 ) {
kleave ( " = -EBADMSG [s >= n] " ) ;
return - EBADMSG ;
}
m = mpi_alloc ( 0 ) ;
if ( ! m )
return - ENOMEM ;
/* (2) m = s^e mod n */
ret = mpi_powm ( m , s , key - > rsa . e , key - > rsa . n ) ;
if ( ret < 0 ) {
mpi_free ( m ) ;
return ret ;
}
* _m = m ;
return 0 ;
}
/*
* Integer to Octet String conversion [ RFC3447 sec 4.1 ]
*/
static int RSA_I2OSP ( MPI x , size_t xLen , u8 * * _X )
{
unsigned X_size , x_size ;
int X_sign ;
u8 * X ;
/* Make sure the string is the right length. The number should begin
* with { 0x00 , 0x01 , . . . } so we have to account for 15 leading zero
* bits not being reported by MPI .
*/
x_size = mpi_get_nbits ( x ) ;
pr_devel ( " size(x)=%u xLen*8=%zu \n " , x_size , xLen * 8 ) ;
if ( x_size ! = xLen * 8 - 15 )
return - ERANGE ;
X = mpi_get_buffer ( x , & X_size , & X_sign ) ;
if ( ! X )
return - ENOMEM ;
if ( X_sign < 0 ) {
kfree ( X ) ;
return - EBADMSG ;
}
if ( X_size ! = xLen - 1 ) {
kfree ( X ) ;
return - EBADMSG ;
}
* _X = X ;
return 0 ;
}
/*
* Perform the RSA signature verification .
* @ H : Value of hash of data and metadata
* @ EM : The computed signature value
* @ k : The size of EM ( EM [ 0 ] is an invalid location but should hold 0x00 )
* @ hash_size : The size of H
* @ asn1_template : The DigestInfo ASN .1 template
* @ asn1_size : Size of asm1_template [ ]
*/
static int RSA_verify ( const u8 * H , const u8 * EM , size_t k , size_t hash_size ,
const u8 * asn1_template , size_t asn1_size )
{
unsigned PS_end , T_offset , i ;
kenter ( " ,,%zu,%zu,%zu " , k , hash_size , asn1_size ) ;
if ( k < 2 + 1 + asn1_size + hash_size )
return - EBADMSG ;
/* Decode the EMSA-PKCS1-v1_5 */
if ( EM [ 1 ] ! = 0x01 ) {
kleave ( " = -EBADMSG [EM[1] == %02u] " , EM [ 1 ] ) ;
return - EBADMSG ;
}
T_offset = k - ( asn1_size + hash_size ) ;
PS_end = T_offset - 1 ;
if ( EM [ PS_end ] ! = 0x00 ) {
kleave ( " = -EBADMSG [EM[T-1] == %02u] " , EM [ PS_end ] ) ;
return - EBADMSG ;
}
for ( i = 2 ; i < PS_end ; i + + ) {
if ( EM [ i ] ! = 0xff ) {
kleave ( " = -EBADMSG [EM[PS%x] == %02u] " , i - 2 , EM [ i ] ) ;
return - EBADMSG ;
}
}
crypto: crypto_memneq - add equality testing of memory regions w/o timing leaks
When comparing MAC hashes, AEAD authentication tags, or other hash
values in the context of authentication or integrity checking, it
is important not to leak timing information to a potential attacker,
i.e. when communication happens over a network.
Bytewise memory comparisons (such as memcmp) are usually optimized so
that they return a nonzero value as soon as a mismatch is found. E.g,
on x86_64/i5 for 512 bytes this can be ~50 cyc for a full mismatch
and up to ~850 cyc for a full match (cold). This early-return behavior
can leak timing information as a side channel, allowing an attacker to
iteratively guess the correct result.
This patch adds a new method crypto_memneq ("memory not equal to each
other") to the crypto API that compares memory areas of the same length
in roughly "constant time" (cache misses could change the timing, but
since they don't reveal information about the content of the strings
being compared, they are effectively benign). Iow, best and worst case
behaviour take the same amount of time to complete (in contrast to
memcmp).
Note that crypto_memneq (unlike memcmp) can only be used to test for
equality or inequality, NOT for lexicographical order. This, however,
is not an issue for its use-cases within the crypto API.
We tried to locate all of the places in the crypto API where memcmp was
being used for authentication or integrity checking, and convert them
over to crypto_memneq.
crypto_memneq is declared noinline, placed in its own source file,
and compiled with optimizations that might increase code size disabled
("Os") because a smart compiler (or LTO) might notice that the return
value is always compared against zero/nonzero, and might then
reintroduce the same early-return optimization that we are trying to
avoid.
Using #pragma or __attribute__ optimization annotations of the code
for disabling optimization was avoided as it seems to be considered
broken or unmaintained for long time in GCC [1]. Therefore, we work
around that by specifying the compile flag for memneq.o directly in
the Makefile. We found that this seems to be most appropriate.
As we use ("Os"), this patch also provides a loop-free "fast-path" for
frequently used 16 byte digests. Similarly to kernel library string
functions, leave an option for future even further optimized architecture
specific assembler implementations.
This was a joint work of James Yonan and Daniel Borkmann. Also thanks
for feedback from Florian Weimer on this and earlier proposals [2].
[1] http://gcc.gnu.org/ml/gcc/2012-07/msg00211.html
[2] https://lkml.org/lkml/2013/2/10/131
Signed-off-by: James Yonan <james@openvpn.net>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Florian Weimer <fw@deneb.enyo.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2013-09-26 12:20:39 +04:00
if ( crypto_memneq ( asn1_template , EM + T_offset , asn1_size ) ! = 0 ) {
2012-09-22 02:25:40 +04:00
kleave ( " = -EBADMSG [EM[T] ASN.1 mismatch] " ) ;
return - EBADMSG ;
}
crypto: crypto_memneq - add equality testing of memory regions w/o timing leaks
When comparing MAC hashes, AEAD authentication tags, or other hash
values in the context of authentication or integrity checking, it
is important not to leak timing information to a potential attacker,
i.e. when communication happens over a network.
Bytewise memory comparisons (such as memcmp) are usually optimized so
that they return a nonzero value as soon as a mismatch is found. E.g,
on x86_64/i5 for 512 bytes this can be ~50 cyc for a full mismatch
and up to ~850 cyc for a full match (cold). This early-return behavior
can leak timing information as a side channel, allowing an attacker to
iteratively guess the correct result.
This patch adds a new method crypto_memneq ("memory not equal to each
other") to the crypto API that compares memory areas of the same length
in roughly "constant time" (cache misses could change the timing, but
since they don't reveal information about the content of the strings
being compared, they are effectively benign). Iow, best and worst case
behaviour take the same amount of time to complete (in contrast to
memcmp).
Note that crypto_memneq (unlike memcmp) can only be used to test for
equality or inequality, NOT for lexicographical order. This, however,
is not an issue for its use-cases within the crypto API.
We tried to locate all of the places in the crypto API where memcmp was
being used for authentication or integrity checking, and convert them
over to crypto_memneq.
crypto_memneq is declared noinline, placed in its own source file,
and compiled with optimizations that might increase code size disabled
("Os") because a smart compiler (or LTO) might notice that the return
value is always compared against zero/nonzero, and might then
reintroduce the same early-return optimization that we are trying to
avoid.
Using #pragma or __attribute__ optimization annotations of the code
for disabling optimization was avoided as it seems to be considered
broken or unmaintained for long time in GCC [1]. Therefore, we work
around that by specifying the compile flag for memneq.o directly in
the Makefile. We found that this seems to be most appropriate.
As we use ("Os"), this patch also provides a loop-free "fast-path" for
frequently used 16 byte digests. Similarly to kernel library string
functions, leave an option for future even further optimized architecture
specific assembler implementations.
This was a joint work of James Yonan and Daniel Borkmann. Also thanks
for feedback from Florian Weimer on this and earlier proposals [2].
[1] http://gcc.gnu.org/ml/gcc/2012-07/msg00211.html
[2] https://lkml.org/lkml/2013/2/10/131
Signed-off-by: James Yonan <james@openvpn.net>
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Cc: Florian Weimer <fw@deneb.enyo.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2013-09-26 12:20:39 +04:00
if ( crypto_memneq ( H , EM + T_offset + asn1_size , hash_size ) ! = 0 ) {
2012-09-22 02:25:40 +04:00
kleave ( " = -EKEYREJECTED [EM[T] hash mismatch] " ) ;
return - EKEYREJECTED ;
}
kleave ( " = 0 " ) ;
return 0 ;
}
/*
* Perform the verification step [ RFC3447 sec 8.2 .2 ] .
*/
static int RSA_verify_signature ( const struct public_key * key ,
const struct public_key_signature * sig )
{
size_t tsize ;
int ret ;
/* Variables as per RFC3447 sec 8.2.2 */
const u8 * H = sig - > digest ;
u8 * EM = NULL ;
MPI m = NULL ;
size_t k ;
kenter ( " " ) ;
if ( ! RSA_ASN1_templates [ sig - > pkey_hash_algo ] . data )
return - ENOTSUPP ;
/* (1) Check the signature size against the public key modulus size */
2012-09-22 02:28:05 +04:00
k = mpi_get_nbits ( key - > rsa . n ) ;
tsize = mpi_get_nbits ( sig - > rsa . s ) ;
2012-09-22 02:25:40 +04:00
2012-09-22 02:28:05 +04:00
/* According to RFC 4880 sec 3.2, length of MPI is computed starting
* from most significant bit . So the RFC 3447 sec 8.2 .2 size check
* must be relaxed to conform with shorter signatures - so we fail here
* only if signature length is longer than modulus size .
*/
2012-09-22 02:25:40 +04:00
pr_devel ( " step 1: k=%zu size(S)=%zu \n " , k , tsize ) ;
2012-09-22 02:28:05 +04:00
if ( k < tsize ) {
2012-09-22 02:25:40 +04:00
ret = - EBADMSG ;
goto error ;
}
2012-09-22 02:28:05 +04:00
/* Round up and convert to octets */
k = ( k + 7 ) / 8 ;
2012-09-22 02:25:40 +04:00
/* (2b) Apply the RSAVP1 verification primitive to the public key */
ret = RSAVP1 ( key , sig - > rsa . s , & m ) ;
if ( ret < 0 )
goto error ;
/* (2c) Convert the message representative (m) to an encoded message
* ( EM ) of length k octets .
*
* NOTE ! The leading zero byte is suppressed by MPI , so we pass a
* pointer to the _preceding_ byte to RSA_verify ( ) !
*/
ret = RSA_I2OSP ( m , k , & EM ) ;
if ( ret < 0 )
goto error ;
ret = RSA_verify ( H , EM - 1 , k , sig - > digest_size ,
RSA_ASN1_templates [ sig - > pkey_hash_algo ] . data ,
RSA_ASN1_templates [ sig - > pkey_hash_algo ] . size ) ;
error :
kfree ( EM ) ;
mpi_free ( m ) ;
kleave ( " = %d " , ret ) ;
return ret ;
}
const struct public_key_algorithm RSA_public_key_algorithm = {
. name = " RSA " ,
. n_pub_mpi = 2 ,
. n_sec_mpi = 3 ,
. n_sig_mpi = 1 ,
. verify_signature = RSA_verify_signature ,
} ;
EXPORT_SYMBOL_GPL ( RSA_public_key_algorithm ) ;