2015-05-25 21:10:20 +08:00
/*
2015-06-23 22:18:54 +08:00
* Non - physical true random number generator based on timing jitter - -
* Jitter RNG standalone code .
2015-05-25 21:10:20 +08:00
*
2020-04-18 03:33:33 +08:00
* Copyright Stephan Mueller < smueller @ chronox . de > , 2015 - 2020
2015-05-25 21:10:20 +08:00
*
* Design
* = = = = = =
*
2020-07-20 00:49:59 +08:00
* See https : //www.chronox.de/jent.html
2015-05-25 21:10:20 +08:00
*
* License
* = = = = = = =
*
* Redistribution and use in source and binary forms , with or without
* modification , are permitted provided that the following conditions
* are met :
* 1. Redistributions of source code must retain the above copyright
* notice , and the entire permission notice in its entirety ,
* including the disclaimer of warranties .
* 2. Redistributions in binary form must reproduce the above copyright
* notice , this list of conditions and the following disclaimer in the
* documentation and / or other materials provided with the distribution .
* 3. The name of the author may not be used to endorse or promote
* products derived from this software without specific prior
* written permission .
*
* ALTERNATIVELY , this product may be distributed under the terms of
* the GNU General Public License , in which case the provisions of the GPL2 are
* required INSTEAD OF the above restrictions . ( This clause is
* necessary due to a potential bad interaction between the GPL and
* the restrictions contained in a BSD - style copyright . )
*
* THIS SOFTWARE IS PROVIDED ` ` AS IS ' ' AND ANY EXPRESS OR IMPLIED
* WARRANTIES , INCLUDING , BUT NOT LIMITED TO , THE IMPLIED WARRANTIES
* OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE , ALL OF
* WHICH ARE HEREBY DISCLAIMED . IN NO EVENT SHALL THE AUTHOR BE
* LIABLE FOR ANY DIRECT , INDIRECT , INCIDENTAL , SPECIAL , EXEMPLARY , OR
* CONSEQUENTIAL DAMAGES ( INCLUDING , BUT NOT LIMITED TO , PROCUREMENT
* OF SUBSTITUTE GOODS OR SERVICES ; LOSS OF USE , DATA , OR PROFITS ; OR
* BUSINESS INTERRUPTION ) HOWEVER CAUSED AND ON ANY THEORY OF
* LIABILITY , WHETHER IN CONTRACT , STRICT LIABILITY , OR TORT
* ( INCLUDING NEGLIGENCE OR OTHERWISE ) ARISING IN ANY WAY OUT OF THE
* USE OF THIS SOFTWARE , EVEN IF NOT ADVISED OF THE POSSIBILITY OF SUCH
* DAMAGE .
*/
/*
* This Jitterentropy RNG is based on the jitterentropy library
2020-07-20 00:49:59 +08:00
* version 2.2 .0 provided at https : //www.chronox.de/jent.html
2015-05-25 21:10:20 +08:00
*/
2015-06-23 22:18:54 +08:00
# ifdef __OPTIMIZE__
# error "The CPU Jitter random number generator must not be compiled with optimizations. See documentation. Use the compiler switch -O0 for compiling jitterentropy.c."
# endif
typedef unsigned long long __u64 ;
typedef long long __s64 ;
typedef unsigned int __u32 ;
# define NULL ((void *) 0)
2015-05-25 21:10:20 +08:00
/* The entropy pool */
struct rand_data {
/* all data values that are vital to maintain the security
* of the RNG are marked as SENSITIVE . A user must not
* access that information while the RNG executes its loops to
* calculate the next random value . */
__u64 data ; /* SENSITIVE Actual random number */
__u64 old_data ; /* SENSITIVE Previous random number */
__u64 prev_time ; /* SENSITIVE Previous time stamp */
# define DATA_SIZE_BITS ((sizeof(__u64)) * 8)
__u64 last_delta ; /* SENSITIVE stuck test */
__s64 last_delta2 ; /* SENSITIVE stuck test */
unsigned int osr ; /* Oversample rate */
# define JENT_MEMORY_BLOCKS 64
# define JENT_MEMORY_BLOCKSIZE 32
# define JENT_MEMORY_ACCESSLOOPS 128
# define JENT_MEMORY_SIZE (JENT_MEMORY_BLOCKS*JENT_MEMORY_BLOCKSIZE)
unsigned char * mem ; /* Memory access location with size of
* memblocks * memblocksize */
unsigned int memlocation ; /* Pointer to byte in *mem */
unsigned int memblocks ; /* Number of memory blocks in *mem */
unsigned int memblocksize ; /* Size of one memory block in bytes */
unsigned int memaccessloops ; /* Number of memory accesses per random
* bit generation */
2020-04-18 03:33:33 +08:00
/* Repetition Count Test */
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
unsigned int rct_count ; /* Number of stuck values */
2020-04-18 03:33:33 +08:00
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
/* Intermittent health test failure threshold of 2^-30 */
# define JENT_RCT_CUTOFF 30 /* Taken from SP800-90B sec 4.4.1 */
2020-04-18 03:33:33 +08:00
# define JENT_APT_CUTOFF 325 /* Taken from SP800-90B sec 4.4.2 */
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
/* Permanent health test failure threshold of 2^-60 */
# define JENT_RCT_CUTOFF_PERMANENT 60
# define JENT_APT_CUTOFF_PERMANENT 355
2020-04-18 03:33:33 +08:00
# define JENT_APT_WINDOW_SIZE 512 /* Data window size */
/* LSB of time stamp to process */
# define JENT_APT_LSB 16
# define JENT_APT_WORD_MASK (JENT_APT_LSB - 1)
unsigned int apt_observations ; /* Number of collected observations */
unsigned int apt_count ; /* APT counter */
unsigned int apt_base ; /* APT base reference */
unsigned int apt_base_set : 1 ; /* APT base reference set? */
2015-05-25 21:10:20 +08:00
} ;
/* Flags that can be used to initialize the RNG */
# define JENT_DISABLE_MEMORY_ACCESS (1<<2) / * Disable memory access for more
* entropy , saves MEMORY_SIZE RAM for
* entropy collector */
/* -- error codes for init function -- */
# define JENT_ENOTIME 1 /* Timer service not available */
# define JENT_ECOARSETIME 2 /* Timer too coarse for RNG */
# define JENT_ENOMONOTONIC 3 /* Timer is not monotonic increasing */
# define JENT_EVARVAR 5 / * Timer does not produce variations of
* variations ( 2 nd derivation of time is
* zero ) . */
2019-05-30 03:24:25 +08:00
# define JENT_ESTUCK 8 /* Too many stuck results during init. */
2020-04-18 03:33:33 +08:00
# define JENT_EHEALTH 9 /* Health test failed during initialization */
# define JENT_ERCT 10 /* RCT failed during initialization */
crypto: jitter - add oversampling of noise source
The output n bits can receive more than n bits of min entropy, of course,
but the fixed output of the conditioning function can only asymptotically
approach the output size bits of min entropy, not attain that bound.
Random maps will tend to have output collisions, which reduces the
creditable output entropy (that is what SP 800-90B Section 3.1.5.1.2
attempts to bound).
The value "64" is justified in Appendix A.4 of the current 90C draft,
and aligns with NIST's in "epsilon" definition in this document, which is
that a string can be considered "full entropy" if you can bound the min
entropy in each bit of output to at least 1-epsilon, where epsilon is
required to be <= 2^(-32).
Note, this patch causes the Jitter RNG to cut its performance in half in
FIPS mode because the conditioning function of the LFSR produces 64 bits
of entropy in one block. The oversampling requires that additionally 64
bits of entropy are sampled from the noise source. If the conditioner is
changed, such as using SHA-256, the impact of the oversampling is only
one fourth, because for the 256 bit block of the conditioner, only 64
additional bits from the noise source must be sampled.
This patch is derived from the user space jitterentropy-library.
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Simo Sorce <simo@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2021-12-20 14:21:53 +08:00
/*
* The output n bits can receive more than n bits of min entropy , of course ,
* but the fixed output of the conditioning function can only asymptotically
* approach the output size bits of min entropy , not attain that bound . Random
* maps will tend to have output collisions , which reduces the creditable
* output entropy ( that is what SP 800 - 90 B Section 3.1 .5 .1 .2 attempts to bound ) .
*
* The value " 64 " is justified in Appendix A .4 of the current 90 C draft ,
* and aligns with NIST ' s in " epsilon " definition in this document , which is
* that a string can be considered " full entropy " if you can bound the min
* entropy in each bit of output to at least 1 - epsilon , where epsilon is
* required to be < = 2 ^ ( - 32 ) .
*/
# define JENT_ENTROPY_SAFETY_FACTOR 64
# include <linux/fips.h>
2020-04-18 03:33:33 +08:00
# include "jitterentropy.h"
2015-05-25 21:10:20 +08:00
/***************************************************************************
2020-04-18 03:33:33 +08:00
* Adaptive Proportion Test
*
* This test complies with SP800 - 90 B section 4.4 .2 .
2015-05-25 21:10:20 +08:00
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2021-08-25 05:05:13 +08:00
/*
2020-04-18 03:33:33 +08:00
* Reset the APT counter
*
* @ ec [ in ] Reference to entropy collector
*/
static void jent_apt_reset ( struct rand_data * ec , unsigned int delta_masked )
{
/* Reset APT counter */
ec - > apt_count = 0 ;
ec - > apt_base = delta_masked ;
ec - > apt_observations = 0 ;
}
2021-08-25 05:05:13 +08:00
/*
2020-04-18 03:33:33 +08:00
* Insert a new entropy event into APT
*
* @ ec [ in ] Reference to entropy collector
* @ delta_masked [ in ] Masked time delta to process
*/
static void jent_apt_insert ( struct rand_data * ec , unsigned int delta_masked )
{
/* Initialize the base reference */
if ( ! ec - > apt_base_set ) {
ec - > apt_base = delta_masked ;
ec - > apt_base_set = 1 ;
return ;
}
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
if ( delta_masked = = ec - > apt_base )
2020-04-18 03:33:33 +08:00
ec - > apt_count + + ;
ec - > apt_observations + + ;
if ( ec - > apt_observations > = JENT_APT_WINDOW_SIZE )
jent_apt_reset ( ec , delta_masked ) ;
}
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
/* APT health test failure detection */
static int jent_apt_permanent_failure ( struct rand_data * ec )
{
return ( ec - > apt_count > = JENT_APT_CUTOFF_PERMANENT ) ? 1 : 0 ;
}
static int jent_apt_failure ( struct rand_data * ec )
{
return ( ec - > apt_count > = JENT_APT_CUTOFF ) ? 1 : 0 ;
}
2020-04-18 03:33:33 +08:00
/***************************************************************************
* Stuck Test and its use as Repetition Count Test
*
* The Jitter RNG uses an enhanced version of the Repetition Count Test
* ( RCT ) specified in SP800 - 90 B section 4.4 .1 . Instead of counting identical
* back - to - back values , the input to the RCT is the counting of the stuck
* values during the generation of one Jitter RNG output block .
*
* The RCT is applied with an alpha of 2 ^ { - 30 } compliant to FIPS 140 - 2 IG 9.8 .
*
* During the counting operation , the Jitter RNG always calculates the RCT
* cut - off value of C . If that value exceeds the allowed cut - off value ,
* the Jitter RNG output block will be calculated completely but discarded at
* the end . The caller of the Jitter RNG is informed with an error code .
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2021-08-25 05:05:13 +08:00
/*
2020-04-18 03:33:33 +08:00
* Repetition Count Test as defined in SP800 - 90 B section 4.4 .1
*
* @ ec [ in ] Reference to entropy collector
* @ stuck [ in ] Indicator whether the value is stuck
*/
static void jent_rct_insert ( struct rand_data * ec , int stuck )
{
if ( stuck ) {
ec - > rct_count + + ;
} else {
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
/* Reset RCT */
2020-04-18 03:33:33 +08:00
ec - > rct_count = 0 ;
}
}
static inline __u64 jent_delta ( __u64 prev , __u64 next )
{
# define JENT_UINT64_MAX (__u64)(~((__u64) 0))
return ( prev < next ) ? ( next - prev ) :
( JENT_UINT64_MAX - prev + 1 + next ) ;
}
2021-08-25 05:05:13 +08:00
/*
2020-04-18 03:33:33 +08:00
* Stuck test by checking the :
* 1 st derivative of the jitter measurement ( time delta )
* 2 nd derivative of the jitter measurement ( delta of time deltas )
* 3 rd derivative of the jitter measurement ( delta of delta of time deltas )
*
* All values must always be non - zero .
*
* @ ec [ in ] Reference to entropy collector
* @ current_delta [ in ] Jitter time delta
*
* @ return
* 0 jitter measurement not stuck ( good bit )
* 1 jitter measurement stuck ( reject bit )
*/
static int jent_stuck ( struct rand_data * ec , __u64 current_delta )
{
__u64 delta2 = jent_delta ( ec - > last_delta , current_delta ) ;
__u64 delta3 = jent_delta ( ec - > last_delta2 , delta2 ) ;
ec - > last_delta = current_delta ;
ec - > last_delta2 = delta2 ;
/*
* Insert the result of the comparison of two back - to - back time
* deltas .
*/
2021-11-21 22:14:20 +08:00
jent_apt_insert ( ec , current_delta ) ;
2020-04-18 03:33:33 +08:00
if ( ! current_delta | | ! delta2 | | ! delta3 ) {
/* RCT with a stuck bit */
jent_rct_insert ( ec , 1 ) ;
return 1 ;
}
/* RCT with a non-stuck bit */
jent_rct_insert ( ec , 0 ) ;
return 0 ;
}
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
/* RCT health test failure detection */
static int jent_rct_permanent_failure ( struct rand_data * ec )
{
return ( ec - > rct_count > = JENT_RCT_CUTOFF_PERMANENT ) ? 1 : 0 ;
}
static int jent_rct_failure ( struct rand_data * ec )
{
return ( ec - > rct_count > = JENT_RCT_CUTOFF ) ? 1 : 0 ;
}
/* Report of health test failures */
2020-04-18 03:33:33 +08:00
static int jent_health_failure ( struct rand_data * ec )
{
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
return jent_rct_failure ( ec ) | jent_apt_failure ( ec ) ;
}
static int jent_permanent_health_failure ( struct rand_data * ec )
{
return jent_rct_permanent_failure ( ec ) | jent_apt_permanent_failure ( ec ) ;
2020-04-18 03:33:33 +08:00
}
/***************************************************************************
* Noise sources
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2015-05-25 21:10:20 +08:00
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* Update of the loop count used for the next round of
* an entropy collection .
*
* Input :
* @ ec entropy collector struct - - may be NULL
* @ bits is the number of low bits of the timer to consider
* @ min is the number of bits we shift the timer value to the right at
* the end to make sure we have a guaranteed minimum value
*
* @ return Newly calculated loop counter
*/
static __u64 jent_loop_shuffle ( struct rand_data * ec ,
unsigned int bits , unsigned int min )
{
__u64 time = 0 ;
__u64 shuffle = 0 ;
unsigned int i = 0 ;
unsigned int mask = ( 1 < < bits ) - 1 ;
jent_get_nstime ( & time ) ;
/*
2019-05-30 03:24:25 +08:00
* Mix the current state of the random number into the shuffle
* calculation to balance that shuffle a bit more .
2015-05-25 21:10:20 +08:00
*/
if ( ec )
time ^ = ec - > data ;
/*
2019-05-30 03:24:25 +08:00
* We fold the time value as much as possible to ensure that as many
* bits of the time stamp are included as possible .
2015-05-25 21:10:20 +08:00
*/
2019-05-30 03:24:25 +08:00
for ( i = 0 ; ( ( DATA_SIZE_BITS + bits - 1 ) / bits ) > i ; i + + ) {
2015-05-25 21:10:20 +08:00
shuffle ^ = time & mask ;
time = time > > bits ;
}
/*
* We add a lower boundary value to ensure we have a minimum
* RNG loop count .
*/
return ( shuffle + ( 1 < < min ) ) ;
}
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* CPU Jitter noise source - - this is the noise source based on the CPU
* execution time jitter
*
2019-05-30 03:24:25 +08:00
* This function injects the individual bits of the time value into the
* entropy pool using an LFSR .
2015-05-25 21:10:20 +08:00
*
2019-05-30 03:24:25 +08:00
* The code is deliberately inefficient with respect to the bit shifting
* and shall stay that way . This function is the root cause why the code
* shall be compiled without optimization . This function not only acts as
* folding operation , but this function ' s execution is used to measure
* the CPU execution time jitter . Any change to the loop in this function
* implies that careful retesting must be done .
2015-05-25 21:10:20 +08:00
*
2020-04-18 03:33:33 +08:00
* @ ec [ in ] entropy collector struct
* @ time [ in ] time stamp to be injected
* @ loop_cnt [ in ] if a value not equal to 0 is set , use the given value as
* number of loops to perform the folding
* @ stuck [ in ] Is the time stamp identified as stuck ?
2015-05-25 21:10:20 +08:00
*
* Output :
2019-05-30 03:24:25 +08:00
* updated ec - > data
2015-05-25 21:10:20 +08:00
*
* @ return Number of loops the folding operation is performed
*/
2020-04-18 03:33:33 +08:00
static void jent_lfsr_time ( struct rand_data * ec , __u64 time , __u64 loop_cnt ,
int stuck )
2015-05-25 21:10:20 +08:00
{
unsigned int i ;
__u64 j = 0 ;
__u64 new = 0 ;
# define MAX_FOLD_LOOP_BIT 4
# define MIN_FOLD_LOOP_BIT 0
__u64 fold_loop_cnt =
jent_loop_shuffle ( ec , MAX_FOLD_LOOP_BIT , MIN_FOLD_LOOP_BIT ) ;
/*
* testing purposes - - allow test app to set the counter , not
* needed during runtime
*/
if ( loop_cnt )
fold_loop_cnt = loop_cnt ;
for ( j = 0 ; j < fold_loop_cnt ; j + + ) {
2019-05-30 03:24:25 +08:00
new = ec - > data ;
2015-05-25 21:10:20 +08:00
for ( i = 1 ; ( DATA_SIZE_BITS ) > = i ; i + + ) {
__u64 tmp = time < < ( DATA_SIZE_BITS - i ) ;
tmp = tmp > > ( DATA_SIZE_BITS - 1 ) ;
2019-05-30 03:24:25 +08:00
/*
* Fibonacci LSFR with polynomial of
* x ^ 64 + x ^ 61 + x ^ 56 + x ^ 31 + x ^ 28 + x ^ 23 + 1 which is
* primitive according to
* http : //poincare.matf.bg.ac.rs/~ezivkovm/publications/primpol1.pdf
* ( the shift values are the polynomial values minus one
* due to counting bits from 0 to 63 ) . As the current
* position is always the LSB , the polynomial only needs
* to shift data in from the left without wrap .
*/
tmp ^ = ( ( new > > 63 ) & 1 ) ;
tmp ^ = ( ( new > > 60 ) & 1 ) ;
tmp ^ = ( ( new > > 55 ) & 1 ) ;
tmp ^ = ( ( new > > 30 ) & 1 ) ;
tmp ^ = ( ( new > > 27 ) & 1 ) ;
tmp ^ = ( ( new > > 22 ) & 1 ) ;
new < < = 1 ;
2015-05-25 21:10:20 +08:00
new ^ = tmp ;
}
}
2019-05-30 03:24:25 +08:00
2020-04-18 03:33:33 +08:00
/*
* If the time stamp is stuck , do not finally insert the value into
* the entropy pool . Although this operation should not do any harm
* even when the time stamp has no entropy , SP800 - 90 B requires that
* any conditioning operation ( SP800 - 90 B considers the LFSR to be a
* conditioning operation ) to have an identical amount of input
* data according to section 3.1 .5 .
*/
if ( ! stuck )
ec - > data = new ;
2015-05-25 21:10:20 +08:00
}
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* Memory Access noise source - - this is a noise source based on variations in
* memory access times
*
* This function performs memory accesses which will add to the timing
* variations due to an unknown amount of CPU wait states that need to be
* added when accessing memory . The memory size should be larger than the L1
* caches as outlined in the documentation and the associated testing .
*
* The L1 cache has a very high bandwidth , albeit its access rate is usually
* slower than accessing CPU registers . Therefore , L1 accesses only add minimal
* variations as the CPU has hardly to wait . Starting with L2 , significant
* variations are added because L2 typically does not belong to the CPU any more
* and therefore a wider range of CPU wait states is necessary for accesses .
* L3 and real memory accesses have even a wider range of wait states . However ,
* to reliably access either L3 or memory , the ec - > mem memory must be quite
* large which is usually not desirable .
*
2020-04-18 03:33:33 +08:00
* @ ec [ in ] Reference to the entropy collector with the memory access data - - if
* the reference to the memory block to be accessed is NULL , this noise
* source is disabled
* @ loop_cnt [ in ] if a value not equal to 0 is set , use the given value
* number of loops to perform the LFSR
2015-05-25 21:10:20 +08:00
*/
2020-04-18 03:33:33 +08:00
static void jent_memaccess ( struct rand_data * ec , __u64 loop_cnt )
2015-05-25 21:10:20 +08:00
{
unsigned int wrap = 0 ;
__u64 i = 0 ;
# define MAX_ACC_LOOP_BIT 7
# define MIN_ACC_LOOP_BIT 0
__u64 acc_loop_cnt =
jent_loop_shuffle ( ec , MAX_ACC_LOOP_BIT , MIN_ACC_LOOP_BIT ) ;
if ( NULL = = ec | | NULL = = ec - > mem )
2020-04-18 03:33:33 +08:00
return ;
2015-05-25 21:10:20 +08:00
wrap = ec - > memblocksize * ec - > memblocks ;
/*
* testing purposes - - allow test app to set the counter , not
* needed during runtime
*/
if ( loop_cnt )
acc_loop_cnt = loop_cnt ;
for ( i = 0 ; i < ( ec - > memaccessloops + acc_loop_cnt ) ; i + + ) {
2019-05-30 03:24:25 +08:00
unsigned char * tmpval = ec - > mem + ec - > memlocation ;
2015-05-25 21:10:20 +08:00
/*
* memory access : just add 1 to one byte ,
* wrap at 255 - - memory access implies read
* from and write to memory location
*/
* tmpval = ( * tmpval + 1 ) & 0xff ;
/*
* Addition of memblocksize - 1 to pointer
* with wrap around logic to ensure that every
* memory location is hit evenly
*/
ec - > memlocation = ec - > memlocation + ec - > memblocksize - 1 ;
ec - > memlocation = ec - > memlocation % wrap ;
}
}
/***************************************************************************
* Start of entropy processing logic
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* This is the heart of the entropy generation : calculate time deltas and
2019-05-30 03:24:25 +08:00
* use the CPU jitter in the time deltas . The jitter is injected into the
* entropy pool .
2015-05-25 21:10:20 +08:00
*
* WARNING : ensure that - > prev_time is primed before using the output
* of this function ! This can be done by calling this function
* and not using its result .
*
2020-04-18 03:33:33 +08:00
* @ ec [ in ] Reference to entropy collector
2015-05-25 21:10:20 +08:00
*
2019-05-30 03:24:25 +08:00
* @ return result of stuck test
2015-05-25 21:10:20 +08:00
*/
2019-05-30 03:24:25 +08:00
static int jent_measure_jitter ( struct rand_data * ec )
2015-05-25 21:10:20 +08:00
{
__u64 time = 0 ;
__u64 current_delta = 0 ;
2020-04-18 03:33:33 +08:00
int stuck ;
2015-05-25 21:10:20 +08:00
/* Invoke one noise source before time measurement to add variations */
jent_memaccess ( ec , 0 ) ;
/*
* Get time stamp and calculate time delta to previous
* invocation to measure the timing variations
*/
jent_get_nstime ( & time ) ;
2020-04-18 03:33:33 +08:00
current_delta = jent_delta ( ec - > prev_time , time ) ;
2015-05-25 21:10:20 +08:00
ec - > prev_time = time ;
2020-04-18 03:33:33 +08:00
/* Check whether we have a stuck measurement. */
stuck = jent_stuck ( ec , current_delta ) ;
2019-05-30 03:24:25 +08:00
/* Now call the next noise sources which also injects the data */
2020-04-18 03:33:33 +08:00
jent_lfsr_time ( ec , current_delta , 0 , stuck ) ;
2015-05-25 21:10:20 +08:00
2020-04-18 03:33:33 +08:00
return stuck ;
2015-05-25 21:10:20 +08:00
}
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* Generator of one 64 bit random number
* Function fills rand_data - > data
*
2020-04-18 03:33:33 +08:00
* @ ec [ in ] Reference to entropy collector
2015-05-25 21:10:20 +08:00
*/
static void jent_gen_entropy ( struct rand_data * ec )
{
crypto: jitter - add oversampling of noise source
The output n bits can receive more than n bits of min entropy, of course,
but the fixed output of the conditioning function can only asymptotically
approach the output size bits of min entropy, not attain that bound.
Random maps will tend to have output collisions, which reduces the
creditable output entropy (that is what SP 800-90B Section 3.1.5.1.2
attempts to bound).
The value "64" is justified in Appendix A.4 of the current 90C draft,
and aligns with NIST's in "epsilon" definition in this document, which is
that a string can be considered "full entropy" if you can bound the min
entropy in each bit of output to at least 1-epsilon, where epsilon is
required to be <= 2^(-32).
Note, this patch causes the Jitter RNG to cut its performance in half in
FIPS mode because the conditioning function of the LFSR produces 64 bits
of entropy in one block. The oversampling requires that additionally 64
bits of entropy are sampled from the noise source. If the conditioner is
changed, such as using SHA-256, the impact of the oversampling is only
one fourth, because for the 256 bit block of the conditioner, only 64
additional bits from the noise source must be sampled.
This patch is derived from the user space jitterentropy-library.
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Simo Sorce <simo@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2021-12-20 14:21:53 +08:00
unsigned int k = 0 , safety_factor = 0 ;
if ( fips_enabled )
safety_factor = JENT_ENTROPY_SAFETY_FACTOR ;
2015-05-25 21:10:20 +08:00
/* priming of the ->prev_time value */
jent_measure_jitter ( ec ) ;
2021-11-30 22:10:09 +08:00
while ( ! jent_health_failure ( ec ) ) {
2019-05-30 03:24:25 +08:00
/* If a stuck measurement is received, repeat measurement */
if ( jent_measure_jitter ( ec ) )
2015-05-25 21:10:20 +08:00
continue ;
/*
* We multiply the loop value with - > osr to obtain the
* oversampling rate requested by the caller
*/
crypto: jitter - add oversampling of noise source
The output n bits can receive more than n bits of min entropy, of course,
but the fixed output of the conditioning function can only asymptotically
approach the output size bits of min entropy, not attain that bound.
Random maps will tend to have output collisions, which reduces the
creditable output entropy (that is what SP 800-90B Section 3.1.5.1.2
attempts to bound).
The value "64" is justified in Appendix A.4 of the current 90C draft,
and aligns with NIST's in "epsilon" definition in this document, which is
that a string can be considered "full entropy" if you can bound the min
entropy in each bit of output to at least 1-epsilon, where epsilon is
required to be <= 2^(-32).
Note, this patch causes the Jitter RNG to cut its performance in half in
FIPS mode because the conditioning function of the LFSR produces 64 bits
of entropy in one block. The oversampling requires that additionally 64
bits of entropy are sampled from the noise source. If the conditioner is
changed, such as using SHA-256, the impact of the oversampling is only
one fourth, because for the 256 bit block of the conditioner, only 64
additional bits from the noise source must be sampled.
This patch is derived from the user space jitterentropy-library.
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Simo Sorce <simo@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2021-12-20 14:21:53 +08:00
if ( + + k > = ( ( DATA_SIZE_BITS + safety_factor ) * ec - > osr ) )
2015-05-25 21:10:20 +08:00
break ;
}
}
2021-08-25 05:05:13 +08:00
/*
2015-05-25 21:10:20 +08:00
* Entry function : Obtain entropy for the caller .
*
* This function invokes the entropy gathering logic as often to generate
* as many bytes as requested by the caller . The entropy gathering logic
* creates 64 bit per invocation .
*
* This function truncates the last 64 bit entropy value output to the exact
* size specified by the caller .
*
2020-04-18 03:33:33 +08:00
* @ ec [ in ] Reference to entropy collector
* @ data [ in ] pointer to buffer for storing random data - - buffer must already
* exist
* @ len [ in ] size of the buffer , specifying also the requested number of random
* in bytes
2015-05-25 21:10:20 +08:00
*
* @ return 0 when request is fulfilled or an error
*
* The following error codes can occur :
* - 1 entropy_collector is NULL
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
* - 2 Intermittent health failure
* - 3 Permanent health failure
2015-05-25 21:10:20 +08:00
*/
2015-06-23 22:18:54 +08:00
int jent_read_entropy ( struct rand_data * ec , unsigned char * data ,
unsigned int len )
2015-05-25 21:10:20 +08:00
{
2015-06-23 22:18:54 +08:00
unsigned char * p = data ;
2015-05-25 21:10:20 +08:00
if ( ! ec )
2015-06-23 22:18:54 +08:00
return - 1 ;
2015-05-25 21:10:20 +08:00
2021-03-17 09:44:03 +08:00
while ( len > 0 ) {
2015-06-23 22:18:54 +08:00
unsigned int tocopy ;
2015-05-25 21:10:20 +08:00
jent_gen_entropy ( ec ) ;
2020-04-18 03:33:33 +08:00
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
if ( jent_permanent_health_failure ( ec ) ) {
2020-04-18 03:33:33 +08:00
/*
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
* At this point , the Jitter RNG instance is considered
* as a failed instance . There is no rerun of the
* startup test any more , because the caller
* is assumed to not further use this instance .
2020-04-18 03:33:33 +08:00
*/
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
return - 3 ;
} else if ( jent_health_failure ( ec ) ) {
2020-04-18 03:33:33 +08:00
/*
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
* Perform startup health tests and return permanent
* error if it fails .
2020-04-18 03:33:33 +08:00
*/
crypto: jitter - permanent and intermittent health errors
According to SP800-90B, two health failures are allowed: the intermittend
and the permanent failure. So far, only the intermittent failure was
implemented. The permanent failure was achieved by resetting the entire
entropy source including its health test state and waiting for two or
more back-to-back health errors.
This approach is appropriate for RCT, but not for APT as APT has a
non-linear cutoff value. Thus, this patch implements 2 cutoff values
for both RCT/APT. This implies that the health state is left untouched
when an intermittent failure occurs. The noise source is reset
and a new APT powerup-self test is performed. Yet, whith the unchanged
health test state, the counting of failures continues until a permanent
failure is reached.
Any non-failing raw entropy value causes the health tests to reset.
The intermittent error has an unchanged significance level of 2^-30.
The permanent error has a significance level of 2^-60. Considering that
this level also indicates a false-positive rate (see SP800-90B section 4.2)
a false-positive must only be incurred with a low probability when
considering a fleet of Linux kernels as a whole. Hitting the permanent
error may cause a panic(), the following calculation applies: Assuming
that a fleet of 10^9 Linux kernels run concurrently with this patch in
FIPS mode and on each kernel 2 health tests are performed every minute
for one year, the chances of a false positive is about 1:1000
based on the binomial distribution.
In addition, any power-up health test errors triggered with
jent_entropy_init are treated as permanent errors.
A permanent failure causes the entire entropy source to permanently
return an error. This implies that a caller can only remedy the situation
by re-allocating a new instance of the Jitter RNG. In a subsequent
patch, a transparent re-allocation will be provided which also changes
the implied heuristic entropy assessment.
In addition, when the kernel is booted with fips=1, the Jitter RNG
is defined to be part of a FIPS module. The permanent error of the
Jitter RNG is translated as a FIPS module error. In this case, the entire
FIPS module must cease operation. This is implemented in the kernel by
invoking panic().
The patch also fixes an off-by-one in the RCT cutoff value which is now
set to 30 instead of 31. This is because the counting of the values
starts with 0.
Reviewed-by: Vladis Dronov <vdronov@redhat.com>
Signed-off-by: Stephan Mueller <smueller@chronox.de>
Reviewed-by: Marcelo Henrique Cerri <marcelo.cerri@canonical.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
2023-03-27 15:03:52 +08:00
if ( jent_entropy_init ( ) )
return - 3 ;
return - 2 ;
2020-04-18 03:33:33 +08:00
}
2015-05-25 21:10:20 +08:00
if ( ( DATA_SIZE_BITS / 8 ) < len )
tocopy = ( DATA_SIZE_BITS / 8 ) ;
else
tocopy = len ;
2015-06-23 22:18:54 +08:00
jent_memcpy ( p , & ec - > data , tocopy ) ;
2015-05-25 21:10:20 +08:00
len - = tocopy ;
p + = tocopy ;
}
return 0 ;
}
/***************************************************************************
* Initialization logic
* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
2015-06-23 22:18:54 +08:00
struct rand_data * jent_entropy_collector_alloc ( unsigned int osr ,
unsigned int flags )
2015-05-25 21:10:20 +08:00
{
struct rand_data * entropy_collector ;
2015-06-23 22:18:54 +08:00
entropy_collector = jent_zalloc ( sizeof ( struct rand_data ) ) ;
2015-05-25 21:10:20 +08:00
if ( ! entropy_collector )
return NULL ;
if ( ! ( flags & JENT_DISABLE_MEMORY_ACCESS ) ) {
/* Allocate memory for adding variations based on memory
* access
*/
2015-06-23 22:18:54 +08:00
entropy_collector - > mem = jent_zalloc ( JENT_MEMORY_SIZE ) ;
2015-05-25 21:10:20 +08:00
if ( ! entropy_collector - > mem ) {
2015-06-23 22:18:54 +08:00
jent_zfree ( entropy_collector ) ;
2015-05-25 21:10:20 +08:00
return NULL ;
}
entropy_collector - > memblocksize = JENT_MEMORY_BLOCKSIZE ;
entropy_collector - > memblocks = JENT_MEMORY_BLOCKS ;
entropy_collector - > memaccessloops = JENT_MEMORY_ACCESSLOOPS ;
}
/* verify and set the oversampling rate */
2021-03-17 09:44:03 +08:00
if ( osr = = 0 )
2015-05-25 21:10:20 +08:00
osr = 1 ; /* minimum sampling rate is 1 */
entropy_collector - > osr = osr ;
/* fill the data pad with non-zero values */
jent_gen_entropy ( entropy_collector ) ;
return entropy_collector ;
}
2015-06-23 22:18:54 +08:00
void jent_entropy_collector_free ( struct rand_data * entropy_collector )
2015-05-25 21:10:20 +08:00
{
2015-06-24 04:30:21 +08:00
jent_zfree ( entropy_collector - > mem ) ;
2015-05-25 21:10:20 +08:00
entropy_collector - > mem = NULL ;
2015-06-24 04:30:21 +08:00
jent_zfree ( entropy_collector ) ;
2015-05-25 21:10:20 +08:00
}
2015-06-23 22:18:54 +08:00
int jent_entropy_init ( void )
2015-05-25 21:10:20 +08:00
{
int i ;
__u64 delta_sum = 0 ;
__u64 old_delta = 0 ;
2020-04-18 03:33:33 +08:00
unsigned int nonstuck = 0 ;
2015-05-25 21:10:20 +08:00
int time_backwards = 0 ;
int count_mod = 0 ;
2019-05-30 03:24:25 +08:00
int count_stuck = 0 ;
struct rand_data ec = { 0 } ;
2015-05-25 21:10:20 +08:00
2020-04-18 03:33:33 +08:00
/* Required for RCT */
ec . osr = 1 ;
2015-05-25 21:10:20 +08:00
/* We could perform statistical tests here, but the problem is
* that we only have a few loop counts to do testing . These
* loop counts may show some slight skew and we produce
* false positives .
*
* Moreover , only old systems show potentially problematic
* jitter entropy that could potentially be caught here . But
* the RNG is intended for hardware that is available or widely
* used , but not old systems that are long out of favor . Thus ,
* no statistical tests .
*/
/*
* We could add a check for system capabilities such as clock_getres or
* check for CONFIG_X86_TSC , but it does not make much sense as the
* following sanity checks verify that we have a high - resolution
* timer .
*/
/*
* TESTLOOPCOUNT needs some loops to identify edge systems . 100 is
* definitely too little .
2020-04-18 03:33:33 +08:00
*
* SP800 - 90 B requires at least 1024 initial test cycles .
2015-05-25 21:10:20 +08:00
*/
2020-04-18 03:33:33 +08:00
# define TESTLOOPCOUNT 1024
2015-05-25 21:10:20 +08:00
# define CLEARCACHE 100
for ( i = 0 ; ( TESTLOOPCOUNT + CLEARCACHE ) > i ; i + + ) {
__u64 time = 0 ;
__u64 time2 = 0 ;
__u64 delta = 0 ;
unsigned int lowdelta = 0 ;
2019-05-30 03:24:25 +08:00
int stuck ;
2015-05-25 21:10:20 +08:00
2019-05-30 03:24:25 +08:00
/* Invoke core entropy collection logic */
2015-05-25 21:10:20 +08:00
jent_get_nstime ( & time ) ;
2019-05-30 03:24:25 +08:00
ec . prev_time = time ;
2020-04-18 03:33:33 +08:00
jent_lfsr_time ( & ec , time , 0 , 0 ) ;
2015-05-25 21:10:20 +08:00
jent_get_nstime ( & time2 ) ;
/* test whether timer works */
if ( ! time | | ! time2 )
return JENT_ENOTIME ;
2020-04-18 03:33:33 +08:00
delta = jent_delta ( time , time2 ) ;
2015-05-25 21:10:20 +08:00
/*
* test whether timer is fine grained enough to provide
* delta even when called shortly after each other - - this
* implies that we also have a high resolution timer
*/
if ( ! delta )
return JENT_ECOARSETIME ;
2019-05-30 03:24:25 +08:00
stuck = jent_stuck ( & ec , delta ) ;
2015-05-25 21:10:20 +08:00
/*
* up to here we did not modify any variable that will be
* evaluated later , but we already performed some work . Thus we
* already have had an impact on the caches , branch prediction ,
* etc . with the goal to clear it to get the worst case
* measurements .
*/
2021-03-17 09:44:03 +08:00
if ( i < CLEARCACHE )
2015-05-25 21:10:20 +08:00
continue ;
2019-05-30 03:24:25 +08:00
if ( stuck )
count_stuck + + ;
2020-04-18 03:33:33 +08:00
else {
nonstuck + + ;
/*
* Ensure that the APT succeeded .
*
* With the check below that count_stuck must be less
* than 10 % of the overall generated raw entropy values
* it is guaranteed that the APT is invoked at
* floor ( ( TESTLOOPCOUNT * 0.9 ) / 64 ) = = 14 times .
*/
if ( ( nonstuck % JENT_APT_WINDOW_SIZE ) = = 0 ) {
jent_apt_reset ( & ec ,
delta & JENT_APT_WORD_MASK ) ;
if ( jent_health_failure ( & ec ) )
return JENT_EHEALTH ;
}
}
/* Validate RCT */
if ( jent_rct_failure ( & ec ) )
return JENT_ERCT ;
2019-05-30 03:24:25 +08:00
2015-05-25 21:10:20 +08:00
/* test whether we have an increasing timer */
if ( ! ( time2 > time ) )
time_backwards + + ;
2019-05-30 03:24:25 +08:00
/* use 32 bit value to ensure compilation on 32 bit arches */
2015-05-25 21:10:20 +08:00
lowdelta = time2 - time ;
if ( ! ( lowdelta % 100 ) )
count_mod + + ;
/*
* ensure that we have a varying delta timer which is necessary
* for the calculation of entropy - - perform this check
* only after the first loop is executed as we need to prime
* the old_data value
*/
2019-05-30 03:24:25 +08:00
if ( delta > old_delta )
delta_sum + = ( delta - old_delta ) ;
else
delta_sum + = ( old_delta - delta ) ;
2015-05-25 21:10:20 +08:00
old_delta = delta ;
}
/*
* we allow up to three times the time running backwards .
* CLOCK_REALTIME is affected by adjtime and NTP operations . Thus ,
* if such an operation just happens to interfere with our test , it
* should not fail . The value of 3 should cover the NTP case being
* performed during our test run .
*/
2021-03-17 09:44:03 +08:00
if ( time_backwards > 3 )
2015-05-25 21:10:20 +08:00
return JENT_ENOMONOTONIC ;
/*
* Variations of deltas of time must on average be larger
* than 1 to ensure the entropy estimation
* implied with 1 is preserved
*/
2019-05-30 03:24:25 +08:00
if ( ( delta_sum ) < = 1 )
return JENT_EVARVAR ;
2015-05-25 21:10:20 +08:00
/*
* Ensure that we have variations in the time stamp below 10 for at
2019-05-30 03:24:25 +08:00
* least 10 % of all checks - - on some platforms , the counter increments
* in multiples of 100 , but not always
2015-05-25 21:10:20 +08:00
*/
if ( ( TESTLOOPCOUNT / 10 * 9 ) < count_mod )
return JENT_ECOARSETIME ;
2019-05-30 03:24:25 +08:00
/*
* If we have more than 90 % stuck results , then this Jitter RNG is
* likely to not work well .
*/
if ( ( TESTLOOPCOUNT / 10 * 9 ) < count_stuck )
return JENT_ESTUCK ;
2015-05-25 21:10:20 +08:00
return 0 ;
}
