crypto: jitter - add oversampling of noise source
PATCH 1/1 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 resurrects the function jent_fips_enabled as the oversampling support is only enabled in FIPS mode.
This patch is derived from the user space jitterentropy-library.
--- crypto/jitterentropy-kcapi.c | 6 ++++++ crypto/jitterentropy.c | 22 ++++++++++++++++++++-- crypto/jitterentropy.h | 1 + 3 files changed, 27 insertions(+), 2 deletions(-)
crypto/jitterentropy-kcapi.c +6 -0
@@ -37,6 +37,7 @@* DAMAGE. */ +#include <linux/fips.h> #include <linux/kernel.h> #include <linux/module.h> #include <linux/slab.h>
@@ -59,6 +60,11 @@ void jent_zfree(void *ptr)kfree_sensitive (ptr); } +int jent_fips_enabled (void ) +{ +return fips_enabled; +} +void jent_panic (char *s) {panic ("%s" , s);
crypto/jitterentropy.c +20 -2
@@ -117,6 +117,21 @@ struct rand_data {#define JENT_EHEALTH 9/* Health test failed during initialization */ #define JENT_ERCT 10/* RCT failed during initialization */ +/* +* 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). +*/ +#define JENT_ENTROPY_SAFETY_FACTOR 64 +#include "jitterentropy.h" /***************************************************************************
@@ -542,7 +557,10 @@ static int jent_measure_jitter(struct rand_data *ec)*/ static void jent_gen_entropy (struct rand_data *ec) { -unsigned int k =0 ; +unsigned int k =0 , safety_factor =JENT_ENTROPY_SAFETY_FACTOR ; + +if (!jent_fips_enabled ()) + safety_factor =0 ;/* priming of the ->prev_time value */ jent_measure_jitter (ec);
@@ -556,7 +574,7 @@ static void jent_gen_entropy(struct rand_data *ec)* We multiply the loop value with ->osr to obtain the * oversampling rate requested by the caller */ -if (++k >= (DATA_SIZE_BITS * ec->osr)) +if (++k >= ((DATA_SIZE_BITS + safety_factor) * ec->osr))break ; } }
crypto/jitterentropy.h +1 -0
@@ -2,6 +2,7 @@extern void *jent_zalloc (unsigned int len);extern void jent_zfree (void *ptr); +extern int jent_fips_enabled (void );extern void jent_panic (char *s);extern void jent_memcpy (void *dest,const void *src,unsigned int n);extern void jent_get_nstime (__u64 *out);
Signature · 1 line
2.33.1
Stephan,
one comment inline.
Quoted 84 lines · Stephan Müller
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 resurrects the function
jent_fips_enabledas the oversampling support is only enabled in FIPS mode.This patch is derived from the user space jitterentropy-library.
Signed-off-by Stephan Mueller <[email protected]>--- crypto/jitterentropy-kcapi.c | 6 ++++++ crypto/jitterentropy.c | 22 ++++++++++++++++++++-- crypto/jitterentropy.h | 1 + 3 files changed, 27 insertions(+), 2 deletions(-)crypto/jitterentropy-kcapi.c +6
-0@@ -37,6 +37,7 @@* DAMAGE. */ +#include <linux/fips.h> #include <linux/kernel.h> #include <linux/module.h> #include <linux/slab.h> @@ -59,6 +60,11 @@ void jent_zfree(void *ptr)kfree_sensitive (ptr); } +int jent_fips_enabled (void ) +{ +return fips_enabled; +} +void jent_panic (char *s) {panic ("%s" , s);crypto/jitterentropy.c +15
-0@@ -117,6 +117,21 @@ struct rand_data {#define JENT_EHEALTH 9/* Health test failed during initialization */ #define JENT_ERCT 10/* RCT failed during initialization */ +/* +* 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). +*/ +#define JENT_ENTROPY_SAFETY_FACTOR 64 +#include "jitterentropy.h" /***************************************************************************
crypto/jitterentropy.c static int jent_measure_jitter(struct rand_data *ec) · line 563
Quoted 8 lines
crypto/jitterentropy.c +4
-1@@ -542,7 +557,10 @@ static int jent_measure_jitter(struct rand_data *ec)*/ static void jent_gen_entropy (struct rand_data *ec) { -unsigned int k =0 ; +unsigned int k =0 , safety_factor =JENT_ENTROPY_SAFETY_FACTOR ; + +if (!jent_fips_enabled ()) + safety_factor =0 ;
I would find this more readable if safety_factor is initialized to 0, and then in the code:
if (jent_fips_enabled()) safety_factor = JENT_ENTROPY_SAFETY_FACTOR;
However this is just readability for me, either option is perfectly identicaly IMO, so
Quoted 23 lines
crypto/jitterentropy.c +1
-1/* priming of the ->prev_time value */ jent_measure_jitter (ec);@@ -556,7 +574,7 @@ static void jent_gen_entropy(struct rand_data *ec)* We multiply the loop value with ->osr to obtain the * oversampling rate requested by the caller */ -if (++k >= (DATA_SIZE_BITS * ec->osr)) +if (++k >= ((DATA_SIZE_BITS + safety_factor) * ec->osr))break ; } }crypto/jitterentropy.h +1
-0@@ -2,6 +2,7 @@extern void *jent_zalloc (unsigned int len);extern void jent_zfree (void *ptr); +extern int jent_fips_enabled (void );extern void jent_panic (char *s);extern void jent_memcpy (void *dest,const void *src,unsigned int n);extern void jent_get_nstime (__u64 *out);
Signature · 3 lines
Simo Sorce RHEL Crypto Team Red Hat, Inc
Am Freitag, 17. Dezember 2021, 13:06:07 CET schrieb Simo Sorce:
Hi Simo,
Quoted 11 lines
crypto/jitterentropy.c static int jent_measure_jitter(struct rand_data *ec) · line 563
Quoted 8 lines
static void
jent_gen_entropy(struct rand_data *ec){crypto/jitterentropy.c +4
-1-unsigned int k =0 ; +unsigned int k =0 , safety_factor =JENT_ENTROPY_SAFETY_FACTOR ; + +if (!jent_fips_enabled ()) + safety_factor =0 ;I would find this more readable if
safety_factoris initialized to 0, and then in the code:
Quoted 5 lines
if (jent_fips_enabled()) safety_factor = JENT_ENTROPY_SAFETY_FACTOR;However this is just readability for me, either option is perfectly identicaly IMO, so
Thank you, I will send an updated patch.
Ciao
Stephan
Herbert Xu
maintainercrypto/jitterentropy-kcapi.c void jent_zfree(void *ptr) · line 66
Quoted 13 lines · Stephan Müller
crypto/jitterentropy-kcapi.c +4
-0@@ -59,6 +60,11 @@ void jent_zfree(void *ptr)kfree_sensitive (ptr); } +int jent_fips_enabled (void ) +{ +return fips_enabled; +}
Why do you need this function? And why can't it be inlined?
Normally fips_enabled is entirely optimised away if FIPS is disabled in Kconfig. This function breaks this.
Thanks,
Signature · 3 lines
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Am Samstag, 18. Dezember 2021, 04:27:21 CET schrieb Herbert Xu:
Hi Herbert,
Quoted 18 lines · Stephan Müller
Quoted 5 lines
crypto/jitterentropy-kcapi.c +0
-0crypto/jitterentropy-kcapi.c void jent_zfree(void *ptr) · line 68
Quoted 9 lines
crypto/jitterentropy-kcapi.c +4
-0@@ -59,6 +60,11 @@ void jent_zfree(void *ptr)kfree_sensitive (ptr); } +int jent_fips_enabled (void ) +{ +return fips_enabled; +}Why do you need this function? And why can't it be inlined?
Quoted 2 lines
Normally
fips_enabledis entirely optimised away if FIPS is disabled inKconfig. This function breaks this.
You are right, fips.h can be included into jitterentropy.c and this helper function can be eliminated.
Thanks
Quoted 2 lines
Thanks,
Ciao
Stephan
PATCH v2 1/1 crypto: jitter - add oversampling of noise source
In reply to Stephan MüllerThe 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.
--- crypto/jitterentropy.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-)
crypto/jitterentropy.c +21 -2
@@ -117,6 +117,22 @@ struct rand_data {#define JENT_EHEALTH 9/* Health test failed during initialization */ #define JENT_ERCT 10/* RCT failed during initialization */ +/* +* 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). +*/ +#define JENT_ENTROPY_SAFETY_FACTOR 64 + +#include "linux/fips.h" #include "jitterentropy.h" /***************************************************************************
@@ -542,7 +558,10 @@ static int jent_measure_jitter(struct rand_data *ec)*/ static void jent_gen_entropy (struct rand_data *ec) { -unsigned int k =0 ; +unsigned int k =0 , safety_factor =0 ; + +if (fips_enabled) + safety_factor =JENT_ENTROPY_SAFETY_FACTOR ;/* priming of the ->prev_time value */ jent_measure_jitter (ec);
@@ -556,7 +575,7 @@ static void jent_gen_entropy(struct rand_data *ec)* We multiply the loop value with ->osr to obtain the * oversampling rate requested by the caller */ -if (++k >= (DATA_SIZE_BITS * ec->osr)) +if (++k >= ((DATA_SIZE_BITS + safety_factor) * ec->osr))break ; } }
Signature · 1 line
2.33.1
Herbert Xu
maintainerQuoted 2 lines · Stephan Müller
+#include "
linux/fips.h"
Shouldn't this be <linux/fips.h>?
Thanks,
Signature · 3 lines
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
PATCH v3 1/1 crypto: jitter - add oversampling of noise source
In reply to Stephan MüllerThe 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.
--- crypto/jitterentropy.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-)
crypto/jitterentropy.c +21 -2
@@ -117,6 +117,22 @@ struct rand_data {#define JENT_EHEALTH 9/* Health test failed during initialization */ #define JENT_ERCT 10/* RCT failed during initialization */ +/* +* 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). +*/ +#define JENT_ENTROPY_SAFETY_FACTOR 64 + +#include <linux/fips.h> #include "jitterentropy.h" /***************************************************************************
@@ -542,7 +558,10 @@ static int jent_measure_jitter(struct rand_data *ec)*/ static void jent_gen_entropy (struct rand_data *ec) { -unsigned int k =0 ; +unsigned int k =0 , safety_factor =0 ; + +if (fips_enabled) + safety_factor =JENT_ENTROPY_SAFETY_FACTOR ;/* priming of the ->prev_time value */ jent_measure_jitter (ec);
@@ -556,7 +575,7 @@ static void jent_gen_entropy(struct rand_data *ec)* We multiply the loop value with ->osr to obtain the * oversampling rate requested by the caller */ -if (++k >= (DATA_SIZE_BITS * ec->osr)) +if (++k >= ((DATA_SIZE_BITS + safety_factor) * ec->osr))break ; } }
Signature · 1 line
2.33.1
Herbert Xu
maintainerQuoted 21 lines · Stephan Müller
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.
Quoted 7 lines
This patch is derived from the user space jitterentropy-library.
Signed-off-by Stephan Mueller <[email protected]>Reviewed-by Simo Sorce <[email protected]>--- crypto/jitterentropy.c | 23 +++++++++++++++++++++-- 1 file changed, 21 insertions(+), 2 deletions(-)
Patch applied. Thanks.
Signature · 3 lines
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt