2021-11-21 14:14:28

by Stephan Müller

[permalink] [raw]
Subject: [PATCH] crypto: Jitter RNG - consider 32 LSB for APT

The APT compares the current time stamp with a pre-set value. The
current code only considered the 4 LSB only. Yet, after reviews by
mathematicians of the user space Jitter RNG version >= 3.1.0, it was
concluded that the APT can be calculated on the 32 LSB of the time
delta. Thi change is applied to the kernel.

This fixes a bug where an AMD EPYC fails this test as its RDTSC value
contains zeros in the LSB. The most appropriate fix would have been to
apply a GCD calculation and divide the time stamp by the GCD. Yet, this
is a significant code change that will be considered for a future
update. Note, tests showed that constantly the GCD always was 32 on
these systems, i.e. the 5 LSB were always zero (thus failing the APT
since it only considered the 4 LSB for its calculation).

Signed-off-by: Stephan Mueller <[email protected]>
---
crypto/jitterentropy.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/crypto/jitterentropy.c b/crypto/jitterentropy.c
index a11b3208760f..f6d3a84e3c21 100644
--- a/crypto/jitterentropy.c
+++ b/crypto/jitterentropy.c
@@ -265,7 +265,6 @@ 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);
- unsigned int delta_masked = current_delta & JENT_APT_WORD_MASK;

ec->last_delta = current_delta;
ec->last_delta2 = delta2;
@@ -274,7 +273,7 @@ static int jent_stuck(struct rand_data *ec, __u64 current_delta)
* Insert the result of the comparison of two back-to-back time
* deltas.
*/
- jent_apt_insert(ec, delta_masked);
+ jent_apt_insert(ec, current_delta);

if (!current_delta || !delta2 || !delta3) {
/* RCT with a stuck bit */
--
2.33.1






2021-11-26 05:34:39

by Herbert Xu

[permalink] [raw]
Subject: Re: [PATCH] crypto: Jitter RNG - consider 32 LSB for APT

On Sun, Nov 21, 2021 at 03:14:20PM +0100, Stephan M?ller wrote:
> The APT compares the current time stamp with a pre-set value. The
> current code only considered the 4 LSB only. Yet, after reviews by
> mathematicians of the user space Jitter RNG version >= 3.1.0, it was
> concluded that the APT can be calculated on the 32 LSB of the time
> delta. Thi change is applied to the kernel.
>
> This fixes a bug where an AMD EPYC fails this test as its RDTSC value
> contains zeros in the LSB. The most appropriate fix would have been to
> apply a GCD calculation and divide the time stamp by the GCD. Yet, this
> is a significant code change that will be considered for a future
> update. Note, tests showed that constantly the GCD always was 32 on
> these systems, i.e. the 5 LSB were always zero (thus failing the APT
> since it only considered the 4 LSB for its calculation).
>
> Signed-off-by: Stephan Mueller <[email protected]>
> ---
> crypto/jitterentropy.c | 3 +--
> 1 file changed, 1 insertion(+), 2 deletions(-)

Patch applied. Thanks.
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt