Re: [PATCH v2] stdlib: Simplify arc4random_uniform

[Yann Droneaud <ydroneaud@opteya.com>]

Le 02/08/2022 à 14:14, Adhemerval Zanella Netto a écrit :
>
> On 02/08/22 09:08, Yann Droneaud wrote:
>> Hi,
>>
>> Le 01/08/2022 à 21:20, Noah Goldstein a écrit :
>>>> diff --git a/stdlib/arc4random_uniform.c b/stdlib/arc4random_uniform.c
>>>> index 1326dfa593..5aa98d1c13 100644
>>>> --- a/stdlib/arc4random_uniform.c
>>>> +++ b/stdlib/arc4random_uniform.c
>>>>
>>>>    uint32_t
>>>>    __arc4random_uniform (uint32_t n)
>>>>    {
>>>> @@ -57,83 +38,33 @@ __arc4random_uniform (uint32_t n)
>>>> +  while (1)
>>>>        {
>>>> +      uint32_t value = __arc4random ();
>>>> +
>>>> +      /* Return if the lower power of 2 minus 1 satisfy the condition.  */
>>>> +      uint32_t r = value & mask;
>>>> +      if (r < n)
>>>> +       return r;
>>>> +
>>>> +      /* Otherwise check if remaining bits of entropy provides fits in the
>>>> +        bound.  */
>>>> +      for (int bits_left = z; bits_left >= bits; bits_left -= bits)
>>>> +       {
>>>> +         value >>= bits;
>>> Can this just be shift by 1 and repeat (32 - z) times or does that
>>> introduce bias (not seeing exactly why it would)?
>>
>> That was bothering me too, as I was thinking a rotation would be possible instead of shift.
>>
>> I posted the question https://crypto.stackexchange.com/questions/101325/uniform-rejection-sampling-by-shifting-or-rotating-bits-from-csprng-output-safe
>>
>> The answer: there's indeed a bias.
>>
>> This explains why my attempt with rotation leads to dieharder complaining. It was so obvious ... Damn
>>
>
> Thanks, I will remove it then.  We might evaluate later if using the mask
> and compare is indeed better than the other methods (as using by OpenBSD).
>
It's the proposal to shift by 1 bit which would introduce a bias.

The implementation used in Glibc is fine.


Regards.

-- 

Yann Droneaud

OPTEYA