[PATCH v4 04/11] mips: use fallback for random_get_entropy() instead of zero

Maciej W. Rozycki macro at orcam.me.uk
Sat Apr 16 07:44:53 PDT 2022


Hi Jason,

> >  There are two variants only of the CP0 Random register that we can ever
> > encounter, as it's been de-facto standardised in early 1990s already and
> > then written down in the MIPSr1 architecture specification ~2000.  So I
> > think it may make sense to actually handle them both explictitly with
> > individual calculations, possibly conditionalised on a CONFIG setting or
> > `cpu_has_3kex', because kernels that support the two variants of the MMU
> > architecture are mutually incompatible.
> 
> Okay, I can give this a shot, but this certainly isn't my forté. It
> may ultimately wind up being simpler for you to just send some code of
> what you envision for this, but if I understand your idea correctly,
> what you're saying is something like:
> 
> static inline unsigned long random_get_entropy(void)
> {
>         unsigned int prid = read_c0_prid();
>         unsigned int imp = prid & PRID_IMP_MASK;
>         unsigned int c0_random;
> 
>         if (can_use_mips_counter(prid))
>                 return read_c0_count();
> 
>         if (cpu_has_3kex)
>                 c0_random = (read_c0_random() >> 8) & 0x3f;
>         else
>                 c0_random = read_c0_random() & 0x3f;
>         return (random_get_entropy_fallback() << 6) | (0x3f - c0_random);
> }
> 
> What do you think of that? Some tweak I'm missing?

 It certainly looks good to me.  Do you have a way I could verify how this 
function performs?  If so, then I could put it through my systems as I can 
cover all the cases handled here.

 Any improvements I previously discussed can then be made locally in the 
MIPS port as follow-up changes.

> >  Isn't it going to be an issue for an entropy source that the distribution
> > of values obtained from the CP0 Random bit-field is not even, that is some
> > values from the 6-bit range will never appear?
> 
> It's the same situation without inverting the order: instead of some
> bits on the top never happening, some bits on the bottom never happen
> instead. In general, counters don't form uniform distributions anyway,
> since the lower bits change faster, and neither are they independent,
> since one sample in large part depends on the previous. This is just
> sort of the nature of the beast, and the code that calls
> random_get_entropy() deals with this appropriately (by, at the moment,
> just hashing all the bits).

 OK then, thanks for your clarification.

  Maciej



More information about the linux-riscv mailing list