> It specifically says "the kernel uses this information."  I don't see
> any place to indicate that the kernel just throws it away, or actually
> even that the kernel may ignore the advice, but it's unreasonable to
> expect that the advice will be followed unconditionally.  However,
> there is the bald statement that the kernel _does_ use the
> information, rather than just throw it away.

Hmm, has anyone fixed the random-number pseudodevice yet?

As a practical matter, I suspect that there is a lot more software which
will break in peculiar ways if ENOTSUPP is returned than will break if
the kernel uses the information solely to heat a few DRAM chips.