Re: adjust prefetch in free_one_pgd()

[Linus Torvalds]

On Fri, 2 Aug 2002, David Mosberger wrote:
>
> I thought the prefetches API intended this to be a safe operation?

Well, any _sane_ prefetch API would be safe.

However, there is known-broken hardware out there, in which a prefetch
from IO space will kill the machine.

Personally, I would just say that we should disable prefetch on such
clearly broken hardware, but since it's Alans favourite machine (some
early AMD Athlon if I remember correctly), I think Alan will disagree ;)

> It's definitely not an issue on ia64: there, prefetches against
> uncached memory translations are automatically canceled.

That's true of just about any other architecture too. I think the AMD case
is an erratum, so even on AMD it is _supposed_ to work.

			Linus