Re: PATCH: EDAC - clean up atomic stuff

[Alan Cox <alan@lxorguk.ukuu.org.uk>]

On Gwe, 2005-10-28 at 10:33 -0600, Eric W. Biederman wrote:
> A couple of questions
> - Why a u32 for length and not just unsigned?

Because it was loading it into a 32bit counter so the input was 32bit.
Just habit really.

> - Why is the x86_64 version clearing 32bit words and not 64bit words,
>   that should be noticeably faster if we ever need to use that
>   code.

I doubt it makes much difference. I kept it 32bit to keep the split
simple. It can certainly be optimised if someone wants to. I'd hope
however ECC scrub is never a hot path!

> - Is KM_BOUNCE_READ a safe atomic_kmap entry to be using?
>   I'm not certain, but my gut feel is that scrubbing probably
>   wants it's own kmap type. 
>   I remember doing some looking when I first wrote this and thinking
>   that KM_BOUNCE_READ looked safe and was good enough until the code
>   got merged into the kernel.

I was looking at that. I think it is but I'm not 100% sure or an expert
on kmaps.