Re: [PATCH] struct page shrinkage

[William Lee Irwin III]

On Mon, Feb 25, 2002 at 06:01:22PM -0800, David S. Miller wrote:
> Rik, not every architecture has a "counter" member of
> atomic_t, that is the problem.  This is a hard bug, please
> fix it.  It is an opaque type, accessing its' implementation
> directly is therefore illegal in the strongest way possible.

> From: Rik van Riel <[e-mail]>
> This exact same code has been in -rmap for a few months
> and went into 2.5 just over a week ago.  It doesn't seem
> to give any problems ...

On Mon, Feb 25, 2002 at 06:01:22PM -0800, David S. Miller wrote:
> Because I haven't pushed my sparc64 changesets yet, I'm doing
> that tonight.

I think I'm to blame for init_page_count(). My bad.

A small bit of analysis seemed to reveal that atomicity wasn't needed
in free_area_init_core(). Apparently the solution I suggested here was
non-portable. Perhaps a better way will crop up later.

Cheers,
Bill