Re: page->offset & PAGE_MASK

chip@atlantic.net
Mon, 2 Feb 1998 17:56:06 -0500 (EST)

According to Linus Torvalds:
> On Sun, 1 Feb 1998 chip@atlantic.net wrote:
> > I'm working on extending Linux's VM mechanism beyond 2G.
> >
> > A comment in include/linux/mm.h says that page->offset need not be
> > a multiple of PAGE_SIZE. However, the code in mm/*.c _seems_ always
> > to set it so.
>
> Look at the exec() functions, especially for a.out. I think it's OMAGIC
> that is not guaranteed to be page-aligned (in fact, it is guaranteed to
> _not_ be page-aligned: the executable will start at offset 1024 in the
> file).

So I can't even count on sb_blocksize to be small enough, since an
a.out may be on a filesystem with a block size of (e.g.) 2048 or 4096.

I guess I'll have to invent a new constant, something like
PAGE_POS_UNIT (and the accompanying PAGE_POS_BITS).

-- 
Chip Salzenberg               - a.k.a. -                <chip@pobox.com>
 "Boy, is it ever dark."  "I hope they don't get a moonburn."  // MST3K