Re: 2.6.5-rc1-aa1

[Andrea Arcangeli <andrea@suse.de>]

On Thu, Mar 18, 2004 at 10:37:55PM +0000, Hugh Dickins wrote:
> On Thu, 18 Mar 2004, Andrea Arcangeli wrote:
> > 
> > The fix is simple: always set and clear PG_anon under the page_map_lock,
> > this will avoid the race since all ClearPageAnon already runs under the
> > page_map_lock. I will implement and test in a few hours.
> > 
> > ... I find this more robust.
> 
> Absolutely, that's what I did too.  My old page_add_rmap had anon flag,
> but the new patches have page_add_anon_rmap and page_add_obj_rmap.

I remebered you had the anon param too, though I didn't realize it was
for this reason :/

the reason why I go with the parameter is that sometime I've a single
call at the end of the function with the same new_page, but sometime
it has to be considered anonymous sometime not. So I couldn't split out
the interface with _anon/_file suffixes with some #define and remove the
0/1 numbering ugliness.