Re: invalidate_inode_pages in 2.5.32/3

[Daniel Phillips <phillips@arcor.de>]

On Friday 13 September 2002 01:23, Rik van Riel wrote:
> On Thu, 12 Sep 2002, Andrew Morton wrote:
> > Rik van Riel wrote:
> 
> > > invalidate_page(struct page * page) {
> 
> > That's the bottom-up approach.  The top-down (vmtruncate) approach
> > would also work, if the locking is suitable.
> 
> The top-down approach will almost certainly be most efficient when
> invalidating a large chunk of a file (truncate, large file locks)
> while the bottom-up approach is probably more efficient when the
> system invalidates very few pages (small file lock, cluster file
> system mmap() support).

The bottom-up approach is the one we want to use when we'd otherwise
skip a page in invalidate_inode_pages.  This is the rare case.  On the
face of it, this works out very well.  Just have to think about
interactions now - I don't see any, but I haven't really gone hunting
yet.

-- 
Daniel