Re: Should PAGE_CACHE_SIZE be discarded?

[Nick Piggin <npiggin@suse.de>]

On Thu, Nov 15, 2007 at 02:46:46PM +0000, David Howells wrote:
> Benny Halevy <bhalevy@panasas.com> wrote:
> 
> > I think that what Nick was trying to say is that PAGE_CACHE_SIZE should
> > always be used properly as the size of the memory struct Page covers (while
> > PAGE_SIZE is the hardware page size and the constraint is that
> > PAGE_CACHE_SIZE == (PAGE_SIZE << k) for some k >= 0).  If everybody does
> > that then "None of the filesystems should really care at all". That said, it
> > doesn't seem like the current usage in fs/ and drivers/ is consistent with
> > this convention.
> 
> Indeed.  One thing you have to consider is kmap().  I would expect it to
> present an area of PAGE_SIZE for access.  However, if the filesystem gets an
> area of PAGE_CACHE_SIZE to fill, then I would have to do multiple kmap() calls
> in the process of filling that 'pagecache page' in AFS.
> 
> Furthermore, if a page struct covers a PAGE_CACHE_SIZE chunk of memory, then I
> suspect the page allocator is also wrong, as it I believe it deals with
> PAGE_SIZE chunks of memory, assuming a struct page for each.

That's because you're thinking about all these difficulties outside the
filesystem. Really, pagecache is in PAGE_CACHE_SIZE. Nobody said pagecache
> PAGE_SIZE will work by changing a single define, but filesystems are
about the least problematic here.