Re: booked-page-flag.patch

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 16 Feb 2007 10:56:55 +0300 Alex Tomas <alex@clusterfs.com> wrote:

> >>>>> Andrew Morton (AM) writes:
> 
>  AM> Well, one could just assume that the page has no disk mapping and go and
>  AM> make the space reservation.  Things will work out OK when we come to do
>  AM> writepage().
> 
>  AM> Or one could do both: call get_block() only if the page was inside i_size.
> 
> well, even so we need to reserve not that block ony, but
> also needed metadata (for the worst case).

Right.  Reserve all the blocks for the page and all the metadata for a page
at that file offset.  Worst case.

> probably this
> is work for get_block or some different method?

umm, when I did this I think I added a new ->reservepage address_space op
and called that from within the VFS's delalloc_prepare_write().

> anyway,
> we have to call it if the page is being written partial. 

Not necessarily.  If we're operating in nobh mode that page is brought
uptodate in prepare_write().