>
>
> On Tue, 14 Dec 1999, Alexander Viro wrote:
> >
> > Question: will you object against switching /dev/loop over regular files
> > to pagecache? Current variant uses bmap()+getblk() for IO and write() for
> > plugging the holes if backing file is sparse - definitely Bad Thing(tm).
>
> Sounds like a capital idea. Mind doing the block device pagecache first,
> though, if you're already looking at this area?
Frankly, I'ld rather start with massaging bmap() out of existence. I will
do block device pagecache, all right, but there is one funny detail - we
have serious code duplication between loopback and swap. I'm trying to
figure out what to do with that and I really suspect that pagecache for
block devices, clean shm implementation and doing anonymous shared
mappings are pieces of the same puzzle. It almost looks like we are
dealing with several partial implementations of page migration from one
address space to another... Hell knows. I think that right design of
pagecache for block devices will become visible when we'll have the second
(regular file) case cleaned up in both existing users - loopback and swap.
In particular, I suspect that we already have most of the needed stuff in
device-related part of brw_page() / rw_swap_page_base(). Comments?
Cheers,
Al
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/