On Wed, 9 Feb 2000, Jeff Garzik wrote:
> al viro was kind enough to e-mail me some of his thoughts on banishing
> the big kernel lock from the VFS. Though my time with the VFS has been
> nil in the past few months, I'd still like to work on this if noone
> beats me to it.
>
> IIRC the two big items are dcache/dentry and inode threading.
s/inode/friggin' POSIX locks shit/
And it's a bit different story - ext2fs needs some serialization of its
own; after all it got some internal structures ;-)
> Any and all ideas for online defrag, please post. I'm very interested.
See below.
> > delayed allocation
>
> this needs to be in the VFS desperately. every new & advanced
> filesystem is winding up implementing their own logic for this...
>
> > Address spaces (viro)
>
> can someone elaborate?
Urgh. It's a long(ish) story. Basically, we are getting address_space
methods. It removes ->readpage/->writepage/->get_block out of
inode_operations, BTW. What it means: we are getting rid of a lots of code
duplication (data semantics; as in normal block fs vs. no-holes fs vs.
extent-based with holes vs. fragments-handling a-la FFS vs. fs with small
files embeddable into inodes vs. ...). Address_space is an MMU. This way
they become separated from filesystems proper (i.e. layout, etc.). If you
want more coherent description - ask and I'll write it. The latest version
of my patch sits on ftp.math.psu.edu/pub/viro/as-patch-26z2 (warning:
needs testing).
