[Ocfs2-devel] [PATCH] xen: cleancache shim to Xen Transcendent Memory

[Ian Campbell]

On Mon, 2011-04-18 at 07:12 -0700, Dan Magenheimer wrote:
> > From: Ian Campbell [mailto:Ian.Campbell at eu.citrix.com]
> > > >
> > > > On Thu, 2011-04-14 at 14:20 -0700, Dan Magenheimer wrote:
> > > >
> > > > There's no need to build this into a kernel which doesn't have
> > > > cleancache (or one of the other frontends), is there? I think there
> > > > should be a Kconfig option (even if its not a user visible one)
> > with
> > > > the appropriate depends/selects.
> > >
> > > Yes, you're right.  It should eventually depend on
> > >
> > > CONFIG_CLEANCACHE || CONFIG_FRONTSWAP
> > >
> > > though there's no sense merging this xen cleancache
> > > shim at all unless/until Linus merges cleancache
> > > (and hopefully later some evolution of frontswap).
> > 
> > Cleancache isn't in already? I thought I saw references to it in
> > drivers/staging?
> 
> Linus said he would review it after 2.6.39-rc1 was released,
> but has neither given thumbs up nor thumbs down so I'm
> assuming he didn't have time and it will be reconsidered
> for 2.6.40.  This latest patchset (V8) is updated in linux-next.
> 
> Yes, zcache is in driver/staging and has references to it.
> I guess that proves the chicken comes before the egg...
> or was it vice versa? :-)

Ah, I didn't realise that the relaxations associated with staging
allowed for unsatisfiable (due to the other half being out of tree)
Kconfig items as well.
 
> > > And once cleancache (and/or frontswap) is merged,
> > > there's very little reason NOT to enable one or
> > > both on a Xen guest kernel.
> > 
> > There are software knobs to allow the host- and guest-admin to opt in
> > or out as they desire though, right?
> 
> Definitely.  Both Xen and a Linux guest have runtime
> options, which currently default to off.

Good to know, thanks!

Ian.