Re: MI Xen drivers (was: Re: [patch] x86_*fence replaced by membar_ops(3))

[Christoph Egger <Christoph_Egger%gmx.de@localhost>]

On Friday 16 January 2009 17:26:39 Manuel Bouyer wrote:
> On Fri, Jan 16, 2009 at 12:25:14AM +0100, Jean-Yves Migeon wrote:
> > Manuel Bouyer wrote:
> > >On Thu, Jan 15, 2009 at 03:13:59PM +0100, Jean-Yves Migeon wrote:
> > >>>>What should I do? Drop it for the x86 generic bus code, and keep it
> > >>>> for the Xen drivers?
> > >>>
> > >>>No, membar_ops won't work for bus_dma/bus_space either. membar_ops are
> > >>>patched
> > >>>to be NOP on UP kernels, still they're needed for interfaces with
> > >>>devices.
> > >>
> > >>Hmm, I was not really clear on that one: I just meant to keep the
> > >>membar_ops(3) routines for the xennet, xbd and xencons drivers: since
> > >> we do not patch the LOCK prefix for UP Xen domains, they are close to
> > >> the x86_*fence() ones.
> > >
> > >Actually yes; but we'll probably want to use x86_patch() in the future.
> >
> > Second attempt, this time for Xen code only, using xen_*mb() macros (the
> > same than those used in the different ring-I/O APIs from Xen).
> >
> > http://www.netbsd.org/~jym/MI-Xen-drivers.diff
> >
> > Next step is to abstract the p2m routines. As most of them already use
> > pmap(9), work is almost done (thanks Manuel). A few exceptions like
> > direct use of the xpmap_phys_to_machine_mapping[] remain, I will have a
> > look at how P2M is handled on other archs (ia64 comes to mind).
> >
> > Please review.
>
> Looks good, but did you try building XEN2 kernels ? I can't see the
> xen_*mb() macros defined for xen2. I think defining them in
> include/xen-public/xen.h should be enough.

Or it is time to remove xen2 in -current. I will do that past when the
first RC of netbsd 5 is out.

Christoph