Re: [Xen-devel] [PATCH 2/3] amd-iommu: disable iommu_hap_pt_share with AMD IOMMUs

[Tim Deegan <tim@xxxxxxx>]

At 09:52 +0100 on 08 Apr (1396947164), Jan Beulich wrote:
> >>> On 07.04.14 at 18:02, <roger.pau@xxxxxxxxxx> wrote:
> > --- a/xen/drivers/passthrough/amd/iommu_init.c
> > +++ b/xen/drivers/passthrough/amd/iommu_init.c
> > @@ -1255,6 +1255,13 @@ int __init amd_iommu_init(void)
> >      if ( iterate_ivrs_mappings(amd_iommu_setup_device_table) != 0 )
> >          goto error_out;
> >  
> > +    /*
> > +     * Disable sharing HAP page tables with AMD IOMMU,
> > +     * since it only supports p2m_ram_rw, and this would
> > +     * prevent doing IO to/from mapped grant frames.
> > +     */
> > +    iommu_hap_pt_share = 0;
> > +
> 
> I guess at the very least when the option was specified on the
> command line you should issue a warning, or perhaps even stay
> away from enforcing this.
> 
> I also think we ought to consider alternatives before taking a harsh
> step like this (and that's independent of my earlier indication that we
> may want to switch away from sharing these tables): It looks like the
> global bit in the host page tables is unused and ignored, and could
> therefore be used to indicate grant entries. The IW bit could be
> used to distinguish p2m_{ram,grant}_{ro,rw} respectively.
> 
> Curious what Tim's thoughts here are...

Happy to use the global bit to mark grant entries (though I haven't
checked the docs to be sure it's actually safe). OTOH it seems
like the hadres IOMMU tables might be going away for other reasons so
I'm not sure it's worth putting a lot of effort into being clever
around this.

Tim.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel