Re: [Xen-devel] [PATCH for-4.9] xen/arm: acpi: Map MMIO on fault in stage-2 page table for the hardware domain

[Julien Grall <julien.grall@xxxxxxx>]

Hi Stefano,

On 30/03/2017 19:55, Stefano Stabellini wrote:
> On Thu, 30 Mar 2017, Julien Grall wrote:
> > Hi Stefano,
> >
> > On 30/03/17 19:37, Stefano Stabellini wrote:
> > > On Thu, 30 Mar 2017, Julien Grall wrote:
> > > > +/*
> > > > + * When using ACPI, most of the MMIO regions will be mapped on-demand
> > > > + * in stage-2 page tables for the hardware domain because Xen is not
> > > > + * able to know from the EFI memory map the MMIO regions.
> > > > + */
> > > > +static bool try_map_mmio(gfn_t gfn)
> > > > +{
> > > > +    struct domain *d = current->domain;
> > > > +
> > > > +    /* For the hardware domain, all MMIOs are mapped with GFN == MFN */
> > > > +    mfn_t mfn = _mfn(gfn_x(gfn));
> > > > +
> > > > +    /*
> > > > +     * Device-Tree should already have everything mapped when building
> > > > +     * the hardware domain.
> > > > +     */
> > > > +    if ( acpi_disabled )
> > > > +        return false;
> > > > +
> > > > +    if ( d != hardware_domain )
> > > is_hardware_domain(d)
> > Will fix it.
> >
> > > > +        return false;
> > > > +
> > > > +    /* The hardware domain can only map permitted MMIO regions */
> > > > +    if ( !iomem_access_permitted(d, mfn_x(mfn), mfn_x(mfn) + 1) )
> > > > +        return false;
> > > Because of the potential of causing problems with DMA to device MMIO,
> > > I would add a warning here, something like:
> > >
> > >    dprintk(XENLOG_WARNING, "Mapping gfn=mfn=%"PRI_gfn" to Dom%d\n",
> > > d->domain_id);
> > >
> > > I hope that it will help us debug issues when/if we'll actually
> > > encounter such a case.
> > I don't think this will help us and it will flood the console in debug-build
> > (there are a lot of MMIO to map).
> >
> > When SMMUs are not used, DMA to device MMIO will work regardless the mapping
> > in stage-2.
> >
> > When SMMUs are in use, you will get a SMMU fault that tells you the address
> > used. With that you could know the region was not mapped.
> Fair enough. Do you think there is any value in adding only one debug
> warning, to say that we are mapping MMIOs on trap on this platform? But
> today, it would basically be the same as saying "ACPI is enabled".
I think we clearly need some documentation in tree to explain the design 
of Xen with ACPI and Device Tree. But this is lacking so far :/.
I can have a go when I find some times.

Cheers,

--
Julien Grall

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