Re: [PATCH 10/12] xen/arm: if is_domain_direct_mapped use native UART address for vPL011

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxx>]

On Sat, 9 May 2020, Julien Grall wrote:
> Hi Stefano,
> 
> On 09/05/2020 01:07, Stefano Stabellini wrote:
> > On Fri, 1 May 2020, Julien Grall wrote:
> > > On 01/05/2020 02:26, Stefano Stabellini wrote:
> > > > On Wed, 15 Apr 2020, Julien Grall wrote:
> > > > > Hi Stefano,
> > > > > 
> > > > > On 15/04/2020 02:02, Stefano Stabellini wrote:
> > > > > > We always use a fix address to map the vPL011 to domains. The
> > > > > > address
> > > > > > could be a problem for domains that are directly mapped.
> > > > > > 
> > > > > > Instead, for domains that are directly mapped, reuse the address of
> > > > > > the
> > > > > > physical UART on the platform to avoid potential clashes.
> > > > > 
> > > > > How do you know the physical UART MMIO region is big enough to fit the
> > > > > PL011?
> > > > 
> > > > That cannot be because the vPL011 MMIO size is 1 page, which is the
> > > > minimum right?
> > > 
> > > No, there are platforms out with multiple UARTs in the same page (see
> > > sunxi
> > > for instance).
> > 
> > But if there are multiple UARTs sharing the same page, and the first one
> > is used by Xen, there is no way to assign one of the secondary UARTs to
> > a domU. So there would be no problem choosing the physical UART address
> > for the virtual PL011.
> 
> AFAICT, nothing prevents a user to assign such UART to a dom0less guest today.
> It would not be safe, but it should work.
>
> If you want to make it safe, then you would need to trap the MMIO access so
> they can be sanitized. For a UART device, I don't think the overhead would be
> too bad.
> 
> Anyway, the only thing I request is to add sanity check in the code to help
> the user diagnostics any potential clash.

OK thanks for clarifying, I'll do that.