Re: [Qemu-devel] [PATCH QEMU v5] hw/arm/sysbus-fdt: Add support for instantiating generic devices

[Geert Uytterhoeven]

Hi Peter,

Thanks for your comments!

On Wed, Jan 9, 2019 at 5:03 PM Peter Maydell <address@hidden> wrote:
> On Wed, 9 Jan 2019 at 15:55, Auger Eric <address@hidden> wrote:
> > On 1/3/19 10:42 AM, Geert Uytterhoeven wrote:
> > > Add a fallback for instantiating generic devices without a type-specific
> > > or compatible-specific instantiation method.  This will be used when no
> > > other match is found.
> > >
> > > Generic device instantiation avoids having to write device-specific
> > > instantiation methods for each and every "simple" device using only a
> > > set of generic properties.  Devices that need more specialized handling
> > > can still provide their own instantiation methods.
>
> > > +    /* Ignoring the following may or may not work, hence the warning */
> > > +    { "gpio-ranges",     PROP_WARN },   /* no support for pinctrl yet */
> > > +    { "dmas",            PROP_WARN },   /* no support for external DMACs 
> > > yet */
> > I would be tempted to simply reject things that may not work.
>
> More generally, this whole feature seems to be "allow things that
> might not work", isn't it? Otherwise we could just have explicit

I can remove the two PROP_WARN properties above from the list, if you prefer.
Exporting rcar-sata still works fine after that.

> whitelists for the devices we want to allow passthrough of and
> that we've tested to work.

In the end, this will become a loooooong list (SoC x devices)...

> I have to say I'm not really very enthusiastic about
> enhancing this to allow random device passthrough,
> because it encourages further use of this. If people
> want hardware that can be passed through they should
> put it behind a bus that can be probed and can go
> behind an IOMMU, ie pci or some equivalent. That
> is a supportable hardware mechanism. All this
> machinery feels very heath-robinson...

As no-iommu suppport is not upstream (in Qemu; it is upstream in Linux,
perhaps it should be removed?), all devices using DMA require being
behind an IOMMU.

Reality is that on embedded, on-SoC devices are usually not on a PCI bus.
But IOMMUs are present, and virtualization is wanted.

Thanks!

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- address@hidden

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds