Re: [virtio] [PATCH 3/3] VIRTIO-110: ARM's feedback for MMIO chapter, clarifications

["Michael S. Tsirkin" <mst@redhat.com>]

On Tue, Aug 05, 2014 at 01:10:46PM +0100, Pawel Moll wrote:
> On Tue, 2014-08-05 at 13:04 +0100, Michael S. Tsirkin wrote:
> > > > > +The driver MAY use other accesses for the device-specific configuration space,
> > > > > +depending on the device type.
> > > > 
> > > > Why are you adding this option?
> > > > If there's a 4 byte register it is a pain for devices to support
> > > > 1 byte accesses, and we gain nothing from it.
> > > 
> > > But 2.3 does not limit the configuration space field size nor alignment,
> > > does it?
> > 
> > No but there is no need: you look at each specific field, you see what
> > the size and alignment is.
> 
> That's what I'm saying. The driver MAY do 8-bit load of the 8-bit field,
> if the field is defined as such in the device spec. Otherwise all
> operations on the configuration space would have to be 32-bit wide and
> aligned. We don't want this, do we? (or do we? :-)

Indeed.  What I am saying is this from PCI part of spec:

	The driver
	MUST access each field using the ``natural'' access method, i.e.
	32-bit accesses for 32-bit fields, 16-bit accesses for 16-bit
	fields and 8-bit accesses for 8-bit fields.

Should apply.  I guess when you say "depending on the device type"
this is what you mean, but maybe being more explicit would be better.



> > Are you supporting transitional devices for MMIO?
> > If no you don't care what happens with existing drivers.
> 
> I don't indeed.
> 
> Paweł

Right so existing drivers doing memcpy is of no concern then.