Re: [virtio] [PATCH] pci: missing documentation for dealing with 64 bit config fields

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

On Thu, Apr 02, 2015 at 11:01:45AM +0200, Cornelia Huck wrote:
> On Wed, 1 Apr 2015 20:42:13 +0200
> "Michael S. Tsirkin" <mst@redhat.com> wrote:
> 
> > pci spec says what width access to use for 32, 16 and 8
> > bit fields, but does not explicitly say what to do for
> > 32 bit fields. As we have text that says driver must
> 
> s/32/64/
> 
> > treat 64 bit accesses as non-atomic, this seems
> > to imply driver should always do two 32 bit wide accesses.
> > 
> > Let's make this an explicit requirement, and require
> > devices to support this.
> > 
> > VIRTIO-139
> > 
> > Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> > ---
> >  conformance.tex |  1 +
> >  content.tex     | 19 ++++++++++++++-----
> >  2 files changed, 15 insertions(+), 5 deletions(-)
> 
> (...)
> 
> > +For device configuration access, the driver MUST use 8-bit wide
> > +accesses for 8-bit wide fields, 16-bit wide and aligned accesses
> > +for 16-bit wide fields and 32-bit wide and aligned accesses for
> > +32-bit and 64-bit wide fields. For 64-bit fields, the driver MAY
> > +access each of the high and low 32-bit parts of the field
> > +independently.
> 
> Doesn't the last sentence follow from "use 32-bit wide accesses"
> already?

Someone might assume that driver must always access low part,
then high part. Last sentence says it does not have to.

-- 
MST