Linux Virtualization - Jan 2015 - [RFC] virtio-mmio: Update the device to OASIS spec version

On Thu, Jan 15, 2015 at 06:51:01PM +0200, Michael S. Tsirkin
wrote:
> Host no longer knows the alignment, so why is it needed?
> In fact, I notice that 4.3.2.3 Virtqueue Layout seems completely wrong:
> it corresponds to legacy devices, and it does not
> say what "align" is.
I created
https://issues.oasis-open.org/browse/VIRTIO-129
to track this. Can you write up a patch please?

-- 
MST

On Thu, 2015-01-15 at 17:12 +0000, Michael S. Tsirkin
wrote:
> On Thu, Jan 15, 2015 at 06:51:01PM +0200, Michael S. Tsirkin wrote:
> > Host no longer knows the alignment, so why is it needed?
> > In fact, I notice that 4.3.2.3 Virtqueue Layout seems completely
wrong:
> > it corresponds to legacy devices, and it does not
> > say what "align" is.
> 
> I created
> https://issues.oasis-open.org/browse/VIRTIO-129
> to track this. Can you write up a patch please?
"4.3 Virtio Over Channel I/O"

(hint: not MMIO)

Pawe?

On Thu, Jan 15, 2015 at 05:15:30PM +0000, Pawel Moll
wrote:
> On Thu, 2015-01-15 at 17:12 +0000, Michael S. Tsirkin wrote:
> > On Thu, Jan 15, 2015 at 06:51:01PM +0200, Michael S. Tsirkin wrote:
> > > Host no longer knows the alignment, so why is it needed?
> > > In fact, I notice that 4.3.2.3 Virtqueue Layout seems completely
wrong:
> > > it corresponds to legacy devices, and it does not
> > > say what "align" is.
> > 
> > I created
> > https://issues.oasis-open.org/browse/VIRTIO-129
> > to track this. Can you write up a patch please?
> 
> "4.3 Virtio Over Channel I/O"
> 
> (hint: not MMIO)
> 
> Pawe?
> 
> 
Ow. So it's s390 actually.
Fixed and assigned to Cornelia.

All the more reason not to have the ALIGN
macro in MMIO version 1 code.

-- 
MST