Re: [virtio-comment] comments for the virtio spec - Introduction section

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

On Fri, Jan 31, 2014 at 04:11:49PM +1030, Rusty Russell wrote:
> "Michael S. Tsirkin" <mst@redhat.com> writes:
> > On Thu, Jan 30, 2014 at 12:38:38PM +0200, Michael S. Tsirkin wrote:
> >> On Wed, Jan 29, 2014 at 03:04:50PM +1030, Rusty Russell wrote:
> >> > diff --git a/introduction.tex b/introduction.tex
> >> > index 5d57f78..98e1e0d 100644
> >> > --- a/introduction.tex
> >> > +++ b/introduction.tex
> >> > @@ -13,14 +13,14 @@ inter-guest communication) requires copying.
> >> >  }
> >> >  
> >> >    Efficient: Virtio devices consist of rings of descriptors
> >> > -  for input and output, which are neatly separated to avoid cache
> >> > +  for both input and output, which are neatly separated to avoid cache
> >> >    effects from both driver and device writing to the same cache
> >> >    lines.
> >> 
> >> missed this one. I agree.
> >
> > Actually, rereading this, it's still IMHO not very clear what is separated
> > from what. descriptors?
> >
> > I think output here means avail and input used?
> > If yes then let's say so:
> > +  for both input and output. Input rings are neatly separated from output rings
> > +  to avoid cache
> >
> >
> > Alternatively if we want to be more precise:
> >
> > Really descriptors are not in a ring at all:
> > they are in the buffer. OTOH used ring is separated from
> > avail ring, but they are not for input and output:
> > input and output are in descriptor buffer.
> > How about
> >
> > +  Efficient: Virtio devices consist of a memory buffer with descriptors
> > +  for both input and output as well as memory rings of indices
> > +  for driver-to-device and device-to-driver communication.
> > +  device-to-driver rings are neatly separated from driver-to-device
> > +  rings to avoid cache
> >
> > Though I think the shorter version is fine for introduction.
> 
> Indeed, good catch.  Hmm, how about we just replace the misleading
> word separated?
> 
>   Virtio devices consist of rings of descriptors
>   for both input and output, which are neatly laid out to avoid cache
>   effects from both driver and device writing to the same cache
>   lines.

Yes, this seems a better fit for the introduction.

> >> I tweaked this one a bit more.
> >> Can you pls tell me whether you agree with the patchset I posted?
> 
> Yep, getting to it now.
> 
> Cheers,
> Rusty.