Re: [RFC PATCH 1/3] content: Add VIRTIO_F_SWIOTLB to negotiate use of SWIOTLB bounce buffers

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

On Thu, Apr 03, 2025 at 08:37:20AM +0100, David Woodhouse wrote:
> On Thu, 2025-04-03 at 00:29 -0700, Christoph Hellwig wrote:
> > On Wed, Apr 02, 2025 at 06:10:53PM +0100, David Woodhouse wrote:
> > > > I know a bit more about PCI, and for PCI I prefer just not saying
> > > > anything. The platform already defines whether it is behind an iommu
> > > > or not, and duplication is not good.
> > > 
> > > Not a hill for me to die on I suppose, but I would personally prefer to
> > > spell it out in words of one syllable or fewer, to make *sure* that
> > > device and driver authors get it right even though it's "obvious".
> > > 
> > > After all, if we could trust them to do their thinking, we would never
> > > have had the awful situation that led to VIRTIO_F_ACCESS_PLATFORM
> > > existing in the first place; the legacy behaviour we get when that bit
> > > *isn't* set would never have happened.
> > 
> > You'll need to define the semanics for VIRTIO_F_ACCESS_PLATFORM only
> > then.  
> > 
> 
> You mean the semantics for VIRTIO_F_ACCESS_PLATFORM only, without
> VIRTIO_F_SWIOTLB? Are those not defined already?
> 
> > An the only sane answer there is: don't allow non-translated
> > regions at all an in a broader sense stop people to use
> > VIRTIO_F_ACCESS_PLATFORM at all or at least for anything that requires
> > a new feature bit.
> > 
> > > > For mmio it is my understanding that the "restricted" does the same
> > > > already? or is it required in the spec for some reason?
> > > 
> > > No, it's exactly the same. But I still don't trust driver authors to
> > > realise the obvious, or VMM implementations either for that matter.
> > > 
> > > I'm not sure I see the *harm* in spelling out explicitly for the hard-
> > > of-thinking.
> > 
> > Write a whitepaper than and explain how it maps to the existing perfectly
> > working features.  Note that VIRTIO_F_ACCESS_PLATFORM just like
> > everything in virtio would actually benefit from being turned into
> > proper spec language, but anecdotes about random use cases are not
> > helpful.
> 
> Hm. I was just trying to point out what seemed obvious, that when a PCI
> device does 'DMA' to an address region which is actually within one of
> its *own* BARs, it isn't going to reach the PCI bus and get translated
> by an IOMMU. If it's causing this much contention, I'll just drop it.
> It didn't *change* anything anyway, except hopefully avoiding bugs in
> implementations.
> 



If you want, I would just spell this in the transport text then.
"Note that on most existing platforms,
 and since the BAR is part of the device itself,
 access controls generally to do not apply to device accesses there,
 and this is true even when VIRTIO_F_ACCESS_PLATFORM has been negotiated".