[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250402124131-mutt-send-email-mst@kernel.org>
Date: Wed, 2 Apr 2025 12:43:47 -0400
From: "Michael S. Tsirkin" <mst@...hat.com>
To: David Woodhouse <dwmw2@...radead.org>
Cc: virtio-comment@...ts.linux.dev, hch@...radead.org,
Claire Chang <tientzu@...omium.org>,
linux-devicetree <devicetree@...r.kernel.org>,
Rob Herring <robh+dt@...nel.org>,
Jörg Roedel <joro@...tes.org>,
iommu@...ts.linux-foundation.org, linux-kernel@...r.kernel.org,
graf@...zon.de
Subject: Re: [RFC PATCH 1/3] content: Add VIRTIO_F_SWIOTLB to negotiate use
of SWIOTLB bounce buffers
On Wed, Apr 02, 2025 at 05:16:28PM +0100, David Woodhouse wrote:
> On Wed, 2025-04-02 at 11:51 -0400, Michael S. Tsirkin wrote:
> > On Wed, Apr 02, 2025 at 04:47:18PM +0100, David Woodhouse wrote:
> > > On Wed, 2025-04-02 at 11:20 -0400, Michael S. Tsirkin wrote:
> > > > On Wed, Apr 02, 2025 at 04:12:39PM +0100, David Woodhouse wrote:
> > > > > On Wed, 2025-04-02 at 10:54 -0400, Michael S. Tsirkin wrote:
> > > > > > > + If a the device transport provides a software IOTLB bounce buffer,
> > > > > > > + addresses within its range are not subject to the requirements of
> > > > > > > + VIRTIO_F_ACCESS_PLATFORM as they are considered to be ``on-device''
>
> ...
>
> > The text you wrote makes it seem that even if the platform says use
> > an IOMMU, it should be bypassed.
>
> It was trying just to state the obvious, that addresses within the
> range of the *on-device* memory buffer are not handled by the IOMMU.
>
> > I would drop this text, and maybe add some clarification in the mmio transport,
> > as needed.
>
> It would be PCI too. I guess we could move the "obvious" comment that
> 'addresses within the range of the SWIOTLB bounce buffer region are
> considered to be "on-device" and are thus not affected by the
> requirements of VIRTIO_F_ACCESS_PLATFORM' into *both* the MMIO and PCI
> transport docs? But then it's basically just saying the same thing in
> two different locations?
>
> I don't think we're debating what the actual implementations should
> *do* ... are we? To me it's obvious that what I'm trying to say here
> *should* always be true.
>
> We're just debating the wording and where to put it, yes?
yes.
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.
For mmio it is my understanding that the "restricted" does the same
already? or is it required in the spec for some reason?
Powered by blists - more mailing lists