Re: [PATCH] VFIO driver: Non-privileged user level PCI drivers

[Joerg Roedel <joro@8bytes.org>]

On Wed, Jun 02, 2010 at 04:17:19PM +0300, Michael S. Tsirkin wrote:
> On Wed, Jun 02, 2010 at 02:50:50PM +0200, Joerg Roedel wrote:
> > On Wed, Jun 02, 2010 at 03:25:11PM +0300, Avi Kivity wrote:
> > > On 06/02/2010 03:19 PM, Joerg Roedel wrote:

> 
> > If its
> > required anyway the binding can happen implicitly. We could allow to do
> > a nop 'ioctl(dev1, SHARE, dev1)' to remove the asymmetry.
> 
> And then when we assign meaning to it we find that half the apps
> are broken because they did not call this ioctl.

The meaning is already assigned and chaning it means changing the
userspace-abi which is a no-go.

> This simple scenario ignores all the real-life corner cases.
> For example, with an explicit iommu open and bind application
> can naturally detect that:
> - we have run out of iommu domains

ioctl(dev, MAP, ...)  will fail in this case.

> - iommu is unsupported

Is best checked by open() anyway because userspace can't do anything
with the device before it is bound to a domain.

> - iommu is in use by another, incompatible device

How should this happen?

> - device is in bad state

How is this checked with your proposal and why can this not be detected
with my one?

> because each is a separate operation, so it is easy to produce meaningful
> errors.

Ok, this is true.

> Another interesting thing that a separate iommu device supports is when
> application A controls the iommu and application B
> controls the device.

Until Linux becomes a micro-kernel the IOMMU itself will _never_ be
controlled by an application.

> This might be good to e.g. improve security (B is run by root, A is
> unpriveledged and passes commands to/from B over a pipe).

Micro-kernel arguments. I hope a userspace controlled IOMMU in Linux
will never happen ;-)

	Joerg