Re: [PATCH 2/5] rust: device: Add a minimal RawDevice trait

[Greg Kroah-Hartman]

On Fri, Feb 24, 2023 at 06:52:04PM +0000, Robin Murphy wrote:
> > But again, let's see about disconnecting the iommu ops entirely from the
> > device and just relying on the bus, that should work better, rigth?
> 
> Um, other way - iommu_ops have almost finished moving *off* the bus to be
> per-device. You acked the patch the other week and seemed pleased with it ;)

Oops, yes, sorry, I meant the other way around in my comment, see, it's
confusing :)

And yes, I want to see your patch series merged, it will make my "clean
up the driver core lifetime rules" work a lot easier, is it still
planned for 6.3-rc1?

> But either way that's orthogonal to this case, since the device that
> io-pgtable deals with cannot have iommu_ops of its own (it *is* the IOMMU,
> or at least a GPU MMU acting as one).

Agreed, and thanks for the detailed explaination, it makes more sense
now.  But I still think the pointer reference needs to be incremented
when saved off to be at least a bit more sane.

thanks,

greg k-h