Re: [Qemu-devel] [RFC 3/3] memory: introduce IOMMU_NOTIFIER_USER_[UN]SET

[Peter Xu <peterx@redhat.com>]

On Wed, Jun 06, 2018 at 02:56:29PM +0800, Peter Xu wrote:

[...]

> > The patches you've
> > sent here seem to me to be adding extra complexity to the
> > notifier API and the core code which isn't required in
> > the patch set that I sent.
> 
> Again I didn't really understand the problem behind before.  Now I
> don't think it's important on whether we'll need to introduce new
> notifier flags, I wanted to know whether we can avoid introducing
> IOMMU index into memory API. Let's just ignore this patch.  Sorry for
> that confusion.

For this part, I'd say of course this is based on the assumption that
the IOMMU index might not help much with non-tcg use cases.  If
someone could help me understand better on how that could help in
general to memory API then I would be much appreciated, since it
really confused me even until today (and I can't see a good way at
least for VT-d to leverage that idea).  Thanks,

-- 
Peter Xu