Re: [PATCH v2 04/25] iommu: Add IOMMU_DOMAIN_PLATFORM for S390

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Jun 01, 2023 at 07:25:32PM +0100, Robin Murphy wrote:
> On 2023-05-16 01:00, Jason Gunthorpe wrote:
> > The PLATFORM domain will be set as the default domain and attached as
> > normal during probe. The driver will ignore the initial attach from a NULL
> > domain to the PLATFORM domain.
> > 
> > After this, the PLATFORM domain's attach_dev will be called whenever we
> > detach from an UNMANAGED domain (eg for VFIO). This is the same time the
> > original design would have called op->detach_dev().
> > 
> > This is temporary until the S390 dma-iommu.c conversion is merged.
> 
> If we do need a stopgap here, can we please just call the current situation
> an identity domain? 

I really rather wouldn't lie, especially since we need something for
PPC more permanently. I definately don't want to call PPC IDENTITY.

> Then similarly for patch #3 - since we already know s390 is temporary, it
> seems an anathema to introduce a whole domain type with its own weird
> ops->default_domain mechanism solely for POWER to not actually use domains
> with.

A main point of this entire series is to remove the NULL default
domain, so power's weirdness has to be accomodated somehow.
 
> In terms of reasoning, I don't see that IOMMU_DOMAIN_PLATFORM is any more
> useful than a NULL default domain, it just renames the problem

Yes! Renaming is the whole point.

We are giving actuall meaningful names to all the things that drivers
did with NULL. The ones that actually implemented IDENTITY correctly
gets to be called IDENTITY, everyone else needs to get labeled
something else.

Then we can tell what is correct and what needs fixing at a glance.

That is the *whole point*.

> and gives us more code to maintain for the privilege.

The PLATFORM bit is only 3 lines of core code. Let's not overstate the
cost of a label please.

> As I say, though, we don't actually need to juggle the semantic of a
> "we don't know what's happening here" domain around any further,
> since it works out that a "we're not influencing anything here"
> domain actually suffices for what we want to reason about, and those
> are already well-defined. Sure, the platform DMA ops *might* be
> doing more, but that's beyond the scope of the IOMMU API either
> way. At that point, lo and behold, s390 and POWER now look just like
> ARM and the core code only needs a single special case for
> arch-specific default identity domains, lovely!

I'm really against mis-labeling things.

That is totally the wrong direction for this series. The point is to
label things correctly. Labeling something that is not an IDENTITY
operation as IDENTITY is just repeating the whole mistake of an
ill-defined NULL all over again.

Labels need to be correct, labels are cheap to add, so lets not be
afraid to use proper labels to actually describe the underlying
behavior.

If you don't like PLATFORM we can choose PRIVATE, OPAQUE or some other
label, but not IDENTITY.

Jason