Re: [PATCH v7 03/15] s390x/pci: Move iommu_mr from S390PCIIOMMU to S390PCIBusDevice

[Matthew Rosato <mjrosato@linux.ibm.com>]

On 8/18/26 1:25 PM, Konstantin Shkolnyy wrote:
> This field is only used when S390PCIBusDevice exists, so it can be moved
> there to simplify S390PCIIOMMU which purpose is just to store the "root"
> AddressSpace.

I would argue that this is not the purpose of S390PCIIOMMU until the end
of this series.  Maybe just re-word this to '... to simplify
S390PCIIOMMU towards a structure that contains only the IOMMU container
information needed by the PCI layer for a given slot.'

And propagate the re-wording through patch 12.

> 
> Reviewed-by: Farhan Ali<alifm@linux.ibm.com>
> Signed-off-by: Konstantin Shkolnyy <kshk@linux.ibm.com>

I'm not sure if it was on-list or off-list, but I had asked previously
about re-visiting when the IOMMU is created as it relates to the
S390PCIBusDevice vs the PCIDevice; the existing lifecycle seems a bit
odd to me in that the IOMMU is a property of the S390PCIBusDevice but
rather than being created when the S390PCIBusDevice is created or
plugged, it's creation is deferred until a point where the PCIDevice is
plugged and associated with the S390PCIBusDevice.

I wonder if really it should be created earlier e.g. as soon as the
S390PCIBusDevice is created but only wired up (e.g. referenced in
table->iommu[]) once the PCIDevice<->S390PCIBusDevice association is made.

I think your approach here of moving many of the fields out of the
S390PCIIOMMU is kind of working around that lifecycle oddity to ensure
migration behaves as expected rather than addressing it outright.

At this point, I'll accept it since I don't think re-arranging the
fields makes the oddity any worse and it gets us useful new
functionality -- but I would really like you to look into this lifecycle
as a follow-on after this series.

So with that said and the commit message change:

Reviewed-by: Matthew Rosato <mjrosato@linux.ibm.com>