On 2/3/26 11:37 PM, Jason Gunthorpe wrote:
On Tue, Feb 03, 2026 at 09:22:13PM +0530, Shivaprasad G Bhat wrote:
Then you'd want to introduce a new domain op to get the apertures
instead of the single range hard coded into the domain struct. The new
op would be able to return a list. We can use this op to return
apertures for sign extension page tables too.
Update iommufd to calculate the reserved regions by evaluating the
whole list.
I think you'll find this pretty straight forward, I'd do it as a
followup patch to this one.
Thanks. I will wait for that patch.
I think you will have to make it :)
Sure. Thanks!
There are ioctl number conflicts like
# grep -n "VFIO_BASE + 1[89]" include/uapi/linux/vfio.h | grep define
940:#defineVFIO_DEVICE_BIND_IOMMUFD_IO(VFIO_TYPE, VFIO_BASE + 18)
976:#defineVFIO_DEVICE_ATTACH_IOMMUFD_PT_IO(VFIO_TYPE, VFIO_BASE + 19)
1833:#defineVFIO_IOMMU_SPAPR_UNREGISTER_MEMORY_IO(VFIO_TYPE, VFIO_BASE + 18)
1856:#defineVFIO_IOMMU_SPAPR_TCE_CREATE_IO(VFIO_TYPE, VFIO_BASE + 19)
# grep -n "VFIO_BASE + 20" include/uapi/linux/vfio.h | grep define
999:#defineVFIO_DEVICE_DETACH_IOMMUFD_PT_IO(VFIO_TYPE, VFIO_BASE + 20)
1870:#defineVFIO_IOMMU_SPAPR_TCE_REMOVE_IO(VFIO_TYPE, VFIO_BASE + 20)
It's Ok the compat codes will know what type it is operating in before
it decodes any of those.
I agree, the additional PPC specific ioctls can be selectively handled based
on arch.
You are right. We do have some use cases beyond VMM, I will consider compat
driver only if it is helpful there.
You can also use the type1 compat mode which will magically start
working with PPC..
You should also implement the BLOCKING domain type to make VFIO work
better
I am not sure how this could help making VFIO better. May be, I am not able
to imagine the advantages with the current platform domain approach
in place. Could you please elaborate more on this?
VFIO always uses a BLOCKED domain when it opens the device, then it
changes to a paging domain. If the driver doesn't support a native
BLOCKED domain then it allocates an empty page table and uses that.
A proper native BLOCKED domain has better error handling
characteristics since it is not allowed to fail attach and it doesn't
require allocation.
I think you will also find what you are doing easier if you push the
iommu_domain down through the PPC iommu ops instead of retaining these
unnecessary historical layers.
I see there is some more history around this being discussed before on
the patch here,
https://lore.kernel.org/all/[email protected]/
Will try to dig more and address this in future.
Thanks,
Shivaprasad
Jason
