On Thu, Feb 23, 2023 at 09:03:14AM +0000, Tian, Kevin wrote:
From: Jason Gunthorpe jgg@nvidia.com Sent: Thursday, February 23, 2023 5:03 AM
The HWPT is always linked to an IOAS and once a HWPT exists its domain should be fully mapped. This ended up being split up into device.c during a two phase creation that was a bit confusing.
Move the iopt_table_add_domain() into iommufd_hw_pagetable_alloc() by having it call back to device.c to complete the domain attach in the required order.
Calling iommufd_hw_pagetable_alloc() with immediate_attach = false will work on most drivers, but notably the SMMU drivers will fail because they can't decide what kind of domain to create until they are attached. This will be fixed when the domain_alloc function can take in a struct device.
Is below understanding correct on how to retire immediate_attach?
- immediate_attach=true in auto domain path for back compat (what this patch does);
yes
- immediate_attach=false when adding hwpt_alloc() uAPI and VT-d support;
yes
- fix domain_alloc() to take struct device so the SMMU drivers can decide the domain info w/o relying on attach;
Yes, this is where Robin's patches are going.
- remove immediate_attach and enable hwpt_alloc() on SMMU.
Sort of, the domain_alloc_user already takes in the device so SMMU can be fixed for that API immediately, in theory, but it looks like some amount of work to get there.
Jason