Re: [PATCH v2 1/2] rust: pci: skip probing VFs if driver doesn't support VFs

[Jason Gunthorpe <jgg@nvidia.com>]

On Thu, Oct 02, 2025 at 11:32:44PM +0200, Danilo Krummrich wrote:

> So, when we call pdev.physfn().drvdata_borrow::<NovaCore>() the checks are
> included already.

I'm not keen on hiding this reasoning inside an physfn() accessor like
this. ie one that returns a Device<Bound>. The reasoning for this is
tricky and special. We have enough cases where physfn won't be a bound
driver. I think it is big stretch just to declare that unconditionally
safe.

There is a reason pci_iov_get_pf_drvdata() has such a big comment..

So I'd rather see you follow the C design and have an explicit helper
function to convert a VF bound device to a PF bound device and check
the owner, basically split up pci_iov_get_pf_drvdata() into a part to
get the struct device and an inline to get the drvdata. Rust still has an
ops pointer it can pass in so it can be consistent with the C code
even if it does another check inside its drvdata_borrow.

This way we keep the reasoning and explanation in one place.

Jason