Re: [PATCH 1/2] PCI: Don't treat ~0 as ready in pci_dev_wait() with RRS SV

[Lukas Wunner]

On Sun, Oct 11, 2026 at 10:24:08AM +0200, Alexander Gruhlke wrote:
> On Sat, Oct 10, 2026 at 08:36:38PM +0200, Lukas Wunner wrote:
> > I think it's sufficient if you just change the if-condition like this:
> >
> > -			if (!pci_bus_rrs_vendor_id(id))
> > +			if (!pci_bus_rrs_vendor_id(id) &&
> > +			    !PCI_POSSIBLE_ERROR(id)) {
> >
> > I don't see the need for the extra read of the Command register
> > you're performing.
> 
> The Vendor ID of a VF always reads ~0, so with that condition, an FLR
> of a VF below a Root Port with RRS SV would poll until the 60 s timeout
> and fail.  That's why the branch without RRS SV reads the Command
> register, too (see the comment above the loop).  I'll mention it in the
> commit message.

Ah okay, thanks for the clarification.  So how about:

-			if (!pci_bus_rrs_vendor_id(id))
+			if (!pci_bus_rrs_vendor_id(id) &&
+			    (!PCI_POSSIBLE_ERROR(id) || dev->is_virtfn))

Per PCIe r7.1 sec 9.4.3.3.1, VFs may respond with RRS Completions
under certain circumstance, so the Vendor ID does not *always* read
as 0xffff, despite sec 7.5.1.1.1 saying so.  The above should handle
that correctly (unless I've botched it again).

Thanks,

Lukas