LinuxKPI: pci: fix pcie_get_speed_cap()

[bz]

Summary

pcie_get_speed_cap() has a hard coded skip of 3 devices at the
beginning. It is either called on a pdev or on a result from
pci_upstream_bridge(). In the latter case skipping another three
devices might get us to acpi0 or nexus, neither of which is a
PCI device still and pci_get_vendor() will panic() on that.

MFC after: 3 days

[emaste]

I've applied this to my local tree for testing

[bz]

I've run with this since Sep 17 apparently even though my amdgpu was only useable for ~4 weeks during that time frame.

@emaste should I just put this into main and stable/15 and we'll see?

[bz]

Also @jhb from looking at the code, on which PCI devices should it be called? I am still confused by the (original) naming of "root". I would assume it's supposedly called on a pcib. Is that correct?