tb_pci: Don't try to attach to PCI buses that aren't below a PCI-PCI brige

[jhb]

Summary

In particular, don't try to probe PCI buses that are children of host
bridges such as the pci0 child of pcib0. This fixes a panic on boot
if tb.ko is loaded at boot time (which the driver recommends for
certain chipsets).

[imp]

So why is this driver only valid in those cases?

[jhb]

In D53202#1215346, @imp wrote:

So why is this driver only valid in those cases?

The panic is that it tries to do pci_get_vendor() on pcib0 (child of acpi0) and that isn't a valid IVAR and the accessor panics. In particular, note that this driver is expecting to call tb_pcib_find_ident on its parent, and tb_pcib_find_ident assumes the device it is operating on is a PCI device with a device ID, etc. That isn't true for host bridges that are children of ACPI or OFW, etc.

[jhb]

Also, FWIW, this just fixes a panic when the probe routine runs against pci0 on my laptop. Later when the driver tries to attach to the actual Thunderbolt slots in my laptop it panics in a different way (the attach still fails as it does post-boot, and I get a different form of a race with the interrupt handler that still results in an interrupt handler raising a GPF, presumably due to dereferencing freed memory field with 0xdeadc0de as in the other review),