Make pci_host_generic and thunderx_pci common

[wma]

Summary

• provided OFW interface for pci_host_generic (for handling devices which are present in DTS under the PCI node)
• removed support for internal pci from qrm64/cavium
• cleaned up and made most of the code common

[br]

very cool - thanks for that !

[br]

sys/dev/pci/pci_host_generic.c

187 ↗

(On Diff #13247)

struct generic_pcie_quirks *quirk;
quirk = g_quirks;

[wma]

Moved qiurk to machine-specific file as suggested by cognet. Also, cleaned up a little and changed bus_probe to be GENERIC.

[wma]

sys/dev/pci/pci_host_generic.c

116 ↗

(On Diff #13487)

will be removed

[andrew]

I may be missing something, but there don't seem to be any quirks defined.

[wma]

Actually, there is one in sys/dev/pci/pci_host_generic.c:232. The generic driver must know the machine-dependent domain (ECAM number) it operates on.

[andrew]

In that case I think it would make sense to create a sub class of the generic driver to handle the ThunderX specific changes.

[andrew]

sys/dev/pci/pci_host_generic.c

171 ↗

(On Diff #13605)

Will this work with ACPI?

629 ↗

(On Diff #13605)

Why didn't you just replace the contents of the old functions?

sys/dev/pci/pci_host_generic.h

2 ↗

(On Diff #13605)

You seem to have misspelled Ruslans name. I only see a few lines in this file that didn't come from pci_host_generic.c.

30 ↗

(On Diff #13605)

Missing tab

[wma]

sys/dev/pci/pci_host_generic.c

171 ↗

(On Diff #13605)

This whole file is meant to be used when FDT is enabled. Are you suggesting surrounding this part with ifdef FDT? Or did you have something else in mind?

629 ↗

(On Diff #13605)

I don't want to mix the code for pci and ofw allocation, it'd make the functions too long, hard to read and to maintain. I'm using the same pattern here as in old thunder_pcie.c.