pci_device::subsystem_device and pci_device::subsystem_vendor have been there since the beginning, but they were never populated. This went unnoticed so far, because they are compared against PCI_ANY_ID everywhere, except for some NS16550-over-PCI quirks. We should either drop the members or populate them unconditionally. Let's do as Linux does and populate them for devices[1]. [1]: https://elixir.bootlin.com/linux/v6.5-rc7/source/drivers/pci/probe.c#L1915 Signed-off-by: Ahmad Fatoum <a.fatoum@xxxxxxxxxxxxxx> --- drivers/pci/pci.c | 3 +++ 1 file changed, 3 insertions(+) diff --git a/drivers/pci/pci.c b/drivers/pci/pci.c index d1b7549d7100..638af9722efc 100644 --- a/drivers/pci/pci.c +++ b/drivers/pci/pci.c @@ -434,6 +434,9 @@ static unsigned int pci_scan_bus(struct pci_bus *bus) goto bad; setup_device(dev, 6); + + pci_read_config_word(dev, PCI_SUBSYSTEM_ID, &dev->subsystem_device); + pci_read_config_word(dev, PCI_SUBSYSTEM_VENDOR_ID, &dev->subsystem_vendor); break; case PCI_HEADER_TYPE_BRIDGE: child_bus = pci_alloc_bus(); -- 2.39.2
