iwx: Look up subdevice rather than device

[thj]

Summary

iwx goes through a couple of steps to match devices to firmware. There is a default look up of pcie device and subdevices and then a secondary pass to find config in a table.
Previously we were not looking up the subdevice
correctly and that led to ax210 cards in some devices (Framework 13)
form being matched to the correct firmware.

Test Plan

This allows my ax210 card to attach, but I still need to check for regressions
in other hardware.

[adrian]

Interesting, do we not have a pci_get_subdevice() call? @jhb

[bz]

PCI accessors are all there @adrian pci_get_(first column)

PCI_ACCESSOR(subvendor,         SUBVENDOR,      uint16_t)
PCI_ACCESSOR(subdevice,         SUBDEVICE,      uint16_t)
PCI_ACCESSOR(vendor,            VENDOR,         uint16_t)
PCI_ACCESSOR(device,            DEVICE,         uint16_t)
PCI_ACCESSOR(devid,             DEVID,          uint32_t)
PCI_ACCESSOR(class,             CLASS,          uint8_t)
PCI_ACCESSOR(subclass,          SUBCLASS,       uint8_t)
...

are all there.

@thj If you look at man iwlwififw.4 you can probably compare what is matched for PNP information. Or read sys/contrib/dev/iwlwifi/pcie/drv.c and you'll find that it's not just matching PCI IDs for some chipsets but also read other things beyond PCI ID to figure out what kind of device it is. It's a pain. I think the 9xxx were especially bad. For the SYSCTL I added I am actually using the entirely processed information from Intel after the driver got loaded (no need to attach to anything) to get the list half-way correct out of it.

[thj]

In D49759#1134392, @bz wrote:

PCI accessors are all there @adrian pci_get_(first column)

PCI_ACCESSOR(subvendor,         SUBVENDOR,      uint16_t)
PCI_ACCESSOR(subdevice,         SUBDEVICE,      uint16_t)
PCI_ACCESSOR(vendor,            VENDOR,         uint16_t)
PCI_ACCESSOR(device,            DEVICE,         uint16_t)
PCI_ACCESSOR(devid,             DEVID,          uint32_t)
PCI_ACCESSOR(class,             CLASS,          uint8_t)
PCI_ACCESSOR(subclass,          SUBCLASS,       uint8_t)
...

are all there.

Excellent, thanks for the reference. I did a quick survey and the drivers I found just seemed to be doing whatever.

@thj If you look at man iwlwififw.4 you can probably compare what is matched for PNP information. Or read sys/contrib/dev/iwlwifi/pcie/drv.c and you'll find that it's not just matching PCI IDs for some chipsets but also read other things beyond PCI ID to figure out what kind of device it is. It's a pain. I think the 9xxx were especially bad. For the SYSCTL I added I am actually using the entirely processed information from Intel after the driver got loaded (no need to attach to anything) to get the list half-way correct out of it.

Great, I'll have a check. This solves the attach for my card, I'm a bit concerned people will get a different firmware now and it'll reproduce the "performance" I get

[thj]

• use the correct pci_get_subdevice macro