On Friday, July 4th, 2025 at 2:58 AM, Mario Limonciello <superm1@xxxxxxxxxx> wrote: > If HID_UNIQ is the right way to get the firmware version but some > drivers encode a serial+firmware and others encode firmware that's going > to make for a very messy "generic" property to read the firmware version > of a device. I think I've also changed my mind on HID_UNIQ -- IIUC the whole point of HID_UNIQ is to be unique comparing devices with the same VID&PID. If I plug in two identical devices with the same firmware version then HID_UNIQ would be the same -- so I think HID_UNIQ should probably always be the serial number. I think HID_UNIQ should also stay the same during the firmware update too, so we don't want to jam two things into one property like "HID_UNIQ=serial:1234,fwver=12.34" Certainly exposing the firmware version as a HID property makes enumerating devices much easier in fwupd; the kernel driver often just knows the firmware version already and for userspace to re-query the firmware version using a per-device custom protocol seems pointless at best and dangerous at worse -- given we're typically doing a SetReport and GetReport with no sequence number available. I don't have much to comment on wrt the implementation, but providing a way for the hidraw kernel driver to export the current firmware version up to userspace makes a lot of sense to me. Richard
