On Thu, Jul 11, 2024 at 3:42 PM Guenter Roeck <linux@xxxxxxxxxxxx> wrote: > > On 7/11/24 14:09, James Hilliard wrote: > > >> The best we could possibly do might be to add a check for the bit in register > >> 0xf1 and warn the user that they might have to use the ACPI driver if the bit > >> is set. I am not sure if that would be helpful or just add noise, though. > > > > Do your systems which work with the it87_wdt driver have that 0xF1 bit not set? > > > > I only have one such system left, and the bit is not set on that system. > I avoid buying hardware with ITE Super-IO chips nowadays since their support > for Linux is non-existent. Yeah, I got stuck with a fleet of these boards, trying to make the best of it. > > > I'm thinking we should check for that bit and prevent loading the > > it87_wdt driver if > > No. That would create the risk of no longer loading the driver on systems where > it currently works. Hmm, any idea how likely it would be that the bit could be set on a board which the driver works on? Or maybe best to have a quirks table with dmi matching to disable the driver on known broken systems? > > > it's set(maybe along with an override param). That way the wdat_wdt driver I > > I prefer the less invasive version of logging a message. The user can then > block the it87_wdt driver if it doesn't work. Hmm, I build multiple watchdog drivers into the same kernel and somewhat rely on the autodetection working correctly as I support multiple boards with the same kernel build. It's not exactly trivial to conditionally prevent drivers from loading when built into the kernel AFAIU. > > Guenter >
