[Ksummit-2013-discuss] [ATTEND] ACPI vs DT

H. Peter Anvin hpa at zytor.com
Fri Jul 26 23:23:59 UTC 2013


On 07/24/2013 02:28 PM, linux at roeck-us.net wrote:
> 
> Every watchdog and hwmon driver for any SuperIO chip out there does
> exactly the same. The SuperIO GPIO driver for PC8763x (for some reason
> located in drivers/char) does the same. The parallel port driver for PCs
> does the same.
> 

The parallel port driver certainly predates ACPI... both the code and
the devices.  This is the kind of stuff that dates back to the original
IBM PC.  Even the IBM PC actually had an ad hoc method for detecting
parallel and serial ports (a table in the BDA), but Linux never used it.

> If there is some better means to implement SuperIO chip detection, it
> would help if you let us (the watchdog and hwmon maintainers) know. Yes,
> I understand from this discussion that there is supposed to be some
> means to detect the chip and its configuration register location through
> ACPI, but how exactly that is supposed to work evades at least me. "Do
> it through ACPI" won't help; you will have to me more specific than that.

I/O probing should certainly only ever be done as a last resort.  Now,
there has been an almost infinite number of sins committed in the PC
ecosystem over the past 32 years, so there will -- and are -- quirks and
exceptions all over the place.  However, we should avoid adding them if
at all possible.

	-hpa



More information about the Ksummit-2013-discuss mailing list