Hi Matthew,

On Tue, 20 Feb 2007 15:18:13 +0000, Matthew Garrett wrote:

On Sun, Feb 18, 2007 at 06:38:05PM +0100, Jean Delvare wrote:

quoted

ACPI is broken here, not k8temp, so let's fix ACPI instead. ACPI
doesn't conflict with only k8temp, but with virtually all hardware
monitoring drivers, all I2C/SMBus drivers, and probably other types of
drivers too. We just can't restrict or blacklist all these drivers
because ACPI misbehaves.

No, the simple fact of the matter is that if you're running on an ACPI 
platform you need to change some of your assumptions. ACPI owns the 
hardware. The OS doesn't. To an extent this has always been true on 

The Linux device driver model assumes that it owns the hardware. If
this is not true, then should we prevent any non-ACPI driver from
loading as soon as ACPI is enabled?

laptops and servers /anyway/ - the BIOS is free to have a wide variety 
of SMM insanity that invalidates basic assumptions like "If I hold this 
lock, nothing can interrupt me between this write and this read". That's 
simply not true.

Yeah, this is correct, and just as unfortunate. It's amazingly sad that
hardware vendors as a whole are still repeating the same design
mistakes over and over again :(

So this isn't about fixing ACPI. It's about trying to find a mechanism 
that allows ACPI and raw hardware drivers to coexist, which is made 

Exactly what I said, you're only rewording it to make it sound nicer ;)

somewhat harder by it not being a situation that the platform designers 
have considered in the slightest. The suggested low-level driver for 
io-port arbitration would certainly be a step forward in making this 
work better.

I sure hope we can find a solution, by as your said yourself, nothing
is going to prevent SMM and similar oddities from messing up the drivers
assumptions.

-- 
Jean Delvare