Re: [PATCH 2/2] hwmon: Allow to compile dell-smm-hwmon driver without /proc/i8k

From: Jean Delvare <jdelvare@suse.de>
To: "Pali Rohár" <pali.rohar@gmail.com>
Cc: Steven Honeyman <stevenhoneyman@gmail.com>,
	Paul Bolle <pebolle@tiscali.nl>,
	Guenter Roeck <linux@roeck-us.net>, Arnd Bergmann <arnd@arndb.de>,
	"Greg Kroah-Hartman" <gregkh@linuxfoundation.org>,
	Valdis Kletnieks <Valdis.Kletnieks@vt.edu>,
	Jochen Eisinger <jochen@penguin-breeder.org>,
	Gabriele Mazzotta <gabriele.mzt@gmail.com>,
	linux-kernel@vger.kernel.org, lm-sensors@lm-sensors.org
Subject: Re: [PATCH 2/2] hwmon: Allow to compile dell-smm-hwmon driver without /proc/i8k
Date: Mon, 30 Mar 2015 09:44:29 +0200	[thread overview]
Message-ID: <20150330094429.53eaacc1@endymion.delvare> (raw)
In-Reply-To: <201503282333.46803@pali>

On Sat, 28 Mar 2015 23:33:46 +0100, Pali Rohár wrote:
> I think it would be better to locate on which bus are sensors 
> connected (smbus? isa? pci?), find out HW chips and wrote native 
> kernel drivers for them...
> 
> I scanned smbus (intel controller connected on pci) on my E6440 
> but there is no additional/unknown device. I have no idea where 
> else could be sensor device connected and accessible (from SMM or 
> kernel mode).

As much as I hate SMM, native access to a BIOS-owned device is not a
good idea, as is it dangerously racy. If anything, we are currently
trying to move _away_ from native drivers on a number of systems
(basically all systems where acpi_enforce_resources=lax is needed.)

On x86, ideally ACPI would offer a standard interface to the hardware
monitoring chip and the OS would need a single driver for all boards
out there. Failing that, ACPI should provide a clean and safe way to
access the chip's registers (read: some mutex to avoid concurrent
access to the registers by the BIOS and the OS.)

-- 
Jean Delvare
SUSE L3 Support