From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mika Westerberg Subject: Re: [PATCH] i2c / ACPI: Pick the first address if device has multiple Date: Tue, 13 Jan 2015 19:00:07 +0200 Message-ID: <20150113170007.GV1386@lahna.fi.intel.com> References: <1419860928-195404-1-git-send-email-mika.westerberg@linux.intel.com> <20150113155059.GN7660@katana> <1421167477.4459.229.camel@spandruv-desktop.jf.intel.com> <20150113164829.GO7660@katana> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: Content-Disposition: inline In-Reply-To: <20150113164829.GO7660@katana> Sender: linux-i2c-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Wolfram Sang Cc: Srinivas Pandruvada , linux-i2c-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-kernel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-i2c@vger.kernel.org On Tue, Jan 13, 2015 at 05:48:29PM +0100, Wolfram Sang wrote: > On Tue, Jan 13, 2015 at 08:44:37AM -0800, Srinivas Pandruvada wrote: > > On Tue, 2015-01-13 at 16:50 +0100, Wolfram Sang wrote: > > > On Mon, Dec 29, 2014 at 03:48:48PM +0200, Mika Westerberg wrote: > > > > ACPI specification allows I2C devices with multiple addresses. The current > > > > implementation goes over all addresses and assigns the last one to the > > > > device. This is typically not the primary address of the device. > > > > > > > > Instead of doing that we assign the first address to the device and then > > > > let the driver handle rest of the addresses as it wishes. > > > > > > > > Signed-off-by: Mika Westerberg > > > > Cc: Srinivas Pandruvada > > > > > > Yes, seems better than what we do know. But maybe taking the lowest > > > address is a bit better heuristic than taking the first address? > > > Not sure, though... > > The problem in taking lowest is that in many cases in current devices, > > the lowest address may end being 0x0C, which is reserved address for > > SMBUS (ARA). This will require different handling. Unfortunately ACPI > > doesn't have a way to distinguish whether SMBUS support is desired or > > not. > > The other option is to skip all reserved addresses for SMBUS also and > > then create on the lowest. > > Well, this makes me think that Mika's approach is probably the sanest > one... Also I think it is more consistent that way.