From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jean Delvare Subject: Re: [PATCH 1/3] MFD: introduce lpc_sch for Intel SCH LPC bridge Date: Tue, 16 Feb 2010 18:29:47 +0100 Message-ID: <20100216182947.61f0e6ff@hyperion.delvare> References: <4B73DAEE.5080400@compulab.co.il> <4B73DB4B.40501@compulab.co.il> <20100216100809.GA2935@sortiz.org> <43F901BD926A4E43B106BF17856F0755A2D58DB5@orsmsx508.amr.corp.intel.com> <20100216164523.5a3e13a5@hyperion.delvare> <43F901BD926A4E43B106BF17856F0755A2D58F23@orsmsx508.amr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <43F901BD926A4E43B106BF17856F0755A2D58F23-osO9UTpF0UQ8Ug9VwtkbtrfspsVTdybXVpNB7YpNyf8@public.gmane.org> Sender: linux-i2c-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: "Pan, Jacob jun" Cc: Samuel Ortiz , Denis Turischev , LKML , David Brownell , "linux-i2c-u79uwXL29TY76Z2rM5mHXA@public.gmane.org" List-Id: linux-i2c@vger.kernel.org On Tue, 16 Feb 2010 09:19:58 -0800, Pan, Jacob jun wrote: > >That shouldn't be a problem. The PCI device is still present, so > >sensors-detect will see it. Then it will load the required driver, and > >that driver will instantiate i2c adapters. The script then probes all > >i2c adapters regardless of who created them, so the exact driver > >implementation doesn't matter. > > > [[JPAN]] thanks for explaining it, all made sense to me. looking at > sensors-detect, it will still load isch driver for the same pci id. > but how would it know it depends on the new lpc driver to set up resources? > i can't see shared symbols. > > }, { > vendid => 0x8086, > devid => 0x8119, > procid => "Intel SCH", > driver => "i2c-isch", > }, On systems with udev, I would expect the mfd driver to load automatically through module aliases, so this should be OK. Actually, even i2c-isch should load automatically if we setup proper aliases for the platform devices it instantiates. For other systems, indeed, the mfd driver won't load in the absence of a shared symbol. Might be worth adding request_module() call in platform drivers? If this isn't acceptable for whatever reason, I can certainly hack sensors-detect to treat this uncommon case properly, but solving this problem at the application level seems wrong. -- Jean Delvare