From mboxrd@z Thu Jan 1 00:00:00 1970 From: Crestez Dan Leonard Subject: Re: [PATCH 2/3] acpi i2c: Initialize info.type from of_compatible Date: Fri, 17 Jun 2016 15:07:39 +0300 Message-ID: <2ca8f355-bb28-8b68-7579-e26eaee3550a@intel.com> References: <13cb5b7766b1b25a2bb61ce0053a38f6f440907b.1466014238.git.leonard.crestez@intel.com> <20160616081145.GU1740@lahna.fi.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=windows-1252 Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20160616081145.GU1740@lahna.fi.intel.com> Sender: linux-kernel-owner@vger.kernel.org To: Mika Westerberg Cc: linux-acpi@vger.kernel.org, "Rafael J. Wysocki" , Len Brown , linux-i2c@vger.kernel.org, Wolfram Sang , linux-spi@vger.kernel.org, Mark Brown , linux-kernel@vger.kernel.org, Octavian Purdila , Jarkko Nikula List-Id: linux-acpi@vger.kernel.org On 06/16/2016 11:11 AM, Mika Westerberg wrote: > On Wed, Jun 15, 2016 at 09:30:28PM +0300, Crestez Dan Leonard wrote: >> When using devicetree i2c_board_info.type is set to the compatible >> string with the vendor prefix removed. For I2C devices described via >> ACPI the i2c_board_info.type string is set to the ACPI device name. When >> using ACPI and DT ids this string ends up something like "PRP0001:00". >> >> If the of_compatible property is present try to use that instead. This >> makes it easier to instantiate i2c drivers through ACPI with DT ids. > > Adding Jarkko -- he was the original author of I2C/SPI device naming in > ACPI. Since this only affects devices with PRP0001 in their _HID/_CID, > it should not break anything. It's possible to have something like this: Name (_HID, "XXX1234") Name (_CID, "PRP0001") Name (_DSD, Package () { ToUUID("daffd814-6eba-4d8c-8a91-bc9bbf4aa301"), Package () { Package () {"compatible", "vendor,x1234"}, } }) With these patches the i2c device name changes from "XXX1234:00" to "x1234". This would happen even if a driver matches the XXX1234 acpi ID. It's not clear if anyone actually uses that or if this change would be harmful. It theory it would be possible to use of_compatible for naming only if there are no other real IDs. I don't think it's worthwhile.