From mboxrd@z Thu Jan 1 00:00:00 1970 From: Rob Herring Subject: Re: [RFC PATCH v1 1/3] dt-bindings: net: Add 'mac-address-lookup' property Date: Wed, 15 Aug 2018 14:45:29 -0600 Message-ID: <20180815204529.GA24830@rob-hp-laptop> References: <20180814223758.117433-1-briannorris@chromium.org> <20180814223758.117433-2-briannorris@chromium.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Greg Kroah-Hartman , "Rafael J. Wysocki" , Andrew Lunn , Florian Fainelli , Dmitry Torokhov , Guenter Roeck , netdev@vger.kernel.org, devicetree@vger.kernel.org, linux-kernel@vger.kernel.org, Julius Werner , Stephen Boyd , Brian Norris To: Brian Norris Return-path: Content-Disposition: inline In-Reply-To: <20180814223758.117433-2-briannorris@chromium.org> Sender: linux-kernel-owner@vger.kernel.org List-Id: netdev.vger.kernel.org On Tue, Aug 14, 2018 at 03:37:56PM -0700, Brian Norris wrote: > Some firmwares present data tables that can be parsed to retrieve > device-specific details, like MAC addresses. While in some cases, one > could teach the firmware to understand the device tree format and insert > a 'mac-address'/'local-mac-address' property into the FDT on its own, > this method can be brittle (e.g., involving memorizing expected FDT > structure), and it's not strictly necessary -- especially when parsers > for such firmware formats are already needed in the OS for other > reasons. If you have an 'ethernetN' alias then you don't need to know the structure. IIRC, that is what u-boot does. > > One such format: the Vital Product Data (VPD) [1] used by Coreboot. It > supports a table of key/value pairs, and some systems keep MAC addresses > there in a well-known format. Allow a device tree to specify > (1) that the MAC address for a given device is stored in the VPD table > and > (2) what key should be used to retrieve the MAC address for said > device (e.g., "ethernet_mac0" or "wifi_mac1"). > > [1] Ref: > https://chromium.googlesource.com/chromiumos/platform/vpd/+/master/README.md > TL;DR: VPD consists of a TLV-like table, with key/value pairs of > strings. This is often stored persistently on the boot flash and > presented via in-memory Coreboot tables, for the operating system to > read. > > Signed-off-by: Brian Norris > --- > Documentation/devicetree/bindings/net/ethernet.txt | 12 ++++++++++++ > 1 file changed, 12 insertions(+) > > diff --git a/Documentation/devicetree/bindings/net/ethernet.txt b/Documentation/devicetree/bindings/net/ethernet.txt > index cfc376bc977a..d3fd1da18bf4 100644 > --- a/Documentation/devicetree/bindings/net/ethernet.txt > +++ b/Documentation/devicetree/bindings/net/ethernet.txt > @@ -4,6 +4,18 @@ NOTE: All 'phy*' properties documented below are Ethernet specific. For the > generic PHY 'phys' property, see > Documentation/devicetree/bindings/phy/phy-bindings.txt. > > +- mac-address-lookup: string, indicating a method by which a MAC address may be > + discovered for this device. Methods may be parameterized by some value, such > + that the method can determine the device's MAC address using that parameter. > + For example, a firmware might store MAC addresses in a table, keyed by some > + predetermined string, and baked in read-only flash. A lookup method "foo" > + with a parameter "bar" should be written "foo:bar". > + Supported values for method: > + "google-vpd" - Google's Vital Product Data (VPD), as used in the Coreboot > + project. Documentation for VPD can be found at: > + https://chromium.googlesource.com/chromiumos/platform/vpd/+/master/README.md > + Example: > + mac-address-lookup = "google-vpd:ethernet_mac0" This doesn't strike me as a very DT style way of describing this. I would expect something like a phandle to the VPD and an identifier. Also, an already common way besides local-mac-address is using nvmem binding which wouldn't use this and perhaps could be used here? This feels very much Google specific, not common (and maybe that is fine). Rob