From mboxrd@z Thu Jan 1 00:00:00 1970 Authentication-Results: mx.groups.io; dkim=missing; spf=none, err=SPF record not found (domain: fensystems.co.uk, ip: 212.13.204.60, mailfrom: mbrown@fensystems.co.uk) Received: from duck.fensystems.co.uk (duck.fensystems.co.uk [212.13.204.60]) by groups.io with SMTP; Tue, 23 Jul 2019 04:00:08 -0700 Received: from pudding.home (unknown [IPv6:2001:470:1dbe:1:ffff::c5f1]) by duck.fensystems.co.uk (Postfix) with ESMTPSA id D2E6F1389; Tue, 23 Jul 2019 12:00:05 +0100 (BST) Subject: Re: [edk2-platforms: PATCH 1/1] Platform/RPi3: Accept "ethernet" or "ethernet0" aliases in device tree To: Leif Lindholm Cc: devel@edk2.groups.io, ard.biesheuvel@linaro.org, pete@akeo.ie References: <20190719172907.788170-1-mbrown@fensystems.co.uk> <20190719172907.788170-2-mbrown@fensystems.co.uk> <20190723103436.GF11541@bivouac.eciton.net> From: Michael Brown Message-ID: <19dd2d90-18b0-7ff2-cc21-a3d33a32d5e1@fensystems.co.uk> Date: Tue, 23 Jul 2019 12:00:04 +0100 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.8.0 MIME-Version: 1.0 In-Reply-To: <20190723103436.GF11541@bivouac.eciton.net> Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-GB Content-Transfer-Encoding: 7bit On 23/07/2019 11:34, Leif Lindholm wrote: > On Fri, Jul 19, 2019 at 06:29:07PM +0100, Michael Brown wrote: >> Older device trees tend to use the alias "ethernet". Newer device >> trees tend to use "ethernet0" since older versions of U-Boot would >> skip aliases that do not include an index number. See, for example, >> Linux kernel commit 10b6c0c ("ARM: dts: bcm2835: add index to the >> ethernet alias") and U-Boot commit f8e57c6 ("fdt_support: Fixup >> 'ethernet' aliases not ending in digits"). >> >> Accept either "ethernet" or "ethernet0" as being the alias to the node >> for which the MAC address should be updated. > > Why is this patch useful? > I understand the problem, but we include the .dtb from our own > edk2-non-osi tree. And it seems that device tree already provides an > alias to support both: > https://github.com/tianocore/edk2-non-osi/blob/master/Platform/RaspberryPi/RPi3/DeviceTree/bcm2710-rpi-3-b.dts > > Would it be more useful to warn loudly if "ethernet" isn't present? > (And just *replace* ethernet[0] with ethernet below.) The device tree can also be provided by the VC4 firmware, e.g. with a config.txt containing e.g. device_tree_address=0x10000 device_tree_end=0x20000 device_tree=bcm2710-rpi-3-b-plus.dtb It is also possible to omit the explicit "device_tree=..." and have just device_tree_address=0x10000 device_tree_end=0x20000 This configuration will cause the VC4 firmware to automatically load the appropriate device tree file. I have tested that this allows me to boot from a 3B or 3B+ using a single SD card image that contains all of the .dtb files from https://github.com/raspberrypi/firmware/tree/master/boot These "official" raspberrypi-firmware .dtb files are built from the stock Linux kernel .dts files, and contain only "ethernet0". To address your specific questions: > Why is this patch useful? It allows us to boot using the "official" .dtb files without any unexpected behaviour. This in turn allows us to boot with the VC4 firmware automatically selecting the appropriate .dtb file for the hardware. This is extremely useful when trying to provide a single SD card image that will work on a wide range of hardware. > Would it be more useful to warn loudly if "ethernet" isn't present? > (And just *replace* ethernet[0] with ethernet below.) This would also solve the problem. I would find it mildly surprising behaviour if I were debugging around it, since the "official" .dtb files have moved in the other direction (going from "ethernet" to "ethernet0"), but I would not expect it to break anything. Another option would be to have UpdateMacAddress() ensure that both "ethernet" and "ethernet0" aliases exist. I am happy to work up a new version of the patch: just let me know which option you prefer. Many thanks, Michael