Re: [PATCH 0/3 - V2] Introducing Device Tree Overlays

From: Pantelis Antoniou
Date: Thu Nov 07 2013 - 02:24:06 EST

Hi Sebestian,

On Nov 6, 2013, at 10:31 PM, Sebastian Andrzej Siewior wrote:

> On 06.11.13, Pantelis Antoniou wrote:
>> Hi Sebastian,
> Hi Pantelis,
>> It has been discussed.
>> We are doing it because
>> a) We tried to do it in u-boot and it has been a complete disaster.
>> Regular users just can't handle bootloader updates.
> How so? The "additional" dtb piece was deleted by accident as part of
> the u-boot upgrade? Do you maybe a link which describes the disaster?

You're assuming that bootloaders are anything like u-boot or barebox.
In the field, especially on consumer products, bootloaders are custom one-off
jobs that don't do much besides handing control to the kernel as soon as possible.

Even when using u-boot, end users botch updates related to bootloader in
such a way that systems end up RMAed.

Ask Koen Kooi in the CClist about the messy details.

>> b) It is similar to that. It was originally created for the beaglebone,
>> which has a concept of capes (similar to Arduino shields).
>> Turns out it's really useful to anyone doing reconfigurable hardware too,
>> so that's why FPGA people are thinking of using it.
> I am aware of this. My understanding is that those capes have hardware
> information encoded in an eeprom behind i2c _and_ you can't or should
> not replace capes at runtime.
> Naive as I am I *assume* it should be easy to read this eeprom in u-boot
> as part of the boot setup and extend the dtb before passing it to the
> kernel. In case this works well, the problem here is a) ?

It is just better system design to have it all done in the kernel.
Other people in the list can chime in, but it's hard to get a feel of the
problem if you haven't dealt with it before.

>> c) There are people that want to tinker with Linux based hardware boards
>> but are not kernel developers. This gives them a way to do so without
>> having to recompile the kernel and/or reboot while tinkering.
> I understand that they want to avoid reboot. But they still need to
> recompile the device tree, don't they? Or does this allow to change the
> HW description without knowing the syntax of .dts?

They understand the syntax of the DTS (barely).
They can't hack compiling the kernel and updating it, not by a long shot.
Not everyone is a kernel hacker (neither it needs be).

Imagine people coming over from Arduino trying to hack a 4K board file to
add support for the thing they're working on.

>> Regards
>> -- Pantelis
> Sebastian


-- Pantelis

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at