[fedora-arm] ARM and shipping of various binary firmware / boot bits

[Gordan Bobic]

Tom Callaway wrote:
> On 03/08/2012 10:04 AM, Peter Robinson wrote:
>> On Thu, Mar 8, 2012 at 3:03 PM, Tom Callaway <tcallawa at redhat.com> wrote:
>>> On 03/08/2012 09:52 AM, Peter Robinson wrote:
>>>> On Thu, Mar 8, 2012 at 2:42 PM, Tom Callaway <tcallawa at redhat.com> wrote:
>>>>> On 03/07/2012 07:14 PM, Peter Robinson wrote:
>>>>>> Hey spot,
>>>>>>
>>>>>> On our weekly call today we discussed the always fun bit of binary
>>>>>> blobs. ARM has the usual wireless and associated blobs most of which i
>>>>>> think are already upstream (and already in Fedora).
>>>>>>
>>>>>> The bits that came up is uboot, MLO (X-Loader) [1]  and what ever some
>>>>>> of the other devices use such as the Raspberry Pi. In the first
>>>>>> example the source code is available but forked from upstream, in the
>>>>>> later it's a binary blob not that dissimilar presumably to a wifi
>>>>>> firmware. For the binary blobs is the process the same as per wifi or
>>>>>> any other binary? What about the MLO/uboot, is it enough to package
>>>>>> the binaries and include details in COPYING/spec where the source code
>>>>>> is?
>>>>>>
>>>>>> I'm sure there's some other cases I've not thought of that you might
>>>>>> be aware of too. Can you advise of the best and easiest way for us to
>>>>>> deal with these?
>>>>> We need to review each of the binary firmware items individually. Just
>>>>> open review request tickets and block FE-Legal immediately.
>>>>>
>>>>> As for uboot, is there any good reason not to build from the available
>>>>> source code? And MLO? I'm not sure we can consider a bootloader to be
>>>>> firmware. That one might not be able to go into Fedora.
>>>> Well we probably might well be able to but there's dozens of branches
>>>> and forks etc for initiated every different SOC in their millions of
>>>> different configurations, it's closer to a BIOS than a bootloader I
>>>> believe, Linaro is in the process of adding grub2 support for ARM so
>>>> grub will eventually run as a bootloader just like on x86.
>>> In that situation, will we still need the uboot/MLO stuff?
>> Yes, I believe so.
> 
> Okay, so I'm admittedly a bit lost here, because we don't normally ship
> BIOS code for any other platform (apart from qemu). Why is ARM different
> in that regard? Why don't the vendors deliver BIOS (or equivalent)?

ARM SoCs come from a completely different, unstandardised background, 
unlike x86. This is the same reason why you cannot have a kernel for one 
SoC booting on another (at least not at the moment).

Linaro are trying to converge the basic features toward a standard that 
will allow a common kernel to boot on multiple SoCs, but this is not 
going to happen overnight.

Gordan