Re: Kernel ELF entry point

[Masao Uebayashi <uebayasi%gmail.com@localhost>]

On Mon, Nov 10, 2014 at 9:53 PM, Nick Hudson <nick.hudson%gmx.co.uk@localhost> wrote:
> On 11/10/14 07:10, Masao Uebayashi wrote:
>>
>> Unfortunately NetBSD kernel build infrastructure can't differentiate
>> kernel configuration from how it's loaded (load address, entry
>> address).  Consequently you have to use workarounds, not a solution.
>
> Defining ENTRYPOINT seems perfectly reasonable to me.

"-e start" is unconditionally overriding ENTRY() in ldscript,
so he needed to override "-e start" by defining ENTRYPOINT in mk.integrator.
Is this a intended design?

>> I'm wondering that kernel should be always provided as a relocatable,
>> so that it can be linked against different load addresses.  I've
>> received only negative reactions, though.
>
> You have had mostly questions about the value of using a relocatable stage
> to solve
> the MI linker script as far as I can tell.
>
> Are there any others? For example, how about generating a linker script from
> MI and MD parts
> at for use at final link time?

Not sure if it's possible, but should be better than "don't touch my
linker script for your reasons".