Re: Questions regarding ld

[Christophe Lyon <christophe.lyon@linaro.org>]

On Wed, 18 Mar 2020 at 19:45, Dan Ciliske <dciliske@netburner.com> wrote:
>
> We are at this moment focused on ARM only. That said, I think this is
> the first time in history I've gone looking for a piece of functionality
> and it was _just_ added.
>

"_just_" as in "I committed it last Friday" :-) (with minor changes today)

There were several threads about this, a good starting point for some
context is probably:
https://sourceware.org/legacy-ml/binutils/2019-11/msg00402.html
(it has a pointer to earlier discussions)

Thanks

Christophe


> I guess it's time to work with the build team to look into being on the
> bleeding edge...
>
> Thanks,
> - Dan
>
> On 3/18/2020 11:19 AM, Jozef Lawrynowicz wrote:
> > Hi,
> >
> > On Wed, 18 Mar 2020 10:55:38 -0700
> > Dan Ciliske <dciliske@netburner.com> wrote:
> >
> >> I do a lot of work at the bottom of embedded systems and I keep running
> >> into the same roadblocks that I'm finally asking about. Since space is
> >> often tight in these systems, I was wondering if there are any known
> >> approaches that I'm missing to solving some common problems.
> >>
> >> My questions are:
> >>
> >> 1. Is it possible to automatically spill across sections if the
> >> underlying region fills? If not, is there some way after the fact move
> >> part of a Section to another Section that is located in a different Region?
> > Assuming that you are talking about moving input sections between output
> > sections (and not splitting up input sections into parts), then the recently
> > added "non-contiguous memory region" functionality sounds like it handles this
> > problem for you. See the documentation on the "--enable-non-contiguous-regions"
> > configure option in ld/ld.texi.
> >
> > If you are not only concerned with ARM, the MSP430 linker has had functionality
> > to shuffle input sections named ".either.*" between upper and lower memory
> > regions for a while. The --{code,data}-region=either LD option also transforms
> > section names which don't have the ".either" prefix (e.g. from libraries) to try
> > to get the program to fit in the available space.
> >
> > Jozef
> >
> >> 2. With ARM Cortex-M, is there a way to split the ex_tab and exidx
> >> tables into two parts such that they are accessible within the 31 bit
> >> limit imposed by ARM_PREL31? I am well aware of the complex implications
> >> of this question, and realize that this is also partly a question for
> >> the GCC lists.
> >>
> --
> Dan Ciliske
> Principal Architect
> NetBurner, Inc.
> 858-558-0293 x216
>