Deprecation/removal of nios2 target support

[Arnd Bergmann]

On Mon, Mar 24, 2025, at 10:26, Simon Schuster wrote:
> On Thu, 18 Apr 2024, Arnd Bergmann wrote:
>>
>> We should probably remove nios2 from the kernel in the near future,
>> but even if we decide not to, I think deprecating it from gcc is the
>> right idea: If there are a few remaining users that still plan
>> to update their kernels, gcc-14 will still be able to build new
>> kernels for several years.
>
> Sorry to dig this up after nearly one year, but this thread has only now
> come to my attention. I'm a colleague of Andreas Oetken at Siemens
> Energy (unfortunately the original mail got lost somewhere between our
> carveout from Siemens and daily business).
>
> First of: We continue to actively use the linux kernel (and the
> associated gcc/binutils-based toolchain) on the NiosII platform, perform
> tests regularly on hardware and inspect any failures we encounter.
> Furthermore, we are also in the (so far internal) process of upstreaming
> two stability fixes related to NiosII memory-management facilities in
> the kernel. As soon as they have passed their required degree of testing
> on our side, we plan to send them to the LKML mailing list.

Generally speaking, it probably won't hurt to just post what you
have after you have identified a problem and fix, so it can be
discussed while you are still testing.

> We are aware that the architecture was removed from binutils, GDB, gcc
> and QEMU as developing new features for a deprecated architecture makes
> no sense. However, with the relatively recent deprecation (announced by
> Intel on 2023-06-09), we, and probably others, still see high value in
> continued support by runtime components (especially in the linux kernel
> and glibc) for the foreseeable future to support existing deployments;
> at least on the level of minimal lifesupport. We also would be glad to
> continue providing periodic testing support to this end for the NiosII
> architecture on hardware (along with bugfixing as required).

I think that is fine: while reducing the number of architectures
in the kernel is something that will help reduce the amount of work
for cross-architecture patches in the long run, nios2 is particularly
small, and keeping it around for another few years is not going to
make a huge difference. There are a few other architectures that
are in a similar position, with only a single person or team actually
interested in keeping it around after the product is no longer
actively marketed for Linux based products (arc, csky, hexagon,
microblaze, nios2, openrisc, sparc32 and xtensa). Pretty much all
of these have been replaced by 64-bit risc-v cores in products,

There are a few things you can do to help keep nios2 in the kernel
for a while longer, as I'm sure we will keep getting requests to
remove support otherwise:

- add yourself to the MAINTAINERS file (R: or M:) in order to
  get Cc's on cross-architecture patches and test them before
  the make it into the kernel.

- let us know about your product support life, i.e. how much
  longer you will need to upgrade to newer LTS kernel releases

- add a dts file for each of your products to arch/nios2/boot/dts/

       Arnd