[GIT PULL] nommu fixes for 3.17-rc1

[Arnd Bergmann]

On Monday 30 June 2014 16:43:55 Uwe Kleine-König wrote:
> Hello Russell,
> 
> the following changes since commit 4c834452aad01531db949414f94f817a86348d59:
> 
>   Linux 3.16-rc3 (2014-06-29 14:11:36 -0700)
> 
> are available in the git repository at:
> 
>   git://git.pengutronix.de/git/ukl/linux.git tags/nommu-for-rmk
> 
> for you to fetch changes up to 7343ab2ad6772f29d6bbb286a7aa0265a7aa5d3b:
> 
>   ARM: make user_addr_max more robust (2014-06-30 15:48:30 +0200)
> 
> ----------------------------------------------------------------
> A small collection of ARM-nommu related fixes and cleanups.
> 
> Regarding the small user base ("Hello, anyone out there apart from me?") and
> the problem being fixed exists since 3.6-rc1 I don't consider the patches
> serious enough to justify them going in before the next merge window.
> 
> This series makes efm32 boot again.
> ----------------------------------------------------------------
> 
> I have chosen -rc3 as base because -rc1 and -rc2 don't work for me as
> they don't have 6980c3e2514e (ARM: 8086/1: Set memblock limit for
> nommu).
> 
> Uwe Kleine-König (5):
>       ARM: nommu: drop support for ARM740T, ARM940T and ARM946E-S processors
>       ARM: nommu: don't limit TASK_SIZE
>       ARM: nommu: simplify definition of MODULES_END
>       ARM: nommu: drop unused Kconfig symbol DRAM_SIZE
>       ARM: make user_addr_max more robust
> 

I seem to have missed the patches when you posted them a couple of weeks
ago. I do have one last-minute comment about dropping ARM740T/940T/946T:

Your reasoning is that mach-integrator now doesn't work with !MMU any more
since it selects ARM_PATCH_PHYS_VIRT, and that no other platform uses
these CPUs. The change to integrator was done as a preparation for
multiplatform support and probably did not mean to remove support for !MMU
integrator. We do plan to remove support for at91x40 in the future, which
is the last ARM7TDMI user, and there is no upstream ARM9TDMI user nor has
there been at least for a couple of years.

I don't think it makes any sense to remove ARM740T and ARM94x but to leave
the option to build other ARMv4/v5 cores with MMU disabled. It would
be more consistent to either fix the build for these (ondoing the
ARM_PATCH_PHYS_VIRT change) or to remove all support for ARMv4/v5 !MMU,
possible also all !MMU support other than ARMv7-M.

Either way, I think it should be a conscious decision, not just a matter
of "this broke a while ago, let's assume nobody cares about it" as we
tend to do it for odd device drivers. There are still people using
arm7tdmi (in particular Russell still maintains a port for oki/lapis
ml67xxx), but at some point the last production deployment will have
upgraded to a new kernel or moved on to a cortex-m based chip.

	Arnd