On Mon, Jul 05, 2021 at 01:33:39PM +0200, Geert Uytterhoeven wrote:
No, at least not yet. Coldfire and co still rely on set_fs for various
things in the m68k arch code, and also don't provide
__{get,put}_kernel_nofault yet. So wile m68k with CPU_HAS_ADDRESS_SPACES
will probably one of the hardest ports to get rid of set_fs() for,
m68k/mmu without CPU_HAS_ADDRESS_SPACES would be a logical next step.
I could try to whiteboard code some of it, but I'd need help from dedicated
M68k/mmu with CPU_HAS_ADDRESS_SPACES needs to use the address spaces
to access user space from kernel space using the "moves" instruction.
Coldfire doesn't.
See arch/m68k/include/asm/uaccess.h.
I know. That doesn't change that:
a) even getting rid for MMMU && !CPU_HAS_ADDRESS_SPACES requires further
work
b) getting rid of set_fs is possible even for CPU_HAS_ADDRESS_SPACES,
although it requires a lot of work.
