TODO

LM32 port TODO:

- add some libkern features
- add crtstuff in libcsu
- verify frame.h
- make _spl() functions less nazi (don't mask all IRQs when it's not necessary)
- make sure it's fine to hardwire CLKF_USERMOD to 0 or 1 (this and the following one seem to be used for statistics purposes only)
- make sure it's fine to hardwire CLKF_INTR to 0 or 1
- Do I need to implement cpu_did_resched() ?
- implement copyin/copyout/copyinstr
  - amd64: sys/arch/amd64/amd64/copy.S:ENTRY(copyout)
  - arm: sys/arch/arm/arm/bcopyinout_xscale.S:ENTRY(copyout)
  - hppa: sys/arch/hppa/hppa/copy.S: * copyin/copyout, fuword/suword, etc.
  - i386: sys/arch/i386/i386/copy.S:ENTRY(copyout)
  - ia64: sys/arch/ia64/ia64/support.S:ENTRY(copyoutstr, 4)
  - m68k: sys/arch/m68k/060sp/dist/os.s:_copyout: | sys/arch/m68k/m68k/copy.s:ENTRY(copyout)
  - ppc: sys/arch/powerpc/booke/copyout.c
  - sh3: sys/arch/sh3/sh3/locore_subr.S:ENTRY(copyout)
  - sparc: sys/arch/sparc/sparc/locore.s:ENTRY(copyout)
  - sparc64: sys/arch/sparc64/sparc64/copy.S:ENTRY(copyout)
  - usermod: sys/arch/usermode/usermode/copy.c
  - vax: sys/arch/vax/boot/boot/if_le.c: copyout(void *f, int dest, int len)

- Decide what to do with SH_HAS_VIRTUAL_ALIAS stuff
  - It seems that as long as total_cache_size <= page size (with total_cache_size = line size * number of sets * cache associativity)
    there is no cache aliasing issue