1Hollis Blanchard <>
   25 Jun 2002
   4This document describes the system (including self-modifying code) used in the
   5PPC Linux kernel to support a variety of PowerPC CPUs without requiring
   6compile-time selection.
   8Early in the boot process the ppc32 kernel detects the current CPU type and
   9chooses a set of features accordingly. Some examples include Altivec support,
  10split instruction and data caches, and if the CPU supports the DOZE and NAP
  11sleep modes.
  13Detection of the feature set is simple. A list of processors can be found in
  14arch/ppc/kernel/cputable.c. The PVR register is masked and compared with each
  15value in the list. If a match is found, the cpu_features of cur_cpu_spec is
  16assigned to the feature bitmask for this processor and a __setup_cpu function
  17is called.
  19C code may test 'cur_cpu_spec[smp_processor_id()]->cpu_features' for a
  20particular feature bit. This is done in quite a few places, for example
  21in ppc_setup_l2cr().
  23Implementing cpufeatures in assembly is a little more involved. There are
  24several paths that are performance-critical and would suffer if an array
  25index, structure dereference, and conditional branch were added. To avoid the
  26performance penalty but still allow for runtime (rather than compile-time) CPU
  27selection, unused code is replaced by 'nop' instructions. This nop'ing is
  28based on CPU 0's capabilities, so a multi-processor system with non-identical
  29processors will not work (but such a system would likely have other problems
  32After detecting the processor type, the kernel patches out sections of code
  33that shouldn't be used by writing nop's over it. Using cpufeatures requires
  34just 2 macros (found in arch/powerpc/include/asm/cputable.h), as seen in head.S
  37        #ifdef CONFIG_ALTIVEC
  39                mfspr   r22,SPRN_VRSAVE         /* if G4, save vrsave register value */
  40                stw     r22,THREAD_VRSAVE(r23)
  42        #endif /* CONFIG_ALTIVEC */
  44If CPU 0 supports Altivec, the code is left untouched. If it doesn't, both
  45instructions are replaced with nop's.
  47The END_FTR_SECTION macro has two simpler variations: END_FTR_SECTION_IFSET
  48and END_FTR_SECTION_IFCLR. These simply test if a flag is set (in
  49cur_cpu_spec[0]->cpu_features) or is cleared, respectively. These two macros
  50should be used in the majority of cases.
  52The END_FTR_SECTION macros are implemented by storing information about this
  53code in the '__ftr_fixup' ELF section. When do_cpu_ftr_fixups
  54(arch/ppc/kernel/misc.S) is invoked, it will iterate over the records in
  55__ftr_fixup, and if the required feature is not present it will loop writing
  56nop's from each BEGIN_FTR_SECTION to END_FTR_SECTION.
  57 kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.