---
- branch: netbsd-9
  date: Fri Jan  1 12:38:50 UTC 2021
  files:
  - new: 1.5.4.2
    old: 1.5.4.1
    path: src/sys/arch/aarch64/aarch64/cpufunc.c
    pathrev: src/sys/arch/aarch64/aarch64/cpufunc.c@1.5.4.2
    type: modified
  - new: 1.41.2.7
    old: 1.41.2.6
    path: src/sys/arch/aarch64/aarch64/pmap.c
    pathrev: src/sys/arch/aarch64/aarch64/pmap.c@1.41.2.7
    type: modified
  id: 20210101T123850Z.64de01c21cf40a063bcdd706d55d7e58e73db0ab
  log: "Pull up following revision(s) (requested by rin in ticket #1170):\n\n\tsys/arch/aarch64/aarch64/cpufunc.c:
    revision 1.22 (patch)\n\tsys/arch/aarch64/aarch64/cpufunc.c: revision 1.23 (patch)\n\tsys/arch/aarch64/aarch64/pmap.c:
    revision 1.81\n\nSet uvmexp.ncolors appropriately, which is required for some
    CPU\nmodels with VIPT icache.\n\nOtherwise, alias in virtual address results in
    inconsistent results,\nat least for applications that rewrite text of other process,
    e.g.,\nGDB for arm32.\n\nAlso, this hopefully fixes other unexpected failures
    due to alias.\nConfirmed that there's no observable regression in performance;\ndifference
    in ``time make -j8'' for GENERIC64 kernel on BCM2837\nwith and without setting
    uvmexp.ncolors is within 0.1%.\n\nThanks to ryo@ for discussion.\n\nFix uvmexp.ncolors
    for some big.LITTLE configuration; it is uncertain\nwhich CPU is used as primary,
    and as a result, secondary CPUs can\nrequire larger number of colors.\n\nIn order
    to solve this problem, update uvmexp.ncolors via\nuvm_page_recolor(9) when secondary
    CPUs are attached, as done for\nother ports like x86.\n\nPointed out by jmcneill@,
    and discussed on port-arm@:\nhttp://mail-index.netbsd.org/port-arm/2020/07/03/msg006837.html\nTested
    and OK'd by ryo@.\n\nFix previous; add missing <uvm/uvm.h> include.\n"
  module: src
  subject: 'CVS commit: [netbsd-9] src/sys/arch/aarch64/aarch64'
  unixtime: '1609504730'
  user: martin