releases/3.10.88/arm64-mm-remove-hack-in-mmap-randomize-layout.patch

   1 From d6c763afab142a85e4770b4bc2a5f40f256d5c5d Mon Sep 17 00:00:00 2001
   2 From: Yann Droneaud <ydroneaud@opteya.com>
   3 Date: Mon, 17 Nov 2014 23:02:19 +0000
   4 Subject: arm64/mm: Remove hack in mmap randomize layout
   5
   6 From: Yann Droneaud <ydroneaud@opteya.com>
   7
   8 commit d6c763afab142a85e4770b4bc2a5f40f256d5c5d upstream.
   9
  10 Since commit 8a0a9bd4db63 ('random: make get_random_int() more
  11 random'), get_random_int() returns a random value for each call,
  12 so comment and hack introduced in mmap_rnd() as part of commit
  13 1d18c47c735e ('arm64: MMU fault handling and page table management')
  14 are incorrects.
  15
  16 Commit 1d18c47c735e seems to use the same hack introduced by
  17 commit a5adc91a4b44 ('powerpc: Ensure random space between stack
  18 and mmaps'), latter copied in commit 5a0efea09f42 ('sparc64: Sharpen
  19 address space randomization calculations.').
  20
  21 But both architectures were cleaned up as part of commit
  22 fa8cbaaf5a68 ('powerpc+sparc64/mm: Remove hack in mmap randomize
  23 layout') as hack is no more needed since commit 8a0a9bd4db63.
  24
  25 So the present patch removes the comment and the hack around
  26 get_random_int() on AArch64's mmap_rnd().
  27
  28 Cc: David S. Miller <davem@davemloft.net>
  29 Cc: Anton Blanchard <anton@samba.org>
  30 Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
  31 Acked-by: Will Deacon <will.deacon@arm.com>
  32 Acked-by: Dan McGee <dpmcgee@gmail.com>
  33 Signed-off-by: Yann Droneaud <ydroneaud@opteya.com>
  34 Signed-off-by: Will Deacon <will.deacon@arm.com>
  35 Cc: Matthias Brugger <mbrugger@suse.com>
  36 Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  37
  38 ---
  39  arch/arm64/mm/mmap.c |   12 ++----------
  40  1 file changed, 2 insertions(+), 10 deletions(-)
  41
  42 --- a/arch/arm64/mm/mmap.c
  43 +++ b/arch/arm64/mm/mmap.c
  44 @@ -47,22 +47,14 @@ static int mmap_is_legacy(void)
  45         return sysctl_legacy_va_layout;
  46  }
  47
  48 -/*
  49 - * Since get_random_int() returns the same value within a 1 jiffy window, we
  50 - * will almost always get the same randomisation for the stack and mmap
  51 - * region. This will mean the relative distance between stack and mmap will be
  52 - * the same.
  53 - *
  54 - * To avoid this we can shift the randomness by 1 bit.
  55 - */
  56  static unsigned long mmap_rnd(void)
  57  {
  58         unsigned long rnd = 0;
  59
  60         if (current->flags & PF_RANDOMIZE)
  61 -               rnd = (long)get_random_int() & (STACK_RND_MASK >> 1);
  62 +               rnd = (long)get_random_int() & STACK_RND_MASK;
  63
  64 -       return rnd << (PAGE_SHIFT + 1);
  65 +       return rnd << PAGE_SHIFT;
  66  }
  67
  68  static unsigned long mmap_base(void)