patches/kernel/0017-x86-mm-Document-how-CR4.PCIDE-restore-works.patch

   1 From 20e07f035810f1b2bb3d816e49f48f6b6a37bf64 Mon Sep 17 00:00:00 2001
   2 From: Andy Lutomirski <luto@kernel.org>
   3 Date: Wed, 6 Sep 2017 19:54:54 -0700
   4 Subject: [PATCH 017/233] x86/mm: Document how CR4.PCIDE restore works
   5 MIME-Version: 1.0
   6 Content-Type: text/plain; charset=UTF-8
   7 Content-Transfer-Encoding: 8bit
   8
   9 CVE-2017-5754
  10
  11 While debugging a problem, I thought that using
  12 cr4_set_bits_and_update_boot() to restore CR4.PCIDE would be
  13 helpful.  It turns out to be counterproductive.
  14
  15 Add a comment documenting how this works.
  16
  17 Signed-off-by: Andy Lutomirski <luto@kernel.org>
  18 Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
  19 (cherry picked from commit 1c9fe4409ce3e9c78b1ed96ee8ed699d4f03bf33)
  20 Signed-off-by: Andy Whitcroft <apw@canonical.com>
  21 Signed-off-by: Kleber Sacilotto de Souza <kleber.souza@canonical.com>
  22 (cherry picked from commit 0d69e4c4a2db42a9bac6609a3df15bd91163f8b9)
  23 Signed-off-by: Fabian Grünbichler <f.gruenbichler@proxmox.com>
  24 ---
  25  arch/x86/kernel/cpu/common.c | 13 +++++++++++++
  26  1 file changed, 13 insertions(+)
  27
  28 diff --git a/arch/x86/kernel/cpu/common.c b/arch/x86/kernel/cpu/common.c
  29 index b95cd94ca97b..0b80ed14ff52 100644
  30 --- a/arch/x86/kernel/cpu/common.c
  31 +++ b/arch/x86/kernel/cpu/common.c
  32 @@ -333,6 +333,19 @@ static void setup_pcid(struct cpuinfo_x86 *c)
  33  {
  34         if (cpu_has(c, X86_FEATURE_PCID)) {
  35                 if (cpu_has(c, X86_FEATURE_PGE)) {
  36 +                       /*
  37 +                        * We'd like to use cr4_set_bits_and_update_boot(),
  38 +                        * but we can't.  CR4.PCIDE is special and can only
  39 +                        * be set in long mode, and the early CPU init code
  40 +                        * doesn't know this and would try to restore CR4.PCIDE
  41 +                        * prior to entering long mode.
  42 +                        *
  43 +                        * Instead, we rely on the fact that hotplug, resume,
  44 +                        * etc all fully restore CR4 before they write anything
  45 +                        * that could have nonzero PCID bits to CR3.  CR4.PCIDE
  46 +                        * has no effect on the page tables themselves, so we
  47 +                        * don't need it to be restored early.
  48 +                        */
  49                         cr4_set_bits(X86_CR4_PCIDE);
  50                 } else {
  51                         /*
  52 --
  53 2.14.2
  54