[PATCH] x86/kpti: WARN if kernel memory mapped to userspace is non-global

From: Dave Hansen
Date: Fri Dec 15 2017 - 19:02:16 EST



From: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>

All memory being mapped out to userspace is shared. That means that
it is both safe and _expected_ to have the page table Global bit set.

If a PMD is found without Global set, it either a place that could be
performing better, or something unexpected is being mapped out to
userspace. Both of those are things for which a warning is good.

Signed-off-by: Dave Hansen <dave.hansen@xxxxxxxxxxxxxxx>
Cc: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Cc: Andy Lutomirski <luto@xxxxxxxxxx>
Cc: Borislav Petkov <bp@xxxxxxx>
Cc: Brian Gerst <brgerst@xxxxxxxxx>
Cc: Denys Vlasenko <dvlasenk@xxxxxxxxxx>
Cc: H. Peter Anvin <hpa@xxxxxxxxx>
Cc: Josh Poimboeuf <jpoimboe@xxxxxxxxxx>
Cc: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>
Cc: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
Cc: Rik van Riel <riel@xxxxxxxxxx>
---

b/arch/x86/mm/kpti.c | 3 +++
1 file changed, 3 insertions(+)

diff -puN arch/x86/mm/kpti.c~kaiser-set-global-in-kernel-for-shared arch/x86/mm/kpti.c
--- a/arch/x86/mm/kpti.c~kaiser-set-global-in-kernel-for-shared 2017-12-15 09:47:52.884717268 -0800
+++ b/arch/x86/mm/kpti.c 2017-12-15 10:00:39.134715357 -0800
@@ -157,6 +157,9 @@ kpti_clone_pmds(unsigned long start, uns
if (WARN_ON(!target_pmd))
return;

+ /* Only clone PMDs which we *intend* to share: */
+ WARN_ON_ONCE(!(pmd_flags(*target_pmd) & _PAGE_GLOBAL));
+
/*
* Copy the PMD. That is, the kernelmode and usermode
* tables will share the last-level page tables of this
_