aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAndy Lutomirski <[email protected]>2016-01-29 11:42:59 -0800
committerIngo Molnar <[email protected]>2016-02-09 13:36:11 +0100
commitd8bced79af1db6734f66b42064cc773cada2ce99 (patch)
tree165e9ab752ebf5f44b4d648d5de8b05c6f0e4f68
parentd12a72b844a49d4162f24cefdab30bed3f86730e (diff)
x86/mm: If INVPCID is available, use it to flush global mappings
On my Skylake laptop, INVPCID function 2 (flush absolutely everything) takes about 376ns, whereas saving flags, twiddling CR4.PGE to flush global mappings, and restoring flags takes about 539ns. Signed-off-by: Andy Lutomirski <[email protected]> Reviewed-by: Borislav Petkov <[email protected]> Cc: Andrew Morton <[email protected]> Cc: Andrey Ryabinin <[email protected]> Cc: Andy Lutomirski <[email protected]> Cc: Borislav Petkov <[email protected]> Cc: Brian Gerst <[email protected]> Cc: Dave Hansen <[email protected]> Cc: Denys Vlasenko <[email protected]> Cc: H. Peter Anvin <[email protected]> Cc: Linus Torvalds <[email protected]> Cc: Luis R. Rodriguez <[email protected]> Cc: Oleg Nesterov <[email protected]> Cc: Peter Zijlstra <[email protected]> Cc: Thomas Gleixner <[email protected]> Cc: Toshi Kani <[email protected]> Cc: [email protected] Link: http://lkml.kernel.org/r/ed0ef62581c0ea9c99b9bf6df726015e96d44743.1454096309.git.luto@kernel.org Signed-off-by: Ingo Molnar <[email protected]>
-rw-r--r--arch/x86/include/asm/tlbflush.h9
1 files changed, 9 insertions, 0 deletions
diff --git a/arch/x86/include/asm/tlbflush.h b/arch/x86/include/asm/tlbflush.h
index 8b576832777e..fc9a2fda1404 100644
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -152,6 +152,15 @@ static inline void __native_flush_tlb_global(void)
{
unsigned long flags;
+ if (static_cpu_has(X86_FEATURE_INVPCID)) {
+ /*
+ * Using INVPCID is considerably faster than a pair of writes
+ * to CR4 sandwiched inside an IRQ flag save/restore.
+ */
+ invpcid_flush_all();
+ return;
+ }
+
/*
* Read-modify-write to CR4 - protect it from preemption and
* from interrupts. (Use the raw variant because this code can