From 66d73e12f2786497424cb611b359e3580978cad7 Mon Sep 17 00:00:00 2001
From: Peter Feiner <pfeiner@google.com>
Date: Mon, 26 Sep 2016 10:45:34 -0700
Subject: [PATCH] KVM: X86: MMU: no mmu_notifier_seq++ in kvm_age_hva
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
The MMU notifier sequence number keeps GPA->HPA mappings in sync when
GPA->HPA lookups are done outside of the MMU lock (e.g., in
tdp_page_fault). Since kvm_age_hva doesn't change GPA->HPA, it's
unnecessary to increment the sequence number.
Signed-off-by: Peter Feiner <pfeiner@google.com>
Reviewed-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
---
arch/x86/kvm/mmu.c | 10 +---------
1 file changed, 1 insertion(+), 9 deletions(-)
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index d9c7e986b4e4e..d3a94ea9f6acd 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -1660,17 +1660,9 @@ int kvm_age_hva(struct kvm *kvm, unsigned long start, unsigned long end)
* This has some overhead, but not as much as the cost of swapping
* out actively used pages or breaking up actively used hugepages.
*/
- if (!shadow_accessed_mask) {
- /*
- * We are holding the kvm->mmu_lock, and we are blowing up
- * shadow PTEs. MMU notifier consumers need to be kept at bay.
- * This is correct as long as we don't decouple the mmu_lock
- * protected regions (like invalidate_range_start|end does).
- */
- kvm->mmu_notifier_seq++;
+ if (!shadow_accessed_mask)
return kvm_handle_hva_range(kvm, start, end, 0,
kvm_unmap_rmapp);
- }
return kvm_handle_hva_range(kvm, start, end, 0, kvm_age_rmapp);
}