[5/7] KVM: arm/arm64: Elide both CMOs and TBLIs on freeing the whole Stage-2
diff mbox series

Message ID 20191213182503.14460-6-maz@kernel.org
State New
Headers show
Series
  • KVM: arm/arm64: Help VMs dying quicker
Related show

Commit Message

Marc Zyngier Dec. 13, 2019, 6:25 p.m. UTC
When freeing the whole of a VM's Stage-2 page tables, there is
little point in doing cache maintenance on each and every page
(the guest won't be running anymore, let alone having its MMU
off). As for TLBs, there is no point in invalidating individual
pages, as we can replace the whole thing with a VMALL operation,
which invalidates all the TLBs for this VM in one go.

Signed-off-by: Marc Zyngier <maz@kernel.org>
---
 virt/kvm/arm/mmu.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

Patch
diff mbox series

diff --git a/virt/kvm/arm/mmu.c b/virt/kvm/arm/mmu.c
index 4399866842dc..d7c710491d26 100644
--- a/virt/kvm/arm/mmu.c
+++ b/virt/kvm/arm/mmu.c
@@ -1016,10 +1016,13 @@  void kvm_free_stage2_pgd(struct kvm *kvm)
 
 	spin_lock(&kvm->mmu_lock);
 	if (kvm->arch.pgd) {
-		unmap_stage2_range(kvm, 0, kvm_phys_size(kvm), 0);
+		unmap_stage2_range(kvm, 0, kvm_phys_size(kvm),
+				   KVM_UNMAP_ELIDE_CMO | KVM_UNMAP_ELIDE_TBLI);
 		pgd = READ_ONCE(kvm->arch.pgd);
 		kvm->arch.pgd = NULL;
 		kvm->arch.pgd_phys = 0;
+
+		kvm_flush_remote_tlbs(kvm);
 	}
 	spin_unlock(&kvm->mmu_lock);