[PATCH 4/4] KVM: arm64: vgic-v3: Pause the source vCPU when processing MOVALL cmd

[Oliver Upton <oupton@kernel.org>]

We currently pause the VM to process a MOVALL command to avoid
undesirable contention on the AP list with guest entry/exit. This is no
longer quite as big of an issue, as the migrations of active/pending
IRQs no longer happens in vgic_prune_ap_list().

To cope with this, vgic_update_lpi_affinity() may pause the source vCPU
internally for a pending LPI to remove it from the AP list. Since
keeping the vCPU paused for the duration of the MOVALL command can
reduce the overhead, relax (but preserve) the VM pause to only pause the
source vCPU.

Signed-off-by: Oliver Upton <oupton@kernel.org>
---
 arch/arm64/kvm/vgic/vgic-its.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/arm64/kvm/vgic/vgic-its.c b/arch/arm64/kvm/vgic/vgic-its.c
index 585df69d1ffe..4e4970944ee0 100644
--- a/arch/arm64/kvm/vgic/vgic-its.c
+++ b/arch/arm64/kvm/vgic/vgic-its.c
@@ -1382,11 +1382,11 @@ static int vgic_its_cmd_handle_movall(struct kvm *kvm, struct vgic_its *its,
 	/*
 	 * Bulk operations such as MOVALL are a pain, as they can clash
 	 * badly locking-wise with other vcpus entering and exiting the
-	 * guest, should they be affected by it. Stopping the guest is a
-	 * safer bet to ensure uncontended access and ultimately forward
-	 * progress. Yeah...
+	 * guest, should they be affected by it. Stopping the source vCPU
+	 * is a safer bet to ensure uncontended access and ultimately
+	 * forward progress. Yeah...
 	 */
-	kvm_arm_halt_guest(kvm);
+	kvm_pause_vcpu(vcpu1);
 
 	xa_for_each(&dist->lpi_xa, intid, irq) {
 		irq = vgic_get_irq(kvm, intid);
@@ -1400,7 +1400,7 @@ static int vgic_its_cmd_handle_movall(struct kvm *kvm, struct vgic_its *its,
 
 	vgic_its_invalidate_cache(its);
 
-	kvm_arm_resume_guest(kvm);
+	kvm_resume_vcpu(vcpu1);
 
 	return 0;
 }
-- 
2.47.3