KVM: x86: deliver KVM IOAPIC scan request to target vCPUs

In IOAPIC fixed delivery mode instead of flushing the scan
requests to all vCPUs, we should only send the requests to
vCPUs specified within the destination field.

This patch introduces kvm_get_dest_vcpus_mask() API which
retrieves an array of target vCPUs by using
kvm_apic_map_get_dest_lapic() and then based on the
vcpus_idx, it sets the bit in a bitmap. However, if the above
fails kvm_get_dest_vcpus_mask() finds the target vCPUs by
traversing all available vCPUs. Followed by setting the
bits in the bitmap.

If we had different vCPUs in the previous request for the
same redirection table entry then bits corresponding to
these vCPUs are also set. This to done to keep
ioapic_handled_vectors synchronized.

This bitmap is then eventually passed on to
kvm_make_vcpus_request_mask() to generate a masked request
only for the target vCPUs.

This would enable us to reduce the latency overhead on isolated
vCPUs caused by the IPI to process due to KVM_REQ_IOAPIC_SCAN.

Suggested-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Nitesh Narayan Lal <nitesh@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

6 files changed, 96 insertions, 2 deletions

diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 20bb2fc0883a53..898ab9eb4dc85d 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -1588,6 +1588,8 @@ bool kvm_is_linear_rip(struct kvm_vcpu *vcpu, unsigned long linear_rip);
 
 void kvm_make_mclock_inprogress_request(struct kvm *kvm);
 void kvm_make_scan_ioapic_request(struct kvm *kvm);
+void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
+				       unsigned long *vcpu_bitmap);
 
 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,
 				     struct kvm_async_pf *work);
diff --git a/arch/x86/kvm/ioapic.c b/arch/x86/kvm/ioapic.c
index d859ae8890d0a6..ce30ef23c86bc3 100644
--- a/arch/x86/kvm/ioapic.c
+++ b/arch/x86/kvm/ioapic.c
@@ -271,8 +271,9 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
 {
 	unsigned index;
 	bool mask_before, mask_after;
-	int old_remote_irr, old_delivery_status;
 	union kvm_ioapic_redirect_entry *e;
+	unsigned long vcpu_bitmap;
+	int old_remote_irr, old_delivery_status, old_dest_id, old_dest_mode;
 
 	switch (ioapic->ioregsel) {
 	case IOAPIC_REG_VERSION:
@@ -296,6 +297,8 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
 		/* Preserve read-only fields */
 		old_remote_irr = e->fields.remote_irr;
 		old_delivery_status = e->fields.delivery_status;
+		old_dest_id = e->fields.dest_id;
+		old_dest_mode = e->fields.dest_mode;
 		if (ioapic->ioregsel & 1) {
 			e->bits &= 0xffffffff;
 			e->bits |= (u64) val << 32;
@@ -321,7 +324,33 @@ static void ioapic_write_indirect(struct kvm_ioapic *ioapic, u32 val)
 		if (e->fields.trig_mode == IOAPIC_LEVEL_TRIG
 		    && ioapic->irr & (1 << index))
 			ioapic_service(ioapic, index, false);
-		kvm_make_scan_ioapic_request(ioapic->kvm);
+		if (e->fields.delivery_mode == APIC_DM_FIXED) {
+			struct kvm_lapic_irq irq;
+
+			irq.shorthand = 0;
+			irq.vector = e->fields.vector;
+			irq.delivery_mode = e->fields.delivery_mode << 8;
+			irq.dest_id = e->fields.dest_id;
+			irq.dest_mode = e->fields.dest_mode;
+			kvm_bitmap_or_dest_vcpus(ioapic->kvm, &irq,
+						 &vcpu_bitmap);
+			if (old_dest_mode != e->fields.dest_mode ||
+			    old_dest_id != e->fields.dest_id) {
+				/*
+				 * Update vcpu_bitmap with vcpus specified in
+				 * the previous request as well. This is done to
+				 * keep ioapic_handled_vectors synchronized.
+				 */
+				irq.dest_id = old_dest_id;
+				irq.dest_mode = old_dest_mode;
+				kvm_bitmap_or_dest_vcpus(ioapic->kvm, &irq,
+							 &vcpu_bitmap);
+			}
+			kvm_make_scan_ioapic_request_mask(ioapic->kvm,
+							  &vcpu_bitmap);
+		} else {
+			kvm_make_scan_ioapic_request(ioapic->kvm);
+		}
 		break;
 	}
 }
diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
index 60fb21fe7f4202..452cedd6382bc6 100644
--- a/arch/x86/kvm/lapic.c
+++ b/arch/x86/kvm/lapic.c
@@ -1122,6 +1122,50 @@ static int __apic_accept_irq(struct kvm_lapic *apic, int delivery_mode,
 	return result;
 }
 
+/*
+ * This routine identifies the destination vcpus mask meant to receive the
+ * IOAPIC interrupts. It either uses kvm_apic_map_get_dest_lapic() to find
+ * out the destination vcpus array and set the bitmap or it traverses to
+ * each available vcpu to identify the same.
+ */
+void kvm_bitmap_or_dest_vcpus(struct kvm *kvm, struct kvm_lapic_irq *irq,
+			      unsigned long *vcpu_bitmap)
+{
+	struct kvm_lapic **dest_vcpu = NULL;
+	struct kvm_lapic *src = NULL;
+	struct kvm_apic_map *map;
+	struct kvm_vcpu *vcpu;
+	unsigned long bitmap;
+	int i, vcpu_idx;
+	bool ret;
+
+	rcu_read_lock();
+	map = rcu_dereference(kvm->arch.apic_map);
+
+	ret = kvm_apic_map_get_dest_lapic(kvm, &src, irq, map, &dest_vcpu,
+					  &bitmap);
+	if (ret) {
+		for_each_set_bit(i, &bitmap, 16) {
+			if (!dest_vcpu[i])
+				continue;
+			vcpu_idx = dest_vcpu[i]->vcpu->vcpu_idx;
+			__set_bit(vcpu_idx, vcpu_bitmap);
+		}
+	} else {
+		kvm_for_each_vcpu(i, vcpu, kvm) {
+			if (!kvm_apic_present(vcpu))
+				continue;
+			if (!kvm_apic_match_dest(vcpu, NULL,
+						 irq->delivery_mode,
+						 irq->dest_id,
+						 irq->dest_mode))
+				continue;
+			__set_bit(i, vcpu_bitmap);
+		}
+	}
+	rcu_read_unlock();
+}
+
 int kvm_apic_compare_prio(struct kvm_vcpu *vcpu1, struct kvm_vcpu *vcpu2)
 {
 	return vcpu1->arch.apic_arb_prio - vcpu2->arch.apic_arb_prio;
diff --git a/arch/x86/kvm/lapic.h b/arch/x86/kvm/lapic.h
index 2aad7e226fc016..c1d77436126aca 100644
--- a/arch/x86/kvm/lapic.h
+++ b/arch/x86/kvm/lapic.h
@@ -226,6 +226,9 @@ bool kvm_apic_pending_eoi(struct kvm_vcpu *vcpu, int vector);
 
 void kvm_wait_lapic_expire(struct kvm_vcpu *vcpu);
 
+void kvm_bitmap_or_dest_vcpus(struct kvm *kvm, struct kvm_lapic_irq *irq,
+			      unsigned long *vcpu_bitmap);
+
 bool kvm_intr_is_single_vcpu_fast(struct kvm *kvm, struct kvm_lapic_irq *irq,
 			struct kvm_vcpu **dest_vcpu);
 int kvm_vector_to_index(u32 vector, u32 dest_vcpus,
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 8db7275d313f22..991dd01ba08b09 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -7874,6 +7874,20 @@ static void process_smi(struct kvm_vcpu *vcpu)
 	kvm_make_request(KVM_REQ_EVENT, vcpu);
 }
 
+void kvm_make_scan_ioapic_request_mask(struct kvm *kvm,
+				       unsigned long *vcpu_bitmap)
+{
+	cpumask_var_t cpus;
+	bool called;
+
+	zalloc_cpumask_var(&cpus, GFP_ATOMIC);
+
+	called = kvm_make_vcpus_request_mask(kvm, KVM_REQ_SCAN_IOAPIC,
+					     vcpu_bitmap, cpus);
+
+	free_cpumask_var(cpus);
+}
+
 void kvm_make_scan_ioapic_request(struct kvm *kvm)
 {
 	kvm_make_all_cpus_request(kvm, KVM_REQ_SCAN_IOAPIC);
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 70b2296fb2aef2..bfe6c672998865 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -786,6 +786,8 @@ void kvm_reload_remote_mmus(struct kvm *kvm);
 bool kvm_make_vcpus_request_mask(struct kvm *kvm, unsigned int req,
 				 unsigned long *vcpu_bitmap, cpumask_var_t tmp);
 bool kvm_make_all_cpus_request(struct kvm *kvm, unsigned int req);
+bool kvm_make_cpus_request_mask(struct kvm *kvm, unsigned int req,
+				unsigned long *vcpu_bitmap);
 
 long kvm_arch_dev_ioctl(struct file *filp,
 			unsigned int ioctl, unsigned long arg);