kvm/ppc/mpic: add KVM_CAP_IRQ_MPIC

Enabling this capability connects the vcpu to the designated in-kernel
MPIC.  Using explicit connections between vcpus and irqchips allows
for flexibility, but the main benefit at the moment is that it
simplifies the code -- KVM doesn't need vm-global state to remember
which MPIC object is associated with this vm, and it doesn't need to
care about ordering between irqchip creation and vcpu creation.
Signed-off-by: Scott Wood <scottwood@freescale.com>
[agraf: add stub functions for kvmppc_mpic_{dis,}connect_vcpu]
Signed-off-by: Alexander Graf <agraf@suse.de>

Documentation/virtual/kvm/api.txt

@@ -2744,3 +2744,11 @@ to receive the topmost interrupt vector.
 When disabled (args[0] == 0), behavior is as if this facility is unsupported.
 
 When this capability is enabled, KVM_EXIT_EPR can occur.
+
+6.6 KVM_CAP_IRQ_MPIC
+
+Architectures: ppc
+Parameters: args[0] is the MPIC device fd
+            args[1] is the MPIC CPU number for this vcpu
+
+This capability connects the vcpu to an in-kernel MPIC device.

arch/powerpc/include/asm/kvm_host.h

@@ -377,6 +377,11 @@ struct kvmppc_booke_debug_reg {
 	u64 dac[KVMPPC_BOOKE_MAX_DAC];
 };
 
+#define KVMPPC_IRQ_DEFAULT	0
+#define KVMPPC_IRQ_MPIC		1
+
+struct openpic;
+
 struct kvm_vcpu_arch {
 	ulong host_stack;
 	u32 host_pid;
@@ -558,6 +563,10 @@ struct kvm_vcpu_arch {
 	unsigned long magic_page_pa; /* phys addr to map the magic page to */
 	unsigned long magic_page_ea; /* effect. addr to map the magic page to */
 
+	int irq_type;		/* one of KVM_IRQ_* */
+	int irq_cpu_id;
+	struct openpic *mpic;	/* KVM_IRQ_MPIC */
+
 #ifdef CONFIG_KVM_BOOK3S_64_HV
 	struct kvm_vcpu_arch_shared shregs;
 

arch/powerpc/include/asm/kvm_ppc.h

@@ -248,7 +248,6 @@ int kvmppc_set_one_reg(struct kvm_vcpu *vcpu, u64 id, union kvmppc_one_reg *);
 void kvmppc_set_pid(struct kvm_vcpu *vcpu, u32 pid);
 
 struct openpic;
-void kvmppc_mpic_put(struct openpic *opp);
 
 #ifdef CONFIG_KVM_BOOK3S_64_HV
 static inline void kvmppc_set_xics_phys(int cpu, unsigned long addr)
@@ -278,6 +277,9 @@ static inline void kvmppc_set_epr(struct kvm_vcpu *vcpu, u32 epr)
 #ifdef CONFIG_KVM_MPIC
 
 void kvmppc_mpic_set_epr(struct kvm_vcpu *vcpu);
+int kvmppc_mpic_connect_vcpu(struct kvm_device *dev, struct kvm_vcpu *vcpu,
+			     u32 cpu);
+void kvmppc_mpic_disconnect_vcpu(struct openpic *opp, struct kvm_vcpu *vcpu);
 
 #else
 
@@ -285,6 +287,17 @@ static inline void kvmppc_mpic_set_epr(struct kvm_vcpu *vcpu)
 {
 }
 
+static inline int kvmppc_mpic_connect_vcpu(struct kvm_device *dev,
+		struct kvm_vcpu *vcpu, u32 cpu)
+{
+	return -EINVAL;
+}
+
+static inline void kvmppc_mpic_disconnect_vcpu(struct openpic *opp,
+		struct kvm_vcpu *vcpu)
+{
+}
+
 #endif /* CONFIG_KVM_MPIC */
 
 int kvm_vcpu_ioctl_config_tlb(struct kvm_vcpu *vcpu,

arch/powerpc/kvm/booke.c

@@ -430,6 +430,10 @@ static int kvmppc_booke_irqprio_deliver(struct kvm_vcpu *vcpu,
 		if (update_epr == true) {
 			if (vcpu->arch.epr_flags & KVMPPC_EPR_USER)
 				kvm_make_request(KVM_REQ_EPR_EXIT, vcpu);
+			else if (vcpu->arch.epr_flags & KVMPPC_EPR_KERNEL) {
+				BUG_ON(vcpu->arch.irq_type != KVMPPC_IRQ_MPIC);
+				kvmppc_mpic_set_epr(vcpu);
+			}
 		}
 
 		new_msr &= msr_mask;

arch/powerpc/kvm/mpic.c

@@ -115,7 +115,7 @@ static int get_current_cpu(void)
 {
 #if defined(CONFIG_KVM) && defined(CONFIG_BOOKE)
 	struct kvm_vcpu *vcpu = current->thread.kvm_vcpu;
-	return vcpu ? vcpu->vcpu_id : -1;
+	return vcpu ? vcpu->arch.irq_cpu_id : -1;
 #else
 	/* XXX */
 	return -1;
@@ -249,7 +249,7 @@ static void mpic_irq_raise(struct openpic *opp, struct irq_dest *dst,
 		return;
 	}
 
-	pr_debug("%s: cpu %d output %d\n", __func__, dst->vcpu->vcpu_id,
+	pr_debug("%s: cpu %d output %d\n", __func__, dst->vcpu->arch.irq_cpu_id,
 		output);
 
 	if (output != ILR_INTTGT_INT)	/* TODO */
@@ -267,7 +267,7 @@ static void mpic_irq_lower(struct openpic *opp, struct irq_dest *dst,
 		return;
 	}
 
-	pr_debug("%s: cpu %d output %d\n", __func__, dst->vcpu->vcpu_id,
+	pr_debug("%s: cpu %d output %d\n", __func__, dst->vcpu->arch.irq_cpu_id,
 		output);
 
 	if (output != ILR_INTTGT_INT)	/* TODO */
@@ -1165,6 +1165,20 @@ static uint32_t openpic_iack(struct openpic *opp, struct irq_dest *dst,
 	return retval;
 }
 
+void kvmppc_mpic_set_epr(struct kvm_vcpu *vcpu)
+{
+	struct openpic *opp = vcpu->arch.mpic;
+	int cpu = vcpu->arch.irq_cpu_id;
+	unsigned long flags;
+
+	spin_lock_irqsave(&opp->lock, flags);
+
+	if ((opp->gcr & opp->mpic_mode_mask) == GCR_MODE_PROXY)
+		kvmppc_set_epr(vcpu, openpic_iack(opp, &opp->dst[cpu], cpu));
+
+	spin_unlock_irqrestore(&opp->lock, flags);
+}
+
 static int openpic_cpu_read_internal(void *opaque, gpa_t addr,
 				     u32 *ptr, int idx)
 {
@@ -1431,10 +1445,10 @@ static void map_mmio(struct openpic *opp)
 
 static void unmap_mmio(struct openpic *opp)
 {
-	BUG_ON(opp->mmio_mapped);
+	if (opp->mmio_mapped) {
 		opp->mmio_mapped = false;
-
 		kvm_io_bus_unregister_dev(opp->kvm, KVM_MMIO_BUS, &opp->mmio);
+	}
 }
 
 static int set_base_addr(struct openpic *opp, struct kvm_device_attr *attr)
@@ -1693,3 +1707,57 @@ struct kvm_device_ops kvm_mpic_ops = {
 	.get_attr = mpic_get_attr,
 	.has_attr = mpic_has_attr,
 };
+
+int kvmppc_mpic_connect_vcpu(struct kvm_device *dev, struct kvm_vcpu *vcpu,
+			     u32 cpu)
+{
+	struct openpic *opp = dev->private;
+	int ret = 0;
+
+	if (dev->ops != &kvm_mpic_ops)
+		return -EPERM;
+	if (opp->kvm != vcpu->kvm)
+		return -EPERM;
+	if (cpu < 0 || cpu >= MAX_CPU)
+		return -EPERM;
+
+	spin_lock_irq(&opp->lock);
+
+	if (opp->dst[cpu].vcpu) {
+		ret = -EEXIST;
+		goto out;
+	}
+	if (vcpu->arch.irq_type) {
+		ret = -EBUSY;
+		goto out;
+	}
+
+	opp->dst[cpu].vcpu = vcpu;
+	opp->nb_cpus = max(opp->nb_cpus, cpu + 1);
+
+	vcpu->arch.mpic = opp;
+	vcpu->arch.irq_cpu_id = cpu;
+	vcpu->arch.irq_type = KVMPPC_IRQ_MPIC;
+
+	/* This might need to be changed if GCR gets extended */
+	if (opp->mpic_mode_mask == GCR_MODE_PROXY)
+		vcpu->arch.epr_flags |= KVMPPC_EPR_KERNEL;
+
+	kvm_device_get(dev);
+out:
+	spin_unlock_irq(&opp->lock);
+	return ret;
+}
+
+/*
+ * This should only happen immediately before the mpic is destroyed,
+ * so we shouldn't need to worry about anything still trying to
+ * access the vcpu pointer.
+ */
+void kvmppc_mpic_disconnect_vcpu(struct openpic *opp, struct kvm_vcpu *vcpu)
+{
+	BUG_ON(!opp->dst[vcpu->arch.irq_cpu_id].vcpu);
+
+	opp->dst[vcpu->arch.irq_cpu_id].vcpu = NULL;
+	kvm_device_put(opp->dev);
+}

arch/powerpc/kvm/powerpc.c

@@ -25,6 +25,7 @@
 #include <linux/hrtimer.h>
 #include <linux/fs.h>
 #include <linux/slab.h>
+#include <linux/file.h>
 #include <asm/cputable.h>
 #include <asm/uaccess.h>
 #include <asm/kvm_ppc.h>
@@ -326,6 +327,9 @@ int kvm_dev_ioctl_check_extension(long ext)
 	case KVM_CAP_PPC_GET_PVINFO:
 #if defined(CONFIG_KVM_E500V2) || defined(CONFIG_KVM_E500MC)
 	case KVM_CAP_SW_TLB:
+#endif
+#ifdef CONFIG_KVM_MPIC
+	case KVM_CAP_IRQ_MPIC:
 #endif
 		r = 1;
 		break;
@@ -460,6 +464,13 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 	tasklet_kill(&vcpu->arch.tasklet);
 
 	kvmppc_remove_vcpu_debugfs(vcpu);
+
+	switch (vcpu->arch.irq_type) {
+	case KVMPPC_IRQ_MPIC:
+		kvmppc_mpic_disconnect_vcpu(vcpu->arch.mpic, vcpu);
+		break;
+	}
+
 	kvmppc_core_vcpu_free(vcpu);
 }
 
@@ -786,6 +797,25 @@ static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
 		r = kvm_vcpu_ioctl_config_tlb(vcpu, &cfg);
 		break;
 	}
+#endif
+#ifdef CONFIG_KVM_MPIC
+	case KVM_CAP_IRQ_MPIC: {
+		struct file *filp;
+		struct kvm_device *dev;
+
+		r = -EBADF;
+		filp = fget(cap->args[0]);
+		if (!filp)
+			break;
+
+		r = -EPERM;
+		dev = kvm_device_from_filp(filp);
+		if (dev)
+			r = kvmppc_mpic_connect_vcpu(dev, vcpu, cap->args[1]);
+
+		fput(filp);
+		break;
+	}
 #endif
 	default:
 		r = -EINVAL;

include/uapi/linux/kvm.h

@@ -667,6 +667,7 @@ struct kvm_ppc_smmu_info {
 #define KVM_CAP_ARM_PSCI 87
 #define KVM_CAP_ARM_SET_DEVICE_ADDR 88
 #define KVM_CAP_DEVICE_CTRL 89
+#define KVM_CAP_IRQ_MPIC 90
 
 #ifdef KVM_CAP_IRQ_ROUTING