]> asedeno.scripts.mit.edu Git - linux.git/commitdiff
KVM: arm/arm64: vgic-new: vgic_init: implement vgic_create
authorEric Auger <eric.auger@linaro.org>
Mon, 21 Dec 2015 13:50:50 +0000 (14:50 +0100)
committerChristoffer Dall <christoffer.dall@linaro.org>
Fri, 20 May 2016 13:40:06 +0000 (15:40 +0200)
This patch implements the vgic_creation function which is
called on CREATE_IRQCHIP VM IOCTL (v2 only) or KVM_CREATE_DEVICE

Signed-off-by: Eric Auger <eric.auger@linaro.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
include/kvm/vgic/vgic.h
virt/kvm/arm/vgic/vgic-init.c

index 393489fdc480bbe4c6c7a9357ef8ecc1be1ca2b9..0634d895007a223f6d09fddff7e3c6615f2ff107 100644 (file)
@@ -195,6 +195,7 @@ struct vgic_cpu {
 };
 
 int kvm_vgic_addr(struct kvm *kvm, unsigned long type, u64 *addr, bool write);
+int kvm_vgic_create(struct kvm *kvm, u32 type);
 int kvm_vgic_hyp_init(void);
 
 int kvm_vgic_inject_irq(struct kvm *kvm, int cpuid, unsigned int intid,
index 4523beb029f3e32d86c5984a9fe8dbb14e1dc67c..15d54284305ba668c2ab954c13feb8e7911014f2 100644 (file)
 #include <asm/kvm_mmu.h>
 #include "vgic.h"
 
+/* CREATION */
+
+/**
+ * kvm_vgic_create: triggered by the instantiation of the VGIC device by
+ * user space, either through the legacy KVM_CREATE_IRQCHIP ioctl (v2 only)
+ * or through the generic KVM_CREATE_DEVICE API ioctl.
+ * irqchip_in_kernel() tells you if this function succeeded or not.
+ */
+int kvm_vgic_create(struct kvm *kvm, u32 type)
+{
+       int i, vcpu_lock_idx = -1, ret;
+       struct kvm_vcpu *vcpu;
+
+       mutex_lock(&kvm->lock);
+
+       if (irqchip_in_kernel(kvm)) {
+               ret = -EEXIST;
+               goto out;
+       }
+
+       /*
+        * This function is also called by the KVM_CREATE_IRQCHIP handler,
+        * which had no chance yet to check the availability of the GICv2
+        * emulation. So check this here again. KVM_CREATE_DEVICE does
+        * the proper checks already.
+        */
+       if (type == KVM_DEV_TYPE_ARM_VGIC_V2 &&
+               !kvm_vgic_global_state.can_emulate_gicv2) {
+               ret = -ENODEV;
+               goto out;
+       }
+
+       /*
+        * Any time a vcpu is run, vcpu_load is called which tries to grab the
+        * vcpu->mutex.  By grabbing the vcpu->mutex of all VCPUs we ensure
+        * that no other VCPUs are run while we create the vgic.
+        */
+       ret = -EBUSY;
+       kvm_for_each_vcpu(i, vcpu, kvm) {
+               if (!mutex_trylock(&vcpu->mutex))
+                       goto out_unlock;
+               vcpu_lock_idx = i;
+       }
+
+       kvm_for_each_vcpu(i, vcpu, kvm) {
+               if (vcpu->arch.has_run_once)
+                       goto out_unlock;
+       }
+       ret = 0;
+
+       if (type == KVM_DEV_TYPE_ARM_VGIC_V2)
+               kvm->arch.max_vcpus = VGIC_V2_MAX_CPUS;
+       else
+               kvm->arch.max_vcpus = VGIC_V3_MAX_CPUS;
+
+       if (atomic_read(&kvm->online_vcpus) > kvm->arch.max_vcpus) {
+               ret = -E2BIG;
+               goto out_unlock;
+       }
+
+       kvm->arch.vgic.in_kernel = true;
+       kvm->arch.vgic.vgic_model = type;
+
+       /*
+        * kvm_vgic_global_state.vctrl_base is set on vgic probe (kvm_arch_init)
+        * it is stored in distributor struct for asm save/restore purpose
+        */
+       kvm->arch.vgic.vctrl_base = kvm_vgic_global_state.vctrl_base;
+
+       kvm->arch.vgic.vgic_dist_base = VGIC_ADDR_UNDEF;
+       kvm->arch.vgic.vgic_cpu_base = VGIC_ADDR_UNDEF;
+       kvm->arch.vgic.vgic_redist_base = VGIC_ADDR_UNDEF;
+
+out_unlock:
+       for (; vcpu_lock_idx >= 0; vcpu_lock_idx--) {
+               vcpu = kvm_get_vcpu(kvm, vcpu_lock_idx);
+               mutex_unlock(&vcpu->mutex);
+       }
+
+out:
+       mutex_unlock(&kvm->lock);
+       return ret;
+}
+
 /* GENERIC PROBE */
 
 static void vgic_init_maintenance_interrupt(void *info)