[PATCH v3 4/5] KVM: add KVM_USER_EXIT vcpu ioctl for userspace exit

["Radim Krčmář" <rkrcmar@redhat.com>]

The guest can use KVM_USER_EXIT instead of a signal-based exiting to
userspace.  Availability depends on KVM_CAP_USER_EXIT.
Only x86 is implemented so far.

Signed-off-by: Radim Krčmář <rkrcmar@redhat.com>
---
 v3:
  * use ioctl argument directly (unsigned long as flags) [Paolo]
 v2:
  * use vcpu ioctl instead of vm one [Paolo]
  * shrink kvm_user_exit from 64 to 32 bytes

 Documentation/virtual/kvm/api.txt | 25 +++++++++++++++++++++++++
 arch/x86/kvm/x86.c                | 15 +++++++++++++++
 include/uapi/linux/kvm.h          |  3 +++
 virt/kvm/kvm_main.c               |  5 +++--
 4 files changed, 46 insertions(+), 2 deletions(-)

diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
index 3c714d43a717..df087ff3c5b6 100644
--- a/Documentation/virtual/kvm/api.txt
+++ b/Documentation/virtual/kvm/api.txt
@@ -3020,6 +3020,31 @@ Returns: 0 on success, -1 on error
 
 Queues an SMI on the thread's vcpu.
 
+
+4.97 KVM_USER_EXIT
+
+Capability: KVM_CAP_USER_EXIT
+Architectures: x86
+Type: vcpu ioctl
+Parameters: unsigned long flags (in)
+Returns: 0 on success,
+         -EINVAL if flags is not 0
+
+The ioctl is asynchronous to VCPU execution and can be issued from all threads.
+
+Make vcpu_id exit to userspace as soon as possible.  If the VCPU is not running
+in kernel at the time, it will exit early on the next call to KVM_RUN.
+If the VCPU was going to exit because of other reasons when KVM_USER_EXIT was
+issued, it will keep the original exit reason without exiting early on next
+KVM_RUN.
+If VCPU exited because of KVM_USER_EXIT, the exit reason is KVM_EXIT_REQUEST.
+
+This ioctl has very similar effect (same sans some races on userspace exit) as
+sending a signal (that is blocked in userspace and set in KVM_SET_SIGNAL_MASK)
+to the VCPU thread.
+
+
+
 5. The kvm_run structure
 ------------------------
 
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 37db1b32684a..d985806b17b1 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -2467,6 +2467,7 @@ int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
 	case KVM_CAP_ASSIGN_DEV_IRQ:
 	case KVM_CAP_PCI_2_3:
 #endif
+	case KVM_CAP_USER_EXIT:
 		r = 1;
 		break;
 	case KVM_CAP_X86_SMM:
@@ -3078,6 +3079,17 @@ static int kvm_set_guest_paused(struct kvm_vcpu *vcpu)
 	return 0;
 }
 
+static int kvm_vcpu_ioctl_user_exit(struct kvm_vcpu *vcpu, unsigned long flags)
+{
+	if (flags != 0)
+		return -EINVAL;
+
+	kvm_make_request(KVM_REQ_EXIT, vcpu);
+	kvm_vcpu_kick(vcpu);
+
+	return 0;
+}
+
 long kvm_arch_vcpu_ioctl(struct file *filp,
 			 unsigned int ioctl, unsigned long arg)
 {
@@ -3342,6 +3354,9 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
 		r = kvm_set_guest_paused(vcpu);
 		goto out;
 	}
+	case KVM_USER_EXIT:
+		r = kvm_vcpu_ioctl_user_exit(vcpu, arg);
+		break;
 	default:
 		r = -EINVAL;
 	}
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index d996a7cdb4d2..58b3a07adc81 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -826,6 +826,7 @@ struct kvm_ppc_smmu_info {
 #define KVM_CAP_X86_SMM 117
 #define KVM_CAP_MULTI_ADDRESS_SPACE 118
 #define KVM_CAP_SPLIT_IRQCHIP 119
+#define KVM_CAP_USER_EXIT 120
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
@@ -1213,6 +1214,8 @@ struct kvm_s390_ucas_mapping {
 #define KVM_S390_GET_IRQ_STATE	  _IOW(KVMIO, 0xb6, struct kvm_s390_irq_state)
 /* Available with KVM_CAP_X86_SMM */
 #define KVM_SMI                   _IO(KVMIO,   0xb7)
+/* Available with KVM_CAP_USER_EXIT */
+#define KVM_USER_EXIT             _IO(KVMIO,   0xb8)
 
 #define KVM_DEV_ASSIGN_ENABLE_IOMMU	(1 << 0)
 #define KVM_DEV_ASSIGN_PCI_2_3		(1 << 1)
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 347899966178..dfa2d5f27713 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -2251,15 +2251,16 @@ static long kvm_vcpu_ioctl(struct file *filp,
 	if (unlikely(_IOC_TYPE(ioctl) != KVMIO))
 		return -EINVAL;
 
-#if defined(CONFIG_S390) || defined(CONFIG_PPC) || defined(CONFIG_MIPS)
 	/*
 	 * Special cases: vcpu ioctls that are asynchronous to vcpu execution,
 	 * so vcpu_load() would break it.
 	 */
+#if defined(CONFIG_S390) || defined(CONFIG_PPC) || defined(CONFIG_MIPS)
 	if (ioctl == KVM_S390_INTERRUPT || ioctl == KVM_S390_IRQ || ioctl == KVM_INTERRUPT)
 		return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
 #endif
-
+	if (ioctl == KVM_USER_EXIT)
+		return kvm_arch_vcpu_ioctl(filp, ioctl, arg);
 
 	r = vcpu_load(vcpu);
 	if (r)
-- 
2.5.0