Re: [PATCH v2 4/5] KVM: add KVM_USER_EXIT vcpu ioctl for userspace exit

[Paolo Bonzini <pbonzini@redhat.com>]

On 05/08/2015 18:33, Radim Krčmář wrote:
> 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>
> ---
>  v2:
>   * use vcpu ioctl instead of vm one [4/5]
>   * shrink kvm_user_exit from 64 to 32 bytes [4/5]
> 
>  Documentation/virtual/kvm/api.txt | 30 ++++++++++++++++++++++++++++++
>  arch/x86/kvm/x86.c                | 24 ++++++++++++++++++++++++
>  include/uapi/linux/kvm.h          |  7 +++++++
>  virt/kvm/kvm_main.c               |  5 +++--
>  4 files changed, 64 insertions(+), 2 deletions(-)
> 
> diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/kvm/api.txt
> index 3c714d43a717..c5844f0b8e7c 100644
> --- a/Documentation/virtual/kvm/api.txt
> +++ b/Documentation/virtual/kvm/api.txt
> @@ -3020,6 +3020,36 @@ 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: struct kvm_user_exit (in)
> +Returns: 0 on success,
> +         -EFAULT if the parameter couldn't be read,
> +         -EINVAL if 'reserved' is not zeroed,
> +
> +struct kvm_user_exit {
> +	__u8 reserved[32];
> +};
> +
> +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 and not exit 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.

Can we just return EINVAL if the parameter is not NULL?

Paolo

> +
> +
> +
>  5. The kvm_run structure
>  ------------------------
>  
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 3493457ad0a1..27d777eb34e4 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -2466,6 +2466,7 @@ 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:
> @@ -3077,6 +3078,20 @@ static int kvm_set_guest_paused(struct kvm_vcpu *vcpu)
>  	return 0;
>  }
>  
> +int kvm_vcpu_ioctl_user_exit(struct kvm_vcpu *vcpu, struct kvm_user_exit *info)
> +{
> +	struct kvm_user_exit valid = {};
> +       BUILD_BUG_ON(sizeof(valid) != 32);
> +
> +	if (memcmp(info, &valid, sizeof(valid)))
> +		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)
>  {
> @@ -3341,6 +3356,15 @@ long kvm_arch_vcpu_ioctl(struct file *filp,
>  		r = kvm_set_guest_paused(vcpu);
>  		goto out;
>  	}
> +	case KVM_USER_EXIT: {
> +		struct kvm_user_exit info;
> +
> +		r = -EFAULT;
> +		if (copy_from_user(&info, argp, sizeof(info)))
> +			goto out;
> +		r = kvm_vcpu_ioctl_user_exit(vcpu, &info);
> +		break;
> +	}
>  	default:
>  		r = -EINVAL;
>  	}
> diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
> index d996a7cdb4d2..bc5a1abe9626 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
>  
> @@ -1008,6 +1009,10 @@ struct kvm_device_attr {
>  	__u64	addr;		/* userspace address of attr data */
>  };
>  
> +struct kvm_user_exit {
> +	__u8 reserved[32];
> +};
> +
>  #define  KVM_DEV_VFIO_GROUP			1
>  #define   KVM_DEV_VFIO_GROUP_ADD			1
>  #define   KVM_DEV_VFIO_GROUP_DEL			2
> @@ -1119,6 +1124,8 @@ struct kvm_s390_ucas_mapping {
>  #define KVM_ARM_SET_DEVICE_ADDR	  _IOW(KVMIO,  0xab, struct kvm_arm_device_addr)
>  /* Available with KVM_CAP_PPC_RTAS */
>  #define KVM_PPC_RTAS_DEFINE_TOKEN _IOW(KVMIO,  0xac, struct kvm_rtas_token_args)
> +/* Available with KVM_CAP_USER_EXIT */
> +#define KVM_USER_EXIT             _IOW(KVMIO,  0xad, struct kvm_user_exit)
>  
>  /* ioctl for vm fd */
>  #define KVM_CREATE_DEVICE	  _IOWR(KVMIO,  0xe0, struct kvm_create_device)
> diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
> index b34a328fdac1..d7ffe6090520 100644
> --- a/virt/kvm/kvm_main.c
> +++ b/virt/kvm/kvm_main.c
> @@ -2248,15 +2248,16 @@ 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)
>