From mboxrd@z Thu Jan 1 00:00:00 1970 From: =?UTF-8?q?Radim=20Kr=C4=8Dm=C3=A1=C5=99?= Subject: [PATCH 3/5] KVM: add KVM_USER_EXIT vm ioctl for userspace exit Date: Wed, 5 Aug 2015 15:21:15 +0200 Message-ID: <1438780877-31838-4-git-send-email-rkrcmar@redhat.com> References: <1438780877-31838-1-git-send-email-rkrcmar@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: kvm@vger.kernel.org, Paolo Bonzini To: linux-kernel@vger.kernel.org Return-path: In-Reply-To: <1438780877-31838-1-git-send-email-rkrcmar@redhat.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org 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. It would be cleaner to use 'unsigned long' to store the vcpu_id, but I really don't like its variable size and 'u64' will be same/bigger for few for more years. Signed-off-by: Radim Kr=C4=8Dm=C3=A1=C5=99 --- Documentation/virtual/kvm/api.txt | 30 ++++++++++++++++++++++++++++++ arch/x86/kvm/x86.c | 1 + include/uapi/linux/kvm.h | 8 ++++++++ virt/kvm/kvm_main.c | 35 +++++++++++++++++++++++++++++++= ++++ 4 files changed, 74 insertions(+) diff --git a/Documentation/virtual/kvm/api.txt b/Documentation/virtual/= kvm/api.txt index 3c714d43a717..5cf25a15fc6f 100644 --- a/Documentation/virtual/kvm/api.txt +++ b/Documentation/virtual/kvm/api.txt @@ -3020,6 +3020,36 @@ Returns: 0 on success, -1 on error =20 Queues an SMI on the thread's vcpu. =20 + +4.97 KVM_USER_EXIT + +Capability: KVM_CAP_USER_EXIT +Architectures: x86 +Type: vm 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, + -ENOENT if VCPU with 'vcpu_id' is not present + +struct kvm_user_exit { + __u64 vcpu_id; + __u32 reserved[14]; +}; + +Make vcpu_id exit to userspace as soon as possible. If the VCPU is no= t 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_E= XIT was +issued, it will keep the original exit reason and not exit early on ne= xt +KVM_RUN. +If VCPU exited because of KVM_USER_EXIT, the exit reason is KVM_EXIT_R= EQUEST. + +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_SIGN= AL_MASK) +to the VCPU thread. + + + 5. The kvm_run structure ------------------------ =20 diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c index c5d790fdfc2e..61f35944dd53 100644 --- a/arch/x86/kvm/x86.c +++ b/arch/x86/kvm/x86.c @@ -2465,6 +2465,7 @@ 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 =3D 1; break; case KVM_CAP_X86_SMM: diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h index d996a7cdb4d2..79316489346c 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 =20 #ifdef KVM_CAP_IRQ_ROUTING =20 @@ -1008,6 +1009,11 @@ struct kvm_device_attr { __u64 addr; /* userspace address of attr data */ }; =20 +struct kvm_user_exit { + __u64 vcpu_id; /* the 'unsigned long' used in KVM_CREATE_VCPU = */ + __u32 reserved[14]; +}; + #define KVM_DEV_VFIO_GROUP 1 #define KVM_DEV_VFIO_GROUP_ADD 1 #define KVM_DEV_VFIO_GROUP_DEL 2 @@ -1119,6 +1125,8 @@ struct kvm_s390_ucas_mapping { #define KVM_ARM_SET_DEVICE_ADDR _IOW(KVMIO, 0xab, struct kvm_arm_de= vice_addr) /* Available with KVM_CAP_PPC_RTAS */ #define KVM_PPC_RTAS_DEFINE_TOKEN _IOW(KVMIO, 0xac, struct kvm_rtas_t= oken_args) +/* Available with KVM_CAP_USER_EXIT */ +#define KVM_USER_EXIT _IOW(KVMIO, 0xad, struct kvm_user_e= xit) =20 /* ioctl for vm fd */ #define KVM_CREATE_DEVICE _IOWR(KVMIO, 0xe0, struct kvm_create_devi= ce) diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c index b34a328fdac1..024428b64812 100644 --- a/virt/kvm/kvm_main.c +++ b/virt/kvm/kvm_main.c @@ -2644,6 +2644,32 @@ static long kvm_vm_ioctl_check_extension_generic= (struct kvm *kvm, long arg) return kvm_vm_ioctl_check_extension(kvm, arg); } =20 +int kvm_vm_ioctl_user_exit(struct kvm *kvm, struct kvm_user_exit *info= ) +{ + /* Casting vcpu_id to int is intended and matches the behavior of + * KVM_CREATE_VCPU, where we cast from unsigned long. + */ + int vcpu_id =3D info->vcpu_id; + int idx; + struct kvm_vcpu *vcpu; + const struct kvm_user_exit valid =3D {.vcpu_id =3D info->vcpu_id}; + + BUILD_BUG_ON(sizeof(struct kvm_user_exit) !=3D 64); + + if (memcmp(info, &valid, sizeof(valid))) + return -EINVAL; + + kvm_for_each_vcpu(idx, vcpu, kvm) + if (vcpu->vcpu_id =3D=3D vcpu_id) { + kvm_make_request(KVM_REQ_EXIT, vcpu); + kvm_vcpu_kick(vcpu); + + return 0; + } + + return -ENOENT; +} + static long kvm_vm_ioctl(struct file *filp, unsigned int ioctl, unsigned long arg) { @@ -2779,6 +2805,15 @@ out_free_irq_routing: vfree(entries); break; } + case KVM_USER_EXIT: { + struct kvm_user_exit info; + + r =3D -EFAULT; + if (copy_from_user(&info, argp, sizeof(info))) + goto out; + r =3D kvm_vm_ioctl_user_exit(kvm, &info); + break; + } #endif /* CONFIG_HAVE_KVM_IRQ_ROUTING */ case KVM_CREATE_DEVICE: { struct kvm_create_device cd; --=20 2.5.0