Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 22 Jul 2022 16:02:38 -0700
In-Reply-To: <20220722230241.1944655-1-avagin@google.com>
Message-Id: <20220722230241.1944655-3-avagin@google.com>
Mime-Version: 1.0
References: <20220722230241.1944655-1-avagin@google.com>
Subject: [PATCH 2/5] kvm/x86: add controls to enable/disable paravirtualized
 system calls
From:   Andrei Vagin <avagin@google.com>
To:     Paolo Bonzini <pbonzini@redhat.com>
Cc:     linux-kernel@vger.kernel.org, kvm@vger.kernel.org,
        Andrei Vagin <avagin@google.com>,
        Sean Christopherson <seanjc@google.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Jianfeng Tan <henry.tjf@antfin.com>,
        Adin Scannell <ascannell@google.com>,
        Konstantin Bogomolov <bogomolov@google.com>,
        Etienne Perot <eperot@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

The following change will add a new hypercall to execute host syscalls.

This hypercall is helpful for user-mode kernel solutions such as gVisor
that needs to manage multiple address spaces.

The new hypercall is a backdoor for most KVM users, so it must be
disabled by default. This change introduces a new capability that has to
be set to enable the hypercall. There is another standard way to allow
hypercalls by using KVM_SET_CPUID2. It isn't suitable in this case
because one of the common ways of using it is to request all available
features (KVM_GET_SUPPORTED_CPUID) and let them all together. In this
case, it is a hard requirement that the new hypercall can be enabled
only intentionally.

Signed-off-by: Andrei Vagin <avagin@google.com>
---
 arch/x86/include/uapi/asm/kvm_para.h |  3 +++
 arch/x86/kvm/cpuid.c                 | 25 +++++++++++++++++++++++++
 arch/x86/kvm/cpuid.h                 |  8 +++++++-
 arch/x86/kvm/x86.c                   |  4 ++++
 include/uapi/linux/kvm.h             |  1 +
 5 files changed, 40 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/uapi/asm/kvm_para.h b/arch/x86/include/uapi/asm/kvm_para.h
index 6e64b27b2c1e..84ad13ffc23c 100644
--- a/arch/x86/include/uapi/asm/kvm_para.h
+++ b/arch/x86/include/uapi/asm/kvm_para.h
@@ -37,6 +37,9 @@
 #define KVM_FEATURE_HC_MAP_GPA_RANGE	16
 #define KVM_FEATURE_MIGRATION_CONTROL	17
 
+/* Features that are not controlled by KVM_SET_CPUID2. */
+#define KVM_FEATURE_PV_HOST_SYSCALL	31
+
 #define KVM_HINTS_REALTIME      0
 
 /* The last 8 bits are used to indicate how to interpret the flags field
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index de6d44e07e34..4fdfe9409506 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -104,6 +104,10 @@ static int kvm_check_cpuid(struct kvm_vcpu *vcpu,
 			return -EINVAL;
 	}
 
+	best = cpuid_entry2_find(entries, nent, KVM_CPUID_FEATURES, 0);
+	if (best && (best->eax & (1<<KVM_FEATURE_PV_HOST_SYSCALL)))
+		return -EINVAL;
+
 	/*
 	 * Exposing dynamic xfeatures to the guest requires additional
 	 * enabling in the FPU, e.g. to expand the guest XSAVE state size.
@@ -273,6 +277,27 @@ static void __kvm_update_cpuid_runtime(struct kvm_vcpu *vcpu, struct kvm_cpuid_e
 	}
 }
 
+int kvm_vcpu_pv_set_host_syscall(struct kvm_vcpu *vcpu, bool set)
+{
+	struct kvm_cpuid_entry2 *best;
+
+	if (!vcpu->arch.pv_cpuid.enforce)
+		return -EINVAL;
+
+	best = kvm_find_cpuid_entry(vcpu, KVM_CPUID_FEATURES, 0);
+	if (!best)
+		return -EINVAL;
+
+	if (set)
+		best->eax |= 1 << KVM_FEATURE_PV_HOST_SYSCALL;
+	else
+		best->eax &= ~(1 << KVM_FEATURE_PV_HOST_SYSCALL);
+
+	kvm_update_pv_runtime(vcpu);
+
+	return 0;
+}
+
 void kvm_update_cpuid_runtime(struct kvm_vcpu *vcpu)
 {
 	__kvm_update_cpuid_runtime(vcpu, vcpu->arch.cpuid_entries, vcpu->arch.cpuid_nent);
diff --git a/arch/x86/kvm/cpuid.h b/arch/x86/kvm/cpuid.h
index 8a770b481d9d..80721093b82b 100644
--- a/arch/x86/kvm/cpuid.h
+++ b/arch/x86/kvm/cpuid.h
@@ -219,10 +219,16 @@ static __always_inline void kvm_cpu_cap_check_and_set(unsigned int x86_feature)
 static __always_inline bool guest_pv_has(struct kvm_vcpu *vcpu,
 					 unsigned int kvm_feature)
 {
-	if (!vcpu->arch.pv_cpuid.enforce)
+	if (!vcpu->arch.pv_cpuid.enforce) {
+		if (kvm_feature == KVM_FEATURE_PV_HOST_SYSCALL)
+			return false;
+
 		return true;
+	}
 
 	return vcpu->arch.pv_cpuid.features & (1u << kvm_feature);
 }
 
+int kvm_vcpu_pv_set_host_syscall(struct kvm_vcpu *vcpu, bool set);
+
 #endif
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index e5fa335a4ea7..19e634768161 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -5306,6 +5306,10 @@ static int kvm_vcpu_ioctl_enable_cap(struct kvm_vcpu *vcpu,
 			kvm_update_pv_runtime(vcpu);
 
 		return 0;
+
+	case KVM_CAP_PV_HOST_SYSCALL:
+		return kvm_vcpu_pv_set_host_syscall(vcpu, cap->args[0]);
+
 	default:
 		return -EINVAL;
 	}
diff --git a/include/uapi/linux/kvm.h b/include/uapi/linux/kvm.h
index 860f867c50c0..89ed59d13877 100644
--- a/include/uapi/linux/kvm.h
+++ b/include/uapi/linux/kvm.h
@@ -1157,6 +1157,7 @@ struct kvm_ppc_resize_hpt {
 #define KVM_CAP_VM_TSC_CONTROL 214
 #define KVM_CAP_SYSTEM_EVENT_DATA 215
 #define KVM_CAP_ARM_SYSTEM_SUSPEND 216
+#define KVM_CAP_PV_HOST_SYSCALL 217
 
 #ifdef KVM_CAP_IRQ_ROUTING
 
-- 
2.37.1.359.gd136c6c3e2-goog