Merge branch 'kvm-redo-enable-virt' into HEAD

Register KVM's cpuhp and syscore callbacks when enabling virtualization in
hardware, as the sole purpose of said callbacks is to disable and re-enable
virtualization as needed.

The primary motivation for this series is to simplify dealing with enabling
virtualization for Intel's TDX, which needs to enable virtualization
when kvm-intel.ko is loaded, i.e. long before the first VM is created.

That said, this is a nice cleanup on its own.  By registering the callbacks
on-demand, the callbacks themselves don't need to check kvm_usage_count,
because their very existence implies a non-zero count.

Patch 1 (re)adds a dedicated lock for kvm_usage_count.  This avoids a
lock ordering issue between cpus_read_lock() and kvm_lock.  The lock
ordering issue still exist in very rare cases, and will be fixed for
good by switching vm_list to an (S)RCU-protected list.

Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

Documentation/admin-guide/kernel-parameters.txt

@@ -2648,6 +2648,23 @@
 
 			Default is Y (on).
 
+	kvm.enable_virt_at_load=[KVM,ARM64,LOONGARCH,MIPS,RISCV,X86]
+			If enabled, KVM will enable virtualization in hardware
+			when KVM is loaded, and disable virtualization when KVM
+			is unloaded (if KVM is built as a module).
+
+			If disabled, KVM will dynamically enable and disable
+			virtualization on-demand when creating and destroying
+			VMs, i.e. on the 0=>1 and 1=>0 transitions of the
+			number of VMs.
+
+			Enabling virtualization at module lode avoids potential
+			latency for creation of the 0=>1 VM, as KVM serializes
+			virtualization enabling across all online CPUs.  The
+			"cost" of enabling virtualization when KVM is loaded,
+			is that doing so may interfere with using out-of-tree
+			hypervisors that want to "own" virtualization hardware.
+
 	kvm.enable_vmware_backdoor=[KVM] Support VMware backdoor PV interface.
 				   Default is false (don't support).
 

Documentation/virt/kvm/locking.rst

@@ -11,6 +11,8 @@ The acquisition orders for mutexes are as follows:
 
 - cpus_read_lock() is taken outside kvm_lock
 
+- kvm_usage_lock is taken outside cpus_read_lock()
+
 - kvm->lock is taken outside vcpu->mutex
 
 - kvm->lock is taken outside kvm->slots_lock and kvm->irq_lock
@@ -24,6 +26,12 @@ The acquisition orders for mutexes are as follows:
   are taken on the waiting side when modifying memslots, so MMU notifiers
   must not take either kvm->slots_lock or kvm->slots_arch_lock.
 
+cpus_read_lock() vs kvm_lock:
+- Taking cpus_read_lock() outside of kvm_lock is problematic, despite that
+  being the official ordering, as it is quite easy to unknowingly trigger
+  cpus_read_lock() while holding kvm_lock.  Use caution when walking vm_list,
+  e.g. avoid complex operations when possible.
+
 For SRCU:
 
 - ``synchronize_srcu(&kvm->srcu)`` is called inside critical sections
@@ -227,10 +235,16 @@ time it will be set using the Dirty tracking mechanism described above.
 :Type:		mutex
 :Arch:		any
 :Protects:	- vm_list
-		- kvm_usage_count
+
+``kvm_usage_lock``
+^^^^^^^^^^^^^^^^^^
+
+:Type:		mutex
+:Arch:		any
+:Protects:	- kvm_usage_count
 		- hardware virtualization enable/disable
-:Comment:	KVM also disables CPU hotplug via cpus_read_lock() during
+:Comment:	Exists to allow taking cpus_read_lock() while kvm_usage_count is
-		enable/disable.
+		protected, which simplifies the virtualization enabling logic.
 
 ``kvm->mn_invalidate_lock``
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^
@@ -290,11 +304,12 @@ time it will be set using the Dirty tracking mechanism described above.
 		wakeup.
 
 ``vendor_module_lock``
-^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+^^^^^^^^^^^^^^^^^^^^^^
 :Type:		mutex
 :Arch:		x86
 :Protects:	loading a vendor module (kvm_amd or kvm_intel)
-:Comment:	Exists because using kvm_lock leads to deadlock.  cpu_hotplug_lock is
+:Comment:	Exists because using kvm_lock leads to deadlock.  kvm_lock is taken
-    taken outside of kvm_lock, e.g. in KVM's CPU online/offline callbacks, and
+    in notifiers, e.g. __kvmclock_cpufreq_notifier(), that may be invoked while
-    many operations need to take cpu_hotplug_lock when loading a vendor module,
+    cpu_hotplug_lock is held, e.g. from cpufreq_boost_trigger_state(), and many
-    e.g. updating static calls.
+    operations need to take cpu_hotplug_lock when loading a vendor module, e.g.
+    updating static calls.

arch/arm64/kvm/arm.c

@@ -2164,7 +2164,7 @@ static void cpu_hyp_uninit(void *discard)
 	}
 }
 
-int kvm_arch_hardware_enable(void)
+int kvm_arch_enable_virtualization_cpu(void)
 {
 	/*
 	 * Most calls to this function are made with migration
@@ -2184,7 +2184,7 @@ int kvm_arch_hardware_enable(void)
 	return 0;
 }
 
-void kvm_arch_hardware_disable(void)
+void kvm_arch_disable_virtualization_cpu(void)
 {
 	kvm_timer_cpu_down();
 	kvm_vgic_cpu_down();
@@ -2380,7 +2380,7 @@ static int __init do_pkvm_init(u32 hyp_va_bits)
 
 	/*
 	 * The stub hypercalls are now disabled, so set our local flag to
-	 * prevent a later re-init attempt in kvm_arch_hardware_enable().
+	 * prevent a later re-init attempt in kvm_arch_enable_virtualization_cpu().
 	 */
 	__this_cpu_write(kvm_hyp_initialized, 1);
 	preempt_enable();

arch/loongarch/kvm/main.c

@@ -261,7 +261,7 @@ long kvm_arch_dev_ioctl(struct file *filp,
 	return -ENOIOCTLCMD;
 }
 
-int kvm_arch_hardware_enable(void)
+int kvm_arch_enable_virtualization_cpu(void)
 {
 	unsigned long env, gcfg = 0;
 
@@ -300,7 +300,7 @@ int kvm_arch_hardware_enable(void)
 	return 0;
 }
 
-void kvm_arch_hardware_disable(void)
+void kvm_arch_disable_virtualization_cpu(void)
 {
 	write_csr_gcfg(0);
 	write_csr_gstat(0);

arch/mips/include/asm/kvm_host.h

@@ -728,8 +728,8 @@ struct kvm_mips_callbacks {
 	int (*handle_fpe)(struct kvm_vcpu *vcpu);
 	int (*handle_msa_disabled)(struct kvm_vcpu *vcpu);
 	int (*handle_guest_exit)(struct kvm_vcpu *vcpu);
-	int (*hardware_enable)(void);
+	int (*enable_virtualization_cpu)(void);
-	void (*hardware_disable)(void);
+	void (*disable_virtualization_cpu)(void);
 	int (*check_extension)(struct kvm *kvm, long ext);
 	int (*vcpu_init)(struct kvm_vcpu *vcpu);
 	void (*vcpu_uninit)(struct kvm_vcpu *vcpu);

arch/mips/kvm/mips.c

@@ -125,14 +125,14 @@ int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
 	return 1;
 }
 
-int kvm_arch_hardware_enable(void)
+int kvm_arch_enable_virtualization_cpu(void)
 {
-	return kvm_mips_callbacks->hardware_enable();
+	return kvm_mips_callbacks->enable_virtualization_cpu();
 }
 
-void kvm_arch_hardware_disable(void)
+void kvm_arch_disable_virtualization_cpu(void)
 {
-	kvm_mips_callbacks->hardware_disable();
+	kvm_mips_callbacks->disable_virtualization_cpu();
 }
 
 int kvm_arch_init_vm(struct kvm *kvm, unsigned long type)

arch/mips/kvm/vz.c

@@ -2869,7 +2869,7 @@ static unsigned int kvm_vz_resize_guest_vtlb(unsigned int size)
 	return ret + 1;
 }
 
-static int kvm_vz_hardware_enable(void)
+static int kvm_vz_enable_virtualization_cpu(void)
 {
 	unsigned int mmu_size, guest_mmu_size, ftlb_size;
 	u64 guest_cvmctl, cvmvmconfig;
@@ -2983,7 +2983,7 @@ static int kvm_vz_hardware_enable(void)
 	return 0;
 }
 
-static void kvm_vz_hardware_disable(void)
+static void kvm_vz_disable_virtualization_cpu(void)
 {
 	u64 cvmvmconfig;
 	unsigned int mmu_size;
@@ -3280,8 +3280,8 @@ static struct kvm_mips_callbacks kvm_vz_callbacks = {
 	.handle_msa_disabled = kvm_trap_vz_handle_msa_disabled,
 	.handle_guest_exit = kvm_trap_vz_handle_guest_exit,
 
-	.hardware_enable = kvm_vz_hardware_enable,
+	.enable_virtualization_cpu = kvm_vz_enable_virtualization_cpu,
-	.hardware_disable = kvm_vz_hardware_disable,
+	.disable_virtualization_cpu = kvm_vz_disable_virtualization_cpu,
 	.check_extension = kvm_vz_check_extension,
 	.vcpu_init = kvm_vz_vcpu_init,
 	.vcpu_uninit = kvm_vz_vcpu_uninit,

arch/riscv/kvm/main.c

@@ -20,7 +20,7 @@ long kvm_arch_dev_ioctl(struct file *filp,
 	return -EINVAL;
 }
 
-int kvm_arch_hardware_enable(void)
+int kvm_arch_enable_virtualization_cpu(void)
 {
 	csr_write(CSR_HEDELEG, KVM_HEDELEG_DEFAULT);
 	csr_write(CSR_HIDELEG, KVM_HIDELEG_DEFAULT);
@@ -35,7 +35,7 @@ int kvm_arch_hardware_enable(void)
 	return 0;
 }
 
-void kvm_arch_hardware_disable(void)
+void kvm_arch_disable_virtualization_cpu(void)
 {
 	kvm_riscv_aia_disable();
 

arch/x86/include/asm/kvm-x86-ops.h

@@ -14,8 +14,8 @@ BUILD_BUG_ON(1)
  * be __static_call_return0.
  */
 KVM_X86_OP(check_processor_compatibility)
-KVM_X86_OP(hardware_enable)
+KVM_X86_OP(enable_virtualization_cpu)
-KVM_X86_OP(hardware_disable)
+KVM_X86_OP(disable_virtualization_cpu)
 KVM_X86_OP(hardware_unsetup)
 KVM_X86_OP(has_emulated_msr)
 KVM_X86_OP(vcpu_after_set_cpuid)

arch/x86/include/asm/kvm_host.h

@@ -36,6 +36,7 @@
 #include <asm/kvm_page_track.h>
 #include <asm/kvm_vcpu_regs.h>
 #include <asm/hyperv-tlfs.h>
+#include <asm/reboot.h>
 
 #define __KVM_HAVE_ARCH_VCPU_DEBUGFS
 
@@ -1629,8 +1630,10 @@ struct kvm_x86_ops {
 
 	int (*check_processor_compatibility)(void);
 
-	int (*hardware_enable)(void);
+	int (*enable_virtualization_cpu)(void);
-	void (*hardware_disable)(void);
+	void (*disable_virtualization_cpu)(void);
+	cpu_emergency_virt_cb *emergency_disable_virtualization_cpu;
+
 	void (*hardware_unsetup)(void);
 	bool (*has_emulated_msr)(struct kvm *kvm, u32 index);
 	void (*vcpu_after_set_cpuid)(struct kvm_vcpu *vcpu);

arch/x86/include/asm/reboot.h

@@ -25,8 +25,8 @@ void __noreturn machine_real_restart(unsigned int type);
 #define MRR_BIOS	0
 #define MRR_APM		1
 
-#if IS_ENABLED(CONFIG_KVM_INTEL) || IS_ENABLED(CONFIG_KVM_AMD)
 typedef void (cpu_emergency_virt_cb)(void);
+#if IS_ENABLED(CONFIG_KVM_INTEL) || IS_ENABLED(CONFIG_KVM_AMD)
 void cpu_emergency_register_virt_callback(cpu_emergency_virt_cb *callback);
 void cpu_emergency_unregister_virt_callback(cpu_emergency_virt_cb *callback);
 void cpu_emergency_disable_virtualization(void);

arch/x86/kvm/svm/svm.c

@@ -592,14 +592,14 @@ static inline void kvm_cpu_svm_disable(void)
 	}
 }
 
-static void svm_emergency_disable(void)
+static void svm_emergency_disable_virtualization_cpu(void)
 {
 	kvm_rebooting = true;
 
 	kvm_cpu_svm_disable();
 }
 
-static void svm_hardware_disable(void)
+static void svm_disable_virtualization_cpu(void)
 {
 	/* Make sure we clean up behind us */
 	if (tsc_scaling)
@@ -610,7 +610,7 @@ static void svm_hardware_disable(void)
 	amd_pmu_disable_virt();
 }
 
-static int svm_hardware_enable(void)
+static int svm_enable_virtualization_cpu(void)
 {
 
 	struct svm_cpu_data *sd;
@@ -1533,7 +1533,7 @@ static void svm_prepare_switch_to_guest(struct kvm_vcpu *vcpu)
 	 * TSC_AUX is always virtualized for SEV-ES guests when the feature is
 	 * available. The user return MSR support is not required in this case
 	 * because TSC_AUX is restored on #VMEXIT from the host save area
-	 * (which has been initialized in svm_hardware_enable()).
+	 * (which has been initialized in svm_enable_virtualization_cpu()).
 	 */
 	if (likely(tsc_aux_uret_slot >= 0) &&
 	    (!boot_cpu_has(X86_FEATURE_V_TSC_AUX) || !sev_es_guest(vcpu->kvm)))
@@ -3144,7 +3144,7 @@ static int svm_set_msr(struct kvm_vcpu *vcpu, struct msr_data *msr)
 		 * feature is available. The user return MSR support is not
 		 * required in this case because TSC_AUX is restored on #VMEXIT
 		 * from the host save area (which has been initialized in
-		 * svm_hardware_enable()).
+		 * svm_enable_virtualization_cpu()).
 		 */
 		if (boot_cpu_has(X86_FEATURE_V_TSC_AUX) && sev_es_guest(vcpu->kvm))
 			break;
@@ -4992,8 +4992,9 @@ static struct kvm_x86_ops svm_x86_ops __initdata = {
 	.check_processor_compatibility = svm_check_processor_compat,
 
 	.hardware_unsetup = svm_hardware_unsetup,
-	.hardware_enable = svm_hardware_enable,
+	.enable_virtualization_cpu = svm_enable_virtualization_cpu,
-	.hardware_disable = svm_hardware_disable,
+	.disable_virtualization_cpu = svm_disable_virtualization_cpu,
+	.emergency_disable_virtualization_cpu = svm_emergency_disable_virtualization_cpu,
 	.has_emulated_msr = svm_has_emulated_msr,
 
 	.vcpu_create = svm_vcpu_create,
@@ -5425,8 +5426,6 @@ static struct kvm_x86_init_ops svm_init_ops __initdata = {
 static void __svm_exit(void)
 {
 	kvm_x86_vendor_exit();
-
-	cpu_emergency_unregister_virt_callback(svm_emergency_disable);
 }
 
 static int __init svm_init(void)
@@ -5442,8 +5441,6 @@ static int __init svm_init(void)
 	if (r)
 		return r;
 
-	cpu_emergency_register_virt_callback(svm_emergency_disable);
-
 	/*
 	 * Common KVM initialization _must_ come last, after this, /dev/kvm is
 	 * exposed to userspace!

arch/x86/kvm/vmx/main.c

@@ -23,8 +23,10 @@ struct kvm_x86_ops vt_x86_ops __initdata = {
 
 	.hardware_unsetup = vmx_hardware_unsetup,
 
-	.hardware_enable = vmx_hardware_enable,
+	.enable_virtualization_cpu = vmx_enable_virtualization_cpu,
-	.hardware_disable = vmx_hardware_disable,
+	.disable_virtualization_cpu = vmx_disable_virtualization_cpu,
+	.emergency_disable_virtualization_cpu = vmx_emergency_disable_virtualization_cpu,
+
 	.has_emulated_msr = vmx_has_emulated_msr,
 
 	.vm_size = sizeof(struct kvm_vmx),

arch/x86/kvm/vmx/vmx.c

@@ -755,7 +755,7 @@ fault:
 	return -EIO;
 }
 
-static void vmx_emergency_disable(void)
+void vmx_emergency_disable_virtualization_cpu(void)
 {
 	int cpu = raw_smp_processor_id();
 	struct loaded_vmcs *v;
@@ -2844,7 +2844,7 @@ fault:
 	return -EFAULT;
 }
 
-int vmx_hardware_enable(void)
+int vmx_enable_virtualization_cpu(void)
 {
 	int cpu = raw_smp_processor_id();
 	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
@@ -2881,7 +2881,7 @@ static void vmclear_local_loaded_vmcss(void)
 		__loaded_vmcs_clear(v);
 }
 
-void vmx_hardware_disable(void)
+void vmx_disable_virtualization_cpu(void)
 {
 	vmclear_local_loaded_vmcss();
 
@@ -8584,8 +8584,6 @@ static void __vmx_exit(void)
 {
 	allow_smaller_maxphyaddr = false;
 
-	cpu_emergency_unregister_virt_callback(vmx_emergency_disable);
-
 	vmx_cleanup_l1d_flush();
 }
 
@@ -8632,8 +8630,6 @@ static int __init vmx_init(void)
 		pi_init_cpu(cpu);
 	}
 
-	cpu_emergency_register_virt_callback(vmx_emergency_disable);
-
 	vmx_check_vmcs12_offsets();
 
 	/*

arch/x86/kvm/vmx/x86_ops.h

@@ -13,8 +13,9 @@ extern struct kvm_x86_init_ops vt_init_ops __initdata;
 
 void vmx_hardware_unsetup(void);
 int vmx_check_processor_compat(void);
-int vmx_hardware_enable(void);
+int vmx_enable_virtualization_cpu(void);
-void vmx_hardware_disable(void);
+void vmx_disable_virtualization_cpu(void);
+void vmx_emergency_disable_virtualization_cpu(void);
 int vmx_vm_init(struct kvm *kvm);
 void vmx_vm_destroy(struct kvm *kvm);
 int vmx_vcpu_precreate(struct kvm *kvm);

arch/x86/kvm/x86.c

@@ -355,7 +355,7 @@ static void kvm_on_user_return(struct user_return_notifier *urn)
 
 	/*
 	 * Disabling irqs at this point since the following code could be
-	 * interrupted and executed through kvm_arch_hardware_disable()
+	 * interrupted and executed through kvm_arch_disable_virtualization_cpu()
 	 */
 	local_irq_save(flags);
 	if (msrs->registered) {
@@ -9753,7 +9753,7 @@ int kvm_x86_vendor_init(struct kvm_x86_init_ops *ops)
 
 	guard(mutex)(&vendor_module_lock);
 
-	if (kvm_x86_ops.hardware_enable) {
+	if (kvm_x86_ops.enable_virtualization_cpu) {
 		pr_err("already loaded vendor module '%s'\n", kvm_x86_ops.name);
 		return -EEXIST;
 	}
@@ -9880,7 +9880,7 @@ int kvm_x86_vendor_init(struct kvm_x86_init_ops *ops)
 	return 0;
 
 out_unwind_ops:
-	kvm_x86_ops.hardware_enable = NULL;
+	kvm_x86_ops.enable_virtualization_cpu = NULL;
 	kvm_x86_call(hardware_unsetup)();
 out_mmu_exit:
 	kvm_mmu_vendor_module_exit();
@@ -9921,7 +9921,7 @@ void kvm_x86_vendor_exit(void)
 	WARN_ON(static_branch_unlikely(&kvm_xen_enabled.key));
 #endif
 	mutex_lock(&vendor_module_lock);
-	kvm_x86_ops.hardware_enable = NULL;
+	kvm_x86_ops.enable_virtualization_cpu = NULL;
 	mutex_unlock(&vendor_module_lock);
 }
 EXPORT_SYMBOL_GPL(kvm_x86_vendor_exit);
@@ -12516,7 +12516,17 @@ void kvm_vcpu_deliver_sipi_vector(struct kvm_vcpu *vcpu, u8 vector)
 }
 EXPORT_SYMBOL_GPL(kvm_vcpu_deliver_sipi_vector);
 
-int kvm_arch_hardware_enable(void)
+void kvm_arch_enable_virtualization(void)
+{
+	cpu_emergency_register_virt_callback(kvm_x86_ops.emergency_disable_virtualization_cpu);
+}
+
+void kvm_arch_disable_virtualization(void)
+{
+	cpu_emergency_unregister_virt_callback(kvm_x86_ops.emergency_disable_virtualization_cpu);
+}
+
+int kvm_arch_enable_virtualization_cpu(void)
 {
 	struct kvm *kvm;
 	struct kvm_vcpu *vcpu;
@@ -12532,7 +12542,7 @@ int kvm_arch_hardware_enable(void)
 	if (ret)
 		return ret;
 
-	ret = kvm_x86_call(hardware_enable)();
+	ret = kvm_x86_call(enable_virtualization_cpu)();
 	if (ret != 0)
 		return ret;
 
@@ -12612,9 +12622,9 @@ int kvm_arch_hardware_enable(void)
 	return 0;
 }
 
-void kvm_arch_hardware_disable(void)
+void kvm_arch_disable_virtualization_cpu(void)
 {
-	kvm_x86_call(hardware_disable)();
+	kvm_x86_call(disable_virtualization_cpu)();
 	drop_user_return_notifiers();
 }
 

include/linux/kvm_host.h

@@ -1529,8 +1529,22 @@ static inline void kvm_create_vcpu_debugfs(struct kvm_vcpu *vcpu) {}
 #endif
 
 #ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
-int kvm_arch_hardware_enable(void);
+/*
-void kvm_arch_hardware_disable(void);
+ * kvm_arch_{enable,disable}_virtualization() are called on one CPU, under
+ * kvm_usage_lock, immediately after/before 0=>1 and 1=>0 transitions of
+ * kvm_usage_count, i.e. at the beginning of the generic hardware enabling
+ * sequence, and at the end of the generic hardware disabling sequence.
+ */
+void kvm_arch_enable_virtualization(void);
+void kvm_arch_disable_virtualization(void);
+/*
+ * kvm_arch_{enable,disable}_virtualization_cpu() are called on "every" CPU to
+ * do the actual twiddling of hardware bits.  The hooks are called on all
+ * online CPUs when KVM enables/disabled virtualization, and on a single CPU
+ * when that CPU is onlined/offlined (including for Resume/Suspend).
+ */
+int kvm_arch_enable_virtualization_cpu(void);
+void kvm_arch_disable_virtualization_cpu(void);
 #endif
 int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu);
 bool kvm_arch_vcpu_in_kernel(struct kvm_vcpu *vcpu);

virt/kvm/kvm_main.c

@@ -136,8 +136,8 @@ static int kvm_no_compat_open(struct inode *inode, struct file *file)
 #define KVM_COMPAT(c)	.compat_ioctl	= kvm_no_compat_ioctl,	\
 			.open		= kvm_no_compat_open
 #endif
-static int hardware_enable_all(void);
+static int kvm_enable_virtualization(void);
-static void hardware_disable_all(void);
+static void kvm_disable_virtualization(void);
 
 static void kvm_io_bus_destroy(struct kvm_io_bus *bus);
 
@@ -1220,7 +1220,7 @@ static struct kvm *kvm_create_vm(unsigned long type, const char *fdname)
 	if (r)
 		goto out_err_no_arch_destroy_vm;
 
-	r = hardware_enable_all();
+	r = kvm_enable_virtualization();
 	if (r)
 		goto out_err_no_disable;
 
@@ -1263,7 +1263,7 @@ out_no_coalesced_mmio:
 		mmu_notifier_unregister(&kvm->mmu_notifier, current->mm);
 #endif
 out_err_no_mmu_notifier:
-	hardware_disable_all();
+	kvm_disable_virtualization();
 out_err_no_disable:
 	kvm_arch_destroy_vm(kvm);
 out_err_no_arch_destroy_vm:
@@ -1360,7 +1360,7 @@ static void kvm_destroy_vm(struct kvm *kvm)
 #endif
 	kvm_arch_free_vm(kvm);
 	preempt_notifier_dec();
-	hardware_disable_all();
+	kvm_disable_virtualization();
 	mmdrop(mm);
 }
 
@@ -5571,137 +5571,67 @@ static struct miscdevice kvm_dev = {
 };
 
 #ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
+static bool enable_virt_at_load = true;
+module_param(enable_virt_at_load, bool, 0444);
+
 __visible bool kvm_rebooting;
 EXPORT_SYMBOL_GPL(kvm_rebooting);
 
-static DEFINE_PER_CPU(bool, hardware_enabled);
+static DEFINE_PER_CPU(bool, virtualization_enabled);
+static DEFINE_MUTEX(kvm_usage_lock);
 static int kvm_usage_count;
 
-static int __hardware_enable_nolock(void)
+__weak void kvm_arch_enable_virtualization(void)
 {
-	if (__this_cpu_read(hardware_enabled))
+
+}
+
+__weak void kvm_arch_disable_virtualization(void)
+{
+
+}
+
+static int kvm_enable_virtualization_cpu(void)
+{
+	if (__this_cpu_read(virtualization_enabled))
 		return 0;
 
-	if (kvm_arch_hardware_enable()) {
+	if (kvm_arch_enable_virtualization_cpu()) {
 		pr_info("kvm: enabling virtualization on CPU%d failed\n",
 			raw_smp_processor_id());
 		return -EIO;
 	}
 
-	__this_cpu_write(hardware_enabled, true);
+	__this_cpu_write(virtualization_enabled, true);
 	return 0;
 }
 
-static void hardware_enable_nolock(void *failed)
-{
-	if (__hardware_enable_nolock())
-		atomic_inc(failed);
-}
-
 static int kvm_online_cpu(unsigned int cpu)
 {
-	int ret = 0;
-
 	/*
 	 * Abort the CPU online process if hardware virtualization cannot
 	 * be enabled. Otherwise running VMs would encounter unrecoverable
 	 * errors when scheduled to this CPU.
 	 */
-	mutex_lock(&kvm_lock);
+	return kvm_enable_virtualization_cpu();
-	if (kvm_usage_count)
-		ret = __hardware_enable_nolock();
-	mutex_unlock(&kvm_lock);
-	return ret;
 }
 
-static void hardware_disable_nolock(void *junk)
+static void kvm_disable_virtualization_cpu(void *ign)
 {
-	/*
+	if (!__this_cpu_read(virtualization_enabled))
-	 * Note, hardware_disable_all_nolock() tells all online CPUs to disable
-	 * hardware, not just CPUs that successfully enabled hardware!
-	 */
-	if (!__this_cpu_read(hardware_enabled))
 		return;
 
-	kvm_arch_hardware_disable();
+	kvm_arch_disable_virtualization_cpu();
 
-	__this_cpu_write(hardware_enabled, false);
+	__this_cpu_write(virtualization_enabled, false);
 }
 
 static int kvm_offline_cpu(unsigned int cpu)
 {
-	mutex_lock(&kvm_lock);
+	kvm_disable_virtualization_cpu(NULL);
-	if (kvm_usage_count)
-		hardware_disable_nolock(NULL);
-	mutex_unlock(&kvm_lock);
 	return 0;
 }
 
-static void hardware_disable_all_nolock(void)
-{
-	BUG_ON(!kvm_usage_count);
-
-	kvm_usage_count--;
-	if (!kvm_usage_count)
-		on_each_cpu(hardware_disable_nolock, NULL, 1);
-}
-
-static void hardware_disable_all(void)
-{
-	cpus_read_lock();
-	mutex_lock(&kvm_lock);
-	hardware_disable_all_nolock();
-	mutex_unlock(&kvm_lock);
-	cpus_read_unlock();
-}
-
-static int hardware_enable_all(void)
-{
-	atomic_t failed = ATOMIC_INIT(0);
-	int r;
-
-	/*
-	 * Do not enable hardware virtualization if the system is going down.
-	 * If userspace initiated a forced reboot, e.g. reboot -f, then it's
-	 * possible for an in-flight KVM_CREATE_VM to trigger hardware enabling
-	 * after kvm_reboot() is called.  Note, this relies on system_state
-	 * being set _before_ kvm_reboot(), which is why KVM uses a syscore ops
-	 * hook instead of registering a dedicated reboot notifier (the latter
-	 * runs before system_state is updated).
-	 */
-	if (system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF ||
-	    system_state == SYSTEM_RESTART)
-		return -EBUSY;
-
-	/*
-	 * When onlining a CPU, cpu_online_mask is set before kvm_online_cpu()
-	 * is called, and so on_each_cpu() between them includes the CPU that
-	 * is being onlined.  As a result, hardware_enable_nolock() may get
-	 * invoked before kvm_online_cpu(), which also enables hardware if the
-	 * usage count is non-zero.  Disable CPU hotplug to avoid attempting to
-	 * enable hardware multiple times.
-	 */
-	cpus_read_lock();
-	mutex_lock(&kvm_lock);
-
-	r = 0;
-
-	kvm_usage_count++;
-	if (kvm_usage_count == 1) {
-		on_each_cpu(hardware_enable_nolock, &failed, 1);
-
-		if (atomic_read(&failed)) {
-			hardware_disable_all_nolock();
-			r = -EBUSY;
-		}
-	}
-
-	mutex_unlock(&kvm_lock);
-	cpus_read_unlock();
-
-	return r;
-}
-
 static void kvm_shutdown(void)
 {
 	/*
@@ -5717,34 +5647,32 @@ static void kvm_shutdown(void)
 	 */
 	pr_info("kvm: exiting hardware virtualization\n");
 	kvm_rebooting = true;
-	on_each_cpu(hardware_disable_nolock, NULL, 1);
+	on_each_cpu(kvm_disable_virtualization_cpu, NULL, 1);
 }
 
 static int kvm_suspend(void)
 {
 	/*
 	 * Secondary CPUs and CPU hotplug are disabled across the suspend/resume
-	 * callbacks, i.e. no need to acquire kvm_lock to ensure the usage count
+	 * callbacks, i.e. no need to acquire kvm_usage_lock to ensure the usage
-	 * is stable.  Assert that kvm_lock is not held to ensure the system
+	 * count is stable.  Assert that kvm_usage_lock is not held to ensure
-	 * isn't suspended while KVM is enabling hardware.  Hardware enabling
+	 * the system isn't suspended while KVM is enabling hardware.  Hardware
-	 * can be preempted, but the task cannot be frozen until it has dropped
+	 * enabling can be preempted, but the task cannot be frozen until it has
-	 * all locks (userspace tasks are frozen via a fake signal).
+	 * dropped all locks (userspace tasks are frozen via a fake signal).
 	 */
-	lockdep_assert_not_held(&kvm_lock);
+	lockdep_assert_not_held(&kvm_usage_lock);
 	lockdep_assert_irqs_disabled();
 
-	if (kvm_usage_count)
+	kvm_disable_virtualization_cpu(NULL);
-		hardware_disable_nolock(NULL);
 	return 0;
 }
 
 static void kvm_resume(void)
 {
-	lockdep_assert_not_held(&kvm_lock);
+	lockdep_assert_not_held(&kvm_usage_lock);
 	lockdep_assert_irqs_disabled();
 
-	if (kvm_usage_count)
+	WARN_ON_ONCE(kvm_enable_virtualization_cpu());
-		WARN_ON_ONCE(__hardware_enable_nolock());
 }
 
 static struct syscore_ops kvm_syscore_ops = {
@@ -5752,13 +5680,95 @@ static struct syscore_ops kvm_syscore_ops = {
 	.resume = kvm_resume,
 	.shutdown = kvm_shutdown,
 };
+
+static int kvm_enable_virtualization(void)
+{
+	int r;
+
+	guard(mutex)(&kvm_usage_lock);
+
+	if (kvm_usage_count++)
+		return 0;
+
+	kvm_arch_enable_virtualization();
+
+	r = cpuhp_setup_state(CPUHP_AP_KVM_ONLINE, "kvm/cpu:online",
+			      kvm_online_cpu, kvm_offline_cpu);
+	if (r)
+		goto err_cpuhp;
+
+	register_syscore_ops(&kvm_syscore_ops);
+
+	/*
+	 * Undo virtualization enabling and bail if the system is going down.
+	 * If userspace initiated a forced reboot, e.g. reboot -f, then it's
+	 * possible for an in-flight operation to enable virtualization after
+	 * syscore_shutdown() is called, i.e. without kvm_shutdown() being
+	 * invoked.  Note, this relies on system_state being set _before_
+	 * kvm_shutdown(), e.g. to ensure either kvm_shutdown() is invoked
+	 * or this CPU observes the impending shutdown.  Which is why KVM uses
+	 * a syscore ops hook instead of registering a dedicated reboot
+	 * notifier (the latter runs before system_state is updated).
+	 */
+	if (system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF ||
+	    system_state == SYSTEM_RESTART) {
+		r = -EBUSY;
+		goto err_rebooting;
+	}
+
+	return 0;
+
+err_rebooting:
+	unregister_syscore_ops(&kvm_syscore_ops);
+	cpuhp_remove_state(CPUHP_AP_KVM_ONLINE);
+err_cpuhp:
+	kvm_arch_disable_virtualization();
+	--kvm_usage_count;
+	return r;
+}
+
+static void kvm_disable_virtualization(void)
+{
+	guard(mutex)(&kvm_usage_lock);
+
+	if (--kvm_usage_count)
+		return;
+
+	unregister_syscore_ops(&kvm_syscore_ops);
+	cpuhp_remove_state(CPUHP_AP_KVM_ONLINE);
+	kvm_arch_disable_virtualization();
+}
+
+static int kvm_init_virtualization(void)
+{
+	if (enable_virt_at_load)
+		return kvm_enable_virtualization();
+
+	return 0;
+}
+
+static void kvm_uninit_virtualization(void)
+{
+	if (enable_virt_at_load)
+		kvm_disable_virtualization();
+}
 #else /* CONFIG_KVM_GENERIC_HARDWARE_ENABLING */
-static int hardware_enable_all(void)
+static int kvm_enable_virtualization(void)
 {
 	return 0;
 }
 
-static void hardware_disable_all(void)
+static int kvm_init_virtualization(void)
+{
+	return 0;
+}
+
+static void kvm_disable_virtualization(void)
+{
+
+}
+
+static void kvm_uninit_virtualization(void)
 {
 
 }
@@ -6460,15 +6470,6 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module)
 	int r;
 	int cpu;
 
-#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
-	r = cpuhp_setup_state_nocalls(CPUHP_AP_KVM_ONLINE, "kvm/cpu:online",
-				      kvm_online_cpu, kvm_offline_cpu);
-	if (r)
-		return r;
-
-	register_syscore_ops(&kvm_syscore_ops);
-#endif
-
 	/* A kmem cache lets us meet the alignment requirements of fx_save. */
 	if (!vcpu_align)
 		vcpu_align = __alignof__(struct kvm_vcpu);
@@ -6479,10 +6480,8 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module)
 					   offsetofend(struct kvm_vcpu, stats_id)
 					   - offsetof(struct kvm_vcpu, arch),
 					   NULL);
-	if (!kvm_vcpu_cache) {
+	if (!kvm_vcpu_cache)
-		r = -ENOMEM;
+		return -ENOMEM;
-		goto err_vcpu_cache;
-	}
 
 	for_each_possible_cpu(cpu) {
 		if (!alloc_cpumask_var_node(&per_cpu(cpu_kick_mask, cpu),
@@ -6516,6 +6515,10 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module)
 
 	kvm_gmem_init(module);
 
+	r = kvm_init_virtualization();
+	if (r)
+		goto err_virt;
+
 	/*
 	 * Registration _must_ be the very last thing done, as this exposes
 	 * /dev/kvm to userspace, i.e. all infrastructure must be setup!
@@ -6529,6 +6532,8 @@ int kvm_init(unsigned vcpu_size, unsigned vcpu_align, struct module *module)
 	return 0;
 
 err_register:
+	kvm_uninit_virtualization();
+err_virt:
 	kvm_vfio_ops_exit();
 err_vfio:
 	kvm_async_pf_deinit();
@@ -6539,11 +6544,6 @@ err_cpu_kick_mask:
 	for_each_possible_cpu(cpu)
 		free_cpumask_var(per_cpu(cpu_kick_mask, cpu));
 	kmem_cache_destroy(kvm_vcpu_cache);
-err_vcpu_cache:
-#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
-	unregister_syscore_ops(&kvm_syscore_ops);
-	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_ONLINE);
-#endif
 	return r;
 }
 EXPORT_SYMBOL_GPL(kvm_init);
@@ -6559,16 +6559,14 @@ void kvm_exit(void)
 	 */
 	misc_deregister(&kvm_dev);
 
+	kvm_uninit_virtualization();
+
 	debugfs_remove_recursive(kvm_debugfs_dir);
 	for_each_possible_cpu(cpu)
 		free_cpumask_var(per_cpu(cpu_kick_mask, cpu));
 	kmem_cache_destroy(kvm_vcpu_cache);
 	kvm_vfio_ops_exit();
 	kvm_async_pf_deinit();
-#ifdef CONFIG_KVM_GENERIC_HARDWARE_ENABLING
-	unregister_syscore_ops(&kvm_syscore_ops);
-	cpuhp_remove_state_nocalls(CPUHP_AP_KVM_ONLINE);
-#endif
 	kvm_irqfd_exit();
 }
 EXPORT_SYMBOL_GPL(kvm_exit);