KVM: suspend/resume support

Index: linux-2.6/include/linux/cpu.h
===================================================================
@@ -49,10 +49,20 @@ struct notifier_block;
 
 #ifdef CONFIG_SMP
 /* Need to know about CPUs going up/down? */
-extern int register_cpu_notifier(struct notifier_block *nb);
 #ifdef CONFIG_HOTPLUG_CPU
+extern int register_cpu_notifier(struct notifier_block *nb);
 extern void unregister_cpu_notifier(struct notifier_block *nb);
 #else
+
+#ifndef MODULE
+extern int register_cpu_notifier(struct notifier_block *nb);
+#else
+static inline int register_cpu_notifier(struct notifier_block *nb)
+{
+	return 0;
+}
+#endif
+
 static inline void unregister_cpu_notifier(struct notifier_block *nb)
 {
 }

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
@@ -41,6 +41,9 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+static spinlock_t kvm_lock = SPIN_LOCK_UNLOCKED;
+static struct list_head vm_list = LIST_HEAD_INIT(vm_list);
+
 struct kvm_arch_ops *kvm_arch_ops;
 struct kvm_stat kvm_stat;
 EXPORT_SYMBOL_GPL(kvm_stat);

@@ -230,9 +233,13 @@ static int kvm_dev_open(struct inode *in
 		struct kvm_vcpu *vcpu = &kvm->vcpus[i];
 
 		mutex_init(&vcpu->mutex);
+		vcpu->cpu = -1;
 		vcpu->kvm = kvm;
 		vcpu->mmu.root_hpa = INVALID_PAGE;
 		INIT_LIST_HEAD(&vcpu->free_pages);
+		spin_lock(&kvm_lock);
+		list_add(&kvm->vm_list, &vm_list);
+		spin_unlock(&kvm_lock);
 	}
 	filp->private_data = kvm;
 	return 0;

@@ -292,6 +299,9 @@ static int kvm_dev_release(struct inode 
 {
 	struct kvm *kvm = filp->private_data;
 
+	spin_lock(&kvm_lock);
+	list_del(&kvm->vm_list);
+	spin_unlock(&kvm_lock);
 	kvm_free_vcpus(kvm);
 	kvm_free_physmem(kvm);
 	kfree(kvm);

@@ -546,7 +556,6 @@ static int kvm_dev_ioctl_create_vcpu(str
 					   FX_IMAGE_ALIGN);
 	vcpu->guest_fx_image = vcpu->host_fx_image + FX_IMAGE_SIZE;
 
-	vcpu->cpu = -1;  /* First load will set up TR */
 	r = kvm_arch_ops->vcpu_create(vcpu);
 	if (r < 0)
 		goto out_free_vcpus;

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
@@ -304,6 +304,7 @@ struct kvm {
 	int memory_config_version;
 	int busy;
 	unsigned long rmap_overflow;
+	struct list_head vm_list;
 };
 
 struct kvm_stat {

Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
@@ -125,6 +125,15 @@ static void __vcpu_clear(void *arg)
 		per_cpu(current_vmcs, cpu) = NULL;
 }
 
+static void vcpu_clear(struct kvm_vcpu *vcpu)
+{
+	if (vcpu->cpu != raw_smp_processor_id() && vcpu->cpu != -1)
+		smp_call_function_single(vcpu->cpu, __vcpu_clear, vcpu, 0, 1);
+	else
+		__vcpu_clear(vcpu);
+	vcpu->launched = 0;
+}
+
 static unsigned long vmcs_readl(unsigned long field)
 {
 	unsigned long value;

@@ -202,10 +211,8 @@ static struct kvm_vcpu *vmx_vcpu_load(st
 
 	cpu = get_cpu();
 
-	if (vcpu->cpu != cpu) {
-		smp_call_function(__vcpu_clear, vcpu, 0, 1);
-		vcpu->launched = 0;
-	}
+	if (vcpu->cpu != cpu)
+		vcpu_clear(vcpu);
 
 	if (per_cpu(current_vmcs, cpu) != vcpu->vmcs) {
 		u8 error;

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
@@ -34,6 +34,7 @@
 #include <linux/highmem.h>
 #include <linux/file.h>
 #include <asm/desc.h>
+#include <linux/cpu.h>
 
 #include "x86_emulate.h"
 #include "segment_descriptor.h"

@@ -2038,6 +2039,64 @@ static struct notifier_block kvm_reboot_
 	.priority = 0,
 };
 
+/*
+ * Make sure that a cpu that is being hot-unplugged does not have any vcpus
+ * cached on it.
+ */
+static void decache_vcpus_on_cpu(int cpu)
+{
+	struct kvm *vm;
+	struct kvm_vcpu *vcpu;
+	int i;
+
+	spin_lock(&kvm_lock);
+	list_for_each_entry(vm, &vm_list, vm_list)
+		for (i = 0; i < KVM_MAX_VCPUS; ++i) {
+			vcpu = &vm->vcpus[i];
+			/*
+			 * If the vcpu is locked, then it is running on some
+			 * other cpu and therefore it is not cached on the
+			 * cpu in question.
+			 *
+			 * If it's not locked, check the last cpu it executed
+			 * on.
+			 */
+			if (mutex_trylock(&vcpu->mutex)) {
+				if (vcpu->cpu == cpu) {
+					kvm_arch_ops->vcpu_decache(vcpu);
+					vcpu->cpu = -1;
+				}
+				mutex_unlock(&vcpu->mutex);
+			}
+		}
+	spin_unlock(&kvm_lock);
+}
+
+static int kvm_cpu_hotplug(struct notifier_block *notifier, unsigned long val,
+			   void *v)
+{
+	int cpu = (long)v;
+
+	switch (val) {
+	case CPU_DEAD:
+	case CPU_UP_CANCELED:
+		decache_vcpus_on_cpu(cpu);
+		smp_call_function_single(cpu, kvm_arch_ops->hardware_disable,
+					 NULL, 0, 1);
+		break;
+	case CPU_UP_PREPARE:
+		smp_call_function_single(cpu, kvm_arch_ops->hardware_enable,
+					 NULL, 0, 1);
+		break;
+	}
+	return NOTIFY_OK;
+}
+
+static struct notifier_block kvm_cpu_notifier = {
+	.notifier_call = kvm_cpu_hotplug,
+	.priority = 20, /* must be > scheduler priority */
+};
+
 static __init void kvm_init_debug(void)
 {
 	struct kvm_stats_debugfs_item *p;

@@ -2084,6 +2143,9 @@ int kvm_init_arch(struct kvm_arch_ops *o
 	    return r;
 
 	on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
+	r = register_cpu_notifier(&kvm_cpu_notifier);
+	if (r)
+		goto out_free_1;
 	register_reboot_notifier(&kvm_reboot_notifier);
 
 	kvm_chardev_ops.owner = module;

@@ -2098,6 +2160,8 @@ int kvm_init_arch(struct kvm_arch_ops *o
 
 out_free:
 	unregister_reboot_notifier(&kvm_reboot_notifier);
+	unregister_cpu_notifier(&kvm_cpu_notifier);
+out_free_1:
 	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
 	kvm_arch_ops->hardware_unsetup();
 	return r;

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
@@ -341,6 +341,7 @@ struct kvm_arch_ops {
 
 	struct kvm_vcpu *(*vcpu_load)(struct kvm_vcpu *vcpu);
 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
+	void (*vcpu_decache)(struct kvm_vcpu *vcpu);
 
 	int (*set_guest_debug)(struct kvm_vcpu *vcpu,
 			       struct kvm_debug_guest *dbg);

Index: linux-2.6/drivers/kvm/svm.c
===================================================================
@@ -609,6 +609,10 @@ static void svm_vcpu_put(struct kvm_vcpu
 	put_cpu();
 }
 
+static void svm_vcpu_decache(struct kvm_vcpu *vcpu)
+{
+}
+
 static void svm_cache_regs(struct kvm_vcpu *vcpu)
 {
 	vcpu->regs[VCPU_REGS_RAX] = vcpu->svm->vmcb->save.rax;

@@ -1677,6 +1681,7 @@ static struct kvm_arch_ops svm_arch_ops 
 
 	.vcpu_load = svm_vcpu_load,
 	.vcpu_put = svm_vcpu_put,
+	.vcpu_decache = svm_vcpu_decache,
 
 	.set_guest_debug = svm_guest_debug,
 	.get_msr = svm_get_msr,

Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
@@ -250,6 +250,11 @@ static void vmx_vcpu_put(struct kvm_vcpu
 	put_cpu();
 }
 
+static void vmx_vcpu_decache(struct kvm_vcpu *vcpu)
+{
+	vcpu_clear(vcpu);
+}
+
 static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu)
 {
 	return vmcs_readl(GUEST_RFLAGS);

@@ -509,7 +514,7 @@ static __init int vmx_disabled_by_bios(v
 	return (msr & 5) == 1; /* locked but not enabled */
 }
 
-static __init void hardware_enable(void *garbage)
+static void hardware_enable(void *garbage)
 {
 	int cpu = raw_smp_processor_id();
 	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));

@@ -2021,6 +2026,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 
 	.vcpu_load = vmx_vcpu_load,
 	.vcpu_put = vmx_vcpu_put,
+	.vcpu_decache = vmx_vcpu_decache,
 
 	.set_guest_debug = set_guest_debug,
 	.get_msr = vmx_get_msr,

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
@@ -34,6 +34,7 @@
 #include <linux/highmem.h>
 #include <linux/file.h>
 #include <asm/desc.h>
+#include <linux/sysdev.h>
 #include <linux/cpu.h>
 
 #include "x86_emulate.h"

@@ -2116,6 +2117,30 @@ static void kvm_exit_debug(void)
 	debugfs_remove(debugfs_dir);
 }
 
+static int kvm_suspend(struct sys_device *dev, pm_message_t state)
+{
+	decache_vcpus_on_cpu(raw_smp_processor_id());
+	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+	return 0;
+}
+
+static int kvm_resume(struct sys_device *dev)
+{
+	on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
+	return 0;
+}
+
+static struct sysdev_class kvm_sysdev_class = {
+	set_kset_name("kvm"),
+	.suspend = kvm_suspend,
+	.resume = kvm_resume,
+};
+
+static struct sys_device kvm_sysdev = {
+	.id = 0,
+	.cls = &kvm_sysdev_class,
+};
+
 hpa_t bad_page_address;
 
 int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)

@@ -2148,6 +2173,14 @@ int kvm_init_arch(struct kvm_arch_ops *o
 		goto out_free_1;
 	register_reboot_notifier(&kvm_reboot_notifier);
 
+	r = sysdev_class_register(&kvm_sysdev_class);
+	if (r)
+		goto out_free_2;
+
+	r = sysdev_register(&kvm_sysdev);
+	if (r)
+		goto out_free_3;
+
 	kvm_chardev_ops.owner = module;
 
 	r = misc_register(&kvm_dev);

@@ -2159,6 +2192,10 @@ int kvm_init_arch(struct kvm_arch_ops *o
 	return r;
 
 out_free:
+	sysdev_unregister(&kvm_sysdev);
+out_free_3:
+	sysdev_class_unregister(&kvm_sysdev_class);
+out_free_2:
 	unregister_reboot_notifier(&kvm_reboot_notifier);
 	unregister_cpu_notifier(&kvm_cpu_notifier);
 out_free_1:

@@ -2170,8 +2207,10 @@ out_free_1:
 void kvm_exit_arch(void)
 {
 	misc_deregister(&kvm_dev);
-
+	sysdev_unregister(&kvm_sysdev);
+	sysdev_class_unregister(&kvm_sysdev_class);
 	unregister_reboot_notifier(&kvm_reboot_notifier);
+	unregister_cpu_notifier(&kvm_cpu_notifier);
 	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
 	kvm_arch_ops->hardware_unsetup();
 	kvm_arch_ops = NULL;