Merge branch 'master' into for_paulus
[linux-drm-fsl-dcu.git] / drivers / kvm / kvm_main.c
index 67c1154960f0a308f72e41c7a5e2f54f2afc2894..af866147ff25613ef1e1c4a7e0d0edccca0dc27a 100644 (file)
@@ -34,6 +34,8 @@
 #include <linux/highmem.h>
 #include <linux/file.h>
 #include <asm/desc.h>
+#include <linux/sysdev.h>
+#include <linux/cpu.h>
 
 #include "x86_emulate.h"
 #include "segment_descriptor.h"
@@ -41,6 +43,9 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+static DEFINE_SPINLOCK(kvm_lock);
+static LIST_HEAD(vm_list);
+
 struct kvm_arch_ops *kvm_arch_ops;
 struct kvm_stat kvm_stat;
 EXPORT_SYMBOL_GPL(kvm_stat);
@@ -62,7 +67,7 @@ static struct kvm_stats_debugfs_item {
        { "halt_exits", &kvm_stat.halt_exits },
        { "request_irq", &kvm_stat.request_irq_exits },
        { "irq_exits", &kvm_stat.irq_exits },
-       { 0, 0 }
+       { NULL, NULL }
 };
 
 static struct dentry *debugfs_dir;
@@ -205,7 +210,7 @@ static struct kvm_vcpu *vcpu_load(struct kvm *kvm, int vcpu_slot)
        mutex_lock(&vcpu->mutex);
        if (unlikely(!vcpu->vmcs)) {
                mutex_unlock(&vcpu->mutex);
-               return 0;
+               return NULL;
        }
        return kvm_arch_ops->vcpu_load(vcpu);
 }
@@ -230,9 +235,13 @@ static int kvm_dev_open(struct inode *inode, struct file *filp)
                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;
@@ -257,9 +266,9 @@ static void kvm_free_physmem_slot(struct kvm_memory_slot *free,
        if (!dont || free->dirty_bitmap != dont->dirty_bitmap)
                vfree(free->dirty_bitmap);
 
-       free->phys_mem = 0;
+       free->phys_mem = NULL;
        free->npages = 0;
-       free->dirty_bitmap = 0;
+       free->dirty_bitmap = NULL;
 }
 
 static void kvm_free_physmem(struct kvm *kvm)
@@ -267,12 +276,16 @@ static void kvm_free_physmem(struct kvm *kvm)
        int i;
 
        for (i = 0; i < kvm->nmemslots; ++i)
-               kvm_free_physmem_slot(&kvm->memslots[i], 0);
+               kvm_free_physmem_slot(&kvm->memslots[i], NULL);
 }
 
 static void kvm_free_vcpu(struct kvm_vcpu *vcpu)
 {
+       if (!vcpu_load(vcpu->kvm, vcpu_slot(vcpu)))
+               return;
+
        kvm_mmu_destroy(vcpu);
+       vcpu_put(vcpu);
        kvm_arch_ops->vcpu_free(vcpu);
 }
 
@@ -288,6 +301,9 @@ static int kvm_dev_release(struct inode *inode, struct file *filp)
 {
        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);
@@ -542,7 +558,6 @@ static int kvm_dev_ioctl_create_vcpu(struct kvm *kvm, int n)
                                           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;
@@ -638,11 +653,11 @@ raced:
 
        /* Deallocate if slot is being removed */
        if (!npages)
-               new.phys_mem = 0;
+               new.phys_mem = NULL;
 
        /* Free page dirty bitmap if unneeded */
        if (!(new.flags & KVM_MEM_LOG_DIRTY_PAGES))
-               new.dirty_bitmap = 0;
+               new.dirty_bitmap = NULL;
 
        r = -ENOMEM;
 
@@ -797,14 +812,14 @@ struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn)
                    && gfn < memslot->base_gfn + memslot->npages)
                        return memslot;
        }
-       return 0;
+       return NULL;
 }
 EXPORT_SYMBOL_GPL(gfn_to_memslot);
 
 void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
 {
        int i;
-       struct kvm_memory_slot *memslot = 0;
+       struct kvm_memory_slot *memslot = NULL;
        unsigned long rel_gfn;
 
        for (i = 0; i < kvm->nmemslots; ++i) {
@@ -1224,6 +1239,9 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata)
        case MSR_IA32_APICBASE:
                data = vcpu->apic_base;
                break;
+       case MSR_IA32_MISC_ENABLE:
+               data = vcpu->ia32_misc_enable_msr;
+               break;
 #ifdef CONFIG_X86_64
        case MSR_EFER:
                data = vcpu->shadow_efer;
@@ -1295,6 +1313,9 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data)
        case MSR_IA32_APICBASE:
                vcpu->apic_base = data;
                break;
+       case MSR_IA32_MISC_ENABLE:
+               vcpu->ia32_misc_enable_msr = data;
+               break;
        default:
                printk(KERN_ERR "kvm: unhandled wrmsr: 0x%x\n", msr);
                return 1;
@@ -1352,6 +1373,9 @@ static int kvm_dev_ioctl_run(struct kvm *kvm, struct kvm_run *kvm_run)
        if (!vcpu)
                return -ENOENT;
 
+       /* re-sync apic's tpr */
+       vcpu->cr8 = kvm_run->cr8;
+
        if (kvm_run->emulated) {
                kvm_arch_ops->skip_emulated_instruction(vcpu);
                kvm_run->emulated = 0;
@@ -1598,6 +1622,10 @@ static u32 msrs_to_save[] = {
 
 static unsigned num_msrs_to_save;
 
+static u32 emulated_msrs[] = {
+       MSR_IA32_MISC_ENABLE,
+};
+
 static __init void kvm_init_msr_list(void)
 {
        u32 dummy[2];
@@ -1766,6 +1794,7 @@ static long kvm_dev_ioctl(struct file *filp,
                          unsigned int ioctl, unsigned long arg)
 {
        struct kvm *kvm = filp->private_data;
+       void __user *argp = (void __user *)arg;
        int r = -EINVAL;
 
        switch (ioctl) {
@@ -1782,12 +1811,12 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_run kvm_run;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_run, (void *)arg, sizeof kvm_run))
+               if (copy_from_user(&kvm_run, argp, sizeof kvm_run))
                        goto out;
                r = kvm_dev_ioctl_run(kvm, &kvm_run);
                if (r < 0 &&  r != -EINTR)
                        goto out;
-               if (copy_to_user((void *)arg, &kvm_run, sizeof kvm_run)) {
+               if (copy_to_user(argp, &kvm_run, sizeof kvm_run)) {
                        r = -EFAULT;
                        goto out;
                }
@@ -1797,13 +1826,13 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_regs kvm_regs;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_regs, (void *)arg, sizeof kvm_regs))
+               if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs))
                        goto out;
                r = kvm_dev_ioctl_get_regs(kvm, &kvm_regs);
                if (r)
                        goto out;
                r = -EFAULT;
-               if (copy_to_user((void *)arg, &kvm_regs, sizeof kvm_regs))
+               if (copy_to_user(argp, &kvm_regs, sizeof kvm_regs))
                        goto out;
                r = 0;
                break;
@@ -1812,7 +1841,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_regs kvm_regs;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_regs, (void *)arg, sizeof kvm_regs))
+               if (copy_from_user(&kvm_regs, argp, sizeof kvm_regs))
                        goto out;
                r = kvm_dev_ioctl_set_regs(kvm, &kvm_regs);
                if (r)
@@ -1824,13 +1853,13 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_sregs kvm_sregs;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_sregs, (void *)arg, sizeof kvm_sregs))
+               if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs))
                        goto out;
                r = kvm_dev_ioctl_get_sregs(kvm, &kvm_sregs);
                if (r)
                        goto out;
                r = -EFAULT;
-               if (copy_to_user((void *)arg, &kvm_sregs, sizeof kvm_sregs))
+               if (copy_to_user(argp, &kvm_sregs, sizeof kvm_sregs))
                        goto out;
                r = 0;
                break;
@@ -1839,7 +1868,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_sregs kvm_sregs;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_sregs, (void *)arg, sizeof kvm_sregs))
+               if (copy_from_user(&kvm_sregs, argp, sizeof kvm_sregs))
                        goto out;
                r = kvm_dev_ioctl_set_sregs(kvm, &kvm_sregs);
                if (r)
@@ -1851,13 +1880,13 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_translation tr;
 
                r = -EFAULT;
-               if (copy_from_user(&tr, (void *)arg, sizeof tr))
+               if (copy_from_user(&tr, argp, sizeof tr))
                        goto out;
                r = kvm_dev_ioctl_translate(kvm, &tr);
                if (r)
                        goto out;
                r = -EFAULT;
-               if (copy_to_user((void *)arg, &tr, sizeof tr))
+               if (copy_to_user(argp, &tr, sizeof tr))
                        goto out;
                r = 0;
                break;
@@ -1866,7 +1895,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_interrupt irq;
 
                r = -EFAULT;
-               if (copy_from_user(&irq, (void *)arg, sizeof irq))
+               if (copy_from_user(&irq, argp, sizeof irq))
                        goto out;
                r = kvm_dev_ioctl_interrupt(kvm, &irq);
                if (r)
@@ -1878,7 +1907,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_debug_guest dbg;
 
                r = -EFAULT;
-               if (copy_from_user(&dbg, (void *)arg, sizeof dbg))
+               if (copy_from_user(&dbg, argp, sizeof dbg))
                        goto out;
                r = kvm_dev_ioctl_debug_guest(kvm, &dbg);
                if (r)
@@ -1890,7 +1919,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_memory_region kvm_mem;
 
                r = -EFAULT;
-               if (copy_from_user(&kvm_mem, (void *)arg, sizeof kvm_mem))
+               if (copy_from_user(&kvm_mem, argp, sizeof kvm_mem))
                        goto out;
                r = kvm_dev_ioctl_set_memory_region(kvm, &kvm_mem);
                if (r)
@@ -1901,7 +1930,7 @@ static long kvm_dev_ioctl(struct file *filp,
                struct kvm_dirty_log log;
 
                r = -EFAULT;
-               if (copy_from_user(&log, (void *)arg, sizeof log))
+               if (copy_from_user(&log, argp, sizeof log))
                        goto out;
                r = kvm_dev_ioctl_get_dirty_log(kvm, &log);
                if (r)
@@ -1909,13 +1938,13 @@ static long kvm_dev_ioctl(struct file *filp,
                break;
        }
        case KVM_GET_MSRS:
-               r = msr_io(kvm, (void __user *)arg, get_msr, 1);
+               r = msr_io(kvm, argp, get_msr, 1);
                break;
        case KVM_SET_MSRS:
-               r = msr_io(kvm, (void __user *)arg, do_set_msr, 0);
+               r = msr_io(kvm, argp, do_set_msr, 0);
                break;
        case KVM_GET_MSR_INDEX_LIST: {
-               struct kvm_msr_list __user *user_msr_list = (void __user *)arg;
+               struct kvm_msr_list __user *user_msr_list = argp;
                struct kvm_msr_list msr_list;
                unsigned n;
 
@@ -1923,7 +1952,7 @@ static long kvm_dev_ioctl(struct file *filp,
                if (copy_from_user(&msr_list, user_msr_list, sizeof msr_list))
                        goto out;
                n = msr_list.nmsrs;
-               msr_list.nmsrs = num_msrs_to_save;
+               msr_list.nmsrs = num_msrs_to_save + ARRAY_SIZE(emulated_msrs);
                if (copy_to_user(user_msr_list, &msr_list, sizeof msr_list))
                        goto out;
                r = -E2BIG;
@@ -1933,6 +1962,11 @@ static long kvm_dev_ioctl(struct file *filp,
                if (copy_to_user(user_msr_list->indices, &msrs_to_save,
                                 num_msrs_to_save * sizeof(u32)))
                        goto out;
+               if (copy_to_user(user_msr_list->indices
+                                + num_msrs_to_save * sizeof(u32),
+                                &emulated_msrs,
+                                ARRAY_SIZE(emulated_msrs) * sizeof(u32)))
+                       goto out;
                r = 0;
                break;
        }
@@ -1997,7 +2031,7 @@ static int kvm_reboot(struct notifier_block *notifier, unsigned long val,
                 * in vmx root mode.
                 */
                printk(KERN_INFO "kvm: exiting hardware virtualization\n");
-               on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+               on_each_cpu(kvm_arch_ops->hardware_disable, NULL, 0, 1);
        }
        return NOTIFY_OK;
 }
@@ -2007,11 +2041,69 @@ static struct notifier_block kvm_reboot_notifier = {
        .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;
 
-       debugfs_dir = debugfs_create_dir("kvm", 0);
+       debugfs_dir = debugfs_create_dir("kvm", NULL);
        for (p = debugfs_entries; p->name; ++p)
                p->dentry = debugfs_create_u32(p->name, 0444, debugfs_dir,
                                               p->data);
@@ -2026,6 +2118,30 @@ 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)
@@ -2052,9 +2168,20 @@ int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)
        if (r < 0)
            return r;
 
-       on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
+       on_each_cpu(kvm_arch_ops->hardware_enable, NULL, 0, 1);
+       r = register_cpu_notifier(&kvm_cpu_notifier);
+       if (r)
+               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);
@@ -2066,8 +2193,14 @@ int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)
        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);
-       on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+       unregister_cpu_notifier(&kvm_cpu_notifier);
+out_free_1:
+       on_each_cpu(kvm_arch_ops->hardware_disable, NULL, 0, 1);
        kvm_arch_ops->hardware_unsetup();
        return r;
 }
@@ -2075,9 +2208,11 @@ out_free:
 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);
-       on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+       unregister_cpu_notifier(&kvm_cpu_notifier);
+       on_each_cpu(kvm_arch_ops->hardware_disable, NULL, 0, 1);
        kvm_arch_ops->hardware_unsetup();
        kvm_arch_ops = NULL;
 }