[PATCH 0/38] KVM: Decouple Intel VT implementation from base kvm

[PATCH 0/38] KVM: Decouple Intel VT implementation from base kvm

[Avi Kivity]

This patchset prepares kvm for multiple architecture implementations.  
The kvm.ko module is split into kvm.ko and kvm-intel.ko, and a function 
vector is used to dispatch arch specific operations to the arch module.


-- 
error compiling committee.c: too many arguments to function

[PATCH 1/38] KVM: Create kvm-intel.ko module

[Avi Kivity]

This patch lays the foundation for arch-independent kvm:  it creates the
kvm-intel.ko module and prepares an arch function vector for the arch-specific
implementation.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -235,11 +235,18 @@ struct kvm_stat {
 	u32 irq_exits;
 };
 
+struct kvm_arch_ops {
+};
+
 extern struct kvm_stat kvm_stat;
+extern struct kvm_arch_ops *kvm_arch_ops;
 
 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
 
+void kvm_init_arch(struct kvm_arch_ops *ops);
+void kvm_exit_arch(void);
+
 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
 int kvm_mmu_init(struct kvm_vcpu *vcpu);
 
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -41,6 +41,7 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+struct kvm_arch_ops *kvm_arch_ops;
 struct kvm_stat kvm_stat;
 
 static struct kvm_stats_debugfs_item {
@@ -3611,6 +3612,15 @@ static void kvm_exit_debug(void)
 
 hpa_t bad_page_address;
 
+void kvm_init_arch(struct kvm_arch_ops *ops)
+{
+	kvm_arch_ops = ops;
+}
+
+void kvm_exit_arch(void)
+{
+}
+
 static __init int kvm_init(void)
 {
 	static struct page *bad_page;
@@ -3670,3 +3680,6 @@ static __exit void kvm_exit(void)
 
 module_init(kvm_init)
 module_exit(kvm_exit)
+
+EXPORT_SYMBOL_GPL(kvm_init_arch);
+EXPORT_SYMBOL_GPL(kvm_exit_arch);
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- /dev/null
+++ linux-2.6/drivers/kvm/vmx.c
@@ -0,0 +1,39 @@
+/*
+ * Kernel-based Virtual Machine driver for Linux
+ *
+ * This module enables machines with Intel VT-x extensions to run virtual
+ * machines without emulation or binary translation.
+ *
+ * Copyright (C) 2006 Qumranet, Inc.
+ *
+ * Authors:
+ *   Avi Kivity   <avi@qumranet.com>
+ *   Yaniv Kamay  <yaniv@qumranet.com>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.  See
+ * the COPYING file in the top-level directory.
+ *
+ */
+
+#include "kvm.h"
+#include <linux/module.h>
+
+MODULE_AUTHOR("Qumranet");
+MODULE_LICENSE("GPL");
+
+static struct kvm_arch_ops vmx_arch_ops = {
+};
+
+static int __init vmx_init(void)
+{
+	kvm_init_arch(&vmx_arch_ops);
+	return 0;
+}
+
+static void __exit vmx_exit(void)
+{
+	kvm_exit_arch();
+}
+
+module_init(vmx_init)
+module_exit(vmx_exit)
Index: linux-2.6/drivers/kvm/Kconfig
===================================================================
--- linux-2.6.orig/drivers/kvm/Kconfig
+++ linux-2.6/drivers/kvm/Kconfig
@@ -6,8 +6,9 @@ config KVM
 	depends on X86 && EXPERIMENTAL
 	---help---
 	  Support hosting fully virtualized guest machines using hardware
-	  virtualization extensions.  You will need a fairly recent Intel
-	  processor equipped with VT extensions.
+	  virtualization extensions.  You will need a fairly recent
+	  processor equipped with virtualization extensions. You will also
+	  need to select one or more of the processor modules below.
 
 	  This module provides access to the hardware capabilities through
 	  a character device node named /dev/kvm.
@@ -16,3 +17,10 @@ config KVM
 	  will be called kvm.
 
 	  If unsure, say N.
+
+config KVM_INTEL
+	tristate "KVM for Intel processors support"
+	depends on KVM
+	---help---
+	  Provides support for KVM on Intel processors equipped with the VT
+	  extensions.
Index: linux-2.6/drivers/kvm/Makefile
===================================================================
--- linux-2.6.orig/drivers/kvm/Makefile
+++ linux-2.6/drivers/kvm/Makefile
@@ -4,3 +4,5 @@
 
 kvm-objs := kvm_main.o mmu.o x86_emulate.o
 obj-$(CONFIG_KVM) += kvm.o
+kvm-intel-objs = vmx.o
+obj-$(CONFIG_KVM_INTEL) += kvm-intel.o

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Ingo Molnar]

* Avi Kivity <avi@qumranet.com> wrote:

> --- linux-2.6.orig/drivers/kvm/kvm.h
> +++ linux-2.6/drivers/kvm/kvm.h

please move this from drivers/kvm/ to kernel/kvm/ [or even into a 
toplevel kvm/ directory] - KVM is not a "driver", KVM enhances the core 
Linux kernel with hypervisor functionality.

Guest paravirtualization drivers (once KVM implements them) can go into 
drivers/kvm/.

	Ingo

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Christoph Hellwig]

On Mon, Nov 27, 2006 at 01:36:06PM +0100, Ingo Molnar wrote:
> 
> * Avi Kivity <avi@qumranet.com> wrote:
> 
> > --- linux-2.6.orig/drivers/kvm/kvm.h
> > +++ linux-2.6/drivers/kvm/kvm.h
> 
> please move this from drivers/kvm/ to kernel/kvm/ [or even into a 
> toplevel kvm/ directory] - KVM is not a "driver", KVM enhances the core 
> Linux kernel with hypervisor functionality.

Actually it's exactly a driver.  It's a character driver that exposes
the virtualization features of modern x86 hardware.  Pretty similar to
things like the msr or mtrr driver that expose cpu features as character
drivers aswell.

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Ingo Molnar]

* Christoph Hellwig <hch@infradead.org> wrote:

> > please move this from drivers/kvm/ to kernel/kvm/ [or even into a 
> > toplevel kvm/ directory] - KVM is not a "driver", KVM enhances the 
> > core Linux kernel with hypervisor functionality.
> 
> Actually it's exactly a driver.  It's a character driver that exposes 
> the virtualization features of modern x86 hardware. [...]

you are fundamentally wrong. In the end KVM is a fundamental and complex 
infrastructure that enables Linux to provide full hardware capabilities 
to another OS via the resources of this OS. This concept justifies a 
system call and a place in linux/kernel/. It's not fundamentally limited 
to x86 either. Full virtualization (and paravirtualization) makes sense 
on any platform. And there's no reason KVM be limited to full 
virtualization alone - both paravirtualization and accelerated guest 
drivers need a sane hypercall API.

> [...] Pretty similar to things like the msr or mtrr driver that expose 
> cpu features as character drivers aswell.

you can expose everything as character drivers and ioctls, but that 
doesnt make it the right solution. It might /start out/ as a driver, 
because that's an easy to hack model, but the moment something becomes 
important enough (and virtualization certainly is such a model) it 
demands a system call.

Just like inotify started out as an ioctl hack, but then was 
(rightfully) moved to the system-call space. [ Which btw. was on your 
request ;-) ]

	Ingo

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Andrew Morton]

On Thu, 30 Nov 2006 16:44:25 +0100
Ingo Molnar <mingo@elte.hu> wrote:

> > [...] Pretty similar to things like the msr or mtrr driver that expose 
> > cpu features as character drivers aswell.
> 
> you can expose everything as character drivers and ioctls, but that 
> doesnt make it the right solution. It might /start out/ as a driver, 
> because that's an easy to hack model, but the moment something becomes 
> important enough (and virtualization certainly is such a model) it 
> demands a system call.

Actually fourteen syscalls and counting, and some of those have `mode'
arguments.

It's a fat, complex, presumably arch-specific, presumably frequently-changing
API.  So whatever we do will be unpleasant - that's unavoidable in this case,
I suspect.

(hmm, the interface isn't versioned at present - should it be?)

Maybe, perhaps, one day it _should_ be a syscall API.  But right now if we
did that it would become a versioned syscall API with obsolete slots and
various other warts.

I get the feeling we'd be best off if we were to revisit this in a year or
so.

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Ingo Molnar]

* Andrew Morton <akpm@osdl.org> wrote:

> It's a fat, complex, presumably arch-specific, presumably 
> frequently-changing API.  So whatever we do will be unpleasant - 
> that's unavoidable in this case, I suspect.
> 
> (hmm, the interface isn't versioned at present - should it be?)
> 
> Maybe, perhaps, one day it _should_ be a syscall API.  But right now 
> if we did that it would become a versioned syscall API with obsolete 
> slots and various other warts.

yeah, very much agreed. For example the paravirtualization/accelerator 
downcalls/upcalls in KVM dont exist yet, so there's little to 
standardize. Once we see it from lhype & KVM how these things look like 
we can design a sane kernel interface around it. But i'm against the 
notion that KVM is 'just' a device. It's not, and it /will/ grow into 
something fundamental.

> I get the feeling we'd be best off if we were to revisit this in a 
> year or so.

yeah. I'd suggest merging it as-is into v2.6.20. In a year we'll have 
some real APIs to think about.

	Ingo

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Christoph Hellwig]

On Thu, Nov 30, 2006 at 09:19:35PM +0100, Ingo Molnar wrote:
> > I get the feeling we'd be best off if we were to revisit this in a 
> > year or so.
> 
> yeah. I'd suggest merging it as-is into v2.6.20. In a year we'll have 
> some real APIs to think about.

Agreed.  And because of that I think keeping it in drivers/ for now
makes a lot of sense - it's just a driver we can deprecate if/when things
have evolved into a real infrastructure.

Re: [PATCH 1/38] KVM: Create kvm-intel.ko module

[Ingo Molnar]

* Christoph Hellwig <hch@infradead.org> wrote:

> On Thu, Nov 30, 2006 at 09:19:35PM +0100, Ingo Molnar wrote:
> > > I get the feeling we'd be best off if we were to revisit this in a 
> > > year or so.
> > 
> > yeah. I'd suggest merging it as-is into v2.6.20. In a year we'll 
> > have some real APIs to think about.
> 
> Agreed.  And because of that I think keeping it in drivers/ for now 
> makes a lot of sense - it's just a driver we can deprecate if/when 
> things have evolved into a real infrastructure.

yeah, with that understanding there's zero objections from me.

	Ingo

[PATCH 2/38] KVM: Make /dev/registration happen when the arch specific module is loaded

[Avi Kivity]

This avoids exposing the driver capabilities before they are loaded.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -244,7 +244,7 @@ extern struct kvm_arch_ops *kvm_arch_ops
 #define kvm_printf(kvm, fmt ...) printk(KERN_DEBUG fmt)
 #define vcpu_printf(vcpu, fmt...) kvm_printf(vcpu->kvm, fmt)
 
-void kvm_init_arch(struct kvm_arch_ops *ops);
+int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module);
 void kvm_exit_arch(void);
 
 void kvm_mmu_destroy(struct kvm_vcpu *vcpu);
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -3558,7 +3558,6 @@ static int kvm_dev_mmap(struct file *fil
 }
 
 static struct file_operations kvm_chardev_ops = {
-	.owner		= THIS_MODULE,
 	.open		= kvm_dev_open,
 	.release        = kvm_dev_release,
 	.unlocked_ioctl = kvm_dev_ioctl,
@@ -3612,13 +3611,26 @@ static void kvm_exit_debug(void)
 
 hpa_t bad_page_address;
 
-void kvm_init_arch(struct kvm_arch_ops *ops)
+int kvm_init_arch(struct kvm_arch_ops *ops, struct module *module)
 {
+	int r;
+
 	kvm_arch_ops = ops;
+	kvm_chardev_ops.owner = module;
+
+	r = misc_register(&kvm_dev);
+	if (r) {
+		printk (KERN_ERR "kvm: misc device register failed\n");
+		goto out_free;
+	}
+
+out_free:
+	return r;
 }
 
 void kvm_exit_arch(void)
 {
+	misc_deregister(&kvm_dev);
 }
 
 static __init int kvm_init(void)
@@ -3644,13 +3656,6 @@ static __init int kvm_init(void)
 	on_each_cpu(kvm_enable, 0, 0, 1);
 	register_reboot_notifier(&kvm_reboot_notifier);
 
-	r = misc_register(&kvm_dev);
-	if (r) {
-		printk (KERN_ERR "kvm: misc device register failed\n");
-		goto out_free;
-	}
-
-
 	if ((bad_page = alloc_page(GFP_KERNEL)) == NULL) {
 		r = -ENOMEM;
 		goto out_free;
@@ -3671,7 +3676,6 @@ out:
 static __exit void kvm_exit(void)
 {
 	kvm_exit_debug();
-	misc_deregister(&kvm_dev);
 	unregister_reboot_notifier(&kvm_reboot_notifier);
 	on_each_cpu(kvm_disable, 0, 0, 1);
 	free_kvm_area();
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -26,7 +26,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 
 static int __init vmx_init(void)
 {
-	kvm_init_arch(&vmx_arch_ops);
+	kvm_init_arch(&vmx_arch_ops, THIS_MODULE);
 	return 0;
 }
 

Re: [PATCH 0/38] KVM: Decouple Intel VT implementation from base kvm

[Avi Kivity]

Avi Kivity wrote:
> This patchset prepares kvm for multiple architecture implementations.  
> The kvm.ko module is split into kvm.ko and kvm-intel.ko, and a 
> function vector is used to dispatch arch specific operations to the 
> arch module.
>

It should be noted that the patchset is bisect friendly (and will even 
boot a guest for you at any point in the set).

-- 
error compiling committee.c: too many arguments to function

[PATCH 3/38] KV: Make hardware detection an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -236,6 +236,8 @@ struct kvm_stat {
 };
 
 struct kvm_arch_ops {
+	int (*cpu_has_kvm_support)(void);          /* __init */
+	int (*disabled_by_bios)(void);             /* __init */
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -527,12 +527,6 @@ static void free_vmcs(struct vmcs *vmcs)
 	free_pages((unsigned long)vmcs, vmcs_descriptor.order);
 }
 
-static __init int cpu_has_kvm_support(void)
-{
-	unsigned long ecx = cpuid_ecx(1);
-	return test_bit(5, &ecx); /* CPUID.1:ECX.VMX[bit 5] -> VT */
-}
-
 static __exit void free_kvm_area(void)
 {
 	int cpu;
@@ -559,14 +553,6 @@ static __init int alloc_kvm_area(void)
 	return 0;
 }
 
-static __init int vmx_disabled_by_bios(void)
-{
-	u64 msr;
-
-	rdmsrl(MSR_IA32_FEATURE_CONTROL, msr);
-	return (msr & 5) == 1; /* locked but not enabled */
-}
-
 static __init void kvm_enable(void *garbage)
 {
 	int cpu = raw_smp_processor_id();
@@ -3616,6 +3602,16 @@ int kvm_init_arch(struct kvm_arch_ops *o
 	int r;
 
 	kvm_arch_ops = ops;
+
+	if (!kvm_arch_ops->cpu_has_kvm_support()) {
+		printk(KERN_ERR "kvm: no hardware support\n");
+		return -EOPNOTSUPP;
+	}
+	if (kvm_arch_ops->disabled_by_bios()) {
+		printk(KERN_ERR "kvm: disabled by bios\n");
+		return -EOPNOTSUPP;
+	}
+
 	kvm_chardev_ops.owner = module;
 
 	r = misc_register(&kvm_dev);
@@ -3638,15 +3634,6 @@ static __init int kvm_init(void)
 	static struct page *bad_page;
 	int r = 0;
 
-	if (!cpu_has_kvm_support()) {
-		printk(KERN_ERR "kvm: no hardware support\n");
-		return -EOPNOTSUPP;
-	}
-	if (vmx_disabled_by_bios()) {
-		printk(KERN_ERR "kvm: disabled by bios\n");
-		return -EOPNOTSUPP;
-	}
-
 	kvm_init_debug();
 
 	setup_vmcs_descriptor();
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -18,10 +18,28 @@
 #include "kvm.h"
 #include <linux/module.h>
 
+#define MSR_IA32_FEATURE_CONTROL 		0x03a
+
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+static __init int cpu_has_kvm_support(void)
+{
+	unsigned long ecx = cpuid_ecx(1);
+	return test_bit(5, &ecx); /* CPUID.1:ECX.VMX[bit 5] -> VT */
+}
+
+static __init int vmx_disabled_by_bios(void)
+{
+	u64 msr;
+
+	rdmsrl(MSR_IA32_FEATURE_CONTROL, msr);
+	return (msr & 5) == 1; /* locked but not enabled */
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
+	.cpu_has_kvm_support = cpu_has_kvm_support,
+	.disabled_by_bios = vmx_disabled_by_bios,
 };
 
 static int __init vmx_init(void)

[PATCH 4/38] KVM: Make the per-cpu enable/disable functions arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -238,6 +238,8 @@ struct kvm_stat {
 struct kvm_arch_ops {
 	int (*cpu_has_kvm_support)(void);          /* __init */
 	int (*disabled_by_bios)(void);             /* __init */
+	void (*hardware_enable)(void *dummy);      /* __init */
+	void (*hardware_disable)(void *dummy);
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -135,7 +135,6 @@ static const u32 vmx_msr_index[] = {
 #define CR0_RESEVED_BITS 0xffffffff1ffaffc0ULL
 #define LMSW_GUEST_MASK 0x0eULL
 #define CR4_RESEVED_BITS (~((1ULL << 11) - 1))
-#define CR4_VMXE 0x2000
 #define CR8_RESEVED_BITS (~0x0fULL)
 #define EFER_RESERVED_BITS 0xfffffffffffff2fe
 
@@ -300,7 +299,8 @@ static void reload_tss(void)
 #endif
 }
 
-static DEFINE_PER_CPU(struct vmcs *, vmxarea);
+DEFINE_PER_CPU(struct vmcs *, vmxarea);
+EXPORT_SYMBOL_GPL(per_cpu__vmxarea); /* temporary hack */
 static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
 
 static struct vmcs_descriptor {
@@ -553,26 +553,6 @@ static __init int alloc_kvm_area(void)
 	return 0;
 }
 
-static __init void kvm_enable(void *garbage)
-{
-	int cpu = raw_smp_processor_id();
-	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
-	u64 old;
-
-	rdmsrl(MSR_IA32_FEATURE_CONTROL, old);
-	if ((old & 5) == 0)
-		/* enable and lock */
-		wrmsrl(MSR_IA32_FEATURE_CONTROL, old | 5);
-	write_cr4(read_cr4() | CR4_VMXE); /* FIXME: not cpu hotplug safe */
-	asm volatile (ASM_VMX_VMXON_RAX : : "a"(&phys_addr), "m"(phys_addr)
-		      : "memory", "cc");
-}
-
-static void kvm_disable(void *garbage)
-{
-	asm volatile (ASM_VMX_VMXOFF : : : "cc");
-}
-
 static int kvm_dev_open(struct inode *inode, struct file *filp)
 {
 	struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
@@ -3565,8 +3545,8 @@ static int kvm_reboot(struct notifier_bl
 		 * Some (well, at least mine) BIOSes hang on reboot if
 		 * in vmx root mode.
 		 */
-		printk(KERN_INFO "kvm: exiting vmx mode\n");
-		on_each_cpu(kvm_disable, 0, 0, 1);
+		printk(KERN_INFO "kvm: exiting hardware virtualization\n");
+		on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
 	}
 	return NOTIFY_OK;
 }
@@ -3612,6 +3592,9 @@ int kvm_init_arch(struct kvm_arch_ops *o
 		return -EOPNOTSUPP;
 	}
 
+	on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
+	register_reboot_notifier(&kvm_reboot_notifier);
+
 	kvm_chardev_ops.owner = module;
 
 	r = misc_register(&kvm_dev);
@@ -3627,6 +3610,9 @@ out_free:
 void kvm_exit_arch(void)
 {
 	misc_deregister(&kvm_dev);
+
+	unregister_reboot_notifier(&kvm_reboot_notifier);
+	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
 }
 
 static __init int kvm_init(void)
@@ -3640,8 +3626,6 @@ static __init int kvm_init(void)
 	r = alloc_kvm_area();
 	if (r)
 		goto out;
-	on_each_cpu(kvm_enable, 0, 0, 1);
-	register_reboot_notifier(&kvm_reboot_notifier);
 
 	if ((bad_page = alloc_page(GFP_KERNEL)) == NULL) {
 		r = -ENOMEM;
@@ -3663,8 +3647,6 @@ out:
 static __exit void kvm_exit(void)
 {
 	kvm_exit_debug();
-	unregister_reboot_notifier(&kvm_reboot_notifier);
-	on_each_cpu(kvm_disable, 0, 0, 1);
 	free_kvm_area();
 	__free_page(pfn_to_page(bad_page_address >> PAGE_SHIFT));
 }
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -23,6 +23,8 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
+DECLARE_PER_CPU(struct vmcs *, vmxarea);
+
 static __init int cpu_has_kvm_support(void)
 {
 	unsigned long ecx = cpuid_ecx(1);
@@ -37,9 +39,31 @@ static __init int vmx_disabled_by_bios(v
 	return (msr & 5) == 1; /* locked but not enabled */
 }
 
+static __init void hardware_enable(void *garbage)
+{
+	int cpu = raw_smp_processor_id();
+	u64 phys_addr = __pa(per_cpu(vmxarea, cpu));
+	u64 old;
+
+	rdmsrl(MSR_IA32_FEATURE_CONTROL, old);
+	if ((old & 5) == 0)
+		/* enable and lock */
+		wrmsrl(MSR_IA32_FEATURE_CONTROL, old | 5);
+	write_cr4(read_cr4() | CR4_VMXE); /* FIXME: not cpu hotplug safe */
+	asm volatile (ASM_VMX_VMXON_RAX : : "a"(&phys_addr), "m"(phys_addr)
+		      : "memory", "cc");
+}
+
+static void hardware_disable(void *garbage)
+{
+	asm volatile (ASM_VMX_VMXOFF : : : "cc");
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
+	.hardware_enable = hardware_enable,
+	.hardware_disable = hardware_disable,
 };
 
 static int __init vmx_init(void)
Index: linux-2.6/drivers/kvm/vmx.h
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.h
+++ linux-2.6/drivers/kvm/vmx.h
@@ -284,4 +284,6 @@ enum vmcs_field {
 
 #define AR_RESERVD_MASK 0xfffe0f00
 
+#define CR4_VMXE 0x2000
+
 #endif

[PATCH 5/38] KVM: Make the hardware setup operations (non-percpu) arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -240,6 +240,8 @@ struct kvm_arch_ops {
 	int (*disabled_by_bios)(void);             /* __init */
 	void (*hardware_enable)(void *dummy);      /* __init */
 	void (*hardware_disable)(void *dummy);
+	int (*hardware_setup)(void);               /* __init */
+	void (*hardware_unsetup)(void);            /* __exit */
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -124,7 +124,6 @@ static const u32 vmx_msr_index[] = {
 
 #define MSR_IA32_TIME_STAMP_COUNTER		0x010
 #define MSR_IA32_FEATURE_CONTROL 		0x03a
-#define MSR_IA32_VMX_BASIC_MSR   		0x480
 #define MSR_IA32_VMX_PINBASED_CTLS_MSR		0x481
 #define MSR_IA32_VMX_PROCBASED_CTLS_MSR		0x482
 #define MSR_IA32_VMX_EXIT_CTLS_MSR		0x483
@@ -303,11 +302,12 @@ DEFINE_PER_CPU(struct vmcs *, vmxarea);
 EXPORT_SYMBOL_GPL(per_cpu__vmxarea); /* temporary hack */
 static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
 
-static struct vmcs_descriptor {
+struct vmcs_descriptor {
 	int size;
 	int order;
 	u32 revision_id;
 } vmcs_descriptor;
+EXPORT_SYMBOL_GPL(vmcs_descriptor);
 
 #ifdef __x86_64__
 static unsigned long read_msr(unsigned long msr)
@@ -394,16 +394,6 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 	return req_size - size;
 }
 
-static __init void setup_vmcs_descriptor(void)
-{
-	u32 vmx_msr_low, vmx_msr_high;
-
-	rdmsr(MSR_IA32_VMX_BASIC_MSR, vmx_msr_low, vmx_msr_high);
-	vmcs_descriptor.size = vmx_msr_high & 0x1fff;
-	vmcs_descriptor.order = get_order(vmcs_descriptor.size);
-	vmcs_descriptor.revision_id = vmx_msr_low;
-};
-
 static void vmcs_clear(struct vmcs *vmcs)
 {
 	u64 phys_addr = __pa(vmcs);
@@ -501,8 +491,7 @@ static void vcpu_put(struct kvm_vcpu *vc
 	mutex_unlock(&vcpu->mutex);
 }
 
-
-static struct vmcs *alloc_vmcs_cpu(int cpu)
+struct vmcs *alloc_vmcs_cpu(int cpu)
 {
 	int node = cpu_to_node(cpu);
 	struct page *pages;
@@ -516,42 +505,18 @@ static struct vmcs *alloc_vmcs_cpu(int c
 	vmcs->revision_id = vmcs_descriptor.revision_id; /* vmcs revision id */
 	return vmcs;
 }
+EXPORT_SYMBOL_GPL(alloc_vmcs_cpu);
 
 static struct vmcs *alloc_vmcs(void)
 {
 	return alloc_vmcs_cpu(smp_processor_id());
 }
 
-static void free_vmcs(struct vmcs *vmcs)
+void free_vmcs(struct vmcs *vmcs)
 {
 	free_pages((unsigned long)vmcs, vmcs_descriptor.order);
 }
-
-static __exit void free_kvm_area(void)
-{
-	int cpu;
-
-	for_each_online_cpu(cpu)
-		free_vmcs(per_cpu(vmxarea, cpu));
-}
-
-static __init int alloc_kvm_area(void)
-{
-	int cpu;
-
-	for_each_online_cpu(cpu) {
-		struct vmcs *vmcs;
-
-		vmcs = alloc_vmcs_cpu(cpu);
-		if (!vmcs) {
-			free_kvm_area();
-			return -ENOMEM;
-		}
-
-		per_cpu(vmxarea, cpu) = vmcs;
-	}
-	return 0;
-}
+EXPORT_SYMBOL_GPL(free_vmcs);
 
 static int kvm_dev_open(struct inode *inode, struct file *filp)
 {
@@ -3592,6 +3557,10 @@ int kvm_init_arch(struct kvm_arch_ops *o
 		return -EOPNOTSUPP;
 	}
 
+	r = kvm_arch_ops->hardware_setup();
+	if (r < 0)
+	    return r;
+
 	on_each_cpu(kvm_arch_ops->hardware_enable, 0, 0, 1);
 	register_reboot_notifier(&kvm_reboot_notifier);
 
@@ -3603,7 +3572,12 @@ int kvm_init_arch(struct kvm_arch_ops *o
 		goto out_free;
 	}
 
+	return r;
+
 out_free:
+	unregister_reboot_notifier(&kvm_reboot_notifier);
+	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+	kvm_arch_ops->hardware_unsetup();
 	return r;
 }
 
@@ -3613,6 +3587,7 @@ void kvm_exit_arch(void)
 
 	unregister_reboot_notifier(&kvm_reboot_notifier);
 	on_each_cpu(kvm_arch_ops->hardware_disable, 0, 0, 1);
+	kvm_arch_ops->hardware_unsetup();
 }
 
 static __init int kvm_init(void)
@@ -3622,14 +3597,9 @@ static __init int kvm_init(void)
 
 	kvm_init_debug();
 
-	setup_vmcs_descriptor();
-	r = alloc_kvm_area();
-	if (r)
-		goto out;
-
 	if ((bad_page = alloc_page(GFP_KERNEL)) == NULL) {
 		r = -ENOMEM;
-		goto out_free;
+		goto out;
 	}
 
 	bad_page_address = page_to_pfn(bad_page) << PAGE_SHIFT;
@@ -3637,8 +3607,6 @@ static __init int kvm_init(void)
 
 	return r;
 
-out_free:
-	free_kvm_area();
 out:
 	kvm_exit_debug();
 	return r;
@@ -3647,7 +3615,6 @@ out:
 static __exit void kvm_exit(void)
 {
 	kvm_exit_debug();
-	free_kvm_area();
 	__free_page(pfn_to_page(bad_page_address >> PAGE_SHIFT));
 }
 
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -25,6 +25,12 @@ MODULE_LICENSE("GPL");
 
 DECLARE_PER_CPU(struct vmcs *, vmxarea);
 
+extern struct vmcs_descriptor {
+	int size;
+	int order;
+	u32 revision_id;
+} vmcs_descriptor;
+
 static __init int cpu_has_kvm_support(void)
 {
 	unsigned long ecx = cpuid_ecx(1);
@@ -59,9 +65,62 @@ static void hardware_disable(void *garba
 	asm volatile (ASM_VMX_VMXOFF : : : "cc");
 }
 
+static __init void setup_vmcs_descriptor(void)
+{
+	u32 vmx_msr_low, vmx_msr_high;
+
+	rdmsr(MSR_IA32_VMX_BASIC_MSR, vmx_msr_low, vmx_msr_high);
+	vmcs_descriptor.size = vmx_msr_high & 0x1fff;
+	vmcs_descriptor.order = get_order(vmcs_descriptor.size);
+	vmcs_descriptor.revision_id = vmx_msr_low;
+};
+
+void free_vmcs(struct vmcs *vmcs);
+
+static __exit void free_kvm_area(void)
+{
+	int cpu;
+
+	for_each_online_cpu(cpu)
+		free_vmcs(per_cpu(vmxarea, cpu));
+}
+
+extern struct vmcs *alloc_vmcs_cpu(int cpu);
+
+static __init int alloc_kvm_area(void)
+{
+	int cpu;
+
+	for_each_online_cpu(cpu) {
+		struct vmcs *vmcs;
+
+		vmcs = alloc_vmcs_cpu(cpu);
+		if (!vmcs) {
+			free_kvm_area();
+			return -ENOMEM;
+		}
+
+		per_cpu(vmxarea, cpu) = vmcs;
+	}
+	return 0;
+}
+
+static __init int hardware_setup(void)
+{
+	setup_vmcs_descriptor();
+	return alloc_kvm_area();
+}
+
+static __exit void hardware_unsetup(void)
+{
+	free_kvm_area();
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
+	.hardware_setup = hardware_setup,
+	.hardware_unsetup = hardware_unsetup,
 	.hardware_enable = hardware_enable,
 	.hardware_disable = hardware_disable,
 };
Index: linux-2.6/drivers/kvm/vmx.h
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.h
+++ linux-2.6/drivers/kvm/vmx.h
@@ -286,4 +286,6 @@ enum vmcs_field {
 
 #define CR4_VMXE 0x2000
 
+#define MSR_IA32_VMX_BASIC_MSR   		0x480
+
 #endif

[PATCH 6/38] KVM: Make the guest debugger an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -242,6 +242,9 @@ struct kvm_arch_ops {
 	void (*hardware_disable)(void *dummy);
 	int (*hardware_setup)(void);               /* __init */
 	void (*hardware_unsetup)(void);            /* __exit */
+
+	int (*set_guest_debug)(struct kvm_vcpu *vcpu,
+			       struct kvm_debug_guest *dbg);
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -611,6 +611,7 @@ unsigned long vmcs_readl(unsigned long f
 		      : "=a"(value) : "d"(field) : "cc");
 	return value;
 }
+EXPORT_SYMBOL_GPL(vmcs_readl);
 
 void vmcs_writel(unsigned long field, unsigned long value)
 {
@@ -622,6 +623,7 @@ void vmcs_writel(unsigned long field, un
 		printk(KERN_ERR "vmwrite error: reg %lx value %lx (err %d)\n",
 		       field, value, vmcs_read32(VM_INSTRUCTION_ERROR));
 }
+EXPORT_SYMBOL_GPL(vmcs_writel);
 
 static void vmcs_write16(unsigned long field, u16 value)
 {
@@ -3230,9 +3232,7 @@ static int kvm_dev_ioctl_debug_guest(str
 				     struct kvm_debug_guest *dbg)
 {
 	struct kvm_vcpu *vcpu;
-	unsigned long dr7 = 0x400;
-	u32 exception_bitmap;
-	int old_singlestep;
+	int r;
 
 	if (dbg->vcpu < 0 || dbg->vcpu >= KVM_MAX_VCPUS)
 		return -EINVAL;
@@ -3240,44 +3240,11 @@ static int kvm_dev_ioctl_debug_guest(str
 	if (!vcpu)
 		return -ENOENT;
 
-	exception_bitmap = vmcs_read32(EXCEPTION_BITMAP);
-	old_singlestep = vcpu->guest_debug.singlestep;
-
-	vcpu->guest_debug.enabled = dbg->enabled;
-	if (vcpu->guest_debug.enabled) {
-		int i;
-
-		dr7 |= 0x200;  /* exact */
-		for (i = 0; i < 4; ++i) {
-			if (!dbg->breakpoints[i].enabled)
-				continue;
-			vcpu->guest_debug.bp[i] = dbg->breakpoints[i].address;
-			dr7 |= 2 << (i*2);    /* global enable */
-			dr7 |= 0 << (i*4+16); /* execution breakpoint */
-		}
-
-		exception_bitmap |= (1u << 1);  /* Trap debug exceptions */
-
-		vcpu->guest_debug.singlestep = dbg->singlestep;
-	} else {
-		exception_bitmap &= ~(1u << 1); /* Ignore debug exceptions */
-		vcpu->guest_debug.singlestep = 0;
-	}
-
-	if (old_singlestep && !vcpu->guest_debug.singlestep) {
-		unsigned long flags;
-
-		flags = vmcs_readl(GUEST_RFLAGS);
-		flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_RF);
-		vmcs_writel(GUEST_RFLAGS, flags);
-	}
-
-	vmcs_write32(EXCEPTION_BITMAP, exception_bitmap);
-	vmcs_writel(GUEST_DR7, dr7);
+	r = kvm_arch_ops->set_guest_debug(vcpu, dbg);
 
 	vcpu_put(vcpu);
 
-	return 0;
+	return r;
 }
 
 static long kvm_dev_ioctl(struct file *filp,
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -31,6 +31,50 @@ extern struct vmcs_descriptor {
 	u32 revision_id;
 } vmcs_descriptor;
 
+static int set_guest_debug(struct kvm_vcpu *vcpu, struct kvm_debug_guest *dbg)
+{
+	unsigned long dr7 = 0x400;
+	u32 exception_bitmap;
+	int old_singlestep;
+
+	exception_bitmap = vmcs_read32(EXCEPTION_BITMAP);
+	old_singlestep = vcpu->guest_debug.singlestep;
+
+	vcpu->guest_debug.enabled = dbg->enabled;
+	if (vcpu->guest_debug.enabled) {
+		int i;
+
+		dr7 |= 0x200;  /* exact */
+		for (i = 0; i < 4; ++i) {
+			if (!dbg->breakpoints[i].enabled)
+				continue;
+			vcpu->guest_debug.bp[i] = dbg->breakpoints[i].address;
+			dr7 |= 2 << (i*2);    /* global enable */
+			dr7 |= 0 << (i*4+16); /* execution breakpoint */
+		}
+
+		exception_bitmap |= (1u << 1);  /* Trap debug exceptions */
+
+		vcpu->guest_debug.singlestep = dbg->singlestep;
+	} else {
+		exception_bitmap &= ~(1u << 1); /* Ignore debug exceptions */
+		vcpu->guest_debug.singlestep = 0;
+	}
+
+	if (old_singlestep && !vcpu->guest_debug.singlestep) {
+		unsigned long flags;
+
+		flags = vmcs_readl(GUEST_RFLAGS);
+		flags &= ~(X86_EFLAGS_TF | X86_EFLAGS_RF);
+		vmcs_writel(GUEST_RFLAGS, flags);
+	}
+
+	vmcs_write32(EXCEPTION_BITMAP, exception_bitmap);
+	vmcs_writel(GUEST_DR7, dr7);
+
+	return 0;
+}
+
 static __init int cpu_has_kvm_support(void)
 {
 	unsigned long ecx = cpuid_ecx(1);
@@ -123,6 +167,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.hardware_unsetup = hardware_unsetup,
 	.hardware_enable = hardware_enable,
 	.hardware_disable = hardware_disable,
+
+	.set_guest_debug = set_guest_debug,
 };
 
 static int __init vmx_init(void)

[PATCH 7/38] KVM: Make msr accessors arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -245,6 +245,8 @@ struct kvm_arch_ops {
 
 	int (*set_guest_debug)(struct kvm_vcpu *vcpu,
 			       struct kvm_debug_guest *dbg);
+	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
+	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
 };
 
 extern struct kvm_stat kvm_stat;
@@ -401,6 +403,8 @@ static inline struct kvm_mmu_page *page_
 #define ASM_VMX_VMXOFF            ".byte 0x0f, 0x01, 0xc4"
 #define ASM_VMX_VMXON_RAX         ".byte 0xf3, 0x0f, 0xc7, 0x30"
 
+#define MSR_IA32_TIME_STAMP_COUNTER		0x010
+
 #ifdef __x86_64__
 
 /*
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -122,7 +122,6 @@ static const u32 vmx_msr_index[] = {
 #define TSS_REDIRECTION_SIZE (256 / 8)
 #define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
 
-#define MSR_IA32_TIME_STAMP_COUNTER		0x010
 #define MSR_IA32_FEATURE_CONTROL 		0x03a
 #define MSR_IA32_VMX_PINBASED_CTLS_MSR		0x481
 #define MSR_IA32_VMX_PROCBASED_CTLS_MSR		0x482
@@ -143,7 +142,7 @@ static const u32 vmx_msr_index[] = {
 #define HOST_IS_64 0
 #endif
 
-static struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr)
+struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr)
 {
 	int i;
 
@@ -152,6 +151,7 @@ static struct vmx_msr_entry *find_msr_en
 			return &vcpu->guest_msrs[i];
 	return 0;
 }
+EXPORT_SYMBOL_GPL(find_msr_entry);
 
 struct descriptor_table {
 	u16 limit;
@@ -657,7 +657,7 @@ static void inject_gp(struct kvm_vcpu *v
  * reads and returns guest's timestamp counter "register"
  * guest_tsc = host_tsc + tsc_offset    -- 21.3
  */
-static u64 guest_read_tsc(void)
+u64 guest_read_tsc(void)
 {
 	u64 host_tsc, tsc_offset;
 
@@ -665,18 +665,20 @@ static u64 guest_read_tsc(void)
 	tsc_offset = vmcs_read64(TSC_OFFSET);
 	return host_tsc + tsc_offset;
 }
+EXPORT_SYMBOL_GPL(guest_read_tsc);
 
 /*
  * writes 'guest_tsc' into guest's timestamp counter "register"
  * guest_tsc = host_tsc + tsc_offset ==> tsc_offset = guest_tsc - host_tsc
  */
-static void guest_write_tsc(u64 guest_tsc)
+void guest_write_tsc(u64 guest_tsc)
 {
 	u64 host_tsc;
 
 	rdtscll(host_tsc);
 	vmcs_write64(TSC_OFFSET, guest_tsc - host_tsc);
 }
+EXPORT_SYMBOL_GPL(guest_write_tsc);
 
 static void update_exception_bitmap(struct kvm_vcpu *vcpu)
 {
@@ -2242,67 +2244,7 @@ static int handle_cpuid(struct kvm_vcpu 
  */
 static int get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
 {
-	u64 data;
-	struct vmx_msr_entry *msr;
-
-	if (!pdata) {
-		printk(KERN_ERR "BUG: get_msr called with NULL pdata\n");
-		return -EINVAL;
-	}
-
-	switch (msr_index) {
-#ifdef __x86_64__
-	case MSR_FS_BASE:
-		data = vmcs_readl(GUEST_FS_BASE);
-		break;
-	case MSR_GS_BASE:
-		data = vmcs_readl(GUEST_GS_BASE);
-		break;
-	case MSR_EFER:
-		data = vcpu->shadow_efer;
-		break;
-#endif
-	case MSR_IA32_TIME_STAMP_COUNTER:
-		data = guest_read_tsc();
-		break;
-	case MSR_IA32_SYSENTER_CS:
-		data = vmcs_read32(GUEST_SYSENTER_CS);
-		break;
-	case MSR_IA32_SYSENTER_EIP:
-		data = vmcs_read32(GUEST_SYSENTER_EIP);
-		break;
-	case MSR_IA32_SYSENTER_ESP:
-		data = vmcs_read32(GUEST_SYSENTER_ESP);
-		break;
-	case MSR_IA32_MC0_CTL:
-	case MSR_IA32_MCG_STATUS:
-	case MSR_IA32_MCG_CAP:
-	case MSR_IA32_MC0_MISC:
-	case MSR_IA32_MC0_MISC+4:
-	case MSR_IA32_MC0_MISC+8:
-	case MSR_IA32_MC0_MISC+12:
-	case MSR_IA32_MC0_MISC+16:
-	case MSR_IA32_UCODE_REV:
-		/* MTRR registers */
-	case 0xfe:
-	case 0x200 ... 0x2ff:
-		data = 0;
-		break;
-	case MSR_IA32_APICBASE:
-		data = vcpu->apic_base;
-		break;
-	default:
-		msr = find_msr_entry(vcpu, msr_index);
-		if (!msr) {
-			printk(KERN_ERR "kvm: unhandled rdmsr: %x\n", msr_index);
-			return 1;
-		}
-		data = msr->data;
-		break;
-	}
-
-	*pdata = data;
-	return 0;
+	return kvm_arch_ops->get_msr(vcpu, msr_index, pdata);
 }
 
 static int handle_rdmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
@@ -2324,7 +2266,7 @@ static int handle_rdmsr(struct kvm_vcpu 
 
 #ifdef __x86_64__
 
-static void set_efer(struct kvm_vcpu *vcpu, u64 efer)
+void set_efer(struct kvm_vcpu *vcpu, u64 efer)
 {
 	struct vmx_msr_entry *msr;
 
@@ -2353,10 +2295,10 @@ static void set_efer(struct kvm_vcpu *vc
 	msr->data = efer;
 	skip_emulated_instruction(vcpu);
 }
+EXPORT_SYMBOL_GPL(set_efer);
 
 #endif
 
-
 /*
  * Writes msr value into into the appropriate "register".
  * Returns 0 on success, non-0 otherwise.
@@ -2364,56 +2306,7 @@ static void set_efer(struct kvm_vcpu *vc
  */
 static int set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
 {
-	struct vmx_msr_entry *msr;
-	switch (msr_index) {
-#ifdef __x86_64__
-	case MSR_FS_BASE:
-		vmcs_writel(GUEST_FS_BASE, data);
-		break;
-	case MSR_GS_BASE:
-		vmcs_writel(GUEST_GS_BASE, data);
-		break;
-#endif
-	case MSR_IA32_SYSENTER_CS:
-		vmcs_write32(GUEST_SYSENTER_CS, data);
-		break;
-	case MSR_IA32_SYSENTER_EIP:
-		vmcs_write32(GUEST_SYSENTER_EIP, data);
-		break;
-	case MSR_IA32_SYSENTER_ESP:
-		vmcs_write32(GUEST_SYSENTER_ESP, data);
-		break;
-#ifdef __x86_64
-	case MSR_EFER:
-		set_efer(vcpu, data);
-		break;
-	case MSR_IA32_MC0_STATUS:
-		printk(KERN_WARNING "%s: MSR_IA32_MC0_STATUS 0x%llx, nop\n"
-			    , __FUNCTION__, data);
-		break;
-#endif
-	case MSR_IA32_TIME_STAMP_COUNTER: {
-		guest_write_tsc(data);
-		break;
-	}
-	case MSR_IA32_UCODE_REV:
-	case MSR_IA32_UCODE_WRITE:
-	case 0x200 ... 0x2ff: /* MTRRs */
-		break;
-	case MSR_IA32_APICBASE:
-		vcpu->apic_base = data;
-		break;
-	default:
-		msr = find_msr_entry(vcpu, msr_index);
-		if (!msr) {
-			printk(KERN_ERR "kvm: unhandled wrmsr: 0x%x\n", msr_index);
-			return 1;
-		}
-		msr->data = data;
-		break;
-	}
-
-	return 0;
+	return kvm_arch_ops->set_msr(vcpu, msr_index, data);
 }
 
 static int handle_wrmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -31,6 +31,146 @@ extern struct vmcs_descriptor {
 	u32 revision_id;
 } vmcs_descriptor;
 
+u64 guest_read_tsc(void);
+void guest_write_tsc(u64 guest_tsc);
+struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
+
+#ifdef __x86_64__
+
+void set_efer(struct kvm_vcpu *vcpu, u64 efer);
+
+#endif
+
+
+/*
+ * Reads an msr value (of 'msr_index') into 'pdata'.
+ * Returns 0 on success, non-0 otherwise.
+ * Assumes vcpu_load() was already called.
+ */
+static int vmx_get_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata)
+{
+	u64 data;
+	struct vmx_msr_entry *msr;
+
+	if (!pdata) {
+		printk(KERN_ERR "BUG: get_msr called with NULL pdata\n");
+		return -EINVAL;
+	}
+
+	switch (msr_index) {
+#ifdef __x86_64__
+	case MSR_FS_BASE:
+		data = vmcs_readl(GUEST_FS_BASE);
+		break;
+	case MSR_GS_BASE:
+		data = vmcs_readl(GUEST_GS_BASE);
+		break;
+	case MSR_EFER:
+		data = vcpu->shadow_efer;
+		break;
+#endif
+	case MSR_IA32_TIME_STAMP_COUNTER:
+		data = guest_read_tsc();
+		break;
+	case MSR_IA32_SYSENTER_CS:
+		data = vmcs_read32(GUEST_SYSENTER_CS);
+		break;
+	case MSR_IA32_SYSENTER_EIP:
+		data = vmcs_read32(GUEST_SYSENTER_EIP);
+		break;
+	case MSR_IA32_SYSENTER_ESP:
+		data = vmcs_read32(GUEST_SYSENTER_ESP);
+		break;
+	case MSR_IA32_MC0_CTL:
+	case MSR_IA32_MCG_STATUS:
+	case MSR_IA32_MCG_CAP:
+	case MSR_IA32_MC0_MISC:
+	case MSR_IA32_MC0_MISC+4:
+	case MSR_IA32_MC0_MISC+8:
+	case MSR_IA32_MC0_MISC+12:
+	case MSR_IA32_MC0_MISC+16:
+	case MSR_IA32_UCODE_REV:
+		/* MTRR registers */
+	case 0xfe:
+	case 0x200 ... 0x2ff:
+		data = 0;
+		break;
+	case MSR_IA32_APICBASE:
+		data = vcpu->apic_base;
+		break;
+	default:
+		msr = find_msr_entry(vcpu, msr_index);
+		if (!msr) {
+			printk(KERN_ERR "kvm: unhandled rdmsr: %x\n", msr_index);
+			return 1;
+		}
+		data = msr->data;
+		break;
+	}
+
+	*pdata = data;
+	return 0;
+}
+
+/*
+ * Writes msr value into into the appropriate "register".
+ * Returns 0 on success, non-0 otherwise.
+ * Assumes vcpu_load() was already called.
+ */
+static int vmx_set_msr(struct kvm_vcpu *vcpu, u32 msr_index, u64 data)
+{
+	struct vmx_msr_entry *msr;
+	switch (msr_index) {
+#ifdef __x86_64__
+	case MSR_FS_BASE:
+		vmcs_writel(GUEST_FS_BASE, data);
+		break;
+	case MSR_GS_BASE:
+		vmcs_writel(GUEST_GS_BASE, data);
+		break;
+#endif
+	case MSR_IA32_SYSENTER_CS:
+		vmcs_write32(GUEST_SYSENTER_CS, data);
+		break;
+	case MSR_IA32_SYSENTER_EIP:
+		vmcs_write32(GUEST_SYSENTER_EIP, data);
+		break;
+	case MSR_IA32_SYSENTER_ESP:
+		vmcs_write32(GUEST_SYSENTER_ESP, data);
+		break;
+#ifdef __x86_64
+	case MSR_EFER:
+		set_efer(vcpu, data);
+		break;
+	case MSR_IA32_MC0_STATUS:
+		printk(KERN_WARNING "%s: MSR_IA32_MC0_STATUS 0x%llx, nop\n"
+			    , __FUNCTION__, data);
+		break;
+#endif
+	case MSR_IA32_TIME_STAMP_COUNTER: {
+		guest_write_tsc(data);
+		break;
+	}
+	case MSR_IA32_UCODE_REV:
+	case MSR_IA32_UCODE_WRITE:
+	case 0x200 ... 0x2ff: /* MTRRs */
+		break;
+	case MSR_IA32_APICBASE:
+		vcpu->apic_base = data;
+		break;
+	default:
+		msr = find_msr_entry(vcpu, msr_index);
+		if (!msr) {
+			printk(KERN_ERR "kvm: unhandled wrmsr: 0x%x\n", msr_index);
+			return 1;
+		}
+		msr->data = data;
+		break;
+	}
+
+	return 0;
+}
+
 static int set_guest_debug(struct kvm_vcpu *vcpu, struct kvm_debug_guest *dbg)
 {
 	unsigned long dr7 = 0x400;
@@ -169,6 +309,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.hardware_disable = hardware_disable,
 
 	.set_guest_debug = set_guest_debug,
+	.get_msr = vmx_get_msr,
+	.set_msr = vmx_set_msr,
 };
 
 static int __init vmx_init(void)

[PATCH 8/38] KVM: Make the segment accessors arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -247,6 +247,10 @@ struct kvm_arch_ops {
 			       struct kvm_debug_guest *dbg);
 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
 	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
+	void (*get_segment)(struct kvm_vcpu *vcpu,
+			    struct kvm_segment *var, int seg);
+	void (*set_segment)(struct kvm_vcpu *vcpu,
+			    struct kvm_segment *var, int seg);
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -36,6 +36,7 @@
 #include <asm/desc.h>
 
 #include "vmx.h"
+#include "kvm_vmx.h"
 #include "x86_emulate.h"
 
 MODULE_AUTHOR("Qumranet");
@@ -82,7 +83,7 @@ enum {
 		GUEST_##seg##_AR_BYTES,			   	\
 	}
 
-static struct kvm_vmx_segment_field {
+struct kvm_vmx_segment_field {
 	unsigned selector;
 	unsigned base;
 	unsigned limit;
@@ -97,6 +98,7 @@ static struct kvm_vmx_segment_field {
 	VMX_SEGMENT_FIELD(TR),
 	VMX_SEGMENT_FIELD(LDTR),
 };
+EXPORT_SYMBOL_GPL(kvm_vmx_segment_fields);
 
 static const u32 vmx_msr_index[] = {
 #ifdef __x86_64__
@@ -625,22 +627,6 @@ void vmcs_writel(unsigned long field, un
 }
 EXPORT_SYMBOL_GPL(vmcs_writel);
 
-static void vmcs_write16(unsigned long field, u16 value)
-{
-	vmcs_writel(field, value);
-}
-
-static void vmcs_write64(unsigned long field, u64 value)
-{
-#ifdef __x86_64__
-	vmcs_writel(field, value);
-#else
-	vmcs_writel(field, value);
-	asm volatile ("");
-	vmcs_writel(field+1, value >> 32);
-#endif
-}
-
 static void inject_gp(struct kvm_vcpu *vcpu)
 {
 	printk(KERN_DEBUG "inject_general_protection: rip 0x%lx\n",
@@ -2821,26 +2807,10 @@ static int kvm_dev_ioctl_set_regs(struct
 	return 0;
 }
 
-static void get_segment(struct kvm_segment *var, int seg)
+static void get_segment(struct kvm_vcpu *vcpu,
+			struct kvm_segment *var, int seg)
 {
-	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
-	u32 ar;
-
-	var->base = vmcs_readl(sf->base);
-	var->limit = vmcs_read32(sf->limit);
-	var->selector = vmcs_read16(sf->selector);
-	ar = vmcs_read32(sf->ar_bytes);
-	if (ar & AR_UNUSABLE_MASK)
-		ar = 0;
-	var->type = ar & 15;
-	var->s = (ar >> 4) & 1;
-	var->dpl = (ar >> 5) & 3;
-	var->present = (ar >> 7) & 1;
-	var->avl = (ar >> 12) & 1;
-	var->l = (ar >> 13) & 1;
-	var->db = (ar >> 14) & 1;
-	var->g = (ar >> 15) & 1;
-	var->unusable = (ar >> 16) & 1;
+	return kvm_arch_ops->get_segment(vcpu, var, seg);
 }
 
 static int kvm_dev_ioctl_get_sregs(struct kvm *kvm, struct kvm_sregs *sregs)
@@ -2853,15 +2823,15 @@ static int kvm_dev_ioctl_get_sregs(struc
 	if (!vcpu)
 		return -ENOENT;
 
-	get_segment(&sregs->cs, VCPU_SREG_CS);
-	get_segment(&sregs->ds, VCPU_SREG_DS);
-	get_segment(&sregs->es, VCPU_SREG_ES);
-	get_segment(&sregs->fs, VCPU_SREG_FS);
-	get_segment(&sregs->gs, VCPU_SREG_GS);
-	get_segment(&sregs->ss, VCPU_SREG_SS);
+	get_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
+	get_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
+	get_segment(vcpu, &sregs->es, VCPU_SREG_ES);
+	get_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
+	get_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
+	get_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
 
-	get_segment(&sregs->tr, VCPU_SREG_TR);
-	get_segment(&sregs->ldt, VCPU_SREG_LDTR);
+	get_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
+	get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
 
 #define get_dtable(var, table) \
 	sregs->var.limit = vmcs_read32(GUEST_##table##_LIMIT), \
@@ -2887,27 +2857,10 @@ static int kvm_dev_ioctl_get_sregs(struc
 	return 0;
 }
 
-static void set_segment(struct kvm_segment *var, int seg)
+static void set_segment(struct kvm_vcpu *vcpu,
+			struct kvm_segment *var, int seg)
 {
-	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
-	u32 ar;
-
-	vmcs_writel(sf->base, var->base);
-	vmcs_write32(sf->limit, var->limit);
-	vmcs_write16(sf->selector, var->selector);
-	if (var->unusable)
-		ar = 1 << 16;
-	else {
-		ar = var->type & 15;
-		ar |= (var->s & 1) << 4;
-		ar |= (var->dpl & 3) << 5;
-		ar |= (var->present & 1) << 7;
-		ar |= (var->avl & 1) << 12;
-		ar |= (var->l & 1) << 13;
-		ar |= (var->db & 1) << 14;
-		ar |= (var->g & 1) << 15;
-	}
-	vmcs_write32(sf->ar_bytes, ar);
+	return kvm_arch_ops->set_segment(vcpu, var, seg);
 }
 
 static int kvm_dev_ioctl_set_sregs(struct kvm *kvm, struct kvm_sregs *sregs)
@@ -2922,15 +2875,15 @@ static int kvm_dev_ioctl_set_sregs(struc
 	if (!vcpu)
 		return -ENOENT;
 
-	set_segment(&sregs->cs, VCPU_SREG_CS);
-	set_segment(&sregs->ds, VCPU_SREG_DS);
-	set_segment(&sregs->es, VCPU_SREG_ES);
-	set_segment(&sregs->fs, VCPU_SREG_FS);
-	set_segment(&sregs->gs, VCPU_SREG_GS);
-	set_segment(&sregs->ss, VCPU_SREG_SS);
+	set_segment(vcpu, &sregs->cs, VCPU_SREG_CS);
+	set_segment(vcpu, &sregs->ds, VCPU_SREG_DS);
+	set_segment(vcpu, &sregs->es, VCPU_SREG_ES);
+	set_segment(vcpu, &sregs->fs, VCPU_SREG_FS);
+	set_segment(vcpu, &sregs->gs, VCPU_SREG_GS);
+	set_segment(vcpu, &sregs->ss, VCPU_SREG_SS);
 
-	set_segment(&sregs->tr, VCPU_SREG_TR);
-	set_segment(&sregs->ldt, VCPU_SREG_LDTR);
+	set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
+	set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
 
 #define set_dtable(var, table) \
 	vmcs_write32(GUEST_##table##_LIMIT, sregs->var.limit), \
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -17,6 +17,8 @@
 
 #include "kvm.h"
 #include <linux/module.h>
+#include "vmx.h"
+#include "kvm_vmx.h"
 
 #define MSR_IA32_FEATURE_CONTROL 		0x03a
 
@@ -31,6 +33,13 @@ extern struct vmcs_descriptor {
 	u32 revision_id;
 } vmcs_descriptor;
 
+extern struct kvm_vmx_segment_field {
+	unsigned selector;
+	unsigned base;
+	unsigned limit;
+	unsigned ar_bytes;
+} kvm_vmx_segment_fields[];
+
 u64 guest_read_tsc(void);
 void guest_write_tsc(u64 guest_tsc);
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
@@ -300,6 +309,53 @@ static __exit void hardware_unsetup(void
 	free_kvm_area();
 }
 
+static void vmx_get_segment(struct kvm_vcpu *vcpu,
+			    struct kvm_segment *var, int seg)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+	u32 ar;
+
+	var->base = vmcs_readl(sf->base);
+	var->limit = vmcs_read32(sf->limit);
+	var->selector = vmcs_read16(sf->selector);
+	ar = vmcs_read32(sf->ar_bytes);
+	if (ar & AR_UNUSABLE_MASK)
+		ar = 0;
+	var->type = ar & 15;
+	var->s = (ar >> 4) & 1;
+	var->dpl = (ar >> 5) & 3;
+	var->present = (ar >> 7) & 1;
+	var->avl = (ar >> 12) & 1;
+	var->l = (ar >> 13) & 1;
+	var->db = (ar >> 14) & 1;
+	var->g = (ar >> 15) & 1;
+	var->unusable = (ar >> 16) & 1;
+}
+
+static void vmx_set_segment(struct kvm_vcpu *vcpu,
+			    struct kvm_segment *var, int seg)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+	u32 ar;
+
+	vmcs_writel(sf->base, var->base);
+	vmcs_write32(sf->limit, var->limit);
+	vmcs_write16(sf->selector, var->selector);
+	if (var->unusable)
+		ar = 1 << 16;
+	else {
+		ar = var->type & 15;
+		ar |= (var->s & 1) << 4;
+		ar |= (var->dpl & 3) << 5;
+		ar |= (var->present & 1) << 7;
+		ar |= (var->avl & 1) << 12;
+		ar |= (var->l & 1) << 13;
+		ar |= (var->db & 1) << 14;
+		ar |= (var->g & 1) << 15;
+	}
+	vmcs_write32(sf->ar_bytes, ar);
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -311,6 +367,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_guest_debug = set_guest_debug,
 	.get_msr = vmx_get_msr,
 	.set_msr = vmx_set_msr,
+	.get_segment = vmx_get_segment,
+	.set_segment = vmx_set_segment,
 };
 
 static int __init vmx_init(void)

[PATCH 9/38] KVM: Cache guest cr4 in vcpu structure

[Avi Kivity]

This eliminates needing to have an arch operation to get cr4.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -168,6 +168,7 @@ struct kvm_vcpu {
 
 	unsigned long cr2;
 	unsigned long cr3;
+	unsigned long cr4;
 	unsigned long cr8;
 	u64 shadow_efer;
 	u64 apic_base;
@@ -335,20 +336,14 @@ static inline int is_long_mode(void)
 	return vmcs_read32(VM_ENTRY_CONTROLS) & VM_ENTRY_CONTROLS_IA32E_MASK;
 }
 
-static inline unsigned long guest_cr4(void)
+static inline int is_pae(struct kvm_vcpu *vcpu)
 {
-	return (vmcs_readl(CR4_READ_SHADOW) & KVM_GUEST_CR4_MASK) |
-		(vmcs_readl(GUEST_CR4) & ~KVM_GUEST_CR4_MASK);
+	return vcpu->cr4 & CR4_PAE_MASK;
 }
 
-static inline int is_pae(void)
+static inline int is_pse(struct kvm_vcpu *vcpu)
 {
-	return guest_cr4() & CR4_PAE_MASK;
-}
-
-static inline int is_pse(void)
-{
-	return guest_cr4() & CR4_PSE_MASK;
+	return vcpu->cr4 & CR4_PSE_MASK;
 }
 
 static inline unsigned long guest_cr0(void)
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -940,7 +940,7 @@ static void set_cr0(struct kvm_vcpu *vcp
 #ifdef __x86_64__
 		if ((vcpu->shadow_efer & EFER_LME)) {
 			u32 guest_cs_ar;
-			if (!is_pae()) {
+			if (!is_pae(vcpu)) {
 				printk(KERN_DEBUG "set_cr0: #GP, start paging "
 				       "in long mode while PAE is disabled\n");
 				inject_gp(vcpu);
@@ -956,7 +956,7 @@ static void set_cr0(struct kvm_vcpu *vcp
 			}
 		} else
 #endif
-		if (is_pae() &&
+		if (is_pae(vcpu) &&
 			    pdptrs_have_reserved_bits_set(vcpu, vcpu->cr3)) {
 			printk(KERN_DEBUG "set_cr0: #GP, pdptrs "
 			       "reserved bits\n");
@@ -993,6 +993,7 @@ static void __set_cr4(struct kvm_vcpu *v
 	vmcs_writel(CR4_READ_SHADOW, cr4);
 	vmcs_writel(GUEST_CR4, cr4 | (vcpu->rmode.active ?
 		    KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON));
+	vcpu->cr4 = cr4;
 }
 
 static void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
@@ -1010,7 +1011,7 @@ static void set_cr4(struct kvm_vcpu *vcp
 			inject_gp(vcpu);
 			return;
 		}
-	} else if (is_paging() && !is_pae() && (cr4 & CR4_PAE_MASK)
+	} else if (is_paging() && !is_pae(vcpu) && (cr4 & CR4_PAE_MASK)
 		   && pdptrs_have_reserved_bits_set(vcpu, vcpu->cr3)) {
 		printk(KERN_DEBUG "set_cr4: #GP, pdptrs reserved bits\n");
 		inject_gp(vcpu);
@@ -1041,7 +1042,7 @@ static void set_cr3(struct kvm_vcpu *vcp
 			inject_gp(vcpu);
 			return;
 		}
-		if (is_paging() && is_pae() &&
+		if (is_paging() && is_pae(vcpu) &&
 		    pdptrs_have_reserved_bits_set(vcpu, cr3)) {
 			printk(KERN_DEBUG "set_cr3: #GP, pdptrs "
 			       "reserved bits\n");
@@ -1902,7 +1903,7 @@ unsigned long realmode_get_cr(struct kvm
 	case 3:
 		return vcpu->cr3;
 	case 4:
-		return guest_cr4();
+		return vcpu->cr4;
 	default:
 		vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr);
 		return 0;
@@ -1924,7 +1925,7 @@ void realmode_set_cr(struct kvm_vcpu *vc
 		set_cr3(vcpu, val);
 		break;
 	case 4:
-		set_cr4(vcpu, mk_cr_64(guest_cr4(), val));
+		set_cr4(vcpu, mk_cr_64(vcpu->cr4, val));
 		break;
 	default:
 		vcpu_printf(vcpu, "%s: unexpected cr %u\n", __FUNCTION__, cr);
@@ -2844,7 +2845,7 @@ static int kvm_dev_ioctl_get_sregs(struc
 	sregs->cr0 = guest_cr0();
 	sregs->cr2 = vcpu->cr2;
 	sregs->cr3 = vcpu->cr3;
-	sregs->cr4 = guest_cr4();
+	sregs->cr4 = vcpu->cr4;
 	sregs->cr8 = vcpu->cr8;
 	sregs->efer = vcpu->shadow_efer;
 	sregs->apic_base = vcpu->apic_base;
@@ -2912,7 +2913,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 	vmcs_writel(GUEST_CR0,
 		    (sregs->cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
 
-	mmu_reset_needed |=  guest_cr4() != sregs->cr4;
+	mmu_reset_needed |= vcpu->cr4 != sregs->cr4;
 	__set_cr4(vcpu, sregs->cr4);
 
 	if (mmu_reset_needed)
Index: linux-2.6/drivers/kvm/mmu.c
===================================================================
--- linux-2.6.orig/drivers/kvm/mmu.c
+++ linux-2.6/drivers/kvm/mmu.c
@@ -564,7 +564,7 @@ static int paging64_init_context(struct 
 {
 	struct kvm_mmu *context = &vcpu->mmu;
 
-	ASSERT(is_pae());
+	ASSERT(is_pae(vcpu));
 	context->new_cr3 = paging_new_cr3;
 	context->page_fault = paging64_page_fault;
 	context->inval_page = paging_inval_page;
@@ -618,7 +618,7 @@ static int init_kvm_mmu(struct kvm_vcpu 
 		return nonpaging_init_context(vcpu);
 	else if (is_long_mode())
 		return paging64_init_context(vcpu);
-	else if (is_pae())
+	else if (is_pae(vcpu))
 		return paging32E_init_context(vcpu);
 	else
 		return paging32_init_context(vcpu);
Index: linux-2.6/drivers/kvm/paging_tmpl.h
===================================================================
--- linux-2.6.orig/drivers/kvm/paging_tmpl.h
+++ linux-2.6/drivers/kvm/paging_tmpl.h
@@ -70,7 +70,7 @@ static void FNAME(init_walker)(struct gu
 	hpa = safe_gpa_to_hpa(vcpu, vcpu->cr3 & PT64_BASE_ADDR_MASK);
 	walker->table = kmap_atomic(pfn_to_page(hpa >> PAGE_SHIFT), KM_USER0);
 
-	ASSERT((!is_long_mode() && is_pae()) ||
+	ASSERT((!is_long_mode() && is_pae(vcpu)) ||
 	       (vcpu->cr3 & ~(PAGE_MASK | CR3_FLAGS_MASK)) == 0);
 
 	walker->table = (pt_element_t *)( (unsigned long)walker->table |
@@ -133,7 +133,7 @@ static pt_element_t *FNAME(fetch_guest)(
 		    !is_present_pte(walker->table[index]) ||
 		    (walker->level == PT_DIRECTORY_LEVEL &&
 		     (walker->table[index] & PT_PAGE_SIZE_MASK) &&
-		     (PTTYPE == 64 || is_pse())))
+		     (PTTYPE == 64 || is_pse(vcpu))))
 			return &walker->table[index];
 		if (walker->level != 3 || is_long_mode())
 			walker->inherited_ar &= walker->table[index];
@@ -369,7 +369,7 @@ static gpa_t FNAME(gva_to_gpa)(struct kv
 
 	if (walker.level == PT_DIRECTORY_LEVEL) {
 		ASSERT((guest_pte & PT_PAGE_SIZE_MASK));
-		ASSERT(PTTYPE == 64 || is_pse());
+		ASSERT(PTTYPE == 64 || is_pse(vcpu));
 
 		gpa = (guest_pte & PT_DIR_BASE_ADDR_MASK) | (vaddr &
 			(PT_LEVEL_MASK(PT_PAGE_TABLE_LEVEL) | ~PAGE_MASK));
Index: linux-2.6/drivers/kvm/kvm_vmx.h
===================================================================
--- /dev/null
+++ linux-2.6/drivers/kvm/kvm_vmx.h
@@ -0,0 +1,20 @@
+#ifndef __KVM_VMX_H
+#define __KVM_VMX_H
+
+static inline void vmcs_write16(unsigned long field, u16 value)
+{
+	vmcs_writel(field, value);
+}
+
+static inline void vmcs_write64(unsigned long field, u64 value)
+{
+#ifdef __x86_64__
+	vmcs_writel(field, value);
+#else
+	vmcs_writel(field, value);
+	asm volatile ("");
+	vmcs_writel(field+1, value >> 32);
+#endif
+}
+
+#endif

[PATCH 10/38] KVM: Cache guest cr0 in vcpu structure

[Avi Kivity]

This eliminates needing to have an arch operation to get cr0.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -41,7 +41,6 @@
 	 | CR0_NW_MASK | CR0_CD_MASK)
 #define KVM_VM_CR0_ALWAYS_ON \
 	(CR0_PG_MASK | CR0_PE_MASK | CR0_WP_MASK | CR0_NE_MASK)
-
 #define KVM_GUEST_CR4_MASK \
 	(CR4_PSE_MASK | CR4_PAE_MASK | CR4_PGE_MASK | CR4_VMXE_MASK | CR4_VME_MASK)
 #define KVM_PMODE_VM_CR4_ALWAYS_ON (CR4_VMXE_MASK | CR4_PAE_MASK)
@@ -166,6 +165,7 @@ struct kvm_vcpu {
 	unsigned long regs[NR_VCPU_REGS]; /* for rsp: vcpu_load_rsp_rip() */
 	unsigned long rip;      /* needs vcpu_load_rsp_rip() */
 
+	unsigned long cr0;
 	unsigned long cr2;
 	unsigned long cr3;
 	unsigned long cr4;
@@ -346,20 +346,14 @@ static inline int is_pse(struct kvm_vcpu
 	return vcpu->cr4 & CR4_PSE_MASK;
 }
 
-static inline unsigned long guest_cr0(void)
-{
-	return (vmcs_readl(CR0_READ_SHADOW) & KVM_GUEST_CR0_MASK) |
-		(vmcs_readl(GUEST_CR0) & ~KVM_GUEST_CR0_MASK);
-}
-
 static inline unsigned guest_cpl(void)
 {
 	return vmcs_read16(GUEST_CS_SELECTOR) & SELECTOR_RPL_MASK;
 }
 
-static inline int is_paging(void)
+static inline int is_paging(struct kvm_vcpu *vcpu)
 {
-	return guest_cr0() & CR0_PG_MASK;
+	return vcpu->cr0 & CR0_PG_MASK;
 }
 
 static inline int is_page_fault(u32 intr_info)
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -875,9 +875,9 @@ static void __set_cr0(struct kvm_vcpu *v
 
 #ifdef __x86_64__
 	if (vcpu->shadow_efer & EFER_LME) {
-		if (!is_paging() && (cr0 & CR0_PG_MASK))
+		if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK))
 			enter_lmode(vcpu);
-		if (is_paging() && !(cr0 & CR0_PG_MASK))
+		if (is_paging(vcpu) && !(cr0 & CR0_PG_MASK))
 			exit_lmode(vcpu);
 	}
 #endif
@@ -885,6 +885,7 @@ static void __set_cr0(struct kvm_vcpu *v
 	vmcs_writel(CR0_READ_SHADOW, cr0);
 	vmcs_writel(GUEST_CR0,
 		    (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
+	vcpu->cr0 = cr0;
 }
 
 static int pdptrs_have_reserved_bits_set(struct kvm_vcpu *vcpu,
@@ -918,7 +919,7 @@ static void set_cr0(struct kvm_vcpu *vcp
 {
 	if (cr0 & CR0_RESEVED_BITS) {
 		printk(KERN_DEBUG "set_cr0: 0x%lx #GP, reserved bits 0x%lx\n",
-		       cr0, guest_cr0());
+		       cr0, vcpu->cr0);
 		inject_gp(vcpu);
 		return;
 	}
@@ -936,7 +937,7 @@ static void set_cr0(struct kvm_vcpu *vcp
 		return;
 	}
 
-	if (!is_paging() && (cr0 & CR0_PG_MASK)) {
+	if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK)) {
 #ifdef __x86_64__
 		if ((vcpu->shadow_efer & EFER_LME)) {
 			u32 guest_cs_ar;
@@ -975,7 +976,7 @@ static void set_cr0(struct kvm_vcpu *vcp
 
 static void lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
 {
-	unsigned long cr0 = guest_cr0();
+	unsigned long cr0 = vcpu->cr0;
 
 	if ((msw & CR0_PE_MASK) && !(cr0 & CR0_PE_MASK)) {
 		enter_pmode(vcpu);
@@ -986,6 +987,7 @@ static void lmsw(struct kvm_vcpu *vcpu, 
 
 	vmcs_writel(GUEST_CR0, (vmcs_readl(GUEST_CR0) & ~LMSW_GUEST_MASK)
 				| (msw & LMSW_GUEST_MASK));
+	vcpu->cr0 = (vcpu->cr0 & ~0xfffful) | msw;
 }
 
 static void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
@@ -1011,7 +1013,7 @@ static void set_cr4(struct kvm_vcpu *vcp
 			inject_gp(vcpu);
 			return;
 		}
-	} else if (is_paging() && !is_pae(vcpu) && (cr4 & CR4_PAE_MASK)
+	} else if (is_paging(vcpu) && !is_pae(vcpu) && (cr4 & CR4_PAE_MASK)
 		   && pdptrs_have_reserved_bits_set(vcpu, vcpu->cr3)) {
 		printk(KERN_DEBUG "set_cr4: #GP, pdptrs reserved bits\n");
 		inject_gp(vcpu);
@@ -1042,7 +1044,7 @@ static void set_cr3(struct kvm_vcpu *vcp
 			inject_gp(vcpu);
 			return;
 		}
-		if (is_paging() && is_pae(vcpu) &&
+		if (is_paging(vcpu) && is_pae(vcpu) &&
 		    pdptrs_have_reserved_bits_set(vcpu, cr3)) {
 			printk(KERN_DEBUG "set_cr3: #GP, pdptrs "
 			       "reserved bits\n");
@@ -1897,7 +1899,7 @@ unsigned long realmode_get_cr(struct kvm
 {
 	switch (cr) {
 	case 0:
-		return guest_cr0();
+		return vcpu->cr0;
 	case 2:
 		return vcpu->cr2;
 	case 3:
@@ -1915,7 +1917,7 @@ void realmode_set_cr(struct kvm_vcpu *vc
 {
 	switch (cr) {
 	case 0:
-		set_cr0(vcpu, mk_cr_64(guest_cr0(), val));
+		set_cr0(vcpu, mk_cr_64(vcpu->cr0, val));
 		*rflags = vmcs_readl(GUEST_RFLAGS);
 		break;
 	case 2:
@@ -2264,7 +2266,8 @@ void set_efer(struct kvm_vcpu *vcpu, u64
 		return;
 	}
 
-	if (is_paging() && (vcpu->shadow_efer & EFER_LME) != (efer & EFER_LME)) {
+	if (is_paging(vcpu)
+	    && (vcpu->shadow_efer & EFER_LME) != (efer & EFER_LME)) {
 		printk(KERN_DEBUG "set_efer: #GP, change LME while paging\n");
 		inject_gp(vcpu);
 		return;
@@ -2842,7 +2845,7 @@ static int kvm_dev_ioctl_get_sregs(struc
 	get_dtable(gdt, GDTR);
 #undef get_dtable
 
-	sregs->cr0 = guest_cr0();
+	sregs->cr0 = vcpu->cr0;
 	sregs->cr2 = vcpu->cr2;
 	sregs->cr3 = vcpu->cr3;
 	sregs->cr4 = vcpu->cr4;
@@ -2906,12 +2909,13 @@ static int kvm_dev_ioctl_set_sregs(struc
 #endif
 	vcpu->apic_base = sregs->apic_base;
 
-	mmu_reset_needed |= guest_cr0() != sregs->cr0;
+	mmu_reset_needed |= vcpu->cr0 != sregs->cr0;
 	vcpu->rmode.active = ((sregs->cr0 & CR0_PE_MASK) == 0);
 	update_exception_bitmap(vcpu);
 	vmcs_writel(CR0_READ_SHADOW, sregs->cr0);
 	vmcs_writel(GUEST_CR0,
 		    (sregs->cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
+	vcpu->cr0 = sregs->cr0;
 
 	mmu_reset_needed |= vcpu->cr4 != sregs->cr4;
 	__set_cr4(vcpu, sregs->cr4);
Index: linux-2.6/drivers/kvm/mmu.c
===================================================================
--- linux-2.6.orig/drivers/kvm/mmu.c
+++ linux-2.6/drivers/kvm/mmu.c
@@ -138,9 +138,9 @@
 #define PT_DIRECTORY_LEVEL 2
 #define PT_PAGE_TABLE_LEVEL 1
 
-static int is_write_protection(void)
+static int is_write_protection(struct kvm_vcpu *vcpu)
 {
-	return guest_cr0() & CR0_WP_MASK;
+	return vcpu->cr0 & CR0_WP_MASK;
 }
 
 static int is_cpuid_PSE36(void)
@@ -321,7 +321,7 @@ static void nonpaging_flush(struct kvm_v
 	root = kvm_mmu_alloc_page(vcpu, 0);
 	ASSERT(VALID_PAGE(root));
 	vcpu->mmu.root_hpa = root;
-	if (is_paging())
+	if (is_paging(vcpu))
 		root |= (vcpu->cr3 & (CR3_PCD_MASK | CR3_WPT_MASK));
 	vmcs_writel(GUEST_CR3, root);
 }
@@ -614,7 +614,7 @@ static int init_kvm_mmu(struct kvm_vcpu 
 	ASSERT(vcpu);
 	ASSERT(!VALID_PAGE(vcpu->mmu.root_hpa));
 
-	if (!is_paging())
+	if (!is_paging(vcpu))
 		return nonpaging_init_context(vcpu);
 	else if (is_long_mode())
 		return paging64_init_context(vcpu);
Index: linux-2.6/drivers/kvm/paging_tmpl.h
===================================================================
--- linux-2.6.orig/drivers/kvm/paging_tmpl.h
+++ linux-2.6/drivers/kvm/paging_tmpl.h
@@ -251,7 +251,7 @@ static int FNAME(fix_write_pf)(struct kv
 		 * supervisor write protection is enabled.
 		 */
 		if (!writable_shadow) {
-			if (is_write_protection())
+			if (is_write_protection(vcpu))
 				return 0;
 			*shadow_ent &= ~PT_USER_MASK;
 		}

[PATCH 11/38] KVM: Add get_segment_base() arch accessor

[Avi Kivity]

This could have been implemented in terms of the existing get_segment(), but
that make unnecessary vmx accesses.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -248,6 +248,7 @@ struct kvm_arch_ops {
 			       struct kvm_debug_guest *dbg);
 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
 	int (*set_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 data);
+	u64 (*get_segment_base)(struct kvm_vcpu *vcpu, int seg);
 	void (*get_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
 	void (*set_segment)(struct kvm_vcpu *vcpu,
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1776,12 +1776,19 @@ static int emulator_cmpxchg_emulated(uns
 	return emulator_write_emulated(addr, new, bytes, ctxt);
 }
 
+static unsigned long get_segment_base(struct kvm_vcpu *vcpu, int seg)
+{
+	return kvm_arch_ops->get_segment_base(vcpu, seg);
+}
+
 static void report_emulation_failure(struct x86_emulate_ctxt *ctxt)
 {
 	static int reported;
 	u8 opcodes[4];
 	unsigned long rip = vmcs_readl(GUEST_RIP);
-	unsigned long rip_linear = rip + vmcs_readl(GUEST_CS_BASE);
+	unsigned long rip_linear;
+
+	rip_linear = rip + get_segment_base(ctxt->vcpu, VCPU_SREG_CS);
 
 	if (reported)
 		return;
@@ -1835,14 +1842,14 @@ static int emulate_instruction(struct kv
 		emulate_ctxt.es_base = 0;
 		emulate_ctxt.ss_base = 0;
 	} else {
-		emulate_ctxt.cs_base = vmcs_readl(GUEST_CS_BASE);
-		emulate_ctxt.ds_base = vmcs_readl(GUEST_DS_BASE);
-		emulate_ctxt.es_base = vmcs_readl(GUEST_ES_BASE);
-		emulate_ctxt.ss_base = vmcs_readl(GUEST_SS_BASE);
+		emulate_ctxt.cs_base = get_segment_base(vcpu, VCPU_SREG_CS);
+		emulate_ctxt.ds_base = get_segment_base(vcpu, VCPU_SREG_DS);
+		emulate_ctxt.es_base = get_segment_base(vcpu, VCPU_SREG_ES);
+		emulate_ctxt.ss_base = get_segment_base(vcpu, VCPU_SREG_SS);
 	}
 
-	emulate_ctxt.gs_base = vmcs_readl(GUEST_GS_BASE);
-	emulate_ctxt.fs_base = vmcs_readl(GUEST_FS_BASE);
+	emulate_ctxt.gs_base = get_segment_base(vcpu, VCPU_SREG_GS);
+	emulate_ctxt.fs_base = get_segment_base(vcpu, VCPU_SREG_FS);
 
 	vcpu->mmio_is_write = 0;
 	r = x86_emulate_memop(&emulate_ctxt, &emulate_ops);
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -309,6 +309,13 @@ static __exit void hardware_unsetup(void
 	free_kvm_area();
 }
 
+static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+
+	return vmcs_readl(sf->base);
+}
+
 static void vmx_get_segment(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg)
 {
@@ -367,6 +374,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_guest_debug = set_guest_debug,
 	.get_msr = vmx_get_msr,
 	.set_msr = vmx_set_msr,
+	.get_segment_base = vmx_get_segment_base,
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
 };

[PATCH 12/38] KVM: Add idt and gdt descriptor accessors

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -236,6 +236,11 @@ struct kvm_stat {
 	u32 irq_exits;
 };
 
+struct descriptor_table {
+	u16 limit;
+	unsigned long base;
+} __attribute__((packed));
+
 struct kvm_arch_ops {
 	int (*cpu_has_kvm_support)(void);          /* __init */
 	int (*disabled_by_bios)(void);             /* __init */
@@ -253,6 +258,10 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
+	void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
+	void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
+	void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
+	void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -155,11 +155,6 @@ struct vmx_msr_entry *find_msr_entry(str
 }
 EXPORT_SYMBOL_GPL(find_msr_entry);
 
-struct descriptor_table {
-	u16 limit;
-	unsigned long base;
-} __attribute__((packed));
-
 static void get_gdt(struct descriptor_table *table)
 {
 	asm ("sgdt %0" : "=m"(*table));
@@ -2827,6 +2822,7 @@ static void get_segment(struct kvm_vcpu 
 static int kvm_dev_ioctl_get_sregs(struct kvm *kvm, struct kvm_sregs *sregs)
 {
 	struct kvm_vcpu *vcpu;
+	struct descriptor_table dt;
 
 	if (sregs->vcpu < 0 || sregs->vcpu >= KVM_MAX_VCPUS)
 		return -EINVAL;
@@ -2844,13 +2840,12 @@ static int kvm_dev_ioctl_get_sregs(struc
 	get_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
 	get_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
 
-#define get_dtable(var, table) \
-	sregs->var.limit = vmcs_read32(GUEST_##table##_LIMIT), \
-		sregs->var.base = vmcs_readl(GUEST_##table##_BASE)
-
-	get_dtable(idt, IDTR);
-	get_dtable(gdt, GDTR);
-#undef get_dtable
+	kvm_arch_ops->get_idt(vcpu, &dt);
+	sregs->idt.limit = dt.limit;
+	sregs->idt.base = dt.base;
+	kvm_arch_ops->get_gdt(vcpu, &dt);
+	sregs->gdt.limit = dt.limit;
+	sregs->gdt.base = dt.base;
 
 	sregs->cr0 = vcpu->cr0;
 	sregs->cr2 = vcpu->cr2;
@@ -2879,6 +2874,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 	struct kvm_vcpu *vcpu;
 	int mmu_reset_needed = 0;
 	int i;
+	struct descriptor_table dt;
 
 	if (sregs->vcpu < 0 || sregs->vcpu >= KVM_MAX_VCPUS)
 		return -EINVAL;
@@ -2896,13 +2892,12 @@ static int kvm_dev_ioctl_set_sregs(struc
 	set_segment(vcpu, &sregs->tr, VCPU_SREG_TR);
 	set_segment(vcpu, &sregs->ldt, VCPU_SREG_LDTR);
 
-#define set_dtable(var, table) \
-	vmcs_write32(GUEST_##table##_LIMIT, sregs->var.limit), \
-	vmcs_writel(GUEST_##table##_BASE, sregs->var.base)
-
-	set_dtable(idt, IDTR);
-	set_dtable(gdt, GDTR);
-#undef set_dtable
+	dt.limit = sregs->idt.limit;
+	dt.base = sregs->idt.base;
+	kvm_arch_ops->set_idt(vcpu, &dt);
+	dt.limit = sregs->gdt.limit;
+	dt.base = sregs->gdt.base;
+	kvm_arch_ops->set_gdt(vcpu, &dt);
 
 	vcpu->cr2 = sregs->cr2;
 	mmu_reset_needed |= vcpu->cr3 != sregs->cr3;
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -363,6 +363,30 @@ static void vmx_set_segment(struct kvm_v
 	vmcs_write32(sf->ar_bytes, ar);
 }
 
+static void vmx_get_idt(struct kvm_vcpu *vcpu, struct descriptor_table *dt)
+{
+	dt->limit = vmcs_read32(GUEST_IDTR_LIMIT);
+	dt->base = vmcs_readl(GUEST_IDTR_BASE);
+}
+
+static void vmx_set_idt(struct kvm_vcpu *vcpu, struct descriptor_table *dt)
+{
+	vmcs_write32(GUEST_IDTR_LIMIT, dt->limit);
+	vmcs_writel(GUEST_IDTR_BASE, dt->base);
+}
+
+static void vmx_get_gdt(struct kvm_vcpu *vcpu, struct descriptor_table *dt)
+{
+	dt->limit = vmcs_read32(GUEST_GDTR_LIMIT);
+	dt->base = vmcs_readl(GUEST_GDTR_BASE);
+}
+
+static void vmx_set_gdt(struct kvm_vcpu *vcpu, struct descriptor_table *dt)
+{
+	vmcs_write32(GUEST_GDTR_LIMIT, dt->limit);
+	vmcs_writel(GUEST_GDTR_BASE, dt->base);
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -377,6 +401,10 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment_base = vmx_get_segment_base,
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
+	.get_idt = vmx_get_idt,
+	.set_idt = vmx_set_idt,
+	.get_gdt = vmx_get_gdt,
+	.set_gdt = vmx_set_gdt,
 };
 
 static int __init vmx_init(void)

[PATCH 13/38] KVM: Make syncing the register file to the vcpu structure an arch operation

[Avi Kivity]

This copies any general purpose guest registers maintained by the hardware
to the vcpu structure (and back).

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -262,6 +262,8 @@ struct kvm_arch_ops {
 	void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
+	void (*cache_regs)(struct kvm_vcpu *vcpu);
+	void (*decache_regs)(struct kvm_vcpu *vcpu);
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1327,26 +1327,6 @@ out:
 }
 
 /*
- * Sync the rsp and rip registers into the vcpu structure.  This allows
- * registers to be accessed by indexing vcpu->regs.
- */
-static void vcpu_load_rsp_rip(struct kvm_vcpu *vcpu)
-{
-	vcpu->regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP);
-	vcpu->rip = vmcs_readl(GUEST_RIP);
-}
-
-/*
- * Syncs rsp and rip back into the vmcs.  Should be called after possible
- * modification.
- */
-static void vcpu_put_rsp_rip(struct kvm_vcpu *vcpu)
-{
-	vmcs_writel(GUEST_RSP, vcpu->regs[VCPU_REGS_RSP]);
-	vmcs_writel(GUEST_RIP, vcpu->rip);
-}
-
-/*
  * Creates some virtual cpus.  Good luck creating more than one.
  */
 static int kvm_dev_ioctl_create_vcpu(struct kvm *kvm, int n)
@@ -1819,7 +1799,7 @@ static int emulate_instruction(struct kv
 	int r;
 	u32 cs_ar;
 
-	vcpu_load_rsp_rip(vcpu);
+	kvm_arch_ops->cache_regs(vcpu);
 
 	cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
 
@@ -1864,7 +1844,7 @@ static int emulate_instruction(struct kv
 		return EMULATE_DO_MMIO;
 	}
 
-	vcpu_put_rsp_rip(vcpu);
+	kvm_arch_ops->decache_regs(vcpu);
 	vmcs_writel(GUEST_RFLAGS, emulate_ctxt.eflags);
 
 	if (vcpu->mmio_is_write)
@@ -2134,22 +2114,22 @@ static int handle_cr(struct kvm_vcpu *vc
 	case 0: /* mov to cr */
 		switch (cr) {
 		case 0:
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			set_cr0(vcpu, vcpu->regs[reg]);
 			skip_emulated_instruction(vcpu);
 			return 1;
 		case 3:
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			set_cr3(vcpu, vcpu->regs[reg]);
 			skip_emulated_instruction(vcpu);
 			return 1;
 		case 4:
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			set_cr4(vcpu, vcpu->regs[reg]);
 			skip_emulated_instruction(vcpu);
 			return 1;
 		case 8:
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			set_cr8(vcpu, vcpu->regs[reg]);
 			skip_emulated_instruction(vcpu);
 			return 1;
@@ -2158,17 +2138,17 @@ static int handle_cr(struct kvm_vcpu *vc
 	case 1: /*mov from cr*/
 		switch (cr) {
 		case 3:
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			vcpu->regs[reg] = vcpu->cr3;
-			vcpu_put_rsp_rip(vcpu);
+			kvm_arch_ops->decache_regs(vcpu);
 			skip_emulated_instruction(vcpu);
 			return 1;
 		case 8:
 			printk(KERN_DEBUG "handle_cr: read CR8 "
 			       "cpu erratum AA15\n");
-			vcpu_load_rsp_rip(vcpu);
+			kvm_arch_ops->cache_regs(vcpu);
 			vcpu->regs[reg] = vcpu->cr8;
-			vcpu_put_rsp_rip(vcpu);
+			kvm_arch_ops->decache_regs(vcpu);
 			skip_emulated_instruction(vcpu);
 			return 1;
 		}
@@ -2200,7 +2180,7 @@ static int handle_dr(struct kvm_vcpu *vc
 	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
 	dr = exit_qualification & 7;
 	reg = (exit_qualification >> 8) & 15;
-	vcpu_load_rsp_rip(vcpu);
+	kvm_arch_ops->cache_regs(vcpu);
 	if (exit_qualification & 16) {
 		/* mov from dr */
 		switch (dr) {
@@ -2217,7 +2197,7 @@ static int handle_dr(struct kvm_vcpu *vc
 	} else {
 		/* mov to dr */
 	}
-	vcpu_put_rsp_rip(vcpu);
+	kvm_arch_ops->decache_regs(vcpu);
 	skip_emulated_instruction(vcpu);
 	return 1;
 }
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -180,6 +180,26 @@ static int vmx_set_msr(struct kvm_vcpu *
 	return 0;
 }
 
+/*
+ * Sync the rsp and rip registers into the vcpu structure.  This allows
+ * registers to be accessed by indexing vcpu->regs.
+ */
+static void vcpu_load_rsp_rip(struct kvm_vcpu *vcpu)
+{
+	vcpu->regs[VCPU_REGS_RSP] = vmcs_readl(GUEST_RSP);
+	vcpu->rip = vmcs_readl(GUEST_RIP);
+}
+
+/*
+ * Syncs rsp and rip back into the vmcs.  Should be called after possible
+ * modification.
+ */
+static void vcpu_put_rsp_rip(struct kvm_vcpu *vcpu)
+{
+	vmcs_writel(GUEST_RSP, vcpu->regs[VCPU_REGS_RSP]);
+	vmcs_writel(GUEST_RIP, vcpu->rip);
+}
+
 static int set_guest_debug(struct kvm_vcpu *vcpu, struct kvm_debug_guest *dbg)
 {
 	unsigned long dr7 = 0x400;
@@ -405,6 +425,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_idt = vmx_set_idt,
 	.get_gdt = vmx_get_gdt,
 	.set_gdt = vmx_set_gdt,
+	.cache_regs = vcpu_load_rsp_rip,
+	.decache_regs = vcpu_put_rsp_rip,
 };
 
 static int __init vmx_init(void)

[PATCH 14/38] KVM: Make the vcpu execution loop an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -264,6 +264,9 @@ struct kvm_arch_ops {
 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*cache_regs)(struct kvm_vcpu *vcpu);
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
+
+	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
+	unsigned long vmx_return; /* temporary hack */
 };
 
 extern struct kvm_stat kvm_stat;
@@ -306,6 +309,11 @@ unsigned long realmode_get_cr(struct kvm
 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
 		     unsigned long *rflags);
 
+void load_msrs(struct vmx_msr_entry *e, int n);
+void save_msrs(struct vmx_msr_entry *e, int n);
+void kvm_resched(struct kvm_vcpu *vcpu);
+int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
+
 int kvm_read_guest(struct kvm_vcpu *vcpu,
 	       gva_t addr,
 	       unsigned long size,
@@ -398,6 +406,69 @@ static inline struct kvm_mmu_page *page_
 	return (struct kvm_mmu_page *)page->private;
 }
 
+static inline u16 read_fs(void)
+{
+	u16 seg;
+	asm ("mov %%fs, %0" : "=g"(seg));
+	return seg;
+}
+
+static inline u16 read_gs(void)
+{
+	u16 seg;
+	asm ("mov %%gs, %0" : "=g"(seg));
+	return seg;
+}
+
+static inline u16 read_ldt(void)
+{
+	u16 ldt;
+	asm ("sldt %0" : "=g"(ldt));
+	return ldt;
+}
+
+static inline void load_fs(u16 sel)
+{
+	asm ("mov %0, %%fs" : : "rm"(sel));
+}
+
+static inline void load_gs(u16 sel)
+{
+	asm ("mov %0, %%gs" : : "rm"(sel));
+}
+
+#ifndef load_ldt
+static inline void load_ldt(u16 sel)
+{
+	asm ("lldt %0" : : "g"(sel));
+}
+#endif
+
+#ifdef __x86_64__
+static inline unsigned long read_msr(unsigned long msr)
+{
+	u64 value;
+
+	rdmsrl(msr, value);
+	return value;
+}
+#endif
+
+static inline void fx_save(void *image)
+{
+	asm ("fxsave (%0)":: "r" (image));
+}
+
+static inline void fx_restore(void *image)
+{
+	asm ("fxrstor (%0)":: "r" (image));
+}
+
+static inline void fpu_init(void)
+{
+	asm ("finit");
+}
+
 #define ASM_VMX_VMCLEAR_RAX       ".byte 0x66, 0x0f, 0xc7, 0x30"
 #define ASM_VMX_VMLAUNCH          ".byte 0x0f, 0x01, 0xc2"
 #define ASM_VMX_VMRESUME          ".byte 0x0f, 0x01, 0xc3"
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -44,6 +44,7 @@ MODULE_LICENSE("GPL");
 
 struct kvm_arch_ops *kvm_arch_ops;
 struct kvm_stat kvm_stat;
+EXPORT_SYMBOL_GPL(kvm_stat);
 
 static struct kvm_stats_debugfs_item {
 	const char *name;
@@ -108,16 +109,6 @@ static const u32 vmx_msr_index[] = {
 };
 #define NR_VMX_MSR (sizeof(vmx_msr_index) / sizeof(*vmx_msr_index))
 
-#ifdef __x86_64__
-/*
- * avoid save/load MSR_SYSCALL_MASK and MSR_LSTAR by std vt
- * mechanism (cpu bug AA24)
- */
-#define NR_BAD_MSRS 2
-#else
-#define NR_BAD_MSRS 0
-#endif
-
 #define TSS_IOPB_BASE_OFFSET 0x66
 #define TSS_BASE_SIZE 0x68
 #define TSS_IOPB_SIZE (65536 / 8)
@@ -165,59 +156,6 @@ static void get_idt(struct descriptor_ta
 	asm ("sidt %0" : "=m"(*table));
 }
 
-static u16 read_fs(void)
-{
-	u16 seg;
-	asm ("mov %%fs, %0" : "=g"(seg));
-	return seg;
-}
-
-static u16 read_gs(void)
-{
-	u16 seg;
-	asm ("mov %%gs, %0" : "=g"(seg));
-	return seg;
-}
-
-static u16 read_ldt(void)
-{
-	u16 ldt;
-	asm ("sldt %0" : "=g"(ldt));
-	return ldt;
-}
-
-static void load_fs(u16 sel)
-{
-	asm ("mov %0, %%fs" : : "rm"(sel));
-}
-
-static void load_gs(u16 sel)
-{
-	asm ("mov %0, %%gs" : : "rm"(sel));
-}
-
-#ifndef load_ldt
-static void load_ldt(u16 sel)
-{
-	asm ("lldt %0" : : "g"(sel));
-}
-#endif
-
-static void fx_save(void *image)
-{
-	asm ("fxsave (%0)":: "r" (image));
-}
-
-static void fx_restore(void *image)
-{
-	asm ("fxrstor (%0)":: "r" (image));
-}
-
-static void fpu_init(void)
-{
-	asm ("finit");
-}
-
 struct segment_descriptor {
 	u16 limit_low;
 	u16 base_low;
@@ -278,23 +216,6 @@ static unsigned long read_tr_base(void)
 	return segment_base(tr);
 }
 
-static void reload_tss(void)
-{
-#ifndef __x86_64__
-
-	/*
-	 * VT restores TR but not its size.  Useless.
-	 */
-	struct descriptor_table gdt;
-	struct segment_descriptor *descs;
-
-	get_gdt(&gdt);
-	descs = (void *)gdt.base;
-	descs[GDT_ENTRY_TSS].type = 9; /* available TSS */
-	load_TR_desc();
-#endif
-}
-
 DEFINE_PER_CPU(struct vmcs *, vmxarea);
 EXPORT_SYMBOL_GPL(per_cpu__vmxarea); /* temporary hack */
 static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
@@ -306,24 +227,12 @@ struct vmcs_descriptor {
 } vmcs_descriptor;
 EXPORT_SYMBOL_GPL(vmcs_descriptor);
 
-#ifdef __x86_64__
-static unsigned long read_msr(unsigned long msr)
-{
-	u64 value;
-
-	rdmsrl(msr, value);
-	return value;
-}
-#endif
-
 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
 {
 	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
 	return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : 0;
 }
 
-
-
 int kvm_read_guest(struct kvm_vcpu *vcpu,
 			     gva_t addr,
 			     unsigned long size,
@@ -357,6 +266,7 @@ int kvm_read_guest(struct kvm_vcpu *vcpu
 	}
 	return req_size - size;
 }
+EXPORT_SYMBOL_GPL(kvm_read_guest);
 
 int kvm_write_guest(struct kvm_vcpu *vcpu,
 			     gva_t addr,
@@ -390,6 +300,7 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 	}
 	return req_size - size;
 }
+EXPORT_SYMBOL_GPL(kvm_write_guest);
 
 static void vmcs_clear(struct vmcs *vmcs)
 {
@@ -1120,7 +1031,6 @@ static void seg_setup(int seg)
  */
 static int kvm_vcpu_setup(struct kvm_vcpu *vcpu)
 {
-	extern asmlinkage void kvm_vmx_return(void);
 	u32 host_sysenter_cs;
 	u32 junk;
 	unsigned long a;
@@ -1251,7 +1161,7 @@ static int kvm_vcpu_setup(struct kvm_vcp
 	vmcs_writel(HOST_IDTR_BASE, dt.base);   /* 22.2.4 */
 
 
-	vmcs_writel(HOST_RIP, (unsigned long)kvm_vmx_return); /* 22.2.5 */
+	vmcs_writel(HOST_RIP, kvm_arch_ops->vmx_return); /* 22.2.5 */
 
 	rdmsr(MSR_IA32_SYSENTER_CS, host_sysenter_cs, junk);
 	vmcs_write32(HOST_IA32_SYSENTER_CS, host_sysenter_cs);
@@ -2344,7 +2254,7 @@ static const int kvm_vmx_max_exit_handle
  * The guest has exited.  See if we can fix it or if we need userspace
  * assistance.
  */
-static int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
+int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 {
 	u32 vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
 	u32 exit_reason = vmcs_read32(VM_EXIT_REASON);
@@ -2363,127 +2273,39 @@ static int kvm_handle_exit(struct kvm_ru
 	}
 	return 0;
 }
+EXPORT_SYMBOL_GPL(kvm_handle_exit);
 
-static void inject_rmode_irq(struct kvm_vcpu *vcpu, int irq)
+void kvm_resched(struct kvm_vcpu *vcpu)
 {
-	u16 ent[2];
-	u16 cs;
-	u16 ip;
-	unsigned long flags;
-	unsigned long ss_base = vmcs_readl(GUEST_SS_BASE);
-	u16 sp =  vmcs_readl(GUEST_RSP);
-	u32 ss_limit = vmcs_read32(GUEST_SS_LIMIT);
-
-	if (sp > ss_limit || sp - 6 > sp) {
-		vcpu_printf(vcpu, "%s: #SS, rsp 0x%lx ss 0x%lx limit 0x%x\n",
-			    __FUNCTION__,
-			    vmcs_readl(GUEST_RSP),
-			    vmcs_readl(GUEST_SS_BASE),
-			    vmcs_read32(GUEST_SS_LIMIT));
-		return;
-	}
-
-	if (kvm_read_guest(vcpu, irq * sizeof(ent), sizeof(ent), &ent) !=
-								sizeof(ent)) {
-		vcpu_printf(vcpu, "%s: read guest err\n", __FUNCTION__);
-		return;
-	}
-
-	flags =  vmcs_readl(GUEST_RFLAGS);
-	cs =  vmcs_readl(GUEST_CS_BASE) >> 4;
-	ip =  vmcs_readl(GUEST_RIP);
-
-
-	if (kvm_write_guest(vcpu, ss_base + sp - 2, 2, &flags) != 2 ||
-	    kvm_write_guest(vcpu, ss_base + sp - 4, 2, &cs) != 2 ||
-	    kvm_write_guest(vcpu, ss_base + sp - 6, 2, &ip) != 2) {
-		vcpu_printf(vcpu, "%s: write guest err\n", __FUNCTION__);
-		return;
-	}
-
-	vmcs_writel(GUEST_RFLAGS, flags &
-		    ~( X86_EFLAGS_IF | X86_EFLAGS_AC | X86_EFLAGS_TF));
-	vmcs_write16(GUEST_CS_SELECTOR, ent[1]) ;
-	vmcs_writel(GUEST_CS_BASE, ent[1] << 4);
-	vmcs_writel(GUEST_RIP, ent[0]);
-	vmcs_writel(GUEST_RSP, (vmcs_readl(GUEST_RSP) & ~0xffff) | (sp - 6));
-}
-
-static void kvm_do_inject_irq(struct kvm_vcpu *vcpu)
-{
-	int word_index = __ffs(vcpu->irq_summary);
-	int bit_index = __ffs(vcpu->irq_pending[word_index]);
-	int irq = word_index * BITS_PER_LONG + bit_index;
-
-	clear_bit(bit_index, &vcpu->irq_pending[word_index]);
-	if (!vcpu->irq_pending[word_index])
-		clear_bit(word_index, &vcpu->irq_summary);
-
-	if (vcpu->rmode.active) {
-		inject_rmode_irq(vcpu, irq);
-		return;
-	}
-	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
-			irq | INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK);
-}
-
-static void kvm_try_inject_irq(struct kvm_vcpu *vcpu)
-{
-	if ((vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF)
-	    && (vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & 3) == 0)
-		/*
-		 * Interrupts enabled, and not blocked by sti or mov ss. Good.
-		 */
-		kvm_do_inject_irq(vcpu);
-	else
-		/*
-		 * Interrupts blocked.  Wait for unblock.
-		 */
-		vmcs_write32(CPU_BASED_VM_EXEC_CONTROL,
-			     vmcs_read32(CPU_BASED_VM_EXEC_CONTROL)
-			     | CPU_BASED_VIRTUAL_INTR_PENDING);
-}
-
-static void kvm_guest_debug_pre(struct kvm_vcpu *vcpu)
-{
-	struct kvm_guest_debug *dbg = &vcpu->guest_debug;
-
-	set_debugreg(dbg->bp[0], 0);
-	set_debugreg(dbg->bp[1], 1);
-	set_debugreg(dbg->bp[2], 2);
-	set_debugreg(dbg->bp[3], 3);
-
-	if (dbg->singlestep) {
-		unsigned long flags;
-
-		flags = vmcs_readl(GUEST_RFLAGS);
-		flags |= X86_EFLAGS_TF | X86_EFLAGS_RF;
-		vmcs_writel(GUEST_RFLAGS, flags);
-	}
+	vcpu_put(vcpu);
+	cond_resched();
+	/* Cannot fail -  no vcpu unplug yet. */
+	vcpu_load(vcpu->kvm, vcpu_slot(vcpu));
 }
+EXPORT_SYMBOL_GPL(kvm_resched);
 
-static void load_msrs(struct vmx_msr_entry *e, int n)
+void load_msrs(struct vmx_msr_entry *e, int n)
 {
 	int i;
 
 	for (i = 0; i < n; ++i)
 		wrmsrl(e[i].index, e[i].data);
 }
+EXPORT_SYMBOL_GPL(load_msrs);
 
-static void save_msrs(struct vmx_msr_entry *e, int n)
+void save_msrs(struct vmx_msr_entry *e, int n)
 {
 	int i;
 
 	for (i = 0; i < n; ++i)
 		rdmsrl(e[i].index, e[i].data);
 }
+EXPORT_SYMBOL_GPL(save_msrs);
 
 static int kvm_dev_ioctl_run(struct kvm *kvm, struct kvm_run *kvm_run)
 {
 	struct kvm_vcpu *vcpu;
-	u8 fail;
-	u16 fs_sel, gs_sel, ldt_sel;
-	int fs_gs_ldt_reload_needed;
+	int r;
 
 	if (kvm_run->vcpu < 0 || kvm_run->vcpu >= KVM_MAX_VCPUS)
 		return -EINVAL;
@@ -2504,211 +2326,10 @@ static int kvm_dev_ioctl_run(struct kvm 
 
 	vcpu->mmio_needed = 0;
 
-again:
-	/*
-	 * Set host fs and gs selectors.  Unfortunately, 22.2.3 does not
-	 * allow segment selectors with cpl > 0 or ti == 1.
-	 */
-	fs_sel = read_fs();
-	gs_sel = read_gs();
-	ldt_sel = read_ldt();
-	fs_gs_ldt_reload_needed = (fs_sel & 7) | (gs_sel & 7) | ldt_sel;
-	if (!fs_gs_ldt_reload_needed) {
-		vmcs_write16(HOST_FS_SELECTOR, fs_sel);
-		vmcs_write16(HOST_GS_SELECTOR, gs_sel);
-	} else {
-		vmcs_write16(HOST_FS_SELECTOR, 0);
-		vmcs_write16(HOST_GS_SELECTOR, 0);
-	}
-
-#ifdef __x86_64__
-	vmcs_writel(HOST_FS_BASE, read_msr(MSR_FS_BASE));
-	vmcs_writel(HOST_GS_BASE, read_msr(MSR_GS_BASE));
-#endif
-
-	if (vcpu->irq_summary &&
-	    !(vmcs_read32(VM_ENTRY_INTR_INFO_FIELD) & INTR_INFO_VALID_MASK))
-		kvm_try_inject_irq(vcpu);
-
-	if (vcpu->guest_debug.enabled)
-		kvm_guest_debug_pre(vcpu);
-
-	fx_save(vcpu->host_fx_image);
-	fx_restore(vcpu->guest_fx_image);
-
-	save_msrs(vcpu->host_msrs, vcpu->nmsrs);
-	load_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
-
-	asm (
-		/* Store host registers */
-		"pushf \n\t"
-#ifdef __x86_64__
-		"push %%rax; push %%rbx; push %%rdx;"
-		"push %%rsi; push %%rdi; push %%rbp;"
-		"push %%r8;  push %%r9;  push %%r10; push %%r11;"
-		"push %%r12; push %%r13; push %%r14; push %%r15;"
-		"push %%rcx \n\t"
-		ASM_VMX_VMWRITE_RSP_RDX "\n\t"
-#else
-		"pusha; push %%ecx \n\t"
-		ASM_VMX_VMWRITE_RSP_RDX "\n\t"
-#endif
-		/* Check if vmlaunch of vmresume is needed */
-		"cmp $0, %1 \n\t"
-		/* Load guest registers.  Don't clobber flags. */
-#ifdef __x86_64__
-		"mov %c[cr2](%3), %%rax \n\t"
-		"mov %%rax, %%cr2 \n\t"
-		"mov %c[rax](%3), %%rax \n\t"
-		"mov %c[rbx](%3), %%rbx \n\t"
-		"mov %c[rdx](%3), %%rdx \n\t"
-		"mov %c[rsi](%3), %%rsi \n\t"
-		"mov %c[rdi](%3), %%rdi \n\t"
-		"mov %c[rbp](%3), %%rbp \n\t"
-		"mov %c[r8](%3),  %%r8  \n\t"
-		"mov %c[r9](%3),  %%r9  \n\t"
-		"mov %c[r10](%3), %%r10 \n\t"
-		"mov %c[r11](%3), %%r11 \n\t"
-		"mov %c[r12](%3), %%r12 \n\t"
-		"mov %c[r13](%3), %%r13 \n\t"
-		"mov %c[r14](%3), %%r14 \n\t"
-		"mov %c[r15](%3), %%r15 \n\t"
-		"mov %c[rcx](%3), %%rcx \n\t" /* kills %3 (rcx) */
-#else
-		"mov %c[cr2](%3), %%eax \n\t"
-		"mov %%eax,   %%cr2 \n\t"
-		"mov %c[rax](%3), %%eax \n\t"
-		"mov %c[rbx](%3), %%ebx \n\t"
-		"mov %c[rdx](%3), %%edx \n\t"
-		"mov %c[rsi](%3), %%esi \n\t"
-		"mov %c[rdi](%3), %%edi \n\t"
-		"mov %c[rbp](%3), %%ebp \n\t"
-		"mov %c[rcx](%3), %%ecx \n\t" /* kills %3 (ecx) */
-#endif
-		/* Enter guest mode */
-		"jne launched \n\t"
-		ASM_VMX_VMLAUNCH "\n\t"
-		"jmp kvm_vmx_return \n\t"
-		"launched: " ASM_VMX_VMRESUME "\n\t"
-		".globl kvm_vmx_return \n\t"
-		"kvm_vmx_return: "
-		/* Save guest registers, load host registers, keep flags */
-#ifdef __x86_64__
-		"xchg %3,     0(%%rsp) \n\t"
-		"mov %%rax, %c[rax](%3) \n\t"
-		"mov %%rbx, %c[rbx](%3) \n\t"
-		"pushq 0(%%rsp); popq %c[rcx](%3) \n\t"
-		"mov %%rdx, %c[rdx](%3) \n\t"
-		"mov %%rsi, %c[rsi](%3) \n\t"
-		"mov %%rdi, %c[rdi](%3) \n\t"
-		"mov %%rbp, %c[rbp](%3) \n\t"
-		"mov %%r8,  %c[r8](%3) \n\t"
-		"mov %%r9,  %c[r9](%3) \n\t"
-		"mov %%r10, %c[r10](%3) \n\t"
-		"mov %%r11, %c[r11](%3) \n\t"
-		"mov %%r12, %c[r12](%3) \n\t"
-		"mov %%r13, %c[r13](%3) \n\t"
-		"mov %%r14, %c[r14](%3) \n\t"
-		"mov %%r15, %c[r15](%3) \n\t"
-		"mov %%cr2, %%rax   \n\t"
-		"mov %%rax, %c[cr2](%3) \n\t"
-		"mov 0(%%rsp), %3 \n\t"
-
-		"pop  %%rcx; pop  %%r15; pop  %%r14; pop  %%r13; pop  %%r12;"
-		"pop  %%r11; pop  %%r10; pop  %%r9;  pop  %%r8;"
-		"pop  %%rbp; pop  %%rdi; pop  %%rsi;"
-		"pop  %%rdx; pop  %%rbx; pop  %%rax \n\t"
-#else
-		"xchg %3, 0(%%esp) \n\t"
-		"mov %%eax, %c[rax](%3) \n\t"
-		"mov %%ebx, %c[rbx](%3) \n\t"
-		"pushl 0(%%esp); popl %c[rcx](%3) \n\t"
-		"mov %%edx, %c[rdx](%3) \n\t"
-		"mov %%esi, %c[rsi](%3) \n\t"
-		"mov %%edi, %c[rdi](%3) \n\t"
-		"mov %%ebp, %c[rbp](%3) \n\t"
-		"mov %%cr2, %%eax  \n\t"
-		"mov %%eax, %c[cr2](%3) \n\t"
-		"mov 0(%%esp), %3 \n\t"
-
-		"pop %%ecx; popa \n\t"
-#endif
-		"setbe %0 \n\t"
-		"popf \n\t"
-	      : "=g" (fail)
-	      : "r"(vcpu->launched), "d"((unsigned long)HOST_RSP),
-		"c"(vcpu),
-		[rax]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RAX])),
-		[rbx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RBX])),
-		[rcx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RCX])),
-		[rdx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RDX])),
-		[rsi]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RSI])),
-		[rdi]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RDI])),
-		[rbp]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RBP])),
-#ifdef __x86_64__
-		[r8 ]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R8 ])),
-		[r9 ]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R9 ])),
-		[r10]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R10])),
-		[r11]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R11])),
-		[r12]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R12])),
-		[r13]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R13])),
-		[r14]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R14])),
-		[r15]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R15])),
-#endif
-		[cr2]"i"(offsetof(struct kvm_vcpu, cr2))
-	      : "cc", "memory" );
-
-	++kvm_stat.exits;
-
-	save_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
-	load_msrs(vcpu->host_msrs, NR_BAD_MSRS);
-
-	fx_save(vcpu->guest_fx_image);
-	fx_restore(vcpu->host_fx_image);
-
-#ifndef __x86_64__
-	asm ("mov %0, %%ds; mov %0, %%es" : : "r"(__USER_DS));
-#endif
-
-	kvm_run->exit_type = 0;
-	if (fail) {
-		kvm_run->exit_type = KVM_EXIT_TYPE_FAIL_ENTRY;
-		kvm_run->exit_reason = vmcs_read32(VM_INSTRUCTION_ERROR);
-	} else {
-		if (fs_gs_ldt_reload_needed) {
-			load_ldt(ldt_sel);
-			load_fs(fs_sel);
-			/*
-			 * If we have to reload gs, we must take care to
-			 * preserve our gs base.
-			 */
-			local_irq_disable();
-			load_gs(gs_sel);
-#ifdef __x86_64__
-			wrmsrl(MSR_GS_BASE, vmcs_readl(HOST_GS_BASE));
-#endif
-			local_irq_enable();
-
-			reload_tss();
-		}
-		vcpu->launched = 1;
-		kvm_run->exit_type = KVM_EXIT_TYPE_VM_EXIT;
-		if (kvm_handle_exit(kvm_run, vcpu)) {
-			/* Give scheduler a change to reschedule. */
-			vcpu_put(vcpu);
-			if (signal_pending(current)) {
-				++kvm_stat.signal_exits;
-				return -EINTR;
-			}
-			cond_resched();
-			/* Cannot fail -  no vcpu unplug yet. */
-			vcpu_load(kvm, vcpu_slot(vcpu));
-			goto again;
-		}
-	}
+	r = kvm_arch_ops->run(vcpu, kvm_run);
 
 	vcpu_put(vcpu);
-	return 0;
+	return r;
 }
 
 static int kvm_dev_ioctl_get_regs(struct kvm *kvm, struct kvm_regs *regs)
Index: linux-2.6/drivers/kvm/kvm_vmx.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_vmx.h
+++ linux-2.6/drivers/kvm/kvm_vmx.h
@@ -17,4 +17,14 @@ static inline void vmcs_write64(unsigned
 #endif
 }
 
+#ifdef __x86_64__
+/*
+ * avoid save/load MSR_SYSCALL_MASK and MSR_LSTAR by std vt
+ * mechanism (cpu bug AA24)
+ */
+#define NR_BAD_MSRS 2
+#else
+#define NR_BAD_MSRS 0
+#endif
+
 #endif
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -50,6 +50,22 @@ void set_efer(struct kvm_vcpu *vcpu, u64
 
 #endif
 
+static void reload_tss(void)
+{
+#ifndef __x86_64__
+
+	/*
+	 * VT restores TR but not its size.  Useless.
+	 */
+	struct descriptor_table gdt;
+	struct segment_descriptor *descs;
+
+	get_gdt(&gdt);
+	descs = (void *)gdt.base;
+	descs[GDT_ENTRY_TSS].type = 9; /* available TSS */
+	load_TR_desc();
+#endif
+}
 
 /*
  * Reads an msr value (of 'msr_index') into 'pdata'.
@@ -407,6 +423,314 @@ static void vmx_set_gdt(struct kvm_vcpu 
 	vmcs_writel(GUEST_GDTR_BASE, dt->base);
 }
 
+static void inject_rmode_irq(struct kvm_vcpu *vcpu, int irq)
+{
+	u16 ent[2];
+	u16 cs;
+	u16 ip;
+	unsigned long flags;
+	unsigned long ss_base = vmcs_readl(GUEST_SS_BASE);
+	u16 sp =  vmcs_readl(GUEST_RSP);
+	u32 ss_limit = vmcs_read32(GUEST_SS_LIMIT);
+
+	if (sp > ss_limit || sp - 6 > sp) {
+		vcpu_printf(vcpu, "%s: #SS, rsp 0x%lx ss 0x%lx limit 0x%x\n",
+			    __FUNCTION__,
+			    vmcs_readl(GUEST_RSP),
+			    vmcs_readl(GUEST_SS_BASE),
+			    vmcs_read32(GUEST_SS_LIMIT));
+		return;
+	}
+
+	if (kvm_read_guest(vcpu, irq * sizeof(ent), sizeof(ent), &ent) !=
+								sizeof(ent)) {
+		vcpu_printf(vcpu, "%s: read guest err\n", __FUNCTION__);
+		return;
+	}
+
+	flags =  vmcs_readl(GUEST_RFLAGS);
+	cs =  vmcs_readl(GUEST_CS_BASE) >> 4;
+	ip =  vmcs_readl(GUEST_RIP);
+
+
+	if (kvm_write_guest(vcpu, ss_base + sp - 2, 2, &flags) != 2 ||
+	    kvm_write_guest(vcpu, ss_base + sp - 4, 2, &cs) != 2 ||
+	    kvm_write_guest(vcpu, ss_base + sp - 6, 2, &ip) != 2) {
+		vcpu_printf(vcpu, "%s: write guest err\n", __FUNCTION__);
+		return;
+	}
+
+	vmcs_writel(GUEST_RFLAGS, flags &
+		    ~( X86_EFLAGS_IF | X86_EFLAGS_AC | X86_EFLAGS_TF));
+	vmcs_write16(GUEST_CS_SELECTOR, ent[1]) ;
+	vmcs_writel(GUEST_CS_BASE, ent[1] << 4);
+	vmcs_writel(GUEST_RIP, ent[0]);
+	vmcs_writel(GUEST_RSP, (vmcs_readl(GUEST_RSP) & ~0xffff) | (sp - 6));
+}
+
+static void kvm_do_inject_irq(struct kvm_vcpu *vcpu)
+{
+	int word_index = __ffs(vcpu->irq_summary);
+	int bit_index = __ffs(vcpu->irq_pending[word_index]);
+	int irq = word_index * BITS_PER_LONG + bit_index;
+
+	clear_bit(bit_index, &vcpu->irq_pending[word_index]);
+	if (!vcpu->irq_pending[word_index])
+		clear_bit(word_index, &vcpu->irq_summary);
+
+	if (vcpu->rmode.active) {
+		inject_rmode_irq(vcpu, irq);
+		return;
+	}
+	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
+			irq | INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK);
+}
+
+static void kvm_try_inject_irq(struct kvm_vcpu *vcpu)
+{
+	if ((vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF)
+	    && (vmcs_read32(GUEST_INTERRUPTIBILITY_INFO) & 3) == 0)
+		/*
+		 * Interrupts enabled, and not blocked by sti or mov ss. Good.
+		 */
+		kvm_do_inject_irq(vcpu);
+	else
+		/*
+		 * Interrupts blocked.  Wait for unblock.
+		 */
+		vmcs_write32(CPU_BASED_VM_EXEC_CONTROL,
+			     vmcs_read32(CPU_BASED_VM_EXEC_CONTROL)
+			     | CPU_BASED_VIRTUAL_INTR_PENDING);
+}
+
+static void kvm_guest_debug_pre(struct kvm_vcpu *vcpu)
+{
+	struct kvm_guest_debug *dbg = &vcpu->guest_debug;
+
+	set_debugreg(dbg->bp[0], 0);
+	set_debugreg(dbg->bp[1], 1);
+	set_debugreg(dbg->bp[2], 2);
+	set_debugreg(dbg->bp[3], 3);
+
+	if (dbg->singlestep) {
+		unsigned long flags;
+
+		flags = vmcs_readl(GUEST_RFLAGS);
+		flags |= X86_EFLAGS_TF | X86_EFLAGS_RF;
+		vmcs_writel(GUEST_RFLAGS, flags);
+	}
+}
+
+static int vmx_vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u8 fail;
+	u16 fs_sel, gs_sel, ldt_sel;
+	int fs_gs_ldt_reload_needed;
+
+again:
+	/*
+	 * Set host fs and gs selectors.  Unfortunately, 22.2.3 does not
+	 * allow segment selectors with cpl > 0 or ti == 1.
+	 */
+	fs_sel = read_fs();
+	gs_sel = read_gs();
+	ldt_sel = read_ldt();
+	fs_gs_ldt_reload_needed = (fs_sel & 7) | (gs_sel & 7) | ldt_sel;
+	if (!fs_gs_ldt_reload_needed) {
+		vmcs_write16(HOST_FS_SELECTOR, fs_sel);
+		vmcs_write16(HOST_GS_SELECTOR, gs_sel);
+	} else {
+		vmcs_write16(HOST_FS_SELECTOR, 0);
+		vmcs_write16(HOST_GS_SELECTOR, 0);
+	}
+
+#ifdef __x86_64__
+	vmcs_writel(HOST_FS_BASE, read_msr(MSR_FS_BASE));
+	vmcs_writel(HOST_GS_BASE, read_msr(MSR_GS_BASE));
+#endif
+
+	if (vcpu->irq_summary &&
+	    !(vmcs_read32(VM_ENTRY_INTR_INFO_FIELD) & INTR_INFO_VALID_MASK))
+		kvm_try_inject_irq(vcpu);
+
+	if (vcpu->guest_debug.enabled)
+		kvm_guest_debug_pre(vcpu);
+
+	fx_save(vcpu->host_fx_image);
+	fx_restore(vcpu->guest_fx_image);
+
+	save_msrs(vcpu->host_msrs, vcpu->nmsrs);
+	load_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
+
+	asm (
+		/* Store host registers */
+		"pushf \n\t"
+#ifdef __x86_64__
+		"push %%rax; push %%rbx; push %%rdx;"
+		"push %%rsi; push %%rdi; push %%rbp;"
+		"push %%r8;  push %%r9;  push %%r10; push %%r11;"
+		"push %%r12; push %%r13; push %%r14; push %%r15;"
+		"push %%rcx \n\t"
+		ASM_VMX_VMWRITE_RSP_RDX "\n\t"
+#else
+		"pusha; push %%ecx \n\t"
+		ASM_VMX_VMWRITE_RSP_RDX "\n\t"
+#endif
+		/* Check if vmlaunch of vmresume is needed */
+		"cmp $0, %1 \n\t"
+		/* Load guest registers.  Don't clobber flags. */
+#ifdef __x86_64__
+		"mov %c[cr2](%3), %%rax \n\t"
+		"mov %%rax, %%cr2 \n\t"
+		"mov %c[rax](%3), %%rax \n\t"
+		"mov %c[rbx](%3), %%rbx \n\t"
+		"mov %c[rdx](%3), %%rdx \n\t"
+		"mov %c[rsi](%3), %%rsi \n\t"
+		"mov %c[rdi](%3), %%rdi \n\t"
+		"mov %c[rbp](%3), %%rbp \n\t"
+		"mov %c[r8](%3),  %%r8  \n\t"
+		"mov %c[r9](%3),  %%r9  \n\t"
+		"mov %c[r10](%3), %%r10 \n\t"
+		"mov %c[r11](%3), %%r11 \n\t"
+		"mov %c[r12](%3), %%r12 \n\t"
+		"mov %c[r13](%3), %%r13 \n\t"
+		"mov %c[r14](%3), %%r14 \n\t"
+		"mov %c[r15](%3), %%r15 \n\t"
+		"mov %c[rcx](%3), %%rcx \n\t" /* kills %3 (rcx) */
+#else
+		"mov %c[cr2](%3), %%eax \n\t"
+		"mov %%eax,   %%cr2 \n\t"
+		"mov %c[rax](%3), %%eax \n\t"
+		"mov %c[rbx](%3), %%ebx \n\t"
+		"mov %c[rdx](%3), %%edx \n\t"
+		"mov %c[rsi](%3), %%esi \n\t"
+		"mov %c[rdi](%3), %%edi \n\t"
+		"mov %c[rbp](%3), %%ebp \n\t"
+		"mov %c[rcx](%3), %%ecx \n\t" /* kills %3 (ecx) */
+#endif
+		/* Enter guest mode */
+		"jne launched \n\t"
+		ASM_VMX_VMLAUNCH "\n\t"
+		"jmp kvm_vmx_return \n\t"
+		"launched: " ASM_VMX_VMRESUME "\n\t"
+		".globl kvm_vmx_return \n\t"
+		"kvm_vmx_return: "
+		/* Save guest registers, load host registers, keep flags */
+#ifdef __x86_64__
+		"xchg %3,     0(%%rsp) \n\t"
+		"mov %%rax, %c[rax](%3) \n\t"
+		"mov %%rbx, %c[rbx](%3) \n\t"
+		"pushq 0(%%rsp); popq %c[rcx](%3) \n\t"
+		"mov %%rdx, %c[rdx](%3) \n\t"
+		"mov %%rsi, %c[rsi](%3) \n\t"
+		"mov %%rdi, %c[rdi](%3) \n\t"
+		"mov %%rbp, %c[rbp](%3) \n\t"
+		"mov %%r8,  %c[r8](%3) \n\t"
+		"mov %%r9,  %c[r9](%3) \n\t"
+		"mov %%r10, %c[r10](%3) \n\t"
+		"mov %%r11, %c[r11](%3) \n\t"
+		"mov %%r12, %c[r12](%3) \n\t"
+		"mov %%r13, %c[r13](%3) \n\t"
+		"mov %%r14, %c[r14](%3) \n\t"
+		"mov %%r15, %c[r15](%3) \n\t"
+		"mov %%cr2, %%rax   \n\t"
+		"mov %%rax, %c[cr2](%3) \n\t"
+		"mov 0(%%rsp), %3 \n\t"
+
+		"pop  %%rcx; pop  %%r15; pop  %%r14; pop  %%r13; pop  %%r12;"
+		"pop  %%r11; pop  %%r10; pop  %%r9;  pop  %%r8;"
+		"pop  %%rbp; pop  %%rdi; pop  %%rsi;"
+		"pop  %%rdx; pop  %%rbx; pop  %%rax \n\t"
+#else
+		"xchg %3, 0(%%esp) \n\t"
+		"mov %%eax, %c[rax](%3) \n\t"
+		"mov %%ebx, %c[rbx](%3) \n\t"
+		"pushl 0(%%esp); popl %c[rcx](%3) \n\t"
+		"mov %%edx, %c[rdx](%3) \n\t"
+		"mov %%esi, %c[rsi](%3) \n\t"
+		"mov %%edi, %c[rdi](%3) \n\t"
+		"mov %%ebp, %c[rbp](%3) \n\t"
+		"mov %%cr2, %%eax  \n\t"
+		"mov %%eax, %c[cr2](%3) \n\t"
+		"mov 0(%%esp), %3 \n\t"
+
+		"pop %%ecx; popa \n\t"
+#endif
+		"setbe %0 \n\t"
+		"popf \n\t"
+	      : "=g" (fail)
+	      : "r"(vcpu->launched), "d"((unsigned long)HOST_RSP),
+		"c"(vcpu),
+		[rax]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RAX])),
+		[rbx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RBX])),
+		[rcx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RCX])),
+		[rdx]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RDX])),
+		[rsi]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RSI])),
+		[rdi]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RDI])),
+		[rbp]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_RBP])),
+#ifdef __x86_64__
+		[r8 ]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R8 ])),
+		[r9 ]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R9 ])),
+		[r10]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R10])),
+		[r11]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R11])),
+		[r12]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R12])),
+		[r13]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R13])),
+		[r14]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R14])),
+		[r15]"i"(offsetof(struct kvm_vcpu, regs[VCPU_REGS_R15])),
+#endif
+		[cr2]"i"(offsetof(struct kvm_vcpu, cr2))
+	      : "cc", "memory" );
+
+	++kvm_stat.exits;
+
+	save_msrs(vcpu->guest_msrs, NR_BAD_MSRS);
+	load_msrs(vcpu->host_msrs, NR_BAD_MSRS);
+
+	fx_save(vcpu->guest_fx_image);
+	fx_restore(vcpu->host_fx_image);
+
+#ifndef __x86_64__
+	asm ("mov %0, %%ds; mov %0, %%es" : : "r"(__USER_DS));
+#endif
+
+	kvm_run->exit_type = 0;
+	if (fail) {
+		kvm_run->exit_type = KVM_EXIT_TYPE_FAIL_ENTRY;
+		kvm_run->exit_reason = vmcs_read32(VM_INSTRUCTION_ERROR);
+	} else {
+		if (fs_gs_ldt_reload_needed) {
+			load_ldt(ldt_sel);
+			load_fs(fs_sel);
+			/*
+			 * If we have to reload gs, we must take care to
+			 * preserve our gs base.
+			 */
+			local_irq_disable();
+			load_gs(gs_sel);
+#ifdef __x86_64__
+			wrmsrl(MSR_GS_BASE, vmcs_readl(HOST_GS_BASE));
+#endif
+			local_irq_enable();
+
+			reload_tss();
+		}
+		vcpu->launched = 1;
+		kvm_run->exit_type = KVM_EXIT_TYPE_VM_EXIT;
+		if (kvm_handle_exit(kvm_run, vcpu)) {
+			/* Give scheduler a change to reschedule. */
+			if (signal_pending(current)) {
+				++kvm_stat.signal_exits;
+				return -EINTR;
+			}
+			kvm_resched(vcpu);
+			goto again;
+		}
+	}
+	return 0;
+}
+
+extern asmlinkage void kvm_vmx_return(void);
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -427,6 +751,9 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_gdt = vmx_set_gdt,
 	.cache_regs = vcpu_load_rsp_rip,
 	.decache_regs = vcpu_put_rsp_rip,
+
+	.run = vmx_vcpu_run,
+	.vmx_return = (unsigned long)kvm_vmx_return,
 };
 
 static int __init vmx_init(void)

[PATCH 15/38] KVM: Move the vmx exit handlers to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -266,6 +266,7 @@ struct kvm_arch_ops {
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
 
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
+	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
 	unsigned long vmx_return; /* temporary hack */
 };
 
@@ -300,6 +301,14 @@ static inline struct page *gfn_to_page(s
 struct kvm_memory_slot *gfn_to_memslot(struct kvm *kvm, gfn_t gfn);
 void mark_page_dirty(struct kvm *kvm, gfn_t gfn);
 
+enum emulation_result {
+	EMULATE_DONE,       /* no further processing */
+	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
+	EMULATE_FAIL,         /* can't emulate this instruction */
+};
+
+int emulate_instruction(struct kvm_vcpu *vcpu, struct kvm_run *run,
+			unsigned long cr2, u16 error_code);
 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
 void realmode_lidt(struct kvm_vcpu *vcpu, u16 size, unsigned long address);
 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,
@@ -309,10 +318,22 @@ unsigned long realmode_get_cr(struct kvm
 void realmode_set_cr(struct kvm_vcpu *vcpu, int cr, unsigned long value,
 		     unsigned long *rflags);
 
+void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
+void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr0);
+void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
+void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
+void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
+
+void inject_gp(struct kvm_vcpu *vcpu);
+
+#ifdef __x86_64__
+void set_efer(struct kvm_vcpu *vcpu, u64 efer);
+#endif
+
+
 void load_msrs(struct vmx_msr_entry *e, int n);
 void save_msrs(struct vmx_msr_entry *e, int n);
 void kvm_resched(struct kvm_vcpu *vcpu);
-int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu);
 
 int kvm_read_guest(struct kvm_vcpu *vcpu,
 	       gva_t addr,
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -533,7 +533,7 @@ void vmcs_writel(unsigned long field, un
 }
 EXPORT_SYMBOL_GPL(vmcs_writel);
 
-static void inject_gp(struct kvm_vcpu *vcpu)
+void inject_gp(struct kvm_vcpu *vcpu)
 {
 	printk(KERN_DEBUG "inject_general_protection: rip 0x%lx\n",
 	       vmcs_readl(GUEST_RIP));
@@ -544,6 +544,7 @@ static void inject_gp(struct kvm_vcpu *v
 		     INTR_INFO_DELIEVER_CODE_MASK |
 		     INTR_INFO_VALID_MASK);
 }
+EXPORT_SYMBOL_GPL(inject_gp);
 
 /*
  * reads and returns guest's timestamp counter "register"
@@ -821,7 +822,7 @@ static int pdptrs_have_reserved_bits_set
 	return i != 4;
 }
 
-static void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+void set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	if (cr0 & CR0_RESEVED_BITS) {
 		printk(KERN_DEBUG "set_cr0: 0x%lx #GP, reserved bits 0x%lx\n",
@@ -879,8 +880,9 @@ static void set_cr0(struct kvm_vcpu *vcp
 	spin_unlock(&vcpu->kvm->lock);
 	return;
 }
+EXPORT_SYMBOL_GPL(set_cr0);
 
-static void lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
+void lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
 {
 	unsigned long cr0 = vcpu->cr0;
 
@@ -895,6 +897,7 @@ static void lmsw(struct kvm_vcpu *vcpu, 
 				| (msw & LMSW_GUEST_MASK));
 	vcpu->cr0 = (vcpu->cr0 & ~0xfffful) | msw;
 }
+EXPORT_SYMBOL_GPL(lmsw);
 
 static void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
@@ -904,7 +907,7 @@ static void __set_cr4(struct kvm_vcpu *v
 	vcpu->cr4 = cr4;
 }
 
-static void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
+void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
 	if (cr4 & CR4_RESEVED_BITS) {
 		printk(KERN_DEBUG "set_cr4: #GP, reserved bits\n");
@@ -935,8 +938,9 @@ static void set_cr4(struct kvm_vcpu *vcp
 	kvm_mmu_reset_context(vcpu);
 	spin_unlock(&vcpu->kvm->lock);
 }
+EXPORT_SYMBOL_GPL(set_cr4);
 
-static void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
+void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 {
 	if (is_long_mode()) {
 		if ( cr3 & CR3_L_MODE_RESEVED_BITS) {
@@ -964,8 +968,9 @@ static void set_cr3(struct kvm_vcpu *vcp
 	vcpu->mmu.new_cr3(vcpu);
 	spin_unlock(&vcpu->kvm->lock);
 }
+EXPORT_SYMBOL_GPL(set_cr3);
 
-static void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8)
+void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr8)
 {
 	if ( cr8 & CR8_RESEVED_BITS) {
 		printk(KERN_DEBUG "set_cr8: #GP, reserved bits 0x%lx\n", cr8);
@@ -974,6 +979,7 @@ static void set_cr8(struct kvm_vcpu *vcp
 	}
 	vcpu->cr8 = cr8;
 }
+EXPORT_SYMBOL_GPL(set_cr8);
 
 static u32 get_rdx_init_val(void)
 {
@@ -1534,25 +1540,6 @@ void mark_page_dirty(struct kvm *kvm, gf
 	}
 }
 
-static void skip_emulated_instruction(struct kvm_vcpu *vcpu)
-{
-	unsigned long rip;
-	u32 interruptibility;
-
-	rip = vmcs_readl(GUEST_RIP);
-	rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
-	vmcs_writel(GUEST_RIP, rip);
-
-	/*
-	 * We emulated an instruction, so temporary interrupt blocking
-	 * should be removed, if set.
-	 */
-	interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO);
-	if (interruptibility & 3)
-		vmcs_write32(GUEST_INTERRUPTIBILITY_INFO,
-			     interruptibility & ~3);
-}
-
 static int emulator_read_std(unsigned long addr,
 			     unsigned long *val,
 			     unsigned int bytes,
@@ -1694,16 +1681,10 @@ struct x86_emulate_ops emulate_ops = {
 	.cmpxchg_emulated    = emulator_cmpxchg_emulated,
 };
 
-enum emulation_result {
-	EMULATE_DONE,       /* no further processing */
-	EMULATE_DO_MMIO,      /* kvm_run filled with mmio request */
-	EMULATE_FAIL,         /* can't emulate this instruction */
-};
-
-static int emulate_instruction(struct kvm_vcpu *vcpu,
-			       struct kvm_run *run,
-			       unsigned long cr2,
-			       u16 error_code)
+int emulate_instruction(struct kvm_vcpu *vcpu,
+			struct kvm_run *run,
+			unsigned long cr2,
+			u16 error_code)
 {
 	struct x86_emulate_ctxt emulate_ctxt;
 	int r;
@@ -1762,6 +1743,7 @@ static int emulate_instruction(struct kv
 
 	return EMULATE_DONE;
 }
+EXPORT_SYMBOL_GPL(emulate_instruction);
 
 static u64 mk_cr_64(u64 curr_cr, u32 new_val)
 {
@@ -1826,298 +1808,6 @@ void realmode_set_cr(struct kvm_vcpu *vc
 	}
 }
 
-static int handle_rmode_exception(struct kvm_vcpu *vcpu,
-				  int vec, u32 err_code)
-{
-	if (!vcpu->rmode.active)
-		return 0;
-
-	if (vec == GP_VECTOR && err_code == 0)
-		if (emulate_instruction(vcpu, 0, 0, 0) == EMULATE_DONE)
-			return 1;
-	return 0;
-}
-
-static int handle_exception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u32 intr_info, error_code;
-	unsigned long cr2, rip;
-	u32 vect_info;
-	enum emulation_result er;
-
-	vect_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
-	intr_info = vmcs_read32(VM_EXIT_INTR_INFO);
-
-	if ((vect_info & VECTORING_INFO_VALID_MASK) &&
-						!is_page_fault(intr_info)) {
-		printk(KERN_ERR "%s: unexpected, vectoring info 0x%x "
-		       "intr info 0x%x\n", __FUNCTION__, vect_info, intr_info);
-	}
-
-	if (is_external_interrupt(vect_info)) {
-		int irq = vect_info & VECTORING_INFO_VECTOR_MASK;
-		set_bit(irq, vcpu->irq_pending);
-		set_bit(irq / BITS_PER_LONG, &vcpu->irq_summary);
-	}
-
-	if ((intr_info & INTR_INFO_INTR_TYPE_MASK) == 0x200) { /* nmi */
-		asm ("int $2");
-		return 1;
-	}
-	error_code = 0;
-	rip = vmcs_readl(GUEST_RIP);
-	if (intr_info & INTR_INFO_DELIEVER_CODE_MASK)
-		error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE);
-	if (is_page_fault(intr_info)) {
-		cr2 = vmcs_readl(EXIT_QUALIFICATION);
-
-		spin_lock(&vcpu->kvm->lock);
-		if (!vcpu->mmu.page_fault(vcpu, cr2, error_code)) {
-			spin_unlock(&vcpu->kvm->lock);
-			return 1;
-		}
-
-		er = emulate_instruction(vcpu, kvm_run, cr2, error_code);
-		spin_unlock(&vcpu->kvm->lock);
-
-		switch (er) {
-		case EMULATE_DONE:
-			return 1;
-		case EMULATE_DO_MMIO:
-			++kvm_stat.mmio_exits;
-			kvm_run->exit_reason = KVM_EXIT_MMIO;
-			return 0;
-		 case EMULATE_FAIL:
-			vcpu_printf(vcpu, "%s: emulate fail\n", __FUNCTION__);
-			break;
-		default:
-			BUG();
-		}
-	}
-
-	if (vcpu->rmode.active &&
-	    handle_rmode_exception(vcpu, intr_info & INTR_INFO_VECTOR_MASK,
-								error_code))
-		return 1;
-
-	if ((intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK)) == (INTR_TYPE_EXCEPTION | 1)) {
-		kvm_run->exit_reason = KVM_EXIT_DEBUG;
-		return 0;
-	}
-	kvm_run->exit_reason = KVM_EXIT_EXCEPTION;
-	kvm_run->ex.exception = intr_info & INTR_INFO_VECTOR_MASK;
-	kvm_run->ex.error_code = error_code;
-	return 0;
-}
-
-static int handle_external_interrupt(struct kvm_vcpu *vcpu,
-				     struct kvm_run *kvm_run)
-{
-	++kvm_stat.irq_exits;
-	return 1;
-}
-
-
-static int get_io_count(struct kvm_vcpu *vcpu, u64 *count)
-{
-	u64 inst;
-	gva_t rip;
-	int countr_size;
-	int i, n;
-
-	if ((vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_VM)) {
-		countr_size = 2;
-	} else {
-		u32 cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
-
-		countr_size = (cs_ar & AR_L_MASK) ? 8:
-			      (cs_ar & AR_DB_MASK) ? 4: 2;
-	}
-
-	rip =  vmcs_readl(GUEST_RIP);
-	if (countr_size != 8)
-		rip += vmcs_readl(GUEST_CS_BASE);
-
-	n = kvm_read_guest(vcpu, rip, sizeof(inst), &inst);
-
-	for (i = 0; i < n; i++) {
-		switch (((u8*)&inst)[i]) {
-		case 0xf0:
-		case 0xf2:
-		case 0xf3:
-		case 0x2e:
-		case 0x36:
-		case 0x3e:
-		case 0x26:
-		case 0x64:
-		case 0x65:
-		case 0x66:
-			break;
-		case 0x67:
-			countr_size = (countr_size == 2) ? 4: (countr_size >> 1);
-		default:
-			goto done;
-		}
-	}
-	return 0;
-done:
-	countr_size *= 8;
-	*count = vcpu->regs[VCPU_REGS_RCX] & (~0ULL >> (64 - countr_size));
-	return 1;
-}
-
-static int handle_io(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u64 exit_qualification;
-
-	++kvm_stat.io_exits;
-	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
-	kvm_run->exit_reason = KVM_EXIT_IO;
-	if (exit_qualification & 8)
-		kvm_run->io.direction = KVM_EXIT_IO_IN;
-	else
-		kvm_run->io.direction = KVM_EXIT_IO_OUT;
-	kvm_run->io.size = (exit_qualification & 7) + 1;
-	kvm_run->io.string = (exit_qualification & 16) != 0;
-	kvm_run->io.string_down
-		= (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_DF) != 0;
-	kvm_run->io.rep = (exit_qualification & 32) != 0;
-	kvm_run->io.port = exit_qualification >> 16;
-	if (kvm_run->io.string) {
-		if (!get_io_count(vcpu, &kvm_run->io.count))
-			return 1;
-		kvm_run->io.address = vmcs_readl(GUEST_LINEAR_ADDRESS);
-	} else
-		kvm_run->io.value = vcpu->regs[VCPU_REGS_RAX]; /* rax */
-	return 0;
-}
-
-static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u64 address = vmcs_read64(EXIT_QUALIFICATION);
-	int instruction_length = vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
-	spin_lock(&vcpu->kvm->lock);
-	vcpu->mmu.inval_page(vcpu, address);
-	spin_unlock(&vcpu->kvm->lock);
-	vmcs_writel(GUEST_RIP, vmcs_readl(GUEST_RIP) + instruction_length);
-	return 1;
-}
-
-static int handle_cr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u64 exit_qualification;
-	int cr;
-	int reg;
-
-#ifdef KVM_DEBUG
-	if (guest_cpl() != 0) {
-		vcpu_printf(vcpu, "%s: not supervisor\n", __FUNCTION__);
-		inject_gp(vcpu);
-		return 1;
-	}
-#endif
-
-	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
-	cr = exit_qualification & 15;
-	reg = (exit_qualification >> 8) & 15;
-	switch ((exit_qualification >> 4) & 3) {
-	case 0: /* mov to cr */
-		switch (cr) {
-		case 0:
-			kvm_arch_ops->cache_regs(vcpu);
-			set_cr0(vcpu, vcpu->regs[reg]);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		case 3:
-			kvm_arch_ops->cache_regs(vcpu);
-			set_cr3(vcpu, vcpu->regs[reg]);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		case 4:
-			kvm_arch_ops->cache_regs(vcpu);
-			set_cr4(vcpu, vcpu->regs[reg]);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		case 8:
-			kvm_arch_ops->cache_regs(vcpu);
-			set_cr8(vcpu, vcpu->regs[reg]);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		};
-		break;
-	case 1: /*mov from cr*/
-		switch (cr) {
-		case 3:
-			kvm_arch_ops->cache_regs(vcpu);
-			vcpu->regs[reg] = vcpu->cr3;
-			kvm_arch_ops->decache_regs(vcpu);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		case 8:
-			printk(KERN_DEBUG "handle_cr: read CR8 "
-			       "cpu erratum AA15\n");
-			kvm_arch_ops->cache_regs(vcpu);
-			vcpu->regs[reg] = vcpu->cr8;
-			kvm_arch_ops->decache_regs(vcpu);
-			skip_emulated_instruction(vcpu);
-			return 1;
-		}
-		break;
-	case 3: /* lmsw */
-		lmsw(vcpu, (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f);
-
-		skip_emulated_instruction(vcpu);
-		return 1;
-	default:
-		break;
-	}
-	kvm_run->exit_reason = 0;
-	printk(KERN_ERR "kvm: unhandled control register: op %d cr %d\n",
-	       (int)(exit_qualification >> 4) & 3, cr);
-	return 0;
-}
-
-static int handle_dr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u64 exit_qualification;
-	unsigned long val;
-	int dr, reg;
-
-	/*
-	 * FIXME: this code assumes the host is debugging the guest.
-	 *        need to deal with guest debugging itself too.
-	 */
-	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
-	dr = exit_qualification & 7;
-	reg = (exit_qualification >> 8) & 15;
-	kvm_arch_ops->cache_regs(vcpu);
-	if (exit_qualification & 16) {
-		/* mov from dr */
-		switch (dr) {
-		case 6:
-			val = 0xffff0ff0;
-			break;
-		case 7:
-			val = 0x400;
-			break;
-		default:
-			val = 0;
-		}
-		vcpu->regs[reg] = val;
-	} else {
-		/* mov to dr */
-	}
-	kvm_arch_ops->decache_regs(vcpu);
-	skip_emulated_instruction(vcpu);
-	return 1;
-}
-
-static int handle_cpuid(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	kvm_run->exit_reason = KVM_EXIT_CPUID;
-	return 0;
-}
-
 /*
  * Reads an msr value (of 'msr_index') into 'pdata'.
  * Returns 0 on success, non-0 otherwise.
@@ -2128,23 +1818,6 @@ static int get_msr(struct kvm_vcpu *vcpu
 	return kvm_arch_ops->get_msr(vcpu, msr_index, pdata);
 }
 
-static int handle_rdmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u32 ecx = vcpu->regs[VCPU_REGS_RCX];
-	u64 data;
-
-	if (get_msr(vcpu, ecx, &data)) {
-		inject_gp(vcpu);
-		return 1;
-	}
-
-	/* FIXME: handling of bits 32:63 of rax, rdx */
-	vcpu->regs[VCPU_REGS_RAX] = data & -1u;
-	vcpu->regs[VCPU_REGS_RDX] = (data >> 32) & -1u;
-	skip_emulated_instruction(vcpu);
-	return 1;
-}
-
 #ifdef __x86_64__
 
 void set_efer(struct kvm_vcpu *vcpu, u64 efer)
@@ -2175,7 +1848,6 @@ void set_efer(struct kvm_vcpu *vcpu, u64
 	if (!(efer & EFER_LMA))
 	    efer &= ~EFER_LME;
 	msr->data = efer;
-	skip_emulated_instruction(vcpu);
 }
 EXPORT_SYMBOL_GPL(set_efer);
 
@@ -2191,90 +1863,6 @@ static int set_msr(struct kvm_vcpu *vcpu
 	return kvm_arch_ops->set_msr(vcpu, msr_index, data);
 }
 
-static int handle_wrmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	u32 ecx = vcpu->regs[VCPU_REGS_RCX];
-	u64 data = (vcpu->regs[VCPU_REGS_RAX] & -1u)
-		| ((u64)(vcpu->regs[VCPU_REGS_RDX] & -1u) << 32);
-
-	if (set_msr(vcpu, ecx, data) != 0) {
-		inject_gp(vcpu);
-		return 1;
-	}
-
-	if (ecx != MSR_EFER)
-		skip_emulated_instruction(vcpu);
-	return 1;
-}
-
-static int handle_interrupt_window(struct kvm_vcpu *vcpu,
-				   struct kvm_run *kvm_run)
-{
-	/* Turn off interrupt window reporting. */
-	vmcs_write32(CPU_BASED_VM_EXEC_CONTROL,
-		     vmcs_read32(CPU_BASED_VM_EXEC_CONTROL)
-		     & ~CPU_BASED_VIRTUAL_INTR_PENDING);
-	return 1;
-}
-
-static int handle_halt(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
-{
-	skip_emulated_instruction(vcpu);
-	if (vcpu->irq_summary && (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF))
-		return 1;
-
-	kvm_run->exit_reason = KVM_EXIT_HLT;
-	return 0;
-}
-
-/*
- * The exit handlers return 1 if the exit was handled fully and guest execution
- * may resume.  Otherwise they set the kvm_run parameter to indicate what needs
- * to be done to userspace and return 0.
- */
-static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu,
-				      struct kvm_run *kvm_run) = {
-	[EXIT_REASON_EXCEPTION_NMI]           = handle_exception,
-	[EXIT_REASON_EXTERNAL_INTERRUPT]      = handle_external_interrupt,
-	[EXIT_REASON_IO_INSTRUCTION]          = handle_io,
-	[EXIT_REASON_INVLPG]                  = handle_invlpg,
-	[EXIT_REASON_CR_ACCESS]               = handle_cr,
-	[EXIT_REASON_DR_ACCESS]               = handle_dr,
-	[EXIT_REASON_CPUID]                   = handle_cpuid,
-	[EXIT_REASON_MSR_READ]                = handle_rdmsr,
-	[EXIT_REASON_MSR_WRITE]               = handle_wrmsr,
-	[EXIT_REASON_PENDING_INTERRUPT]       = handle_interrupt_window,
-	[EXIT_REASON_HLT]                     = handle_halt,
-};
-
-static const int kvm_vmx_max_exit_handlers =
-	sizeof(kvm_vmx_exit_handlers) / sizeof(*kvm_vmx_exit_handlers);
-
-/*
- * The guest has exited.  See if we can fix it or if we need userspace
- * assistance.
- */
-int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
-{
-	u32 vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
-	u32 exit_reason = vmcs_read32(VM_EXIT_REASON);
-
-	if ( (vectoring_info & VECTORING_INFO_VALID_MASK) &&
-				exit_reason != EXIT_REASON_EXCEPTION_NMI )
-		printk(KERN_WARNING "%s: unexpected, valid vectoring info and "
-		       "exit reason is 0x%x\n", __FUNCTION__, exit_reason);
-	kvm_run->instruction_length = vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
-	if (exit_reason < kvm_vmx_max_exit_handlers
-	    && kvm_vmx_exit_handlers[exit_reason])
-		return kvm_vmx_exit_handlers[exit_reason](vcpu, kvm_run);
-	else {
-		kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
-		kvm_run->hw.hardware_exit_reason = exit_reason;
-	}
-	return 0;
-}
-EXPORT_SYMBOL_GPL(kvm_handle_exit);
-
 void kvm_resched(struct kvm_vcpu *vcpu)
 {
 	vcpu_put(vcpu);
@@ -2315,7 +1903,7 @@ static int kvm_dev_ioctl_run(struct kvm 
 		return -ENOENT;
 
 	if (kvm_run->emulated) {
-		skip_emulated_instruction(vcpu);
+		kvm_arch_ops->skip_emulated_instruction(vcpu);
 		kvm_run->emulated = 0;
 	}
 
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -44,11 +44,24 @@ u64 guest_read_tsc(void);
 void guest_write_tsc(u64 guest_tsc);
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
-#ifdef __x86_64__
+static void skip_emulated_instruction(struct kvm_vcpu *vcpu)
+{
+	unsigned long rip;
+	u32 interruptibility;
 
-void set_efer(struct kvm_vcpu *vcpu, u64 efer);
+	rip = vmcs_readl(GUEST_RIP);
+	rip += vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
+	vmcs_writel(GUEST_RIP, rip);
 
-#endif
+	/*
+	 * We emulated an instruction, so temporary interrupt blocking
+	 * should be removed, if set.
+	 */
+	interruptibility = vmcs_read32(GUEST_INTERRUPTIBILITY_INFO);
+	if (interruptibility & 3)
+		vmcs_write32(GUEST_INTERRUPTIBILITY_INFO,
+			     interruptibility & ~3);
+}
 
 static void reload_tss(void)
 {
@@ -521,6 +534,397 @@ static void kvm_guest_debug_pre(struct k
 	}
 }
 
+static int handle_rmode_exception(struct kvm_vcpu *vcpu,
+				  int vec, u32 err_code)
+{
+	if (!vcpu->rmode.active)
+		return 0;
+
+	if (vec == GP_VECTOR && err_code == 0)
+		if (emulate_instruction(vcpu, 0, 0, 0) == EMULATE_DONE)
+			return 1;
+	return 0;
+}
+
+static int handle_exception(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u32 intr_info, error_code;
+	unsigned long cr2, rip;
+	u32 vect_info;
+	enum emulation_result er;
+
+	vect_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
+	intr_info = vmcs_read32(VM_EXIT_INTR_INFO);
+
+	if ((vect_info & VECTORING_INFO_VALID_MASK) &&
+						!is_page_fault(intr_info)) {
+		printk(KERN_ERR "%s: unexpected, vectoring info 0x%x "
+		       "intr info 0x%x\n", __FUNCTION__, vect_info, intr_info);
+	}
+
+	if (is_external_interrupt(vect_info)) {
+		int irq = vect_info & VECTORING_INFO_VECTOR_MASK;
+		set_bit(irq, vcpu->irq_pending);
+		set_bit(irq / BITS_PER_LONG, &vcpu->irq_summary);
+	}
+
+	if ((intr_info & INTR_INFO_INTR_TYPE_MASK) == 0x200) { /* nmi */
+		asm ("int $2");
+		return 1;
+	}
+	error_code = 0;
+	rip = vmcs_readl(GUEST_RIP);
+	if (intr_info & INTR_INFO_DELIEVER_CODE_MASK)
+		error_code = vmcs_read32(VM_EXIT_INTR_ERROR_CODE);
+	if (is_page_fault(intr_info)) {
+		cr2 = vmcs_readl(EXIT_QUALIFICATION);
+
+		spin_lock(&vcpu->kvm->lock);
+		if (!vcpu->mmu.page_fault(vcpu, cr2, error_code)) {
+			spin_unlock(&vcpu->kvm->lock);
+			return 1;
+		}
+
+		er = emulate_instruction(vcpu, kvm_run, cr2, error_code);
+		spin_unlock(&vcpu->kvm->lock);
+
+		switch (er) {
+		case EMULATE_DONE:
+			return 1;
+		case EMULATE_DO_MMIO:
+			++kvm_stat.mmio_exits;
+			kvm_run->exit_reason = KVM_EXIT_MMIO;
+			return 0;
+		 case EMULATE_FAIL:
+			vcpu_printf(vcpu, "%s: emulate fail\n", __FUNCTION__);
+			break;
+		default:
+			BUG();
+		}
+	}
+
+	if (vcpu->rmode.active &&
+	    handle_rmode_exception(vcpu, intr_info & INTR_INFO_VECTOR_MASK,
+								error_code))
+		return 1;
+
+	if ((intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK)) == (INTR_TYPE_EXCEPTION | 1)) {
+		kvm_run->exit_reason = KVM_EXIT_DEBUG;
+		return 0;
+	}
+	kvm_run->exit_reason = KVM_EXIT_EXCEPTION;
+	kvm_run->ex.exception = intr_info & INTR_INFO_VECTOR_MASK;
+	kvm_run->ex.error_code = error_code;
+	return 0;
+}
+
+static int handle_external_interrupt(struct kvm_vcpu *vcpu,
+				     struct kvm_run *kvm_run)
+{
+	++kvm_stat.irq_exits;
+	return 1;
+}
+
+
+static int get_io_count(struct kvm_vcpu *vcpu, u64 *count)
+{
+	u64 inst;
+	gva_t rip;
+	int countr_size;
+	int i, n;
+
+	if ((vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_VM)) {
+		countr_size = 2;
+	} else {
+		u32 cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
+
+		countr_size = (cs_ar & AR_L_MASK) ? 8:
+			      (cs_ar & AR_DB_MASK) ? 4: 2;
+	}
+
+	rip =  vmcs_readl(GUEST_RIP);
+	if (countr_size != 8)
+		rip += vmcs_readl(GUEST_CS_BASE);
+
+	n = kvm_read_guest(vcpu, rip, sizeof(inst), &inst);
+
+	for (i = 0; i < n; i++) {
+		switch (((u8*)&inst)[i]) {
+		case 0xf0:
+		case 0xf2:
+		case 0xf3:
+		case 0x2e:
+		case 0x36:
+		case 0x3e:
+		case 0x26:
+		case 0x64:
+		case 0x65:
+		case 0x66:
+			break;
+		case 0x67:
+			countr_size = (countr_size == 2) ? 4: (countr_size >> 1);
+		default:
+			goto done;
+		}
+	}
+	return 0;
+done:
+	countr_size *= 8;
+	*count = vcpu->regs[VCPU_REGS_RCX] & (~0ULL >> (64 - countr_size));
+	return 1;
+}
+
+static int handle_io(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u64 exit_qualification;
+
+	++kvm_stat.io_exits;
+	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
+	kvm_run->exit_reason = KVM_EXIT_IO;
+	if (exit_qualification & 8)
+		kvm_run->io.direction = KVM_EXIT_IO_IN;
+	else
+		kvm_run->io.direction = KVM_EXIT_IO_OUT;
+	kvm_run->io.size = (exit_qualification & 7) + 1;
+	kvm_run->io.string = (exit_qualification & 16) != 0;
+	kvm_run->io.string_down
+		= (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_DF) != 0;
+	kvm_run->io.rep = (exit_qualification & 32) != 0;
+	kvm_run->io.port = exit_qualification >> 16;
+	if (kvm_run->io.string) {
+		if (!get_io_count(vcpu, &kvm_run->io.count))
+			return 1;
+		kvm_run->io.address = vmcs_readl(GUEST_LINEAR_ADDRESS);
+	} else
+		kvm_run->io.value = vcpu->regs[VCPU_REGS_RAX]; /* rax */
+	return 0;
+}
+
+static int handle_invlpg(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u64 address = vmcs_read64(EXIT_QUALIFICATION);
+	int instruction_length = vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
+	spin_lock(&vcpu->kvm->lock);
+	vcpu->mmu.inval_page(vcpu, address);
+	spin_unlock(&vcpu->kvm->lock);
+	vmcs_writel(GUEST_RIP, vmcs_readl(GUEST_RIP) + instruction_length);
+	return 1;
+}
+
+static int handle_cr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u64 exit_qualification;
+	int cr;
+	int reg;
+
+#ifdef KVM_DEBUG
+	if (guest_cpl() != 0) {
+		vcpu_printf(vcpu, "%s: not supervisor\n", __FUNCTION__);
+		inject_gp(vcpu);
+		return 1;
+	}
+#endif
+
+	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
+	cr = exit_qualification & 15;
+	reg = (exit_qualification >> 8) & 15;
+	switch ((exit_qualification >> 4) & 3) {
+	case 0: /* mov to cr */
+		switch (cr) {
+		case 0:
+			vcpu_load_rsp_rip(vcpu);
+			set_cr0(vcpu, vcpu->regs[reg]);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		case 3:
+			vcpu_load_rsp_rip(vcpu);
+			set_cr3(vcpu, vcpu->regs[reg]);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		case 4:
+			vcpu_load_rsp_rip(vcpu);
+			set_cr4(vcpu, vcpu->regs[reg]);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		case 8:
+			vcpu_load_rsp_rip(vcpu);
+			set_cr8(vcpu, vcpu->regs[reg]);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		};
+		break;
+	case 1: /*mov from cr*/
+		switch (cr) {
+		case 3:
+			vcpu_load_rsp_rip(vcpu);
+			vcpu->regs[reg] = vcpu->cr3;
+			vcpu_put_rsp_rip(vcpu);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		case 8:
+			printk(KERN_DEBUG "handle_cr: read CR8 "
+			       "cpu erratum AA15\n");
+			vcpu_load_rsp_rip(vcpu);
+			vcpu->regs[reg] = vcpu->cr8;
+			vcpu_put_rsp_rip(vcpu);
+			skip_emulated_instruction(vcpu);
+			return 1;
+		}
+		break;
+	case 3: /* lmsw */
+		lmsw(vcpu, (exit_qualification >> LMSW_SOURCE_DATA_SHIFT) & 0x0f);
+
+		skip_emulated_instruction(vcpu);
+		return 1;
+	default:
+		break;
+	}
+	kvm_run->exit_reason = 0;
+	printk(KERN_ERR "kvm: unhandled control register: op %d cr %d\n",
+	       (int)(exit_qualification >> 4) & 3, cr);
+	return 0;
+}
+
+static int handle_dr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u64 exit_qualification;
+	unsigned long val;
+	int dr, reg;
+
+	/*
+	 * FIXME: this code assumes the host is debugging the guest.
+	 *        need to deal with guest debugging itself too.
+	 */
+	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
+	dr = exit_qualification & 7;
+	reg = (exit_qualification >> 8) & 15;
+	vcpu_load_rsp_rip(vcpu);
+	if (exit_qualification & 16) {
+		/* mov from dr */
+		switch (dr) {
+		case 6:
+			val = 0xffff0ff0;
+			break;
+		case 7:
+			val = 0x400;
+			break;
+		default:
+			val = 0;
+		}
+		vcpu->regs[reg] = val;
+	} else {
+		/* mov to dr */
+	}
+	vcpu_put_rsp_rip(vcpu);
+	skip_emulated_instruction(vcpu);
+	return 1;
+}
+
+static int handle_cpuid(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	kvm_run->exit_reason = KVM_EXIT_CPUID;
+	return 0;
+}
+
+static int handle_rdmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u32 ecx = vcpu->regs[VCPU_REGS_RCX];
+	u64 data;
+
+	if (vmx_get_msr(vcpu, ecx, &data)) {
+		inject_gp(vcpu);
+		return 1;
+	}
+
+	/* FIXME: handling of bits 32:63 of rax, rdx */
+	vcpu->regs[VCPU_REGS_RAX] = data & -1u;
+	vcpu->regs[VCPU_REGS_RDX] = (data >> 32) & -1u;
+	skip_emulated_instruction(vcpu);
+	return 1;
+}
+
+static int handle_wrmsr(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	u32 ecx = vcpu->regs[VCPU_REGS_RCX];
+	u64 data = (vcpu->regs[VCPU_REGS_RAX] & -1u)
+		| ((u64)(vcpu->regs[VCPU_REGS_RDX] & -1u) << 32);
+
+	if (vmx_set_msr(vcpu, ecx, data) != 0) {
+		inject_gp(vcpu);
+		return 1;
+	}
+
+	skip_emulated_instruction(vcpu);
+	return 1;
+}
+
+static int handle_interrupt_window(struct kvm_vcpu *vcpu,
+				   struct kvm_run *kvm_run)
+{
+	/* Turn off interrupt window reporting. */
+	vmcs_write32(CPU_BASED_VM_EXEC_CONTROL,
+		     vmcs_read32(CPU_BASED_VM_EXEC_CONTROL)
+		     & ~CPU_BASED_VIRTUAL_INTR_PENDING);
+	return 1;
+}
+
+static int handle_halt(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
+{
+	skip_emulated_instruction(vcpu);
+	if (vcpu->irq_summary && (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF))
+		return 1;
+
+	kvm_run->exit_reason = KVM_EXIT_HLT;
+	return 0;
+}
+
+/*
+ * The exit handlers return 1 if the exit was handled fully and guest execution
+ * may resume.  Otherwise they set the kvm_run parameter to indicate what needs
+ * to be done to userspace and return 0.
+ */
+static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu,
+				      struct kvm_run *kvm_run) = {
+	[EXIT_REASON_EXCEPTION_NMI]           = handle_exception,
+	[EXIT_REASON_EXTERNAL_INTERRUPT]      = handle_external_interrupt,
+	[EXIT_REASON_IO_INSTRUCTION]          = handle_io,
+	[EXIT_REASON_INVLPG]                  = handle_invlpg,
+	[EXIT_REASON_CR_ACCESS]               = handle_cr,
+	[EXIT_REASON_DR_ACCESS]               = handle_dr,
+	[EXIT_REASON_CPUID]                   = handle_cpuid,
+	[EXIT_REASON_MSR_READ]                = handle_rdmsr,
+	[EXIT_REASON_MSR_WRITE]               = handle_wrmsr,
+	[EXIT_REASON_PENDING_INTERRUPT]       = handle_interrupt_window,
+	[EXIT_REASON_HLT]                     = handle_halt,
+};
+
+static const int kvm_vmx_max_exit_handlers =
+	sizeof(kvm_vmx_exit_handlers) / sizeof(*kvm_vmx_exit_handlers);
+
+/*
+ * The guest has exited.  See if we can fix it or if we need userspace
+ * assistance.
+ */
+static int kvm_handle_exit(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
+{
+	u32 vectoring_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
+	u32 exit_reason = vmcs_read32(VM_EXIT_REASON);
+
+	if ( (vectoring_info & VECTORING_INFO_VALID_MASK) &&
+				exit_reason != EXIT_REASON_EXCEPTION_NMI )
+		printk(KERN_WARNING "%s: unexpected, valid vectoring info and "
+		       "exit reason is 0x%x\n", __FUNCTION__, exit_reason);
+	kvm_run->instruction_length = vmcs_read32(VM_EXIT_INSTRUCTION_LEN);
+	if (exit_reason < kvm_vmx_max_exit_handlers
+	    && kvm_vmx_exit_handlers[exit_reason])
+		return kvm_vmx_exit_handlers[exit_reason](vcpu, kvm_run);
+	else {
+		kvm_run->exit_reason = KVM_EXIT_UNKNOWN;
+		kvm_run->hw.hardware_exit_reason = exit_reason;
+	}
+	return 0;
+}
+
 static int vmx_vcpu_run(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
 {
 	u8 fail;
@@ -753,6 +1157,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.decache_regs = vcpu_put_rsp_rip,
 
 	.run = vmx_vcpu_run,
+	.skip_emulated_instruction = skip_emulated_instruction,
 	.vmx_return = (unsigned long)kvm_vmx_return,
 };
 

[PATCH 16/38] KVM: Make vcpu_setup() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -153,6 +153,17 @@ enum {
 	NR_VCPU_REGS
 };
 
+enum {
+	VCPU_SREG_CS,
+	VCPU_SREG_DS,
+	VCPU_SREG_ES,
+	VCPU_SREG_FS,
+	VCPU_SREG_GS,
+	VCPU_SREG_SS,
+	VCPU_SREG_TR,
+	VCPU_SREG_LDTR,
+};
+
 struct kvm_vcpu {
 	struct kvm *kvm;
 	struct vmcs *vmcs;
@@ -266,8 +277,8 @@ struct kvm_arch_ops {
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
 
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
+	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
-	unsigned long vmx_return; /* temporary hack */
 };
 
 extern struct kvm_stat kvm_stat;
@@ -324,12 +335,18 @@ void set_cr4(struct kvm_vcpu *vcpu, unsi
 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 
+void __set_efer(struct kvm_vcpu *vcpu, u64 efer);
+void __set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
+void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
+
+int rmode_tss_base(struct kvm* kvm); /* temporary hack */
 void inject_gp(struct kvm_vcpu *vcpu);
 
 #ifdef __x86_64__
 void set_efer(struct kvm_vcpu *vcpu, u64 efer);
 #endif
 
+void fx_init(struct kvm_vcpu *vcpu);
 
 void load_msrs(struct vmx_msr_entry *e, int n);
 void save_msrs(struct vmx_msr_entry *e, int n);
@@ -348,6 +365,12 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 void vmcs_writel(unsigned long field, unsigned long value);
 unsigned long vmcs_readl(unsigned long field);
 
+static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
+{
+	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
+	return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : 0;
+}
+
 static inline u16 vmcs_read16(unsigned long field)
 {
 	return vmcs_readl(field);
@@ -465,6 +488,11 @@ static inline void load_ldt(u16 sel)
 }
 #endif
 
+static inline void get_idt(struct descriptor_table *table)
+{
+	asm ("sidt %0" : "=m"(*table));
+}
+
 #ifdef __x86_64__
 static inline unsigned long read_msr(unsigned long msr)
 {
@@ -490,6 +518,11 @@ static inline void fpu_init(void)
 	asm ("finit");
 }
 
+static inline u32 get_rdx_init_val(void)
+{
+	return 0x600; /* P6 family */
+}
+
 #define ASM_VMX_VMCLEAR_RAX       ".byte 0x66, 0x0f, 0xc7, 0x30"
 #define ASM_VMX_VMLAUNCH          ".byte 0x0f, 0x01, 0xc2"
 #define ASM_VMX_VMRESUME          ".byte 0x0f, 0x01, 0xc3"
@@ -502,6 +535,12 @@ static inline void fpu_init(void)
 
 #define MSR_IA32_TIME_STAMP_COUNTER		0x010
 
+#define TSS_IOPB_BASE_OFFSET 0x66
+#define TSS_BASE_SIZE 0x68
+#define TSS_IOPB_SIZE (65536 / 8)
+#define TSS_REDIRECTION_SIZE (256 / 8)
+#define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
+
 #ifdef __x86_64__
 
 /*
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -65,17 +65,6 @@ static struct kvm_stats_debugfs_item {
 
 static struct dentry *debugfs_dir;
 
-enum {
-	VCPU_SREG_CS,
-	VCPU_SREG_DS,
-	VCPU_SREG_ES,
-	VCPU_SREG_FS,
-	VCPU_SREG_GS,
-	VCPU_SREG_SS,
-	VCPU_SREG_TR,
-	VCPU_SREG_LDTR,
-};
-
 #define VMX_SEGMENT_FIELD(seg)					\
 	[VCPU_SREG_##seg] {                                     \
 		GUEST_##seg##_SELECTOR,				\
@@ -101,26 +90,6 @@ struct kvm_vmx_segment_field {
 };
 EXPORT_SYMBOL_GPL(kvm_vmx_segment_fields);
 
-static const u32 vmx_msr_index[] = {
-#ifdef __x86_64__
-	MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, MSR_KERNEL_GS_BASE,
-#endif
-	MSR_EFER, MSR_K6_STAR,
-};
-#define NR_VMX_MSR (sizeof(vmx_msr_index) / sizeof(*vmx_msr_index))
-
-#define TSS_IOPB_BASE_OFFSET 0x66
-#define TSS_BASE_SIZE 0x68
-#define TSS_IOPB_SIZE (65536 / 8)
-#define TSS_REDIRECTION_SIZE (256 / 8)
-#define RMODE_TSS_SIZE (TSS_BASE_SIZE + TSS_REDIRECTION_SIZE + TSS_IOPB_SIZE + 1)
-
-#define MSR_IA32_FEATURE_CONTROL 		0x03a
-#define MSR_IA32_VMX_PINBASED_CTLS_MSR		0x481
-#define MSR_IA32_VMX_PROCBASED_CTLS_MSR		0x482
-#define MSR_IA32_VMX_EXIT_CTLS_MSR		0x483
-#define MSR_IA32_VMX_ENTRY_CTLS_MSR		0x484
-
 #define MAX_IO_MSRS 256
 
 #define CR0_RESEVED_BITS 0xffffffff1ffaffc0ULL
@@ -129,12 +98,6 @@ static const u32 vmx_msr_index[] = {
 #define CR8_RESEVED_BITS (~0x0fULL)
 #define EFER_RESERVED_BITS 0xfffffffffffff2fe
 
-#ifdef __x86_64__
-#define HOST_IS_64 1
-#else
-#define HOST_IS_64 0
-#endif
-
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr)
 {
 	int i;
@@ -151,11 +114,6 @@ static void get_gdt(struct descriptor_ta
 	asm ("sgdt %0" : "=m"(*table));
 }
 
-static void get_idt(struct descriptor_table *table)
-{
-	asm ("sidt %0" : "=m"(*table));
-}
-
 struct segment_descriptor {
 	u16 limit_low;
 	u16 base_low;
@@ -227,12 +185,6 @@ struct vmcs_descriptor {
 } vmcs_descriptor;
 EXPORT_SYMBOL_GPL(vmcs_descriptor);
 
-static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
-{
-	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
-	return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : 0;
-}
-
 int kvm_read_guest(struct kvm_vcpu *vcpu,
 			     gva_t addr,
 			     unsigned long size,
@@ -630,11 +582,12 @@ static void enter_pmode(struct kvm_vcpu 
 	vmcs_write32(GUEST_CS_AR_BYTES, 0x9b);
 }
 
-static int rmode_tss_base(struct kvm* kvm)
+int rmode_tss_base(struct kvm* kvm)
 {
 	gfn_t base_gfn = kvm->memslots[0].base_gfn + kvm->memslots[0].npages - 3;
 	return base_gfn << PAGE_SHIFT;
 }
+EXPORT_SYMBOL_GPL(rmode_tss_base);
 
 static void fix_rmode_seg(int seg, struct kvm_save_segment *save)
 {
@@ -686,41 +639,9 @@ static void enter_rmode(struct kvm_vcpu 
 	fix_rmode_seg(VCPU_SREG_FS, &vcpu->rmode.fs);
 }
 
-static int init_rmode_tss(struct kvm* kvm)
-{
-	struct page *p1, *p2, *p3;
-	gfn_t fn = rmode_tss_base(kvm) >> PAGE_SHIFT;
-	char *page;
-
-	p1 = _gfn_to_page(kvm, fn++);
-	p2 = _gfn_to_page(kvm, fn++);
-	p3 = _gfn_to_page(kvm, fn);
-
-	if (!p1 || !p2 || !p3) {
-		kvm_printf(kvm,"%s: gfn_to_page failed\n", __FUNCTION__);
-		return 0;
-	}
-
-	page = kmap_atomic(p1, KM_USER0);
-	memset(page, 0, PAGE_SIZE);
-	*(u16*)(page + 0x66) = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE;
-	kunmap_atomic(page, KM_USER0);
-
-	page = kmap_atomic(p2, KM_USER0);
-	memset(page, 0, PAGE_SIZE);
-	kunmap_atomic(page, KM_USER0);
-
-	page = kmap_atomic(p3, KM_USER0);
-	memset(page, 0, PAGE_SIZE);
-	*(page + RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1) = ~0;
-	kunmap_atomic(page, KM_USER0);
-
-	return 1;
-}
-
 #ifdef __x86_64__
 
-static void __set_efer(struct kvm_vcpu *vcpu, u64 efer)
+void __set_efer(struct kvm_vcpu *vcpu, u64 efer)
 {
 	struct vmx_msr_entry *msr = find_msr_entry(vcpu, MSR_EFER);
 
@@ -739,6 +660,7 @@ static void __set_efer(struct kvm_vcpu *
 		msr->data = efer & ~EFER_LME;
 	}
 }
+EXPORT_SYMBOL_GPL(__set_efer);
 
 static void enter_lmode(struct kvm_vcpu *vcpu)
 {
@@ -772,7 +694,7 @@ static void exit_lmode(struct kvm_vcpu *
 
 #endif
 
-static void __set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+void __set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
 {
 	if (vcpu->rmode.active && (cr0 & CR0_PE_MASK))
 		enter_pmode(vcpu);
@@ -794,6 +716,7 @@ static void __set_cr0(struct kvm_vcpu *v
 		    (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
 	vcpu->cr0 = cr0;
 }
+EXPORT_SYMBOL_GPL(__set_cr0);
 
 static int pdptrs_have_reserved_bits_set(struct kvm_vcpu *vcpu,
 					 unsigned long cr3)
@@ -899,13 +822,14 @@ void lmsw(struct kvm_vcpu *vcpu, unsigne
 }
 EXPORT_SYMBOL_GPL(lmsw);
 
-static void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
+void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
 	vmcs_writel(CR4_READ_SHADOW, cr4);
 	vmcs_writel(GUEST_CR4, cr4 | (vcpu->rmode.active ?
 		    KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON));
 	vcpu->cr4 = cr4;
 }
+EXPORT_SYMBOL_GPL(__set_cr4);
 
 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
@@ -981,12 +905,7 @@ void set_cr8(struct kvm_vcpu *vcpu, unsi
 }
 EXPORT_SYMBOL_GPL(set_cr8);
 
-static u32 get_rdx_init_val(void)
-{
-	return 0x600; /* P6 family */
-}
-
-static void fx_init(struct kvm_vcpu *vcpu)
+void fx_init(struct kvm_vcpu *vcpu)
 {
 	struct __attribute__ ((__packed__)) fx_image_s {
 		u16 control; //fcw
@@ -1010,237 +929,7 @@ static void fx_init(struct kvm_vcpu *vcp
 	memset(vcpu->guest_fx_image + sizeof(struct fx_image_s),
 	       0, FX_IMAGE_SIZE - sizeof(struct fx_image_s));
 }
-
-static void vmcs_write32_fixedbits(u32 msr, u32 vmcs_field, u32 val)
-{
-	u32 msr_high, msr_low;
-
-	rdmsr(msr, msr_low, msr_high);
-
-	val &= msr_high;
-	val |= msr_low;
-	vmcs_write32(vmcs_field, val);
-}
-
-static void seg_setup(int seg)
-{
-	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
-
-	vmcs_write16(sf->selector, 0);
-	vmcs_writel(sf->base, 0);
-	vmcs_write32(sf->limit, 0xffff);
-	vmcs_write32(sf->ar_bytes, 0x93);
-}
-
-/*
- * Sets up the vmcs for emulated real mode.
- */
-static int kvm_vcpu_setup(struct kvm_vcpu *vcpu)
-{
-	u32 host_sysenter_cs;
-	u32 junk;
-	unsigned long a;
-	struct descriptor_table dt;
-	int i;
-	int ret;
-	int nr_good_msrs;
-
-
-	if (!init_rmode_tss(vcpu->kvm)) {
-		ret = 0;
-		goto out;
-	}
-
-	memset(vcpu->regs, 0, sizeof(vcpu->regs));
-	vcpu->regs[VCPU_REGS_RDX] = get_rdx_init_val();
-	vcpu->cr8 = 0;
-	vcpu->apic_base = 0xfee00000 |
-			/*for vcpu 0*/ MSR_IA32_APICBASE_BSP |
-			MSR_IA32_APICBASE_ENABLE;
-
-	fx_init(vcpu);
-
-	/*
-	 * GUEST_CS_BASE should really be 0xffff0000, but VT vm86 mode
-	 * insists on having GUEST_CS_BASE == GUEST_CS_SELECTOR << 4.  Sigh.
-	 */
-	vmcs_write16(GUEST_CS_SELECTOR, 0xf000);
-	vmcs_writel(GUEST_CS_BASE, 0x000f0000);
-	vmcs_write32(GUEST_CS_LIMIT, 0xffff);
-	vmcs_write32(GUEST_CS_AR_BYTES, 0x9b);
-
-	seg_setup(VCPU_SREG_DS);
-	seg_setup(VCPU_SREG_ES);
-	seg_setup(VCPU_SREG_FS);
-	seg_setup(VCPU_SREG_GS);
-	seg_setup(VCPU_SREG_SS);
-
-	vmcs_write16(GUEST_TR_SELECTOR, 0);
-	vmcs_writel(GUEST_TR_BASE, 0);
-	vmcs_write32(GUEST_TR_LIMIT, 0xffff);
-	vmcs_write32(GUEST_TR_AR_BYTES, 0x008b);
-
-	vmcs_write16(GUEST_LDTR_SELECTOR, 0);
-	vmcs_writel(GUEST_LDTR_BASE, 0);
-	vmcs_write32(GUEST_LDTR_LIMIT, 0xffff);
-	vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082);
-
-	vmcs_write32(GUEST_SYSENTER_CS, 0);
-	vmcs_writel(GUEST_SYSENTER_ESP, 0);
-	vmcs_writel(GUEST_SYSENTER_EIP, 0);
-
-	vmcs_writel(GUEST_RFLAGS, 0x02);
-	vmcs_writel(GUEST_RIP, 0xfff0);
-	vmcs_writel(GUEST_RSP, 0);
-
-	vmcs_writel(GUEST_CR3, 0);
-
-	//todo: dr0 = dr1 = dr2 = dr3 = 0; dr6 = 0xffff0ff0
-	vmcs_writel(GUEST_DR7, 0x400);
-
-	vmcs_writel(GUEST_GDTR_BASE, 0);
-	vmcs_write32(GUEST_GDTR_LIMIT, 0xffff);
-
-	vmcs_writel(GUEST_IDTR_BASE, 0);
-	vmcs_write32(GUEST_IDTR_LIMIT, 0xffff);
-
-	vmcs_write32(GUEST_ACTIVITY_STATE, 0);
-	vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0);
-	vmcs_write32(GUEST_PENDING_DBG_EXCEPTIONS, 0);
-
-	/* I/O */
-	vmcs_write64(IO_BITMAP_A, 0);
-	vmcs_write64(IO_BITMAP_B, 0);
-
-	guest_write_tsc(0);
-
-	vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */
-
-	/* Special registers */
-	vmcs_write64(GUEST_IA32_DEBUGCTL, 0);
-
-	/* Control */
-	vmcs_write32_fixedbits(MSR_IA32_VMX_PINBASED_CTLS_MSR,
-			       PIN_BASED_VM_EXEC_CONTROL,
-			       PIN_BASED_EXT_INTR_MASK   /* 20.6.1 */
-			       | PIN_BASED_NMI_EXITING   /* 20.6.1 */
-			);
-	vmcs_write32_fixedbits(MSR_IA32_VMX_PROCBASED_CTLS_MSR,
-			       CPU_BASED_VM_EXEC_CONTROL,
-			       CPU_BASED_HLT_EXITING         /* 20.6.2 */
-			       | CPU_BASED_CR8_LOAD_EXITING    /* 20.6.2 */
-			       | CPU_BASED_CR8_STORE_EXITING   /* 20.6.2 */
-			       | CPU_BASED_UNCOND_IO_EXITING   /* 20.6.2 */
-			       | CPU_BASED_INVDPG_EXITING
-			       | CPU_BASED_MOV_DR_EXITING
-			       | CPU_BASED_USE_TSC_OFFSETING   /* 21.3 */
-			);
-
-	vmcs_write32(EXCEPTION_BITMAP, 1 << PF_VECTOR);
-	vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0);
-	vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0);
-	vmcs_write32(CR3_TARGET_COUNT, 0);           /* 22.2.1 */
-
-	vmcs_writel(HOST_CR0, read_cr0());  /* 22.2.3 */
-	vmcs_writel(HOST_CR4, read_cr4());  /* 22.2.3, 22.2.5 */
-	vmcs_writel(HOST_CR3, read_cr3());  /* 22.2.3  FIXME: shadow tables */
-
-	vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS);  /* 22.2.4 */
-	vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
-	vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
-	vmcs_write16(HOST_FS_SELECTOR, read_fs());    /* 22.2.4 */
-	vmcs_write16(HOST_GS_SELECTOR, read_gs());    /* 22.2.4 */
-	vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
-#ifdef __x86_64__
-	rdmsrl(MSR_FS_BASE, a);
-	vmcs_writel(HOST_FS_BASE, a); /* 22.2.4 */
-	rdmsrl(MSR_GS_BASE, a);
-	vmcs_writel(HOST_GS_BASE, a); /* 22.2.4 */
-#else
-	vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */
-	vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */
-#endif
-
-	vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8);  /* 22.2.4 */
-
-	get_idt(&dt);
-	vmcs_writel(HOST_IDTR_BASE, dt.base);   /* 22.2.4 */
-
-
-	vmcs_writel(HOST_RIP, kvm_arch_ops->vmx_return); /* 22.2.5 */
-
-	rdmsr(MSR_IA32_SYSENTER_CS, host_sysenter_cs, junk);
-	vmcs_write32(HOST_IA32_SYSENTER_CS, host_sysenter_cs);
-	rdmsrl(MSR_IA32_SYSENTER_ESP, a);
-	vmcs_writel(HOST_IA32_SYSENTER_ESP, a);   /* 22.2.3 */
-	rdmsrl(MSR_IA32_SYSENTER_EIP, a);
-	vmcs_writel(HOST_IA32_SYSENTER_EIP, a);   /* 22.2.3 */
-
-	ret = -ENOMEM;
-	vcpu->guest_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL);
-	if (!vcpu->guest_msrs)
-		goto out;
-	vcpu->host_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL);
-	if (!vcpu->host_msrs)
-		goto out_free_guest_msrs;
-
-	for (i = 0; i < NR_VMX_MSR; ++i) {
-		u32 index = vmx_msr_index[i];
-		u32 data_low, data_high;
-		u64 data;
-		int j = vcpu->nmsrs;
-
-		if (rdmsr_safe(index, &data_low, &data_high) < 0)
-			continue;
-		data = data_low | ((u64)data_high << 32);
-		vcpu->host_msrs[j].index = index;
-		vcpu->host_msrs[j].reserved = 0;
-		vcpu->host_msrs[j].data = data;
-		vcpu->guest_msrs[j] = vcpu->host_msrs[j];
-		++vcpu->nmsrs;
-	}
-	printk("msrs: %d\n", vcpu->nmsrs);
-
-	nr_good_msrs = vcpu->nmsrs - NR_BAD_MSRS;
-	vmcs_writel(VM_ENTRY_MSR_LOAD_ADDR,
-		    virt_to_phys(vcpu->guest_msrs + NR_BAD_MSRS));
-	vmcs_writel(VM_EXIT_MSR_STORE_ADDR,
-		    virt_to_phys(vcpu->guest_msrs + NR_BAD_MSRS));
-	vmcs_writel(VM_EXIT_MSR_LOAD_ADDR,
-		    virt_to_phys(vcpu->host_msrs + NR_BAD_MSRS));
-	vmcs_write32_fixedbits(MSR_IA32_VMX_EXIT_CTLS_MSR, VM_EXIT_CONTROLS,
-		     	       (HOST_IS_64 << 9));  /* 22.2,1, 20.7.1 */
-	vmcs_write32(VM_EXIT_MSR_STORE_COUNT, nr_good_msrs); /* 22.2.2 */
-	vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, nr_good_msrs);  /* 22.2.2 */
-	vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, nr_good_msrs); /* 22.2.2 */
-
-
-	/* 22.2.1, 20.8.1 */
-	vmcs_write32_fixedbits(MSR_IA32_VMX_ENTRY_CTLS_MSR,
-                               VM_ENTRY_CONTROLS, 0);
-	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0);  /* 22.2.1 */
-
-	vmcs_writel(VIRTUAL_APIC_PAGE_ADDR, 0);
-	vmcs_writel(TPR_THRESHOLD, 0);
-
-	vmcs_writel(CR0_GUEST_HOST_MASK, KVM_GUEST_CR0_MASK);
-	vmcs_writel(CR4_GUEST_HOST_MASK, KVM_GUEST_CR4_MASK);
-
-	__set_cr0(vcpu, 0x60000010); // enter rmode
-	__set_cr4(vcpu, 0);
-#ifdef __x86_64__
-	__set_efer(vcpu, 0);
-#endif
-
-	ret = kvm_mmu_init(vcpu);
-
-	return ret;
-
-out_free_guest_msrs:
-	kfree(vcpu->guest_msrs);
-out:
-	return ret;
-}
+EXPORT_SYMBOL_GPL(fx_init);
 
 /*
  * Creates some virtual cpus.  Good luck creating more than one.
@@ -1281,7 +970,9 @@ static int kvm_dev_ioctl_create_vcpu(str
 
 	__vcpu_load(vcpu);
 
-	r = kvm_vcpu_setup(vcpu);
+	r = kvm_arch_ops->vcpu_setup(vcpu);
+	if (r >= 0)
+		r = kvm_mmu_init(vcpu);
 
 	vcpu_put(vcpu);
 
@@ -1514,6 +1205,7 @@ struct kvm_memory_slot *gfn_to_memslot(s
 	}
 	return 0;
 }
+EXPORT_SYMBOL_GPL(gfn_to_memslot);
 
 void mark_page_dirty(struct kvm *kvm, gfn_t gfn)
 {
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -19,6 +19,9 @@
 #include <linux/module.h>
 #include "vmx.h"
 #include "kvm_vmx.h"
+#include <linux/mm.h>
+#include <linux/highmem.h>
+#include <asm/io.h>
 
 #define MSR_IA32_FEATURE_CONTROL 		0x03a
 
@@ -27,6 +30,12 @@ MODULE_LICENSE("GPL");
 
 DECLARE_PER_CPU(struct vmcs *, vmxarea);
 
+#ifdef __x86_64__
+#define HOST_IS_64 1
+#else
+#define HOST_IS_64 0
+#endif
+
 extern struct vmcs_descriptor {
 	int size;
 	int order;
@@ -40,6 +49,14 @@ extern struct kvm_vmx_segment_field {
 	unsigned ar_bytes;
 } kvm_vmx_segment_fields[];
 
+static const u32 vmx_msr_index[] = {
+#ifdef __x86_64__
+	MSR_SYSCALL_MASK, MSR_LSTAR, MSR_CSTAR, MSR_KERNEL_GS_BASE,
+#endif
+	MSR_EFER, MSR_K6_STAR,
+};
+#define NR_VMX_MSR (sizeof(vmx_msr_index) / sizeof(*vmx_msr_index))
+
 u64 guest_read_tsc(void);
 void guest_write_tsc(u64 guest_tsc);
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
@@ -436,6 +453,267 @@ static void vmx_set_gdt(struct kvm_vcpu 
 	vmcs_writel(GUEST_GDTR_BASE, dt->base);
 }
 
+static int init_rmode_tss(struct kvm* kvm)
+{
+	struct page *p1, *p2, *p3;
+	gfn_t fn = rmode_tss_base(kvm) >> PAGE_SHIFT;
+	char *page;
+
+	p1 = _gfn_to_page(kvm, fn++);
+	p2 = _gfn_to_page(kvm, fn++);
+	p3 = _gfn_to_page(kvm, fn);
+
+	if (!p1 || !p2 || !p3) {
+		kvm_printf(kvm,"%s: gfn_to_page failed\n", __FUNCTION__);
+		return 0;
+	}
+
+	page = kmap_atomic(p1, KM_USER0);
+	memset(page, 0, PAGE_SIZE);
+	*(u16*)(page + 0x66) = TSS_BASE_SIZE + TSS_REDIRECTION_SIZE;
+	kunmap_atomic(page, KM_USER0);
+
+	page = kmap_atomic(p2, KM_USER0);
+	memset(page, 0, PAGE_SIZE);
+	kunmap_atomic(page, KM_USER0);
+
+	page = kmap_atomic(p3, KM_USER0);
+	memset(page, 0, PAGE_SIZE);
+	*(page + RMODE_TSS_SIZE - 2 * PAGE_SIZE - 1) = ~0;
+	kunmap_atomic(page, KM_USER0);
+
+	return 1;
+}
+
+static void vmcs_write32_fixedbits(u32 msr, u32 vmcs_field, u32 val)
+{
+	u32 msr_high, msr_low;
+
+	rdmsr(msr, msr_low, msr_high);
+
+	val &= msr_high;
+	val |= msr_low;
+	vmcs_write32(vmcs_field, val);
+}
+
+static void seg_setup(int seg)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+
+	vmcs_write16(sf->selector, 0);
+	vmcs_writel(sf->base, 0);
+	vmcs_write32(sf->limit, 0xffff);
+	vmcs_write32(sf->ar_bytes, 0x93);
+}
+
+/*
+ * Sets up the vmcs for emulated real mode.
+ */
+static int vmx_vcpu_setup(struct kvm_vcpu *vcpu)
+{
+	u32 host_sysenter_cs;
+	u32 junk;
+	unsigned long a;
+	struct descriptor_table dt;
+	int i;
+	int ret = 0;
+	int nr_good_msrs;
+	extern asmlinkage void kvm_vmx_return(void);
+
+	if (!init_rmode_tss(vcpu->kvm)) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	memset(vcpu->regs, 0, sizeof(vcpu->regs));
+	vcpu->regs[VCPU_REGS_RDX] = get_rdx_init_val();
+	vcpu->cr8 = 0;
+	vcpu->apic_base = 0xfee00000 |
+			/*for vcpu 0*/ MSR_IA32_APICBASE_BSP |
+			MSR_IA32_APICBASE_ENABLE;
+
+	fx_init(vcpu);
+
+	/*
+	 * GUEST_CS_BASE should really be 0xffff0000, but VT vm86 mode
+	 * insists on having GUEST_CS_BASE == GUEST_CS_SELECTOR << 4.  Sigh.
+	 */
+	vmcs_write16(GUEST_CS_SELECTOR, 0xf000);
+	vmcs_writel(GUEST_CS_BASE, 0x000f0000);
+	vmcs_write32(GUEST_CS_LIMIT, 0xffff);
+	vmcs_write32(GUEST_CS_AR_BYTES, 0x9b);
+
+	seg_setup(VCPU_SREG_DS);
+	seg_setup(VCPU_SREG_ES);
+	seg_setup(VCPU_SREG_FS);
+	seg_setup(VCPU_SREG_GS);
+	seg_setup(VCPU_SREG_SS);
+
+	vmcs_write16(GUEST_TR_SELECTOR, 0);
+	vmcs_writel(GUEST_TR_BASE, 0);
+	vmcs_write32(GUEST_TR_LIMIT, 0xffff);
+	vmcs_write32(GUEST_TR_AR_BYTES, 0x008b);
+
+	vmcs_write16(GUEST_LDTR_SELECTOR, 0);
+	vmcs_writel(GUEST_LDTR_BASE, 0);
+	vmcs_write32(GUEST_LDTR_LIMIT, 0xffff);
+	vmcs_write32(GUEST_LDTR_AR_BYTES, 0x00082);
+
+	vmcs_write32(GUEST_SYSENTER_CS, 0);
+	vmcs_writel(GUEST_SYSENTER_ESP, 0);
+	vmcs_writel(GUEST_SYSENTER_EIP, 0);
+
+	vmcs_writel(GUEST_RFLAGS, 0x02);
+	vmcs_writel(GUEST_RIP, 0xfff0);
+	vmcs_writel(GUEST_RSP, 0);
+
+	vmcs_writel(GUEST_CR3, 0);
+
+	//todo: dr0 = dr1 = dr2 = dr3 = 0; dr6 = 0xffff0ff0
+	vmcs_writel(GUEST_DR7, 0x400);
+
+	vmcs_writel(GUEST_GDTR_BASE, 0);
+	vmcs_write32(GUEST_GDTR_LIMIT, 0xffff);
+
+	vmcs_writel(GUEST_IDTR_BASE, 0);
+	vmcs_write32(GUEST_IDTR_LIMIT, 0xffff);
+
+	vmcs_write32(GUEST_ACTIVITY_STATE, 0);
+	vmcs_write32(GUEST_INTERRUPTIBILITY_INFO, 0);
+	vmcs_write32(GUEST_PENDING_DBG_EXCEPTIONS, 0);
+
+	/* I/O */
+	vmcs_write64(IO_BITMAP_A, 0);
+	vmcs_write64(IO_BITMAP_B, 0);
+
+	guest_write_tsc(0);
+
+	vmcs_write64(VMCS_LINK_POINTER, -1ull); /* 22.3.1.5 */
+
+	/* Special registers */
+	vmcs_write64(GUEST_IA32_DEBUGCTL, 0);
+
+	/* Control */
+	vmcs_write32_fixedbits(MSR_IA32_VMX_PINBASED_CTLS_MSR,
+			       PIN_BASED_VM_EXEC_CONTROL,
+			       PIN_BASED_EXT_INTR_MASK   /* 20.6.1 */
+			       | PIN_BASED_NMI_EXITING   /* 20.6.1 */
+			);
+	vmcs_write32_fixedbits(MSR_IA32_VMX_PROCBASED_CTLS_MSR,
+			       CPU_BASED_VM_EXEC_CONTROL,
+			       CPU_BASED_HLT_EXITING         /* 20.6.2 */
+			       | CPU_BASED_CR8_LOAD_EXITING    /* 20.6.2 */
+			       | CPU_BASED_CR8_STORE_EXITING   /* 20.6.2 */
+			       | CPU_BASED_UNCOND_IO_EXITING   /* 20.6.2 */
+			       | CPU_BASED_INVDPG_EXITING
+			       | CPU_BASED_MOV_DR_EXITING
+			       | CPU_BASED_USE_TSC_OFFSETING   /* 21.3 */
+			);
+
+	vmcs_write32(EXCEPTION_BITMAP, 1 << PF_VECTOR);
+	vmcs_write32(PAGE_FAULT_ERROR_CODE_MASK, 0);
+	vmcs_write32(PAGE_FAULT_ERROR_CODE_MATCH, 0);
+	vmcs_write32(CR3_TARGET_COUNT, 0);           /* 22.2.1 */
+
+	vmcs_writel(HOST_CR0, read_cr0());  /* 22.2.3 */
+	vmcs_writel(HOST_CR4, read_cr4());  /* 22.2.3, 22.2.5 */
+	vmcs_writel(HOST_CR3, read_cr3());  /* 22.2.3  FIXME: shadow tables */
+
+	vmcs_write16(HOST_CS_SELECTOR, __KERNEL_CS);  /* 22.2.4 */
+	vmcs_write16(HOST_DS_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
+	vmcs_write16(HOST_ES_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
+	vmcs_write16(HOST_FS_SELECTOR, read_fs());    /* 22.2.4 */
+	vmcs_write16(HOST_GS_SELECTOR, read_gs());    /* 22.2.4 */
+	vmcs_write16(HOST_SS_SELECTOR, __KERNEL_DS);  /* 22.2.4 */
+#ifdef __x86_64__
+	rdmsrl(MSR_FS_BASE, a);
+	vmcs_writel(HOST_FS_BASE, a); /* 22.2.4 */
+	rdmsrl(MSR_GS_BASE, a);
+	vmcs_writel(HOST_GS_BASE, a); /* 22.2.4 */
+#else
+	vmcs_writel(HOST_FS_BASE, 0); /* 22.2.4 */
+	vmcs_writel(HOST_GS_BASE, 0); /* 22.2.4 */
+#endif
+
+	vmcs_write16(HOST_TR_SELECTOR, GDT_ENTRY_TSS*8);  /* 22.2.4 */
+
+	get_idt(&dt);
+	vmcs_writel(HOST_IDTR_BASE, dt.base);   /* 22.2.4 */
+
+
+	vmcs_writel(HOST_RIP, (unsigned long)kvm_vmx_return); /* 22.2.5 */
+
+	rdmsr(MSR_IA32_SYSENTER_CS, host_sysenter_cs, junk);
+	vmcs_write32(HOST_IA32_SYSENTER_CS, host_sysenter_cs);
+	rdmsrl(MSR_IA32_SYSENTER_ESP, a);
+	vmcs_writel(HOST_IA32_SYSENTER_ESP, a);   /* 22.2.3 */
+	rdmsrl(MSR_IA32_SYSENTER_EIP, a);
+	vmcs_writel(HOST_IA32_SYSENTER_EIP, a);   /* 22.2.3 */
+
+	ret = -ENOMEM;
+	vcpu->guest_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!vcpu->guest_msrs)
+		goto out;
+	vcpu->host_msrs = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	if (!vcpu->host_msrs)
+		goto out_free_guest_msrs;
+
+	for (i = 0; i < NR_VMX_MSR; ++i) {
+		u32 index = vmx_msr_index[i];
+		u32 data_low, data_high;
+		u64 data;
+		int j = vcpu->nmsrs;
+
+		if (rdmsr_safe(index, &data_low, &data_high) < 0)
+			continue;
+		data = data_low | ((u64)data_high << 32);
+		vcpu->host_msrs[j].index = index;
+		vcpu->host_msrs[j].reserved = 0;
+		vcpu->host_msrs[j].data = data;
+		vcpu->guest_msrs[j] = vcpu->host_msrs[j];
+		++vcpu->nmsrs;
+	}
+	printk("msrs: %d\n", vcpu->nmsrs);
+
+	nr_good_msrs = vcpu->nmsrs - NR_BAD_MSRS;
+	vmcs_writel(VM_ENTRY_MSR_LOAD_ADDR,
+		    virt_to_phys(vcpu->guest_msrs + NR_BAD_MSRS));
+	vmcs_writel(VM_EXIT_MSR_STORE_ADDR,
+		    virt_to_phys(vcpu->guest_msrs + NR_BAD_MSRS));
+	vmcs_writel(VM_EXIT_MSR_LOAD_ADDR,
+		    virt_to_phys(vcpu->host_msrs + NR_BAD_MSRS));
+	vmcs_write32_fixedbits(MSR_IA32_VMX_EXIT_CTLS_MSR, VM_EXIT_CONTROLS,
+		     	       (HOST_IS_64 << 9));  /* 22.2,1, 20.7.1 */
+	vmcs_write32(VM_EXIT_MSR_STORE_COUNT, nr_good_msrs); /* 22.2.2 */
+	vmcs_write32(VM_EXIT_MSR_LOAD_COUNT, nr_good_msrs);  /* 22.2.2 */
+	vmcs_write32(VM_ENTRY_MSR_LOAD_COUNT, nr_good_msrs); /* 22.2.2 */
+
+
+	/* 22.2.1, 20.8.1 */
+	vmcs_write32_fixedbits(MSR_IA32_VMX_ENTRY_CTLS_MSR,
+                               VM_ENTRY_CONTROLS, 0);
+	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD, 0);  /* 22.2.1 */
+
+	vmcs_writel(VIRTUAL_APIC_PAGE_ADDR, 0);
+	vmcs_writel(TPR_THRESHOLD, 0);
+
+	vmcs_writel(CR0_GUEST_HOST_MASK, KVM_GUEST_CR0_MASK);
+	vmcs_writel(CR4_GUEST_HOST_MASK, KVM_GUEST_CR4_MASK);
+
+	__set_cr0(vcpu, 0x60000010); // enter rmode
+	__set_cr4(vcpu, 0);
+#ifdef __x86_64__
+	__set_efer(vcpu, 0);
+#endif
+
+	return 0;
+
+out_free_guest_msrs:
+	kfree(vcpu->guest_msrs);
+out:
+	return ret;
+}
+
 static void inject_rmode_irq(struct kvm_vcpu *vcpu, int irq)
 {
 	u16 ent[2];
@@ -1133,8 +1411,6 @@ again:
 	return 0;
 }
 
-extern asmlinkage void kvm_vmx_return(void);
-
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -1158,7 +1434,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,
-	.vmx_return = (unsigned long)kvm_vmx_return,
+	.vcpu_setup = vmx_vcpu_setup,
 };
 
 static int __init vmx_init(void)
Index: linux-2.6/drivers/kvm/vmx.h
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.h
+++ linux-2.6/drivers/kvm/vmx.h
@@ -287,5 +287,10 @@ enum vmcs_field {
 #define CR4_VMXE 0x2000
 
 #define MSR_IA32_VMX_BASIC_MSR   		0x480
+#define MSR_IA32_FEATURE_CONTROL 		0x03a
+#define MSR_IA32_VMX_PINBASED_CTLS_MSR		0x481
+#define MSR_IA32_VMX_PROCBASED_CTLS_MSR		0x482
+#define MSR_IA32_VMX_EXIT_CTLS_MSR		0x483
+#define MSR_IA32_VMX_ENTRY_CTLS_MSR		0x484
 
 #endif

[PATCH 17/38] KVM: Make __set_cr0() (and dependencies) arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -269,6 +269,7 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
+	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
@@ -279,6 +280,8 @@ struct kvm_arch_ops {
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
 	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
+
+	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu); /* hack */
 };
 
 extern struct kvm_stat kvm_stat;
@@ -336,10 +339,8 @@ void set_cr8(struct kvm_vcpu *vcpu, unsi
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 
 void __set_efer(struct kvm_vcpu *vcpu, u64 efer);
-void __set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0);
 void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
 
-int rmode_tss_base(struct kvm* kvm); /* temporary hack */
 void inject_gp(struct kvm_vcpu *vcpu);
 
 #ifdef __x86_64__
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -525,120 +525,6 @@ void guest_write_tsc(u64 guest_tsc)
 }
 EXPORT_SYMBOL_GPL(guest_write_tsc);
 
-static void update_exception_bitmap(struct kvm_vcpu *vcpu)
-{
-	if (vcpu->rmode.active)
-		vmcs_write32(EXCEPTION_BITMAP, ~0);
-	else
-		vmcs_write32(EXCEPTION_BITMAP, 1 << PF_VECTOR);
-}
-
-static void fix_pmode_dataseg(int seg, struct kvm_save_segment *save)
-{
-	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
-
-	if (vmcs_readl(sf->base) == save->base) {
-		vmcs_write16(sf->selector, save->selector);
-		vmcs_writel(sf->base, save->base);
-		vmcs_write32(sf->limit, save->limit);
-		vmcs_write32(sf->ar_bytes, save->ar);
-	} else {
-		u32 dpl = (vmcs_read16(sf->selector) & SELECTOR_RPL_MASK)
-			<< AR_DPL_SHIFT;
-		vmcs_write32(sf->ar_bytes, 0x93 | dpl);
-	}
-}
-
-static void enter_pmode(struct kvm_vcpu *vcpu)
-{
-	unsigned long flags;
-
-	vcpu->rmode.active = 0;
-
-	vmcs_writel(GUEST_TR_BASE, vcpu->rmode.tr.base);
-	vmcs_write32(GUEST_TR_LIMIT, vcpu->rmode.tr.limit);
-	vmcs_write32(GUEST_TR_AR_BYTES, vcpu->rmode.tr.ar);
-
-	flags = vmcs_readl(GUEST_RFLAGS);
-	flags &= ~(IOPL_MASK | X86_EFLAGS_VM);
-	flags |= (vcpu->rmode.save_iopl << IOPL_SHIFT);
-	vmcs_writel(GUEST_RFLAGS, flags);
-
-	vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~CR4_VME_MASK) |
-			(vmcs_readl(CR4_READ_SHADOW) & CR4_VME_MASK));
-
-	update_exception_bitmap(vcpu);
-
-	fix_pmode_dataseg(VCPU_SREG_ES, &vcpu->rmode.es);
-	fix_pmode_dataseg(VCPU_SREG_DS, &vcpu->rmode.ds);
-	fix_pmode_dataseg(VCPU_SREG_GS, &vcpu->rmode.gs);
-	fix_pmode_dataseg(VCPU_SREG_FS, &vcpu->rmode.fs);
-
-	vmcs_write16(GUEST_SS_SELECTOR, 0);
-	vmcs_write32(GUEST_SS_AR_BYTES, 0x93);
-
-	vmcs_write16(GUEST_CS_SELECTOR,
-		     vmcs_read16(GUEST_CS_SELECTOR) & ~SELECTOR_RPL_MASK);
-	vmcs_write32(GUEST_CS_AR_BYTES, 0x9b);
-}
-
-int rmode_tss_base(struct kvm* kvm)
-{
-	gfn_t base_gfn = kvm->memslots[0].base_gfn + kvm->memslots[0].npages - 3;
-	return base_gfn << PAGE_SHIFT;
-}
-EXPORT_SYMBOL_GPL(rmode_tss_base);
-
-static void fix_rmode_seg(int seg, struct kvm_save_segment *save)
-{
-	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
-
-	save->selector = vmcs_read16(sf->selector);
-	save->base = vmcs_readl(sf->base);
-	save->limit = vmcs_read32(sf->limit);
-	save->ar = vmcs_read32(sf->ar_bytes);
-	vmcs_write16(sf->selector, vmcs_readl(sf->base) >> 4);
-	vmcs_write32(sf->limit, 0xffff);
-	vmcs_write32(sf->ar_bytes, 0xf3);
-}
-
-static void enter_rmode(struct kvm_vcpu *vcpu)
-{
-	unsigned long flags;
-
-	vcpu->rmode.active = 1;
-
-	vcpu->rmode.tr.base = vmcs_readl(GUEST_TR_BASE);
-	vmcs_writel(GUEST_TR_BASE, rmode_tss_base(vcpu->kvm));
-
-	vcpu->rmode.tr.limit = vmcs_read32(GUEST_TR_LIMIT);
-	vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1);
-
-	vcpu->rmode.tr.ar = vmcs_read32(GUEST_TR_AR_BYTES);
-	vmcs_write32(GUEST_TR_AR_BYTES, 0x008b);
-
-	flags = vmcs_readl(GUEST_RFLAGS);
-	vcpu->rmode.save_iopl = (flags & IOPL_MASK) >> IOPL_SHIFT;
-
-	flags |= IOPL_MASK | X86_EFLAGS_VM;
-
-	vmcs_writel(GUEST_RFLAGS, flags);
-	vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | CR4_VME_MASK);
-	update_exception_bitmap(vcpu);
-
-	vmcs_write16(GUEST_SS_SELECTOR, vmcs_readl(GUEST_SS_BASE) >> 4);
-	vmcs_write32(GUEST_SS_LIMIT, 0xffff);
-	vmcs_write32(GUEST_SS_AR_BYTES, 0xf3);
-
-	vmcs_write32(GUEST_CS_AR_BYTES, 0xf3);
-	vmcs_write16(GUEST_CS_SELECTOR, vmcs_readl(GUEST_CS_BASE) >> 4);
-
-	fix_rmode_seg(VCPU_SREG_ES, &vcpu->rmode.es);
-	fix_rmode_seg(VCPU_SREG_DS, &vcpu->rmode.ds);
-	fix_rmode_seg(VCPU_SREG_GS, &vcpu->rmode.gs);
-	fix_rmode_seg(VCPU_SREG_FS, &vcpu->rmode.fs);
-}
-
 #ifdef __x86_64__
 
 void __set_efer(struct kvm_vcpu *vcpu, u64 efer)
@@ -662,62 +548,8 @@ void __set_efer(struct kvm_vcpu *vcpu, u
 }
 EXPORT_SYMBOL_GPL(__set_efer);
 
-static void enter_lmode(struct kvm_vcpu *vcpu)
-{
-	u32 guest_tr_ar;
-
-	guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES);
-	if ((guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) {
-		printk(KERN_DEBUG "%s: tss fixup for long mode. \n",
-		       __FUNCTION__);
-		vmcs_write32(GUEST_TR_AR_BYTES,
-			     (guest_tr_ar & ~AR_TYPE_MASK)
-			     | AR_TYPE_BUSY_64_TSS);
-	}
-
-	vcpu->shadow_efer |= EFER_LMA;
-
-	find_msr_entry(vcpu, MSR_EFER)->data |= EFER_LMA | EFER_LME;
-	vmcs_write32(VM_ENTRY_CONTROLS,
-		     vmcs_read32(VM_ENTRY_CONTROLS)
-		     | VM_ENTRY_CONTROLS_IA32E_MASK);
-}
-
-static void exit_lmode(struct kvm_vcpu *vcpu)
-{
-	vcpu->shadow_efer &= ~EFER_LMA;
-
-	vmcs_write32(VM_ENTRY_CONTROLS,
-		     vmcs_read32(VM_ENTRY_CONTROLS)
-		     & ~VM_ENTRY_CONTROLS_IA32E_MASK);
-}
-
-#endif
-
-void __set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
-{
-	if (vcpu->rmode.active && (cr0 & CR0_PE_MASK))
-		enter_pmode(vcpu);
-
-	if (!vcpu->rmode.active && !(cr0 & CR0_PE_MASK))
-		enter_rmode(vcpu);
-
-#ifdef __x86_64__
-	if (vcpu->shadow_efer & EFER_LME) {
-		if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK))
-			enter_lmode(vcpu);
-		if (is_paging(vcpu) && !(cr0 & CR0_PG_MASK))
-			exit_lmode(vcpu);
-	}
 #endif
 
-	vmcs_writel(CR0_READ_SHADOW, cr0);
-	vmcs_writel(GUEST_CR0,
-		    (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
-	vcpu->cr0 = cr0;
-}
-EXPORT_SYMBOL_GPL(__set_cr0);
-
 static int pdptrs_have_reserved_bits_set(struct kvm_vcpu *vcpu,
 					 unsigned long cr3)
 {
@@ -797,7 +629,9 @@ void set_cr0(struct kvm_vcpu *vcpu, unsi
 
 	}
 
-	__set_cr0(vcpu, cr0);
+	kvm_arch_ops->set_cr0(vcpu, cr0);
+	vcpu->cr0 = cr0;
+
 	spin_lock(&vcpu->kvm->lock);
 	kvm_mmu_reset_context(vcpu);
 	spin_unlock(&vcpu->kvm->lock);
@@ -807,18 +641,7 @@ EXPORT_SYMBOL_GPL(set_cr0);
 
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw)
 {
-	unsigned long cr0 = vcpu->cr0;
-
-	if ((msw & CR0_PE_MASK) && !(cr0 & CR0_PE_MASK)) {
-		enter_pmode(vcpu);
-		vmcs_writel(CR0_READ_SHADOW, cr0 | CR0_PE_MASK);
-
-	} else
-		printk(KERN_DEBUG "lmsw: unexpected\n");
-
-	vmcs_writel(GUEST_CR0, (vmcs_readl(GUEST_CR0) & ~LMSW_GUEST_MASK)
-				| (msw & LMSW_GUEST_MASK));
-	vcpu->cr0 = (vcpu->cr0 & ~0xfffful) | msw;
+	set_cr0(vcpu, (vcpu->cr0 & ~0x0ful) | (msw & 0x0f));
 }
 EXPORT_SYMBOL_GPL(lmsw);
 
@@ -1794,7 +1617,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 
 	mmu_reset_needed |= vcpu->cr0 != sregs->cr0;
 	vcpu->rmode.active = ((sregs->cr0 & CR0_PE_MASK) == 0);
-	update_exception_bitmap(vcpu);
+	kvm_arch_ops->update_exception_bitmap(vcpu);
 	vmcs_writel(CR0_READ_SHADOW, sregs->cr0);
 	vmcs_writel(GUEST_CR0,
 		    (sregs->cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -375,6 +375,176 @@ static __exit void hardware_unsetup(void
 	free_kvm_area();
 }
 
+static void update_exception_bitmap(struct kvm_vcpu *vcpu)
+{
+	if (vcpu->rmode.active)
+		vmcs_write32(EXCEPTION_BITMAP, ~0);
+	else
+		vmcs_write32(EXCEPTION_BITMAP, 1 << PF_VECTOR);
+}
+
+static void fix_pmode_dataseg(int seg, struct kvm_save_segment *save)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+
+	if (vmcs_readl(sf->base) == save->base) {
+		vmcs_write16(sf->selector, save->selector);
+		vmcs_writel(sf->base, save->base);
+		vmcs_write32(sf->limit, save->limit);
+		vmcs_write32(sf->ar_bytes, save->ar);
+	} else {
+		u32 dpl = (vmcs_read16(sf->selector) & SELECTOR_RPL_MASK)
+			<< AR_DPL_SHIFT;
+		vmcs_write32(sf->ar_bytes, 0x93 | dpl);
+	}
+}
+
+static void enter_pmode(struct kvm_vcpu *vcpu)
+{
+	unsigned long flags;
+
+	vcpu->rmode.active = 0;
+
+	vmcs_writel(GUEST_TR_BASE, vcpu->rmode.tr.base);
+	vmcs_write32(GUEST_TR_LIMIT, vcpu->rmode.tr.limit);
+	vmcs_write32(GUEST_TR_AR_BYTES, vcpu->rmode.tr.ar);
+
+	flags = vmcs_readl(GUEST_RFLAGS);
+	flags &= ~(IOPL_MASK | X86_EFLAGS_VM);
+	flags |= (vcpu->rmode.save_iopl << IOPL_SHIFT);
+	vmcs_writel(GUEST_RFLAGS, flags);
+
+	vmcs_writel(GUEST_CR4, (vmcs_readl(GUEST_CR4) & ~CR4_VME_MASK) |
+			(vmcs_readl(CR4_READ_SHADOW) & CR4_VME_MASK));
+
+	update_exception_bitmap(vcpu);
+
+	fix_pmode_dataseg(VCPU_SREG_ES, &vcpu->rmode.es);
+	fix_pmode_dataseg(VCPU_SREG_DS, &vcpu->rmode.ds);
+	fix_pmode_dataseg(VCPU_SREG_GS, &vcpu->rmode.gs);
+	fix_pmode_dataseg(VCPU_SREG_FS, &vcpu->rmode.fs);
+
+	vmcs_write16(GUEST_SS_SELECTOR, 0);
+	vmcs_write32(GUEST_SS_AR_BYTES, 0x93);
+
+	vmcs_write16(GUEST_CS_SELECTOR,
+		     vmcs_read16(GUEST_CS_SELECTOR) & ~SELECTOR_RPL_MASK);
+	vmcs_write32(GUEST_CS_AR_BYTES, 0x9b);
+}
+
+static int rmode_tss_base(struct kvm* kvm)
+{
+	gfn_t base_gfn = kvm->memslots[0].base_gfn + kvm->memslots[0].npages - 3;
+	return base_gfn << PAGE_SHIFT;
+}
+
+static void fix_rmode_seg(int seg, struct kvm_save_segment *save)
+{
+	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
+
+	save->selector = vmcs_read16(sf->selector);
+	save->base = vmcs_readl(sf->base);
+	save->limit = vmcs_read32(sf->limit);
+	save->ar = vmcs_read32(sf->ar_bytes);
+	vmcs_write16(sf->selector, vmcs_readl(sf->base) >> 4);
+	vmcs_write32(sf->limit, 0xffff);
+	vmcs_write32(sf->ar_bytes, 0xf3);
+}
+
+static void enter_rmode(struct kvm_vcpu *vcpu)
+{
+	unsigned long flags;
+
+	vcpu->rmode.active = 1;
+
+	vcpu->rmode.tr.base = vmcs_readl(GUEST_TR_BASE);
+	vmcs_writel(GUEST_TR_BASE, rmode_tss_base(vcpu->kvm));
+
+	vcpu->rmode.tr.limit = vmcs_read32(GUEST_TR_LIMIT);
+	vmcs_write32(GUEST_TR_LIMIT, RMODE_TSS_SIZE - 1);
+
+	vcpu->rmode.tr.ar = vmcs_read32(GUEST_TR_AR_BYTES);
+	vmcs_write32(GUEST_TR_AR_BYTES, 0x008b);
+
+	flags = vmcs_readl(GUEST_RFLAGS);
+	vcpu->rmode.save_iopl = (flags & IOPL_MASK) >> IOPL_SHIFT;
+
+	flags |= IOPL_MASK | X86_EFLAGS_VM;
+
+	vmcs_writel(GUEST_RFLAGS, flags);
+	vmcs_writel(GUEST_CR4, vmcs_readl(GUEST_CR4) | CR4_VME_MASK);
+	update_exception_bitmap(vcpu);
+
+	vmcs_write16(GUEST_SS_SELECTOR, vmcs_readl(GUEST_SS_BASE) >> 4);
+	vmcs_write32(GUEST_SS_LIMIT, 0xffff);
+	vmcs_write32(GUEST_SS_AR_BYTES, 0xf3);
+
+	vmcs_write32(GUEST_CS_AR_BYTES, 0xf3);
+	vmcs_write16(GUEST_CS_SELECTOR, vmcs_readl(GUEST_CS_BASE) >> 4);
+
+	fix_rmode_seg(VCPU_SREG_ES, &vcpu->rmode.es);
+	fix_rmode_seg(VCPU_SREG_DS, &vcpu->rmode.ds);
+	fix_rmode_seg(VCPU_SREG_GS, &vcpu->rmode.gs);
+	fix_rmode_seg(VCPU_SREG_FS, &vcpu->rmode.fs);
+}
+
+#ifdef __x86_64__
+
+static void enter_lmode(struct kvm_vcpu *vcpu)
+{
+	u32 guest_tr_ar;
+
+	guest_tr_ar = vmcs_read32(GUEST_TR_AR_BYTES);
+	if ((guest_tr_ar & AR_TYPE_MASK) != AR_TYPE_BUSY_64_TSS) {
+		printk(KERN_DEBUG "%s: tss fixup for long mode. \n",
+		       __FUNCTION__);
+		vmcs_write32(GUEST_TR_AR_BYTES,
+			     (guest_tr_ar & ~AR_TYPE_MASK)
+			     | AR_TYPE_BUSY_64_TSS);
+	}
+
+	vcpu->shadow_efer |= EFER_LMA;
+
+	find_msr_entry(vcpu, MSR_EFER)->data |= EFER_LMA | EFER_LME;
+	vmcs_write32(VM_ENTRY_CONTROLS,
+		     vmcs_read32(VM_ENTRY_CONTROLS)
+		     | VM_ENTRY_CONTROLS_IA32E_MASK);
+}
+
+static void exit_lmode(struct kvm_vcpu *vcpu)
+{
+	vcpu->shadow_efer &= ~EFER_LMA;
+
+	vmcs_write32(VM_ENTRY_CONTROLS,
+		     vmcs_read32(VM_ENTRY_CONTROLS)
+		     & ~VM_ENTRY_CONTROLS_IA32E_MASK);
+}
+
+#endif
+
+static void vmx_set_cr0(struct kvm_vcpu *vcpu, unsigned long cr0)
+{
+	if (vcpu->rmode.active && (cr0 & CR0_PE_MASK))
+		enter_pmode(vcpu);
+
+	if (!vcpu->rmode.active && !(cr0 & CR0_PE_MASK))
+		enter_rmode(vcpu);
+
+#ifdef __x86_64__
+	if (vcpu->shadow_efer & EFER_LME) {
+		if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK))
+			enter_lmode(vcpu);
+		if (is_paging(vcpu) && !(cr0 & CR0_PG_MASK))
+			exit_lmode(vcpu);
+	}
+#endif
+
+	vmcs_writel(CR0_READ_SHADOW, cr0);
+	vmcs_writel(GUEST_CR0,
+		    (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
+	vcpu->cr0 = cr0;
+}
+
 static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg)
 {
 	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
@@ -700,7 +870,8 @@ static int vmx_vcpu_setup(struct kvm_vcp
 	vmcs_writel(CR0_GUEST_HOST_MASK, KVM_GUEST_CR0_MASK);
 	vmcs_writel(CR4_GUEST_HOST_MASK, KVM_GUEST_CR4_MASK);
 
-	__set_cr0(vcpu, 0x60000010); // enter rmode
+	vcpu->cr0 = 0x60000010;
+	vmx_set_cr0(vcpu, vcpu->cr0); // enter rmode
 	__set_cr4(vcpu, 0);
 #ifdef __x86_64__
 	__set_efer(vcpu, 0);
@@ -1425,6 +1596,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment_base = vmx_get_segment_base,
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
+	.set_cr0 = vmx_set_cr0,
 	.get_idt = vmx_get_idt,
 	.set_idt = vmx_set_idt,
 	.get_gdt = vmx_get_gdt,
@@ -1435,6 +1607,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,
 	.vcpu_setup = vmx_vcpu_setup,
+
+	.update_exception_bitmap = update_exception_bitmap,
 };
 
 static int __init vmx_init(void)

[PATCH 18/38] KVM: Make __set_cr4() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -270,6 +270,7 @@ struct kvm_arch_ops {
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
+	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
@@ -339,7 +340,6 @@ void set_cr8(struct kvm_vcpu *vcpu, unsi
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 
 void __set_efer(struct kvm_vcpu *vcpu, u64 efer);
-void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr0);
 
 void inject_gp(struct kvm_vcpu *vcpu);
 
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -645,15 +645,6 @@ void lmsw(struct kvm_vcpu *vcpu, unsigne
 }
 EXPORT_SYMBOL_GPL(lmsw);
 
-void __set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
-{
-	vmcs_writel(CR4_READ_SHADOW, cr4);
-	vmcs_writel(GUEST_CR4, cr4 | (vcpu->rmode.active ?
-		    KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON));
-	vcpu->cr4 = cr4;
-}
-EXPORT_SYMBOL_GPL(__set_cr4);
-
 void set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
 	if (cr4 & CR4_RESEVED_BITS) {
@@ -680,7 +671,7 @@ void set_cr4(struct kvm_vcpu *vcpu, unsi
 		inject_gp(vcpu);
 		return;
 	}
-	__set_cr4(vcpu, cr4);
+	kvm_arch_ops->set_cr4(vcpu, cr4);
 	spin_lock(&vcpu->kvm->lock);
 	kvm_mmu_reset_context(vcpu);
 	spin_unlock(&vcpu->kvm->lock);
@@ -1624,7 +1615,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 	vcpu->cr0 = sregs->cr0;
 
 	mmu_reset_needed |= vcpu->cr4 != sregs->cr4;
-	__set_cr4(vcpu, sregs->cr4);
+	kvm_arch_ops->set_cr4(vcpu, sregs->cr4);
 
 	if (mmu_reset_needed)
 		kvm_mmu_reset_context(vcpu);
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -545,6 +545,14 @@ static void vmx_set_cr0(struct kvm_vcpu 
 	vcpu->cr0 = cr0;
 }
 
+static void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
+{
+	vmcs_writel(CR4_READ_SHADOW, cr4);
+	vmcs_writel(GUEST_CR4, cr4 | (vcpu->rmode.active ?
+		    KVM_RMODE_VM_CR4_ALWAYS_ON : KVM_PMODE_VM_CR4_ALWAYS_ON));
+	vcpu->cr4 = cr4;
+}
+
 static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg)
 {
 	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
@@ -872,7 +880,7 @@ static int vmx_vcpu_setup(struct kvm_vcp
 
 	vcpu->cr0 = 0x60000010;
 	vmx_set_cr0(vcpu, vcpu->cr0); // enter rmode
-	__set_cr4(vcpu, 0);
+	vmx_set_cr4(vcpu, 0);
 #ifdef __x86_64__
 	__set_efer(vcpu, 0);
 #endif
@@ -1597,6 +1605,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
 	.set_cr0 = vmx_set_cr0,
+	.set_cr4 = vmx_set_cr4,
 	.get_idt = vmx_get_idt,
 	.set_idt = vmx_set_idt,
 	.get_gdt = vmx_get_gdt,

[PATCH 19/38] KVM: Make __set_efer() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -271,6 +271,7 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
+	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
 	void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*set_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*get_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
@@ -339,8 +340,6 @@ void set_cr4(struct kvm_vcpu *vcpu, unsi
 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 
-void __set_efer(struct kvm_vcpu *vcpu, u64 efer);
-
 void inject_gp(struct kvm_vcpu *vcpu);
 
 #ifdef __x86_64__
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -525,31 +525,6 @@ void guest_write_tsc(u64 guest_tsc)
 }
 EXPORT_SYMBOL_GPL(guest_write_tsc);
 
-#ifdef __x86_64__
-
-void __set_efer(struct kvm_vcpu *vcpu, u64 efer)
-{
-	struct vmx_msr_entry *msr = find_msr_entry(vcpu, MSR_EFER);
-
-	vcpu->shadow_efer = efer;
-	if (efer & EFER_LMA) {
-		vmcs_write32(VM_ENTRY_CONTROLS,
-				     vmcs_read32(VM_ENTRY_CONTROLS) |
-				     VM_ENTRY_CONTROLS_IA32E_MASK);
-		msr->data = efer;
-
-	} else {
-		vmcs_write32(VM_ENTRY_CONTROLS,
-				     vmcs_read32(VM_ENTRY_CONTROLS) &
-				     ~VM_ENTRY_CONTROLS_IA32E_MASK);
-
-		msr->data = efer & ~EFER_LME;
-	}
-}
-EXPORT_SYMBOL_GPL(__set_efer);
-
-#endif
-
 static int pdptrs_have_reserved_bits_set(struct kvm_vcpu *vcpu,
 					 unsigned long cr3)
 {
@@ -1602,7 +1577,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 
 	mmu_reset_needed |= vcpu->shadow_efer != sregs->efer;
 #ifdef __x86_64__
-	__set_efer(vcpu, sregs->efer);
+	kvm_arch_ops->set_efer(vcpu, sregs->efer);
 #endif
 	vcpu->apic_base = sregs->apic_base;
 
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -553,6 +553,30 @@ static void vmx_set_cr4(struct kvm_vcpu 
 	vcpu->cr4 = cr4;
 }
 
+#ifdef __x86_64__
+
+static void vmx_set_efer(struct kvm_vcpu *vcpu, u64 efer)
+{
+	struct vmx_msr_entry *msr = find_msr_entry(vcpu, MSR_EFER);
+
+	vcpu->shadow_efer = efer;
+	if (efer & EFER_LMA) {
+		vmcs_write32(VM_ENTRY_CONTROLS,
+				     vmcs_read32(VM_ENTRY_CONTROLS) |
+				     VM_ENTRY_CONTROLS_IA32E_MASK);
+		msr->data = efer;
+
+	} else {
+		vmcs_write32(VM_ENTRY_CONTROLS,
+				     vmcs_read32(VM_ENTRY_CONTROLS) &
+				     ~VM_ENTRY_CONTROLS_IA32E_MASK);
+
+		msr->data = efer & ~EFER_LME;
+	}
+}
+
+#endif
+
 static u64 vmx_get_segment_base(struct kvm_vcpu *vcpu, int seg)
 {
 	struct kvm_vmx_segment_field *sf = &kvm_vmx_segment_fields[seg];
@@ -882,7 +906,7 @@ static int vmx_vcpu_setup(struct kvm_vcp
 	vmx_set_cr0(vcpu, vcpu->cr0); // enter rmode
 	vmx_set_cr4(vcpu, 0);
 #ifdef __x86_64__
-	__set_efer(vcpu, 0);
+	vmx_set_efer(vcpu, 0);
 #endif
 
 	return 0;
@@ -1606,6 +1630,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_segment = vmx_set_segment,
 	.set_cr0 = vmx_set_cr0,
 	.set_cr4 = vmx_set_cr4,
+	.set_efer = vmx_set_efer,
 	.get_idt = vmx_get_idt,
 	.set_idt = vmx_set_idt,
 	.get_gdt = vmx_get_gdt,

Re: [PATCH 19/38] KVM: Make __set_efer() an arch operation

[Christoph Hellwig]

On Mon, Nov 27, 2006 at 12:29:38PM -0000, Avi Kivity wrote:
>  #ifdef __x86_64__
> -	__set_efer(vcpu, sregs->efer);
> +	kvm_arch_ops->set_efer(vcpu, sregs->efer);
>  #endif

I think it would be much better to make ->set_efer a noop for 32bit,
and have different operation vectors for 32 vs 64 bit.

>  #ifdef __x86_64__
> -	__set_efer(vcpu, 0);
> +	vmx_set_efer(vcpu, 0);
>  #endif

Similarly vmx_set_efer should just become a noop on 32bit.

Re: [PATCH 19/38] KVM: Make __set_efer() an arch operation

[Avi Kivity]

Christoph Hellwig wrote:
> On Mon, Nov 27, 2006 at 12:29:38PM -0000, Avi Kivity wrote:
>   
>>  #ifdef __x86_64__
>> -	__set_efer(vcpu, sregs->efer);
>> +	kvm_arch_ops->set_efer(vcpu, sregs->efer);
>>  #endif
>>     
>
> I think it would be much better to make ->set_efer a noop for 32bit,
> and have different operation vectors for 32 vs 64 bit.
>
>   

Okay.  I'll submit an incremental patch as part of a larger cleanup I'm 
planning.


>>  #ifdef __x86_64__
>> -	__set_efer(vcpu, 0);
>> +	vmx_set_efer(vcpu, 0);
>>  #endif
>>     
>
> Similarly vmx_set_efer should just become a noop on 32bit.
>   

Ok.

-- 
error compiling committee.c: too many arguments to function

[PATCH 20/38] KVM: Make set_cr3() and tlb flushing arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -270,6 +270,7 @@ struct kvm_arch_ops {
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
+	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
 	void (*get_idt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
@@ -279,6 +280,8 @@ struct kvm_arch_ops {
 	void (*cache_regs)(struct kvm_vcpu *vcpu);
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
 
+	void (*flush_tlb)(struct kvm_vcpu *vcpu);
+
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
 	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
Index: linux-2.6/drivers/kvm/mmu.c
===================================================================
--- linux-2.6.orig/drivers/kvm/mmu.c
+++ linux-2.6/drivers/kvm/mmu.c
@@ -323,7 +323,7 @@ static void nonpaging_flush(struct kvm_v
 	vcpu->mmu.root_hpa = root;
 	if (is_paging(vcpu))
 		root |= (vcpu->cr3 & (CR3_PCD_MASK | CR3_WPT_MASK));
-	vmcs_writel(GUEST_CR3, root);
+	kvm_arch_ops->set_cr3(vcpu, root);
 }
 
 static gpa_t nonpaging_gva_to_gpa(struct kvm_vcpu *vcpu, gva_t vaddr)
@@ -386,7 +386,7 @@ static int nonpaging_init_context(struct
 	context->shadow_root_level = PT32E_ROOT_LEVEL;
 	context->root_hpa = kvm_mmu_alloc_page(vcpu, 0);
 	ASSERT(VALID_PAGE(context->root_hpa));
-	vmcs_writel(GUEST_CR3, context->root_hpa);
+	kvm_arch_ops->set_cr3(vcpu, context->root_hpa);
 	return 0;
 }
 
@@ -539,9 +539,7 @@ static void paging_inval_page(struct kvm
 			table[index] = 0;
 			release_pt_page_64(vcpu, page_addr, PT_PAGE_TABLE_LEVEL);
 
-			//flush tlb
-			vmcs_writel(GUEST_CR3, vcpu->mmu.root_hpa |
-				    (vcpu->cr3 & (CR3_PCD_MASK | CR3_WPT_MASK)));
+			kvm_arch_ops->flush_tlb(vcpu);
 			return;
 		}
 	}
@@ -574,7 +572,7 @@ static int paging64_init_context(struct 
 	context->shadow_root_level = PT64_ROOT_LEVEL;
 	context->root_hpa = kvm_mmu_alloc_page(vcpu, 0);
 	ASSERT(VALID_PAGE(context->root_hpa));
-	vmcs_writel(GUEST_CR3, context->root_hpa |
+	kvm_arch_ops->set_cr3(vcpu, context->root_hpa |
 		    (vcpu->cr3 & (CR3_PCD_MASK | CR3_WPT_MASK)));
 	return 0;
 }
@@ -592,7 +590,7 @@ static int paging32_init_context(struct 
 	context->shadow_root_level = PT32E_ROOT_LEVEL;
 	context->root_hpa = kvm_mmu_alloc_page(vcpu, 0);
 	ASSERT(VALID_PAGE(context->root_hpa));
-	vmcs_writel(GUEST_CR3, context->root_hpa |
+	kvm_arch_ops->set_cr3(vcpu, context->root_hpa |
 		    (vcpu->cr3 & (CR3_PCD_MASK | CR3_WPT_MASK)));
 	return 0;
 }
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -545,6 +545,11 @@ static void vmx_set_cr0(struct kvm_vcpu 
 	vcpu->cr0 = cr0;
 }
 
+static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
+{
+	vmcs_writel(GUEST_CR3, cr3);
+}
+
 static void vmx_set_cr4(struct kvm_vcpu *vcpu, unsigned long cr4)
 {
 	vmcs_writel(CR4_READ_SHADOW, cr4);
@@ -1614,6 +1619,11 @@ again:
 	return 0;
 }
 
+static void vmx_flush_tlb(struct kvm_vcpu *vcpu)
+{
+	vmcs_writel(GUEST_CR3, vmcs_readl(GUEST_CR3));
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -1629,6 +1639,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
 	.set_cr0 = vmx_set_cr0,
+	.set_cr3 = vmx_set_cr3,
 	.set_cr4 = vmx_set_cr4,
 	.set_efer = vmx_set_efer,
 	.get_idt = vmx_get_idt,
@@ -1638,6 +1649,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.cache_regs = vcpu_load_rsp_rip,
 	.decache_regs = vcpu_put_rsp_rip,
 
+	.flush_tlb = vmx_flush_tlb,
+
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,
 	.vcpu_setup = vmx_vcpu_setup,

[PATCH 21/38] KVM: Make inject_page_fault() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -281,6 +281,8 @@ struct kvm_arch_ops {
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
 
 	void (*flush_tlb)(struct kvm_vcpu *vcpu);
+	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
+				  unsigned long addr, u32 err_code);
 
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
 	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
Index: linux-2.6/drivers/kvm/mmu.c
===================================================================
--- linux-2.6.orig/drivers/kvm/mmu.c
+++ linux-2.6/drivers/kvm/mmu.c
@@ -455,32 +455,7 @@ static void inject_page_fault(struct kvm
 			      u64 addr,
 			      u32 err_code)
 {
-	u32 vect_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
-
-	pgprintk("inject_page_fault: 0x%llx err 0x%x\n", addr, err_code);
-
-	++kvm_stat.pf_guest;
-
-	if (is_page_fault(vect_info)) {
-		printk(KERN_DEBUG "inject_page_fault: "
-		       "double fault 0x%llx @ 0x%lx\n",
-		       addr, vmcs_readl(GUEST_RIP));
-		vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, 0);
-		vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
-			     DF_VECTOR |
-			     INTR_TYPE_EXCEPTION |
-			     INTR_INFO_DELIEVER_CODE_MASK |
-			     INTR_INFO_VALID_MASK);
-		return;
-	}
-	vcpu->cr2 = addr;
-	vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, err_code);
-	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
-		     PF_VECTOR |
-		     INTR_TYPE_EXCEPTION |
-		     INTR_INFO_DELIEVER_CODE_MASK |
-		     INTR_INFO_VALID_MASK);
-
+	kvm_arch_ops->inject_page_fault(vcpu, addr, err_code);
 }
 
 static inline int fix_read_pf(u64 *shadow_ent)
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -1624,6 +1624,36 @@ static void vmx_flush_tlb(struct kvm_vcp
 	vmcs_writel(GUEST_CR3, vmcs_readl(GUEST_CR3));
 }
 
+static void vmx_inject_page_fault(struct kvm_vcpu *vcpu,
+				  unsigned long addr,
+				  u32 err_code)
+{
+	u32 vect_info = vmcs_read32(IDT_VECTORING_INFO_FIELD);
+
+	++kvm_stat.pf_guest;
+
+	if (is_page_fault(vect_info)) {
+		printk(KERN_DEBUG "inject_page_fault: "
+		       "double fault 0x%lx @ 0x%lx\n",
+		       addr, vmcs_readl(GUEST_RIP));
+		vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, 0);
+		vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
+			     DF_VECTOR |
+			     INTR_TYPE_EXCEPTION |
+			     INTR_INFO_DELIEVER_CODE_MASK |
+			     INTR_INFO_VALID_MASK);
+		return;
+	}
+	vcpu->cr2 = addr;
+	vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, err_code);
+	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
+		     PF_VECTOR |
+		     INTR_TYPE_EXCEPTION |
+		     INTR_INFO_DELIEVER_CODE_MASK |
+		     INTR_INFO_VALID_MASK);
+
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -1650,6 +1680,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.decache_regs = vcpu_put_rsp_rip,
 
 	.flush_tlb = vmx_flush_tlb,
+	.inject_page_fault = vmx_inject_page_fault,
 
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,

[PATCH 22/38] KVM: Make inject_gp() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -284,6 +284,8 @@ struct kvm_arch_ops {
 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
 				  unsigned long addr, u32 err_code);
 
+	void (*inject_gp)(struct kvm_vcpu *vcpu, unsigned err_code);
+
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
 	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
@@ -345,8 +347,6 @@ void set_cr4(struct kvm_vcpu *vcpu, unsi
 void set_cr8(struct kvm_vcpu *vcpu, unsigned long cr0);
 void lmsw(struct kvm_vcpu *vcpu, unsigned long msw);
 
-void inject_gp(struct kvm_vcpu *vcpu);
-
 #ifdef __x86_64__
 void set_efer(struct kvm_vcpu *vcpu, u64 efer);
 #endif
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -485,18 +485,10 @@ void vmcs_writel(unsigned long field, un
 }
 EXPORT_SYMBOL_GPL(vmcs_writel);
 
-void inject_gp(struct kvm_vcpu *vcpu)
+static void inject_gp(struct kvm_vcpu *vcpu)
 {
-	printk(KERN_DEBUG "inject_general_protection: rip 0x%lx\n",
-	       vmcs_readl(GUEST_RIP));
-	vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, 0);
-	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
-		     GP_VECTOR |
-		     INTR_TYPE_EXCEPTION |
-		     INTR_INFO_DELIEVER_CODE_MASK |
-		     INTR_INFO_VALID_MASK);
+	kvm_arch_ops->inject_gp(vcpu, 0);
 }
-EXPORT_SYMBOL_GPL(inject_gp);
 
 /*
  * reads and returns guest's timestamp counter "register"
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -80,6 +80,18 @@ static void skip_emulated_instruction(st
 			     interruptibility & ~3);
 }
 
+static void vmx_inject_gp(struct kvm_vcpu *vcpu, unsigned error_code)
+{
+	printk(KERN_DEBUG "inject_general_protection: rip 0x%lx\n",
+	       vmcs_readl(GUEST_RIP));
+	vmcs_write32(VM_ENTRY_EXCEPTION_ERROR_CODE, error_code);
+	vmcs_write32(VM_ENTRY_INTR_INFO_FIELD,
+		     GP_VECTOR |
+		     INTR_TYPE_EXCEPTION |
+		     INTR_INFO_DELIEVER_CODE_MASK |
+		     INTR_INFO_VALID_MASK);
+}
+
 static void reload_tss(void)
 {
 #ifndef __x86_64__
@@ -1318,7 +1330,7 @@ static int handle_rdmsr(struct kvm_vcpu 
 	u64 data;
 
 	if (vmx_get_msr(vcpu, ecx, &data)) {
-		inject_gp(vcpu);
+		vmx_inject_gp(vcpu, 0);
 		return 1;
 	}
 
@@ -1336,7 +1348,7 @@ static int handle_wrmsr(struct kvm_vcpu 
 		| ((u64)(vcpu->regs[VCPU_REGS_RDX] & -1u) << 32);
 
 	if (vmx_set_msr(vcpu, ecx, data) != 0) {
-		inject_gp(vcpu);
+		vmx_inject_gp(vcpu, 0);
 		return 1;
 	}
 
@@ -1682,6 +1694,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.flush_tlb = vmx_flush_tlb,
 	.inject_page_fault = vmx_inject_page_fault,
 
+	.inject_gp = vmx_inject_gp,
+
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,
 	.vcpu_setup = vmx_vcpu_setup,

[PATCH 23/38] KVM: Use the idt and gdt accessors in realmode emulation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1225,14 +1225,16 @@ static u64 mk_cr_64(u64 curr_cr, u32 new
 
 void realmode_lgdt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base)
 {
-	vmcs_writel(GUEST_GDTR_BASE, base);
-	vmcs_write32(GUEST_GDTR_LIMIT, limit);
+	struct descriptor_table dt = { limit, base };
+
+	kvm_arch_ops->set_gdt(vcpu, &dt);
 }
 
 void realmode_lidt(struct kvm_vcpu *vcpu, u16 limit, unsigned long base)
 {
-	vmcs_writel(GUEST_IDTR_BASE, base);
-	vmcs_write32(GUEST_IDTR_LIMIT, limit);
+	struct descriptor_table dt = { limit, base };
+
+	kvm_arch_ops->set_idt(vcpu, &dt);
 }
 
 void realmode_lmsw(struct kvm_vcpu *vcpu, unsigned long msw,

[PATCH 24/38] KVM: Use the general purpose register accessors rather than direct access

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1130,7 +1130,7 @@ static void report_emulation_failure(str
 {
 	static int reported;
 	u8 opcodes[4];
-	unsigned long rip = vmcs_readl(GUEST_RIP);
+	unsigned long rip = ctxt->vcpu->rip;
 	unsigned long rip_linear;
 
 	rip_linear = rip + get_segment_base(ctxt->vcpu, VCPU_SREG_CS);
@@ -1406,13 +1406,15 @@ static int kvm_dev_ioctl_get_regs(struct
 	if (!vcpu)
 		return -ENOENT;
 
+	kvm_arch_ops->cache_regs(vcpu);
+
 	regs->rax = vcpu->regs[VCPU_REGS_RAX];
 	regs->rbx = vcpu->regs[VCPU_REGS_RBX];
 	regs->rcx = vcpu->regs[VCPU_REGS_RCX];
 	regs->rdx = vcpu->regs[VCPU_REGS_RDX];
 	regs->rsi = vcpu->regs[VCPU_REGS_RSI];
 	regs->rdi = vcpu->regs[VCPU_REGS_RDI];
-	regs->rsp = vmcs_readl(GUEST_RSP);
+	regs->rsp = vcpu->regs[VCPU_REGS_RSP];
 	regs->rbp = vcpu->regs[VCPU_REGS_RBP];
 #ifdef __x86_64__
 	regs->r8 = vcpu->regs[VCPU_REGS_R8];
@@ -1425,7 +1427,7 @@ static int kvm_dev_ioctl_get_regs(struct
 	regs->r15 = vcpu->regs[VCPU_REGS_R15];
 #endif
 
-	regs->rip = vmcs_readl(GUEST_RIP);
+	regs->rip = vcpu->rip;
 	regs->rflags = vmcs_readl(GUEST_RFLAGS);
 
 	/*
@@ -1456,7 +1458,7 @@ static int kvm_dev_ioctl_set_regs(struct
 	vcpu->regs[VCPU_REGS_RDX] = regs->rdx;
 	vcpu->regs[VCPU_REGS_RSI] = regs->rsi;
 	vcpu->regs[VCPU_REGS_RDI] = regs->rdi;
-	vmcs_writel(GUEST_RSP, regs->rsp);
+	vcpu->regs[VCPU_REGS_RSP] = regs->rsp;
 	vcpu->regs[VCPU_REGS_RBP] = regs->rbp;
 #ifdef __x86_64__
 	vcpu->regs[VCPU_REGS_R8] = regs->r8;
@@ -1469,9 +1471,11 @@ static int kvm_dev_ioctl_set_regs(struct
 	vcpu->regs[VCPU_REGS_R15] = regs->r15;
 #endif
 
-	vmcs_writel(GUEST_RIP, regs->rip);
+	vcpu->rip = regs->rip;
 	vmcs_writel(GUEST_RFLAGS, regs->rflags);
 
+	kvm_arch_ops->decache_regs(vcpu);
+
 	vcpu_put(vcpu);
 
 	return 0;

[PATCH 25/38] KVM: Move the vmx tsc accessors to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -490,33 +490,6 @@ static void inject_gp(struct kvm_vcpu *v
 	kvm_arch_ops->inject_gp(vcpu, 0);
 }
 
-/*
- * reads and returns guest's timestamp counter "register"
- * guest_tsc = host_tsc + tsc_offset    -- 21.3
- */
-u64 guest_read_tsc(void)
-{
-	u64 host_tsc, tsc_offset;
-
-	rdtscll(host_tsc);
-	tsc_offset = vmcs_read64(TSC_OFFSET);
-	return host_tsc + tsc_offset;
-}
-EXPORT_SYMBOL_GPL(guest_read_tsc);
-
-/*
- * writes 'guest_tsc' into guest's timestamp counter "register"
- * guest_tsc = host_tsc + tsc_offset ==> tsc_offset = guest_tsc - host_tsc
- */
-void guest_write_tsc(u64 guest_tsc)
-{
-	u64 host_tsc;
-
-	rdtscll(host_tsc);
-	vmcs_write64(TSC_OFFSET, guest_tsc - host_tsc);
-}
-EXPORT_SYMBOL_GPL(guest_write_tsc);
-
 static int pdptrs_have_reserved_bits_set(struct kvm_vcpu *vcpu,
 					 unsigned long cr3)
 {
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -57,8 +57,6 @@ static const u32 vmx_msr_index[] = {
 };
 #define NR_VMX_MSR (sizeof(vmx_msr_index) / sizeof(*vmx_msr_index))
 
-u64 guest_read_tsc(void);
-void guest_write_tsc(u64 guest_tsc);
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
 static void skip_emulated_instruction(struct kvm_vcpu *vcpu)
@@ -92,6 +90,31 @@ static void vmx_inject_gp(struct kvm_vcp
 		     INTR_INFO_VALID_MASK);
 }
 
+/*
+ * reads and returns guest's timestamp counter "register"
+ * guest_tsc = host_tsc + tsc_offset    -- 21.3
+ */
+static u64 guest_read_tsc(void)
+{
+	u64 host_tsc, tsc_offset;
+
+	rdtscll(host_tsc);
+	tsc_offset = vmcs_read64(TSC_OFFSET);
+	return host_tsc + tsc_offset;
+}
+
+/*
+ * writes 'guest_tsc' into guest's timestamp counter "register"
+ * guest_tsc = host_tsc + tsc_offset ==> tsc_offset = guest_tsc - host_tsc
+ */
+static void guest_write_tsc(u64 guest_tsc)
+{
+	u64 host_tsc;
+
+	rdtscll(host_tsc);
+	vmcs_write64(TSC_OFFSET, guest_tsc - host_tsc);
+}
+
 static void reload_tss(void)
 {
 #ifndef __x86_64__

[PATCH 26/38] KVM: Access rflags through an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -279,6 +279,8 @@ struct kvm_arch_ops {
 	void (*set_gdt)(struct kvm_vcpu *vcpu, struct descriptor_table *dt);
 	void (*cache_regs)(struct kvm_vcpu *vcpu);
 	void (*decache_regs)(struct kvm_vcpu *vcpu);
+	unsigned long (*get_rflags)(struct kvm_vcpu *vcpu);
+	void (*set_rflags)(struct kvm_vcpu *vcpu, unsigned long rflags);
 
 	void (*flush_tlb)(struct kvm_vcpu *vcpu);
 	void (*inject_page_fault)(struct kvm_vcpu *vcpu,
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1141,7 +1141,7 @@ int emulate_instruction(struct kvm_vcpu 
 	cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
 
 	emulate_ctxt.vcpu = vcpu;
-	emulate_ctxt.eflags = vmcs_readl(GUEST_RFLAGS);
+	emulate_ctxt.eflags = kvm_arch_ops->get_rflags(vcpu);
 	emulate_ctxt.cr2 = cr2;
 	emulate_ctxt.mode = (emulate_ctxt.eflags & X86_EFLAGS_VM)
 		? X86EMUL_MODE_REAL : (cs_ar & AR_L_MASK)
@@ -1182,7 +1182,7 @@ int emulate_instruction(struct kvm_vcpu 
 	}
 
 	kvm_arch_ops->decache_regs(vcpu);
-	vmcs_writel(GUEST_RFLAGS, emulate_ctxt.eflags);
+	kvm_arch_ops->set_rflags(vcpu, emulate_ctxt.eflags);
 
 	if (vcpu->mmio_is_write)
 		return EMULATE_DO_MMIO;
@@ -1214,7 +1214,7 @@ void realmode_lmsw(struct kvm_vcpu *vcpu
 		   unsigned long *rflags)
 {
 	lmsw(vcpu, msw);
-	*rflags = vmcs_readl(GUEST_RFLAGS);
+	*rflags = kvm_arch_ops->get_rflags(vcpu);
 }
 
 unsigned long realmode_get_cr(struct kvm_vcpu *vcpu, int cr)
@@ -1240,7 +1240,7 @@ void realmode_set_cr(struct kvm_vcpu *vc
 	switch (cr) {
 	case 0:
 		set_cr0(vcpu, mk_cr_64(vcpu->cr0, val));
-		*rflags = vmcs_readl(GUEST_RFLAGS);
+		*rflags = kvm_arch_ops->get_rflags(vcpu);
 		break;
 	case 2:
 		vcpu->cr2 = val;
@@ -1401,7 +1401,7 @@ static int kvm_dev_ioctl_get_regs(struct
 #endif
 
 	regs->rip = vcpu->rip;
-	regs->rflags = vmcs_readl(GUEST_RFLAGS);
+	regs->rflags = kvm_arch_ops->get_rflags(vcpu);
 
 	/*
 	 * Don't leak debug flags in case they were set for guest debugging
@@ -1445,7 +1445,7 @@ static int kvm_dev_ioctl_set_regs(struct
 #endif
 
 	vcpu->rip = regs->rip;
-	vmcs_writel(GUEST_RFLAGS, regs->rflags);
+	kvm_arch_ops->set_rflags(vcpu, regs->rflags);
 
 	kvm_arch_ops->decache_regs(vcpu);
 
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -59,6 +59,16 @@ static const u32 vmx_msr_index[] = {
 
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
+static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu)
+{
+	return vmcs_readl(GUEST_RFLAGS);
+}
+
+static void vmx_set_rflags(struct kvm_vcpu *vcpu, unsigned long rflags)
+{
+	vmcs_writel(GUEST_RFLAGS, rflags);
+}
+
 static void skip_emulated_instruction(struct kvm_vcpu *vcpu)
 {
 	unsigned long rip;
@@ -1713,6 +1723,8 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_gdt = vmx_set_gdt,
 	.cache_regs = vcpu_load_rsp_rip,
 	.decache_regs = vcpu_put_rsp_rip,
+	.get_rflags = vmx_get_rflags,
+	.set_rflags = vmx_set_rflags,
 
 	.flush_tlb = vmx_flush_tlb,
 	.inject_page_fault = vmx_inject_page_fault,

[PATCH 27/38] KVM: Move the vmx segment field definitions to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -65,31 +65,6 @@ static struct kvm_stats_debugfs_item {
 
 static struct dentry *debugfs_dir;
 
-#define VMX_SEGMENT_FIELD(seg)					\
-	[VCPU_SREG_##seg] {                                     \
-		GUEST_##seg##_SELECTOR,				\
-		GUEST_##seg##_BASE,			   	\
-		GUEST_##seg##_LIMIT,			   	\
-		GUEST_##seg##_AR_BYTES,			   	\
-	}
-
-struct kvm_vmx_segment_field {
-	unsigned selector;
-	unsigned base;
-	unsigned limit;
-	unsigned ar_bytes;
-} kvm_vmx_segment_fields[] = {
-	VMX_SEGMENT_FIELD(CS),
-	VMX_SEGMENT_FIELD(DS),
-	VMX_SEGMENT_FIELD(ES),
-	VMX_SEGMENT_FIELD(FS),
-	VMX_SEGMENT_FIELD(GS),
-	VMX_SEGMENT_FIELD(SS),
-	VMX_SEGMENT_FIELD(TR),
-	VMX_SEGMENT_FIELD(LDTR),
-};
-EXPORT_SYMBOL_GPL(kvm_vmx_segment_fields);
-
 #define MAX_IO_MSRS 256
 
 #define CR0_RESEVED_BITS 0xffffffff1ffaffc0ULL
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -42,12 +42,29 @@ extern struct vmcs_descriptor {
 	u32 revision_id;
 } vmcs_descriptor;
 
-extern struct kvm_vmx_segment_field {
+#define VMX_SEGMENT_FIELD(seg)					\
+	[VCPU_SREG_##seg] {                                     \
+		GUEST_##seg##_SELECTOR,				\
+		GUEST_##seg##_BASE,			   	\
+		GUEST_##seg##_LIMIT,			   	\
+		GUEST_##seg##_AR_BYTES,			   	\
+	}
+
+static struct kvm_vmx_segment_field {
 	unsigned selector;
 	unsigned base;
 	unsigned limit;
 	unsigned ar_bytes;
-} kvm_vmx_segment_fields[];
+} kvm_vmx_segment_fields[] = {
+	VMX_SEGMENT_FIELD(CS),
+	VMX_SEGMENT_FIELD(DS),
+	VMX_SEGMENT_FIELD(ES),
+	VMX_SEGMENT_FIELD(FS),
+	VMX_SEGMENT_FIELD(GS),
+	VMX_SEGMENT_FIELD(SS),
+	VMX_SEGMENT_FIELD(TR),
+	VMX_SEGMENT_FIELD(LDTR),
+};
 
 static const u32 vmx_msr_index[] = {
 #ifdef __x86_64__

[PATCH 28/38] KVM: Add an arch accessor for cs d/b and l bits

[Avi Kivity]

These are used for detecting the current processor mode.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -269,6 +269,7 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
+	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -517,15 +517,16 @@ void set_cr0(struct kvm_vcpu *vcpu, unsi
 	if (!is_paging(vcpu) && (cr0 & CR0_PG_MASK)) {
 #ifdef __x86_64__
 		if ((vcpu->shadow_efer & EFER_LME)) {
-			u32 guest_cs_ar;
+			int cs_db, cs_l;
+
 			if (!is_pae(vcpu)) {
 				printk(KERN_DEBUG "set_cr0: #GP, start paging "
 				       "in long mode while PAE is disabled\n");
 				inject_gp(vcpu);
 				return;
 			}
-			guest_cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
-			if (guest_cs_ar & SEGMENT_AR_L_MASK) {
+			kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
+			if (cs_l) {
 				printk(KERN_DEBUG "set_cr0: #GP, start paging "
 				       "in long mode while CS.L == 1\n");
 				inject_gp(vcpu);
@@ -1109,18 +1110,18 @@ int emulate_instruction(struct kvm_vcpu 
 {
 	struct x86_emulate_ctxt emulate_ctxt;
 	int r;
-	u32 cs_ar;
+	int cs_db, cs_l;
 
 	kvm_arch_ops->cache_regs(vcpu);
 
-	cs_ar = vmcs_read32(GUEST_CS_AR_BYTES);
+	kvm_arch_ops->get_cs_db_l_bits(vcpu, &cs_db, &cs_l);
 
 	emulate_ctxt.vcpu = vcpu;
 	emulate_ctxt.eflags = kvm_arch_ops->get_rflags(vcpu);
 	emulate_ctxt.cr2 = cr2;
 	emulate_ctxt.mode = (emulate_ctxt.eflags & X86_EFLAGS_VM)
-		? X86EMUL_MODE_REAL : (cs_ar & AR_L_MASK)
-		? X86EMUL_MODE_PROT64 :	(cs_ar & AR_DB_MASK)
+		? X86EMUL_MODE_REAL : cs_l
+		? X86EMUL_MODE_PROT64 :	cs_db
 		? X86EMUL_MODE_PROT32 : X86EMUL_MODE_PROT16;
 
 	if (emulate_ctxt.mode == X86EMUL_MODE_PROT64) {
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -698,6 +698,14 @@ static void vmx_set_segment(struct kvm_v
 	vmcs_write32(sf->ar_bytes, ar);
 }
 
+static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l)
+{
+	u32 ar = vmcs_read32(GUEST_CS_AR_BYTES);
+
+	*db = (ar >> 14) & 1;
+	*l = (ar >> 13) & 1;
+}
+
 static void vmx_get_idt(struct kvm_vcpu *vcpu, struct descriptor_table *dt)
 {
 	dt->limit = vmcs_read32(GUEST_IDTR_LIMIT);
@@ -1730,6 +1738,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment_base = vmx_get_segment_base,
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
+	.get_cs_db_l_bits = vmx_get_cs_db_l_bits,
 	.set_cr0 = vmx_set_cr0,
 	.set_cr3 = vmx_set_cr3,
 	.set_cr4 = vmx_set_cr4,

[PATCH 29/38] KVM: Add a set_cr0_no_modeswitch() arch accessor

[Avi Kivity]

A side effect of using vm86 mode to simulate real mode is that we mangle
the segment registers on mode switching.  The new set_cr0_no_modeswitch()
accessor avoids the mangling when the restore is due to loading the virtual
machine from a file.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -271,6 +271,8 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
+	void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
+				      unsigned long cr0);
 	void (*set_cr3)(struct kvm_vcpu *vcpu, unsigned long cr3);
 	void (*set_cr4)(struct kvm_vcpu *vcpu, unsigned long cr4);
 	void (*set_efer)(struct kvm_vcpu *vcpu, u64 efer);
@@ -292,8 +294,6 @@ struct kvm_arch_ops {
 	int (*run)(struct kvm_vcpu *vcpu, struct kvm_run *run);
 	int (*vcpu_setup)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
-
-	void (*update_exception_bitmap)(struct kvm_vcpu *vcpu); /* hack */
 };
 
 extern struct kvm_stat kvm_stat;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -1529,12 +1529,7 @@ static int kvm_dev_ioctl_set_sregs(struc
 	vcpu->apic_base = sregs->apic_base;
 
 	mmu_reset_needed |= vcpu->cr0 != sregs->cr0;
-	vcpu->rmode.active = ((sregs->cr0 & CR0_PE_MASK) == 0);
-	kvm_arch_ops->update_exception_bitmap(vcpu);
-	vmcs_writel(CR0_READ_SHADOW, sregs->cr0);
-	vmcs_writel(GUEST_CR0,
-		    (sregs->cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
-	vcpu->cr0 = sregs->cr0;
+	kvm_arch_ops->set_cr0_no_modeswitch(vcpu, sregs->cr0);
 
 	mmu_reset_needed |= vcpu->cr4 != sregs->cr4;
 	kvm_arch_ops->set_cr4(vcpu, sregs->cr4);
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -607,6 +607,19 @@ static void vmx_set_cr0(struct kvm_vcpu 
 	vcpu->cr0 = cr0;
 }
 
+/*
+ * Used when restoring the VM to avoid corrupting segment registers
+ */
+static void vmx_set_cr0_no_modeswitch(struct kvm_vcpu *vcpu, unsigned long cr0)
+{
+	vcpu->rmode.active = ((cr0 & CR0_PE_MASK) == 0);
+	update_exception_bitmap(vcpu);
+	vmcs_writel(CR0_READ_SHADOW, cr0);
+	vmcs_writel(GUEST_CR0,
+		    (cr0 & ~KVM_GUEST_CR0_MASK) | KVM_VM_CR0_ALWAYS_ON);
+	vcpu->cr0 = cr0;
+}
+
 static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 {
 	vmcs_writel(GUEST_CR3, cr3);
@@ -1740,6 +1753,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.set_segment = vmx_set_segment,
 	.get_cs_db_l_bits = vmx_get_cs_db_l_bits,
 	.set_cr0 = vmx_set_cr0,
+	.set_cr0_no_modeswitch = vmx_set_cr0_no_modeswitch,
 	.set_cr3 = vmx_set_cr3,
 	.set_cr4 = vmx_set_cr4,
 	.set_efer = vmx_set_efer,
@@ -1760,8 +1774,6 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.run = vmx_vcpu_run,
 	.skip_emulated_instruction = skip_emulated_instruction,
 	.vcpu_setup = vmx_vcpu_setup,
-
-	.update_exception_bitmap = update_exception_bitmap,
 };
 
 static int __init vmx_init(void)

[PATCH 30/38] KVM: Make vcpu_load() and vcpu_put() arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -260,6 +260,9 @@ struct kvm_arch_ops {
 	int (*hardware_setup)(void);               /* __init */
 	void (*hardware_unsetup)(void);            /* __exit */
 
+	struct kvm_vcpu *(*vcpu_load)(struct kvm_vcpu *vcpu);
+	void (*vcpu_put)(struct kvm_vcpu *vcpu);
+
 	int (*set_guest_debug)(struct kvm_vcpu *vcpu,
 			       struct kvm_debug_guest *dbg);
 	int (*get_msr)(struct kvm_vcpu *vcpu, u32 msr_index, u64 *pdata);
@@ -296,6 +299,8 @@ struct kvm_arch_ops {
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
 };
 
+void __vcpu_clear(void *arg); /* temporary hack */
+
 extern struct kvm_stat kvm_stat;
 extern struct kvm_arch_ops *kvm_arch_ops;
 
@@ -373,6 +378,8 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 void vmcs_writel(unsigned long field, unsigned long value);
 unsigned long vmcs_readl(unsigned long field);
 
+unsigned long segment_base(u16 selector);
+
 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
 {
 	struct kvm_memory_slot *slot = gfn_to_memslot(kvm, gfn);
@@ -501,6 +508,18 @@ static inline void get_idt(struct descri
 	asm ("sidt %0" : "=m"(*table));
 }
 
+static inline void get_gdt(struct descriptor_table *table)
+{
+	asm ("sgdt %0" : "=m"(*table));
+}
+
+static inline unsigned long read_tr_base(void)
+{
+	u16 tr;
+	asm ("str %0" : "=g"(tr));
+	return segment_base(tr);
+}
+
 #ifdef __x86_64__
 static inline unsigned long read_msr(unsigned long msr)
 {
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -84,11 +84,6 @@ struct vmx_msr_entry *find_msr_entry(str
 }
 EXPORT_SYMBOL_GPL(find_msr_entry);
 
-static void get_gdt(struct descriptor_table *table)
-{
-	asm ("sgdt %0" : "=m"(*table));
-}
-
 struct segment_descriptor {
 	u16 limit_low;
 	u16 base_low;
@@ -115,7 +110,7 @@ struct segment_descriptor_64 {
 
 #endif
 
-static unsigned long segment_base(u16 selector)
+unsigned long segment_base(u16 selector)
 {
 	struct descriptor_table gdt;
 	struct segment_descriptor *d;
@@ -141,17 +136,12 @@ static unsigned long segment_base(u16 se
 #endif
 	return v;
 }
-
-static unsigned long read_tr_base(void)
-{
-	u16 tr;
-	asm ("str %0" : "=g"(tr));
-	return segment_base(tr);
-}
+EXPORT_SYMBOL_GPL(segment_base);
 
 DEFINE_PER_CPU(struct vmcs *, vmxarea);
 EXPORT_SYMBOL_GPL(per_cpu__vmxarea); /* temporary hack */
-static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
+DEFINE_PER_CPU(struct vmcs *, current_vmcs);
+EXPORT_SYMBOL_GPL(per_cpu__current_vmcs); /* temporary hack */
 
 struct vmcs_descriptor {
 	int size;
@@ -242,7 +232,7 @@ static void vmcs_clear(struct vmcs *vmcs
 		       vmcs, phys_addr);
 }
 
-static void __vcpu_clear(void *arg)
+void __vcpu_clear(void *arg)
 {
 	struct kvm_vcpu *vcpu = arg;
 	int cpu = smp_processor_id();
@@ -252,6 +242,7 @@ static void __vcpu_clear(void *arg)
 	if (per_cpu(current_vmcs, cpu) == vcpu->vmcs)
 		per_cpu(current_vmcs, cpu) = 0;
 }
+EXPORT_SYMBOL_GPL(__vcpu_clear);
 
 static int vcpu_slot(struct kvm_vcpu *vcpu)
 {
@@ -259,53 +250,6 @@ static int vcpu_slot(struct kvm_vcpu *vc
 }
 
 /*
- * Switches to specified vcpu, until a matching vcpu_put(), but assumes
- * vcpu mutex is already taken.
- */
-static struct kvm_vcpu *__vcpu_load(struct kvm_vcpu *vcpu)
-{
-	u64 phys_addr = __pa(vcpu->vmcs);
-	int cpu;
-
-	cpu = get_cpu();
-
-	if (vcpu->cpu != cpu) {
-		smp_call_function(__vcpu_clear, vcpu, 0, 1);
-		vcpu->launched = 0;
-	}
-
-	if (per_cpu(current_vmcs, cpu) != vcpu->vmcs) {
-		u8 error;
-
-		per_cpu(current_vmcs, cpu) = vcpu->vmcs;
-		asm volatile (ASM_VMX_VMPTRLD_RAX "; setna %0"
-			      : "=g"(error) : "a"(&phys_addr), "m"(phys_addr)
-			      : "cc");
-		if (error)
-			printk(KERN_ERR "kvm: vmptrld %p/%llx fail\n",
-			       vcpu->vmcs, phys_addr);
-	}
-
-	if (vcpu->cpu != cpu) {
-		struct descriptor_table dt;
-		unsigned long sysenter_esp;
-
-		vcpu->cpu = cpu;
-		/*
-		 * Linux uses per-cpu TSS and GDT, so set these when switching
-		 * processors.
-		 */
-		vmcs_writel(HOST_TR_BASE, read_tr_base()); /* 22.2.4 */
-		get_gdt(&dt);
-		vmcs_writel(HOST_GDTR_BASE, dt.base);   /* 22.2.4 */
-
-		rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp);
-		vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */
-	}
-	return vcpu;
-}
-
-/*
  * Switches to specified vcpu, until a matching vcpu_put()
  */
 static struct kvm_vcpu *vcpu_load(struct kvm *kvm, int vcpu_slot)
@@ -317,11 +261,12 @@ static struct kvm_vcpu *vcpu_load(struct
 		mutex_unlock(&vcpu->mutex);
 		return 0;
 	}
-	return __vcpu_load(vcpu);
+	return kvm_arch_ops->vcpu_load(vcpu);
 }
 
 static void vcpu_put(struct kvm_vcpu *vcpu)
 {
+	kvm_arch_ops->vcpu_put(vcpu);
 	put_cpu();
 	mutex_unlock(&vcpu->mutex);
 }
@@ -698,7 +643,7 @@ static int kvm_dev_ioctl_create_vcpu(str
 	vcpu->vmcs = vmcs;
 	vcpu->launched = 0;
 
-	__vcpu_load(vcpu);
+	kvm_arch_ops->vcpu_load(vcpu);
 
 	r = kvm_arch_ops->vcpu_setup(vcpu);
 	if (r >= 0)
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -29,6 +29,7 @@ MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
 DECLARE_PER_CPU(struct vmcs *, vmxarea);
+DECLARE_PER_CPU(struct vmcs *, current_vmcs);
 
 #ifdef __x86_64__
 #define HOST_IS_64 1
@@ -76,6 +77,58 @@ static const u32 vmx_msr_index[] = {
 
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
+/*
+ * Switches to specified vcpu, until a matching vcpu_put(), but assumes
+ * vcpu mutex is already taken.
+ */
+static struct kvm_vcpu *vmx_vcpu_load(struct kvm_vcpu *vcpu)
+{
+	u64 phys_addr = __pa(vcpu->vmcs);
+	int cpu;
+
+	cpu = get_cpu();
+
+	if (vcpu->cpu != cpu) {
+		smp_call_function(__vcpu_clear, vcpu, 0, 1);
+		vcpu->launched = 0;
+	}
+
+	if (per_cpu(current_vmcs, cpu) != vcpu->vmcs) {
+		u8 error;
+
+		per_cpu(current_vmcs, cpu) = vcpu->vmcs;
+		asm volatile (ASM_VMX_VMPTRLD_RAX "; setna %0"
+			      : "=g"(error) : "a"(&phys_addr), "m"(phys_addr)
+			      : "cc");
+		if (error)
+			printk(KERN_ERR "kvm: vmptrld %p/%llx fail\n",
+			       vcpu->vmcs, phys_addr);
+	}
+
+	if (vcpu->cpu != cpu) {
+		struct descriptor_table dt;
+		unsigned long sysenter_esp;
+
+		vcpu->cpu = cpu;
+		/*
+		 * Linux uses per-cpu TSS and GDT, so set these when switching
+		 * processors.
+		 */
+		vmcs_writel(HOST_TR_BASE, read_tr_base()); /* 22.2.4 */
+		get_gdt(&dt);
+		vmcs_writel(HOST_GDTR_BASE, dt.base);   /* 22.2.4 */
+
+		rdmsrl(MSR_IA32_SYSENTER_ESP, sysenter_esp);
+		vmcs_writel(HOST_IA32_SYSENTER_ESP, sysenter_esp); /* 22.2.3 */
+	}
+	return vcpu;
+}
+
+static void vmx_vcpu_put(struct kvm_vcpu *vcpu)
+{
+	put_cpu();
+}
+
 static unsigned long vmx_get_rflags(struct kvm_vcpu *vcpu)
 {
 	return vmcs_readl(GUEST_RFLAGS);
@@ -1745,6 +1798,9 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.hardware_enable = hardware_enable,
 	.hardware_disable = hardware_disable,
 
+	.vcpu_load = vmx_vcpu_load,
+	.vcpu_put = vmx_vcpu_put,
+
 	.set_guest_debug = set_guest_debug,
 	.get_msr = vmx_get_msr,
 	.set_msr = vmx_set_msr,

[PATCH 31/38] KVM: Make vcpu creation and destruction arch operations

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -260,6 +260,9 @@ struct kvm_arch_ops {
 	int (*hardware_setup)(void);               /* __init */
 	void (*hardware_unsetup)(void);            /* __exit */
 
+	int (*vcpu_create)(struct kvm_vcpu *vcpu);
+	void (*vcpu_free)(struct kvm_vcpu *vcpu);
+
 	struct kvm_vcpu *(*vcpu_load)(struct kvm_vcpu *vcpu);
 	void (*vcpu_put)(struct kvm_vcpu *vcpu);
 
@@ -299,8 +302,6 @@ struct kvm_arch_ops {
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
 };
 
-void __vcpu_clear(void *arg); /* temporary hack */
-
 extern struct kvm_stat kvm_stat;
 extern struct kvm_arch_ops *kvm_arch_ops;
 
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -219,31 +219,6 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 }
 EXPORT_SYMBOL_GPL(kvm_write_guest);
 
-static void vmcs_clear(struct vmcs *vmcs)
-{
-	u64 phys_addr = __pa(vmcs);
-	u8 error;
-
-	asm volatile (ASM_VMX_VMCLEAR_RAX "; setna %0"
-		      : "=g"(error) : "a"(&phys_addr), "m"(phys_addr)
-		      : "cc", "memory");
-	if (error)
-		printk(KERN_ERR "kvm: vmclear fail: %p/%llx\n",
-		       vmcs, phys_addr);
-}
-
-void __vcpu_clear(void *arg)
-{
-	struct kvm_vcpu *vcpu = arg;
-	int cpu = smp_processor_id();
-
-	if (vcpu->cpu == cpu)
-		vmcs_clear(vcpu->vmcs);
-	if (per_cpu(current_vmcs, cpu) == vcpu->vmcs)
-		per_cpu(current_vmcs, cpu) = 0;
-}
-EXPORT_SYMBOL_GPL(__vcpu_clear);
-
 static int vcpu_slot(struct kvm_vcpu *vcpu)
 {
 	return vcpu - vcpu->kvm->vcpus;
@@ -271,33 +246,6 @@ static void vcpu_put(struct kvm_vcpu *vc
 	mutex_unlock(&vcpu->mutex);
 }
 
-struct vmcs *alloc_vmcs_cpu(int cpu)
-{
-	int node = cpu_to_node(cpu);
-	struct page *pages;
-	struct vmcs *vmcs;
-
-	pages = alloc_pages_node(node, GFP_KERNEL, vmcs_descriptor.order);
-	if (!pages)
-		return 0;
-	vmcs = page_address(pages);
-	memset(vmcs, 0, vmcs_descriptor.size);
-	vmcs->revision_id = vmcs_descriptor.revision_id; /* vmcs revision id */
-	return vmcs;
-}
-EXPORT_SYMBOL_GPL(alloc_vmcs_cpu);
-
-static struct vmcs *alloc_vmcs(void)
-{
-	return alloc_vmcs_cpu(smp_processor_id());
-}
-
-void free_vmcs(struct vmcs *vmcs)
-{
-	free_pages((unsigned long)vmcs, vmcs_descriptor.order);
-}
-EXPORT_SYMBOL_GPL(free_vmcs);
-
 static int kvm_dev_open(struct inode *inode, struct file *filp)
 {
 	struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
@@ -350,18 +298,9 @@ static void kvm_free_physmem(struct kvm 
 		kvm_free_physmem_slot(&kvm->memslots[i], 0);
 }
 
-static void kvm_free_vmcs(struct kvm_vcpu *vcpu)
-{
-	if (vcpu->vmcs) {
-		on_each_cpu(__vcpu_clear, vcpu, 0, 1);
-		free_vmcs(vcpu->vmcs);
-		vcpu->vmcs = 0;
-	}
-}
-
 static void kvm_free_vcpu(struct kvm_vcpu *vcpu)
 {
-	kvm_free_vmcs(vcpu);
+	kvm_arch_ops->vcpu_free(vcpu);
 	kvm_mmu_destroy(vcpu);
 }
 
@@ -613,7 +552,6 @@ static int kvm_dev_ioctl_create_vcpu(str
 {
 	int r;
 	struct kvm_vcpu *vcpu;
-	struct vmcs *vmcs;
 
 	r = -EINVAL;
 	if (n < 0 || n >= KVM_MAX_VCPUS)
@@ -634,14 +572,9 @@ static int kvm_dev_ioctl_create_vcpu(str
 
 	vcpu->cpu = -1;  /* First load will set up TR */
 	vcpu->kvm = kvm;
-	vmcs = alloc_vmcs();
-	if (!vmcs) {
-		mutex_unlock(&vcpu->mutex);
+	r = kvm_arch_ops->vcpu_create(vcpu);
+	if (r < 0)
 		goto out_free_vcpus;
-	}
-	vmcs_clear(vmcs);
-	vcpu->vmcs = vmcs;
-	vcpu->launched = 0;
 
 	kvm_arch_ops->vcpu_load(vcpu);
 
@@ -658,6 +591,7 @@ static int kvm_dev_ioctl_create_vcpu(str
 
 out_free_vcpus:
 	kvm_free_vcpu(vcpu);
+	mutex_unlock(&vcpu->mutex);
 out:
 	return r;
 }
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -77,6 +77,30 @@ static const u32 vmx_msr_index[] = {
 
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
+static void vmcs_clear(struct vmcs *vmcs)
+{
+	u64 phys_addr = __pa(vmcs);
+	u8 error;
+
+	asm volatile (ASM_VMX_VMCLEAR_RAX "; setna %0"
+		      : "=g"(error) : "a"(&phys_addr), "m"(phys_addr)
+		      : "cc", "memory");
+	if (error)
+		printk(KERN_ERR "kvm: vmclear fail: %p/%llx\n",
+		       vmcs, phys_addr);
+}
+
+static void __vcpu_clear(void *arg)
+{
+	struct kvm_vcpu *vcpu = arg;
+	int cpu = smp_processor_id();
+
+	if (vcpu->cpu == cpu)
+		vmcs_clear(vcpu->vmcs);
+	if (per_cpu(current_vmcs, cpu) == vcpu->vmcs)
+		per_cpu(current_vmcs, cpu) = 0;
+}
+
 /*
  * Switches to specified vcpu, until a matching vcpu_put(), but assumes
  * vcpu mutex is already taken.
@@ -449,7 +473,30 @@ static __init void setup_vmcs_descriptor
 	vmcs_descriptor.revision_id = vmx_msr_low;
 };
 
-void free_vmcs(struct vmcs *vmcs);
+static struct vmcs *alloc_vmcs_cpu(int cpu)
+{
+	int node = cpu_to_node(cpu);
+	struct page *pages;
+	struct vmcs *vmcs;
+
+	pages = alloc_pages_node(node, GFP_KERNEL, vmcs_descriptor.order);
+	if (!pages)
+		return 0;
+	vmcs = page_address(pages);
+	memset(vmcs, 0, vmcs_descriptor.size);
+	vmcs->revision_id = vmcs_descriptor.revision_id; /* vmcs revision id */
+	return vmcs;
+}
+
+static struct vmcs *alloc_vmcs(void)
+{
+	return alloc_vmcs_cpu(smp_processor_id());
+}
+
+static void free_vmcs(struct vmcs *vmcs)
+{
+	free_pages((unsigned long)vmcs, vmcs_descriptor.order);
+}
 
 static __exit void free_kvm_area(void)
 {
@@ -1790,6 +1837,33 @@ static void vmx_inject_page_fault(struct
 
 }
 
+static void vmx_free_vmcs(struct kvm_vcpu *vcpu)
+{
+	if (vcpu->vmcs) {
+		on_each_cpu(__vcpu_clear, vcpu, 0, 1);
+		free_vmcs(vcpu->vmcs);
+		vcpu->vmcs = 0;
+	}
+}
+
+static void vmx_free_vcpu(struct kvm_vcpu *vcpu)
+{
+	vmx_free_vmcs(vcpu);
+}
+
+static int vmx_create_vcpu(struct kvm_vcpu *vcpu)
+{
+	struct vmcs *vmcs;
+
+	vmcs = alloc_vmcs();
+	if (!vmcs)
+		return -ENOMEM;
+	vmcs_clear(vmcs);
+	vcpu->vmcs = vmcs;
+	vcpu->launched = 0;
+	return 0;
+}
+
 static struct kvm_arch_ops vmx_arch_ops = {
 	.cpu_has_kvm_support = cpu_has_kvm_support,
 	.disabled_by_bios = vmx_disabled_by_bios,
@@ -1798,6 +1872,9 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.hardware_enable = hardware_enable,
 	.hardware_disable = hardware_disable,
 
+	.vcpu_create = vmx_create_vcpu,
+	.vcpu_free = vmx_free_vcpu,
+
 	.vcpu_load = vmx_vcpu_load,
 	.vcpu_put = vmx_vcpu_put,
 

[PATCH 32/38] KVM: Move vmcs static variables to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -138,18 +138,6 @@ unsigned long segment_base(u16 selector)
 }
 EXPORT_SYMBOL_GPL(segment_base);
 
-DEFINE_PER_CPU(struct vmcs *, vmxarea);
-EXPORT_SYMBOL_GPL(per_cpu__vmxarea); /* temporary hack */
-DEFINE_PER_CPU(struct vmcs *, current_vmcs);
-EXPORT_SYMBOL_GPL(per_cpu__current_vmcs); /* temporary hack */
-
-struct vmcs_descriptor {
-	int size;
-	int order;
-	u32 revision_id;
-} vmcs_descriptor;
-EXPORT_SYMBOL_GPL(vmcs_descriptor);
-
 int kvm_read_guest(struct kvm_vcpu *vcpu,
 			     gva_t addr,
 			     unsigned long size,
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -28,8 +28,8 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
-DECLARE_PER_CPU(struct vmcs *, vmxarea);
-DECLARE_PER_CPU(struct vmcs *, current_vmcs);
+static DEFINE_PER_CPU(struct vmcs *, vmxarea);
+static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
 
 #ifdef __x86_64__
 #define HOST_IS_64 1
@@ -37,7 +37,7 @@ DECLARE_PER_CPU(struct vmcs *, current_v
 #define HOST_IS_64 0
 #endif
 
-extern struct vmcs_descriptor {
+static struct vmcs_descriptor {
 	int size;
 	int order;
 	u32 revision_id;

[PATCH 33/38] KVM: Make is_long_mode() an arch operation

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -275,6 +275,7 @@ struct kvm_arch_ops {
 			    struct kvm_segment *var, int seg);
 	void (*set_segment)(struct kvm_vcpu *vcpu,
 			    struct kvm_segment *var, int seg);
+	int (*is_long_mode)(struct kvm_vcpu *vcpu);
 	void (*get_cs_db_l_bits)(struct kvm_vcpu *vcpu, int *db, int *l);
 	void (*set_cr0)(struct kvm_vcpu *vcpu, unsigned long cr0);
 	void (*set_cr0_no_modeswitch)(struct kvm_vcpu *vcpu,
@@ -411,11 +412,6 @@ static inline void vmcs_write32(unsigned
 	vmcs_writel(field, value);
 }
 
-static inline int is_long_mode(void)
-{
-	return vmcs_read32(VM_ENTRY_CONTROLS) & VM_ENTRY_CONTROLS_IA32E_MASK;
-}
-
 static inline int is_pae(struct kvm_vcpu *vcpu)
 {
 	return vcpu->cr4 & CR4_PAE_MASK;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -441,7 +441,7 @@ void set_cr4(struct kvm_vcpu *vcpu, unsi
 		return;
 	}
 
-	if (is_long_mode()) {
+	if (kvm_arch_ops->is_long_mode(vcpu)) {
 		if (!(cr4 & CR4_PAE_MASK)) {
 			printk(KERN_DEBUG "set_cr4: #GP, clearing PAE while "
 			       "in long mode\n");
@@ -468,7 +468,7 @@ EXPORT_SYMBOL_GPL(set_cr4);
 
 void set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3)
 {
-	if (is_long_mode()) {
+	if (kvm_arch_ops->is_long_mode(vcpu)) {
 		if ( cr3 & CR3_L_MODE_RESEVED_BITS) {
 			printk(KERN_DEBUG "set_cr3: #GP, reserved bits\n");
 			inject_gp(vcpu);
Index: linux-2.6/drivers/kvm/mmu.c
===================================================================
--- linux-2.6.orig/drivers/kvm/mmu.c
+++ linux-2.6/drivers/kvm/mmu.c
@@ -589,7 +589,7 @@ static int init_kvm_mmu(struct kvm_vcpu 
 
 	if (!is_paging(vcpu))
 		return nonpaging_init_context(vcpu);
-	else if (is_long_mode())
+	else if (kvm_arch_ops->is_long_mode(vcpu))
 		return paging64_init_context(vcpu);
 	else if (is_pae(vcpu))
 		return paging32E_init_context(vcpu);
Index: linux-2.6/drivers/kvm/paging_tmpl.h
===================================================================
--- linux-2.6.orig/drivers/kvm/paging_tmpl.h
+++ linux-2.6/drivers/kvm/paging_tmpl.h
@@ -70,7 +70,7 @@ static void FNAME(init_walker)(struct gu
 	hpa = safe_gpa_to_hpa(vcpu, vcpu->cr3 & PT64_BASE_ADDR_MASK);
 	walker->table = kmap_atomic(pfn_to_page(hpa >> PAGE_SHIFT), KM_USER0);
 
-	ASSERT((!is_long_mode() && is_pae(vcpu)) ||
+	ASSERT((!kvm_arch_ops->is_long_mode(vcpu) && is_pae(vcpu)) ||
 	       (vcpu->cr3 & ~(PAGE_MASK | CR3_FLAGS_MASK)) == 0);
 
 	walker->table = (pt_element_t *)( (unsigned long)walker->table |
@@ -135,7 +135,7 @@ static pt_element_t *FNAME(fetch_guest)(
 		     (walker->table[index] & PT_PAGE_SIZE_MASK) &&
 		     (PTTYPE == 64 || is_pse(vcpu))))
 			return &walker->table[index];
-		if (walker->level != 3 || is_long_mode())
+		if (walker->level != 3 || kvm_arch_ops->is_long_mode(vcpu))
 			walker->inherited_ar &= walker->table[index];
 		paddr = safe_gpa_to_hpa(vcpu, walker->table[index] & PT_BASE_ADDR_MASK);
 		kunmap_atomic(walker->table, KM_USER0);
@@ -204,7 +204,7 @@ static u64 *FNAME(fetch)(struct kvm_vcpu
 		shadow_addr = kvm_mmu_alloc_page(vcpu, shadow_ent);
 		if (!VALID_PAGE(shadow_addr))
 			return ERR_PTR(-ENOMEM);
-		if (!is_long_mode() && level == 3)
+		if (!kvm_arch_ops->is_long_mode(vcpu) && level == 3)
 			*shadow_ent = shadow_addr |
 				(*guest_ent & (PT_PRESENT_MASK | PT_PWT_MASK | PT_PCD_MASK));
 		else {
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -811,6 +811,11 @@ static void vmx_set_segment(struct kvm_v
 	vmcs_write32(sf->ar_bytes, ar);
 }
 
+static int vmx_is_long_mode(struct kvm_vcpu *vcpu)
+{
+	return vmcs_read32(VM_ENTRY_CONTROLS) & VM_ENTRY_CONTROLS_IA32E_MASK;
+}
+
 static void vmx_get_cs_db_l_bits(struct kvm_vcpu *vcpu, int *db, int *l)
 {
 	u32 ar = vmcs_read32(GUEST_CS_AR_BYTES);
@@ -1884,6 +1889,7 @@ static struct kvm_arch_ops vmx_arch_ops 
 	.get_segment_base = vmx_get_segment_base,
 	.get_segment = vmx_get_segment,
 	.set_segment = vmx_set_segment,
+	.is_long_mode = vmx_is_long_mode,
 	.get_cs_db_l_bits = vmx_get_cs_db_l_bits,
 	.set_cr0 = vmx_set_cr0,
 	.set_cr0_no_modeswitch = vmx_set_cr0_no_modeswitch,

[PATCH 34/38] KVM: Use the tlb flush arch operation instead of an inline

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -445,11 +445,6 @@ static inline int is_external_interrupt(
 		== (INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK);
 }
 
-static inline void flush_guest_tlb(struct kvm_vcpu *vcpu)
-{
-	vmcs_writel(GUEST_CR3, vmcs_readl(GUEST_CR3));
-}
-
 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
 	return slot - kvm->memslots;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -775,7 +775,7 @@ static int kvm_dev_ioctl_get_dirty_log(s
 
 			if (!vcpu)
 				continue;
-			flush_guest_tlb(vcpu);
+			kvm_arch_ops->flush_tlb(vcpu);
 			vcpu_put(vcpu);
 		}
 	}

[PATCH 35/38] KVM: Remove guest_cpl()

[Avi Kivity]

This was used only in debugging.

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -422,11 +422,6 @@ static inline int is_pse(struct kvm_vcpu
 	return vcpu->cr4 & CR4_PSE_MASK;
 }
 
-static inline unsigned guest_cpl(void)
-{
-	return vmcs_read16(GUEST_CS_SELECTOR) & SELECTOR_RPL_MASK;
-}
-
 static inline int is_paging(struct kvm_vcpu *vcpu)
 {
 	return vcpu->cr0 & CR0_PG_MASK;
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -1391,14 +1391,6 @@ static int handle_cr(struct kvm_vcpu *vc
 	int cr;
 	int reg;
 
-#ifdef KVM_DEBUG
-	if (guest_cpl() != 0) {
-		vcpu_printf(vcpu, "%s: not supervisor\n", __FUNCTION__);
-		inject_gp(vcpu);
-		return 1;
-	}
-#endif
-
 	exit_qualification = vmcs_read64(EXIT_QUALIFICATION);
 	cr = exit_qualification & 15;
 	reg = (exit_qualification >> 8) & 15;

[PATCH 36/38] KVM: Move vmcs accessors to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -377,9 +377,6 @@ int kvm_write_guest(struct kvm_vcpu *vcp
 		unsigned long size,
 		void *data);
 
-void vmcs_writel(unsigned long field, unsigned long value);
-unsigned long vmcs_readl(unsigned long field);
-
 unsigned long segment_base(u16 selector);
 
 static inline struct page *_gfn_to_page(struct kvm *kvm, gfn_t gfn)
@@ -388,30 +385,6 @@ static inline struct page *_gfn_to_page(
 	return (slot) ? slot->phys_mem[gfn - slot->base_gfn] : 0;
 }
 
-static inline u16 vmcs_read16(unsigned long field)
-{
-	return vmcs_readl(field);
-}
-
-static inline u32 vmcs_read32(unsigned long field)
-{
-	return vmcs_readl(field);
-}
-
-static inline u64 vmcs_read64(unsigned long field)
-{
-#ifdef __x86_64__
-	return vmcs_readl(field);
-#else
-	return vmcs_readl(field) | ((u64)vmcs_readl(field+1) << 32);
-#endif
-}
-
-static inline void vmcs_write32(unsigned long field, u32 value)
-{
-	vmcs_writel(field, value);
-}
-
 static inline int is_pae(struct kvm_vcpu *vcpu)
 {
 	return vcpu->cr4 & CR4_PAE_MASK;
Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -310,28 +310,6 @@ static int kvm_dev_release(struct inode 
 	return 0;
 }
 
-unsigned long vmcs_readl(unsigned long field)
-{
-	unsigned long value;
-
-	asm volatile (ASM_VMX_VMREAD_RDX_RAX
-		      : "=a"(value) : "d"(field) : "cc");
-	return value;
-}
-EXPORT_SYMBOL_GPL(vmcs_readl);
-
-void vmcs_writel(unsigned long field, unsigned long value)
-{
-	u8 error;
-
-	asm volatile (ASM_VMX_VMWRITE_RAX_RDX "; setna %0"
-		       : "=q"(error) : "a"(value), "d"(field) : "cc" );
-	if (error)
-		printk(KERN_ERR "vmwrite error: reg %lx value %lx (err %d)\n",
-		       field, value, vmcs_read32(VM_INSTRUCTION_ERROR));
-}
-EXPORT_SYMBOL_GPL(vmcs_writel);
-
 static void inject_gp(struct kvm_vcpu *vcpu)
 {
 	kvm_arch_ops->inject_gp(vcpu, 0);
Index: linux-2.6/drivers/kvm/kvm_vmx.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_vmx.h
+++ linux-2.6/drivers/kvm/kvm_vmx.h
@@ -1,22 +1,6 @@
 #ifndef __KVM_VMX_H
 #define __KVM_VMX_H
 
-static inline void vmcs_write16(unsigned long field, u16 value)
-{
-	vmcs_writel(field, value);
-}
-
-static inline void vmcs_write64(unsigned long field, u64 value)
-{
-#ifdef __x86_64__
-	vmcs_writel(field, value);
-#else
-	vmcs_writel(field, value);
-	asm volatile ("");
-	vmcs_writel(field+1, value >> 32);
-#endif
-}
-
 #ifdef __x86_64__
 /*
  * avoid save/load MSR_SYSCALL_MASK and MSR_LSTAR by std vt
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -101,6 +101,66 @@ static void __vcpu_clear(void *arg)
 		per_cpu(current_vmcs, cpu) = 0;
 }
 
+static unsigned long vmcs_readl(unsigned long field)
+{
+	unsigned long value;
+
+	asm volatile (ASM_VMX_VMREAD_RDX_RAX
+		      : "=a"(value) : "d"(field) : "cc");
+	return value;
+}
+
+static u16 vmcs_read16(unsigned long field)
+{
+	return vmcs_readl(field);
+}
+
+static u32 vmcs_read32(unsigned long field)
+{
+	return vmcs_readl(field);
+}
+
+static u64 vmcs_read64(unsigned long field)
+{
+#ifdef __x86_64__
+	return vmcs_readl(field);
+#else
+	return vmcs_readl(field) | ((u64)vmcs_readl(field+1) << 32);
+#endif
+}
+
+static void vmcs_writel(unsigned long field, unsigned long value)
+{
+	u8 error;
+
+	asm volatile (ASM_VMX_VMWRITE_RAX_RDX "; setna %0"
+		       : "=q"(error) : "a"(value), "d"(field) : "cc" );
+	if (error)
+		printk(KERN_ERR "vmwrite error: reg %lx value %lx (err %d)\n",
+		       field, value, vmcs_read32(VM_INSTRUCTION_ERROR));
+}
+
+static void vmcs_write16(unsigned long field, u16 value)
+{
+	vmcs_writel(field, value);
+}
+
+static void vmcs_write32(unsigned long field, u32 value)
+{
+	vmcs_writel(field, value);
+}
+
+static void vmcs_write64(unsigned long field, u64 value)
+{
+#ifdef __x86_64__
+	vmcs_writel(field, value);
+#else
+	vmcs_writel(field, value);
+	asm volatile ("");
+	vmcs_writel(field+1, value >> 32);
+#endif
+}
+
 /*
  * Switches to specified vcpu, until a matching vcpu_put(), but assumes
  * vcpu mutex is already taken.

[PATCH 37/38] KVM: Move vmx helper inlines to vmx.c

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm.h
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm.h
+++ linux-2.6/drivers/kvm/kvm.h
@@ -400,19 +400,6 @@ static inline int is_paging(struct kvm_v
 	return vcpu->cr0 & CR0_PG_MASK;
 }
 
-static inline int is_page_fault(u32 intr_info)
-{
-	return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK |
-			     INTR_INFO_VALID_MASK)) ==
-		(INTR_TYPE_EXCEPTION | PF_VECTOR | INTR_INFO_VALID_MASK);
-}
-
-static inline int is_external_interrupt(u32 intr_info)
-{
-	return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK))
-		== (INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK);
-}
-
 static inline int memslot_id(struct kvm *kvm, struct kvm_memory_slot *slot)
 {
 	return slot - kvm->memslots;
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -77,6 +77,19 @@ static const u32 vmx_msr_index[] = {
 
 struct vmx_msr_entry *find_msr_entry(struct kvm_vcpu *vcpu, u32 msr);
 
+static inline int is_page_fault(u32 intr_info)
+{
+	return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VECTOR_MASK |
+			     INTR_INFO_VALID_MASK)) ==
+		(INTR_TYPE_EXCEPTION | PF_VECTOR | INTR_INFO_VALID_MASK);
+}
+
+static inline int is_external_interrupt(u32 intr_info)
+{
+	return (intr_info & (INTR_INFO_INTR_TYPE_MASK | INTR_INFO_VALID_MASK))
+		== (INTR_TYPE_EXT_INTR | INTR_INFO_VALID_MASK);
+}
+
 static void vmcs_clear(struct vmcs *vmcs)
 {
 	u64 phys_addr = __pa(vmcs);

[PATCH 38/38] KVM: Remove vmx includes from arch independent code

[Avi Kivity]

Signed-off-by: Avi Kivity <avi@qumranet.com>

Index: linux-2.6/drivers/kvm/kvm_main.c
===================================================================
--- linux-2.6.orig/drivers/kvm/kvm_main.c
+++ linux-2.6/drivers/kvm/kvm_main.c
@@ -35,8 +35,6 @@
 #include <linux/file.h>
 #include <asm/desc.h>
 
-#include "vmx.h"
-#include "kvm_vmx.h"
 #include "x86_emulate.h"
 
 MODULE_AUTHOR("Qumranet");
Index: linux-2.6/drivers/kvm/vmx.c
===================================================================
--- linux-2.6.orig/drivers/kvm/vmx.c
+++ linux-2.6/drivers/kvm/vmx.c
@@ -16,9 +16,9 @@
  */
 
 #include "kvm.h"
-#include <linux/module.h>
 #include "vmx.h"
 #include "kvm_vmx.h"
+#include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/highmem.h>
 #include <asm/io.h>