[KVM TSC emulation 9/9] Add software TSC emulation

From: Zachary Amsden
Date: Mon Jun 20 2011 - 20:00:32 EST

Next message: Zachary Amsden: "[KVM TSC emulation 4/9] Fix last_guest_tsc / tsc_offset semantics"
Previous message: Zachary Amsden: "[KVM TSC emulation 8/9] Track TSC synchronization in generations"
In reply to: Zachary Amsden: "[KVM TSC emulation 8/9] Track TSC synchronization in generations"
Next in thread: Zachary Amsden: "[KVM TSC emulation 4/9] Fix last_guest_tsc / tsc_offset semantics"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

When hardware assistance is unavailable to scale the TSC, or it is
not possible to keep in sync, add a software virtualization mode
where the TSC is trapped and thus guaranteed to always have perfect
synchronization.

Currently this behavior defaults to on; how and when the decision to
use trapping is made is likely to be a matter of debate. For now,
just make it possible.

Signed-off-by: Zachary Amsden <zamsden@xxxxxxxxxx>
---
arch/x86/kvm/svm.c | 26 +++++++++++++++++++++++++-
arch/x86/kvm/vmx.c | 28 +++++++++++++++++++++++++++-
arch/x86/kvm/x86.c | 34 +++++++++++++++++++++++-----------
arch/x86/kvm/x86.h | 5 +++++
4 files changed, 80 insertions(+), 13 deletions(-)

diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index dcab00e..fc4583d 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -185,6 +185,7 @@ module_param(nested, int, S_IRUGO);

static void svm_flush_tlb(struct kvm_vcpu *vcpu);
static void svm_complete_interrupts(struct vcpu_svm *svm);
+static void svm_set_tsc_trapping(struct kvm_vcpu *vcpu, bool trap);

static int nested_svm_exit_handled(struct vcpu_svm *svm);
static int nested_svm_intercept(struct vcpu_svm *svm);
@@ -912,13 +913,18 @@ static void svm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale)
u64 khz;

/* Guest TSC same frequency as host TSC? */
- if (!scale) {
+ if (!scale && !check_tsc_unstable()) {
svm->tsc_ratio = TSC_RATIO_DEFAULT;
return;
}

/* TSC scaling supported? */
if (!boot_cpu_has(X86_FEATURE_TSCRATEMSR)) {
+ if (kvm_software_tsc) {
+ pr_debug("kvm: using TSC trapping\n");
+ svm_set_tsc_trapping(vcpu, true);
+ return;
+ }
if (user_tsc_khz > tsc_khz) {
vcpu->arch.tsc_catchup = 1;
vcpu->arch.tsc_always_catchup = 1;
@@ -1184,6 +1190,7 @@ static struct kvm_vcpu *svm_create_vcpu(struct kvm *kvm, unsigned int id)
svm->vmcb_pa = page_to_pfn(page) << PAGE_SHIFT;
svm->asid_generation = 0;
init_vmcb(svm);
+ kvm_set_tsc_khz(&svm->vcpu, kvm_max_tsc_khz);
kvm_write_tsc(&svm->vcpu, 0);

err = fx_init(&svm->vcpu);
@@ -1303,6 +1310,15 @@ static void svm_clear_vintr(struct vcpu_svm *svm)
clr_intercept(svm, INTERCEPT_VINTR);
}

+static void svm_set_tsc_trapping(struct kvm_vcpu *vcpu, bool trap)
+{
+ struct vcpu_svm *svm = to_svm(vcpu);
+ if (trap)
+ set_intercept(svm, INTERCEPT_RDTSC);
+ else
+ clr_intercept(svm, INTERCEPT_RDTSC);
+}
+
static struct vmcb_seg *svm_seg(struct kvm_vcpu *vcpu, int seg)
{
struct vmcb_save_area *save = &to_svm(vcpu)->vmcb->save;
@@ -2732,6 +2748,13 @@ static int task_switch_interception(struct vcpu_svm *svm)
return 1;
}

+static int rdtsc_interception(struct vcpu_svm *svm)
+{
+ svm->next_rip = kvm_rip_read(&svm->vcpu) + 2;
+ kvm_read_tsc(&svm->vcpu);
+ return 1;
+}
+
static int cpuid_interception(struct vcpu_svm *svm)
{
svm->next_rip = kvm_rip_read(&svm->vcpu) + 2;
@@ -3178,6 +3201,7 @@ static int (*svm_exit_handlers[])(struct vcpu_svm *svm) = {
[SVM_EXIT_SMI] = nop_on_interception,
[SVM_EXIT_INIT] = nop_on_interception,
[SVM_EXIT_VINTR] = interrupt_window_interception,
+ [SVM_EXIT_RDTSC] = rdtsc_interception,
[SVM_EXIT_CPUID] = cpuid_interception,
[SVM_EXIT_IRET] = iret_interception,
[SVM_EXIT_INVD] = emulate_on_interception,
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 780fe12..65066b4 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -606,6 +606,7 @@ static void kvm_cpu_vmxon(u64 addr);
static void kvm_cpu_vmxoff(void);
static void vmx_set_cr3(struct kvm_vcpu *vcpu, unsigned long cr3);
static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr);
+static void vmx_set_tsc_trapping(struct kvm_vcpu *vcpu, bool trap);

static DEFINE_PER_CPU(struct vmcs *, vmxarea);
static DEFINE_PER_CPU(struct vmcs *, current_vmcs);
@@ -1756,9 +1757,14 @@ static u64 guest_read_tsc(void)
*/
static void vmx_set_tsc_khz(struct kvm_vcpu *vcpu, u32 user_tsc_khz, bool scale)
{
- if (!scale)
+ if (!scale && !check_tsc_unstable())
return;

+ if (kvm_software_tsc) {
+ pr_debug("kvm: using TSC trapping\n");
+ vmx_set_tsc_trapping(vcpu, true);
+ return;
+ }
if (user_tsc_khz > tsc_khz) {
vcpu->arch.tsc_catchup = 1;
vcpu->arch.tsc_always_catchup = 1;
@@ -3695,6 +3701,7 @@ static int vmx_vcpu_setup(struct vcpu_vmx *vmx)
vmcs_writel(CR0_GUEST_HOST_MASK, ~0UL);
set_cr4_guest_host_mask(vmx);

+ kvm_set_tsc_khz(&vmx->vcpu, kvm_max_tsc_khz);
kvm_write_tsc(&vmx->vcpu, 0);

return 0;
@@ -3997,6 +4004,18 @@ static int vmx_set_tss_addr(struct kvm *kvm, unsigned int addr)
return 0;
}

+static void vmx_set_tsc_trapping(struct kvm_vcpu *vcpu, bool trap)
+{
+ u32 cpu_based_vm_exec_control;
+
+ cpu_based_vm_exec_control = vmcs_read32(CPU_BASED_VM_EXEC_CONTROL);
+ if (trap)
+ cpu_based_vm_exec_control |= CPU_BASED_RDTSC_EXITING;
+ else
+ cpu_based_vm_exec_control &= ~CPU_BASED_RDTSC_EXITING;
+ vmcs_write32(CPU_BASED_VM_EXEC_CONTROL, cpu_based_vm_exec_control);
+}
+
static int handle_rmode_exception(struct kvm_vcpu *vcpu,
int vec, u32 err_code)
{
@@ -4497,6 +4516,12 @@ static int handle_invlpg(struct kvm_vcpu *vcpu)
return 1;
}

+static int handle_rdtsc(struct kvm_vcpu *vcpu)
+{
+ kvm_read_tsc(vcpu);
+ return 1;
+}
+
static int handle_wbinvd(struct kvm_vcpu *vcpu)
{
skip_emulated_instruction(vcpu);
@@ -5421,6 +5446,7 @@ static int (*kvm_vmx_exit_handlers[])(struct kvm_vcpu *vcpu) = {
[EXIT_REASON_HLT] = handle_halt,
[EXIT_REASON_INVD] = handle_invd,
[EXIT_REASON_INVLPG] = handle_invlpg,
+ [EXIT_REASON_RDTSC] = handle_rdtsc,
[EXIT_REASON_VMCALL] = handle_vmcall,
[EXIT_REASON_VMCLEAR] = handle_vmclear,
[EXIT_REASON_VMLAUNCH] = handle_vmlaunch,
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 09e67fb..1a07796 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -99,6 +99,10 @@ EXPORT_SYMBOL_GPL(kvm_max_guest_tsc_khz);
static u32 tsc_tolerance_ppm = 250;
module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR);

+int kvm_software_tsc = 1;
+module_param_named(software_tsc_emulation, kvm_software_tsc, bool, 0644);
+EXPORT_SYMBOL_GPL(kvm_software_tsc);
+
#define KVM_NR_SHARED_MSRS 16

struct kvm_shared_msrs_global {
@@ -993,7 +997,8 @@ static inline u64 get_kernel_ns(void)
}

static DEFINE_PER_CPU(unsigned long, cpu_tsc_khz);
-unsigned long max_tsc_khz;
+unsigned long kvm_max_tsc_khz;
+EXPORT_SYMBOL_GPL(kvm_max_tsc_khz);

static inline u64 nsec_to_cycles(struct kvm_vcpu *vcpu, u64 nsec)
{
@@ -1001,7 +1006,7 @@ static inline u64 nsec_to_cycles(struct kvm_vcpu *vcpu, u64 nsec)
vcpu->arch.virtual_tsc_shift);
}

-static void kvm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 this_tsc_khz)
+void kvm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 this_tsc_khz)
{
u32 thresh_lo, thresh_hi;
int use_scaling = 0;
@@ -1026,6 +1031,7 @@ static void kvm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 this_tsc_khz)
}
kvm_x86_ops->set_tsc_khz(vcpu, this_tsc_khz, use_scaling);
}
+EXPORT_SYMBOL_GPL(kvm_set_tsc_khz);

static u64 compute_guest_tsc(struct kvm_vcpu *vcpu, s64 kernel_ns)
{
@@ -1117,6 +1123,18 @@ void kvm_write_tsc(struct kvm_vcpu *vcpu, u64 data)

EXPORT_SYMBOL_GPL(kvm_write_tsc);

+void kvm_read_tsc(struct kvm_vcpu *vcpu)
+{
+ u64 tsc;
+ s64 kernel_ns = get_kernel_ns();
+
+ tsc = compute_guest_tsc(vcpu, kernel_ns);
+ kvm_register_write(vcpu, VCPU_REGS_RAX, (u32)tsc);
+ kvm_register_write(vcpu, VCPU_REGS_RDX, tsc >> 32);
+ kvm_x86_ops->skip_emulated_instruction(vcpu);
+}
+EXPORT_SYMBOL_GPL(kvm_read_tsc);
+
static int kvm_guest_time_update(struct kvm_vcpu *v)
{
unsigned long flags;
@@ -4931,7 +4949,7 @@ static void kvm_timer_init(void)
{
int cpu;

- max_tsc_khz = tsc_khz;
+ kvm_max_tsc_khz = tsc_khz;
register_hotcpu_notifier(&kvmclock_cpu_notifier_block);
if (!boot_cpu_has(X86_FEATURE_CONSTANT_TSC)) {
#ifdef CONFIG_CPU_FREQ
@@ -4940,13 +4958,13 @@ static void kvm_timer_init(void)
cpu = get_cpu();
cpufreq_get_policy(&policy, cpu);
if (policy.cpuinfo.max_freq)
- max_tsc_khz = policy.cpuinfo.max_freq;
+ kvm_max_tsc_khz = policy.cpuinfo.max_freq;
put_cpu();
#endif
cpufreq_register_notifier(&kvmclock_cpufreq_notifier_block,
CPUFREQ_TRANSITION_NOTIFIER);
}
- pr_debug("kvm: max_tsc_khz = %ld\n", max_tsc_khz);
+ pr_debug("kvm: max_tsc_khz = %ld\n", kvm_max_tsc_khz);
for_each_online_cpu(cpu)
smp_call_function_single(cpu, tsc_khz_changed, NULL, 1);
}
@@ -6194,10 +6212,6 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
struct kvm_vcpu *kvm_arch_vcpu_create(struct kvm *kvm,
unsigned int id)
{
- if (check_tsc_unstable() && atomic_read(&kvm->online_vcpus) != 0)
- printk_once(KERN_WARNING
- "kvm: SMP vm created on host with unstable TSC; "
- "guest TSC will not be reliable\n");
return kvm_x86_ops->vcpu_create(kvm, id);
}

@@ -6385,8 +6399,6 @@ int kvm_arch_vcpu_init(struct kvm_vcpu *vcpu)
}
vcpu->arch.pio_data = page_address(page);

- kvm_set_tsc_khz(vcpu, max_tsc_khz);
-
r = kvm_mmu_create(vcpu);
if (r < 0)
goto fail_free_pio_data;
diff --git a/arch/x86/kvm/x86.h b/arch/x86/kvm/x86.h
index 256da82..94780df 100644
--- a/arch/x86/kvm/x86.h
+++ b/arch/x86/kvm/x86.h
@@ -80,6 +80,10 @@ void kvm_after_handle_nmi(struct kvm_vcpu *vcpu);
int kvm_inject_realmode_interrupt(struct kvm_vcpu *vcpu, int irq, int inc_eip);

void kvm_write_tsc(struct kvm_vcpu *vcpu, u64 data);
+void kvm_read_tsc(struct kvm_vcpu *vcpu);
+void kvm_set_tsc_khz(struct kvm_vcpu *vcpu, u32 this_tsc_khz);
+extern int kvm_software_tsc;
+extern unsigned long kvm_max_tsc_khz;

int kvm_read_guest_virt(struct x86_emulate_ctxt *ctxt,
gva_t addr, void *val, unsigned int bytes,
@@ -89,4 +93,5 @@ int kvm_write_guest_virt_system(struct x86_emulate_ctxt *ctxt,
gva_t addr, void *val, unsigned int bytes,
struct x86_exception *exception);

+
#endif
--
1.7.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

Next message: Zachary Amsden: "[KVM TSC emulation 4/9] Fix last_guest_tsc / tsc_offset semantics"
Previous message: Zachary Amsden: "[KVM TSC emulation 8/9] Track TSC synchronization in generations"
In reply to: Zachary Amsden: "[KVM TSC emulation 8/9] Track TSC synchronization in generations"
Next in thread: Zachary Amsden: "[KVM TSC emulation 4/9] Fix last_guest_tsc / tsc_offset semantics"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]