[patch 0/3] KVM: add option to advance tscdeadline hrtimer expiration (v5)

[patch 0/3] KVM: add option to advance tscdeadline hrtimer expiration (v5)

[Marcelo Tosatti]

See patches for details.

v2:
- fix email address.

v3:
- use module parameter for configuration of value (Paolo/Radim)

v4:
- fix check for tscdeadline mode while waiting for expiration (Paolo)
- use proper delay function (Radim)
- fix LVTT tscdeadline mode check in hrtimer interrupt handler (Radim)
- add comment regarding PPR and APICv (Paolo)

v5:
- use tscdeadline expiration and guest tsc difference in 
  tracepoint (Paolo)

[patch 1/3] KVM: x86: add method to test PIR bitmap vector

[Marcelo Tosatti]

[-- Attachment #1: add-test-pir --]
[-- Type: text/plain, Size: 2342 bytes --]

kvm_x86_ops->test_posted_interrupt() returns true/false depending
whether 'vector' is set.

Next patch makes use of this interface.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/arch/x86/include/asm/kvm_host.h
@@ -743,6 +743,7 @@ struct kvm_x86_ops {
 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
+	bool (*test_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
 	int (*get_tdp_level)(void);
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -435,6 +435,11 @@ static int pi_test_and_set_pir(int vecto
 	return test_and_set_bit(vector, (unsigned long *)pi_desc->pir);
 }
 
+static int pi_test_pir(int vector, struct pi_desc *pi_desc)
+{
+	return test_bit(vector, (unsigned long *)pi_desc->pir);
+}
+
 struct vcpu_vmx {
 	struct kvm_vcpu       vcpu;
 	unsigned long         host_rsp;
@@ -5939,6 +5944,7 @@ static __init int hardware_setup(void)
 	else {
 		kvm_x86_ops->hwapic_irr_update = NULL;
 		kvm_x86_ops->deliver_posted_interrupt = NULL;
+		kvm_x86_ops->test_posted_interrupt = NULL;
 		kvm_x86_ops->sync_pir_to_irr = vmx_sync_pir_to_irr_dummy;
 	}
 
@@ -6960,6 +6966,13 @@ static int handle_invvpid(struct kvm_vcp
 	return 1;
 }
 
+static bool vmx_test_pir(struct kvm_vcpu *vcpu, int vector)
+{
+	struct vcpu_vmx *vmx = to_vmx(vcpu);
+
+	return pi_test_pir(vector, &vmx->pi_desc);
+}
+
 /*
  * 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
@@ -9374,6 +9387,7 @@ static struct kvm_x86_ops vmx_x86_ops =
 	.hwapic_isr_update = vmx_hwapic_isr_update,
 	.sync_pir_to_irr = vmx_sync_pir_to_irr,
 	.deliver_posted_interrupt = vmx_deliver_posted_interrupt,
+	.test_posted_interrupt = vmx_test_pir,
 
 	.set_tss_addr = vmx_set_tss_addr,
 	.get_tdp_level = get_ept_level,

Re: [patch 1/3] KVM: x86: add method to test PIR bitmap vector

[Radim Krcmar]

2014-12-16 09:08-0500, Marcelo Tosatti:
> kvm_x86_ops->test_posted_interrupt() returns true/false depending
> whether 'vector' is set.
> 
> Next patch makes use of this interface.
> 
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>

[patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Marcelo Tosatti]

[-- Attachment #1: lapic-timer-advance --]
[-- Type: text/plain, Size: 5366 bytes --]

For the hrtimer which emulates the tscdeadline timer in the guest,
add an option to advance expiration, and busy spin on VM-entry waiting
for the actual expiration time to elapse.

This allows achieving low latencies in cyclictest (or any scenario 
which requires strict timing regarding timer expiration).

Reduces average cyclictest latency from 12us to 8us
on Core i5 desktop.

Note: this option requires tuning to find the appropriate value 
for a particular hardware/guest combination. One method is to measure the 
average delay between apic_timer_fn and VM-entry. 
Another method is to start with 1000ns, and increase the value
in say 500ns increments until avg cyclictest numbers stop decreasing.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Index: kvm/arch/x86/kvm/lapic.c
===================================================================
--- kvm.orig/arch/x86/kvm/lapic.c
+++ kvm/arch/x86/kvm/lapic.c
@@ -33,6 +33,7 @@
 #include <asm/page.h>
 #include <asm/current.h>
 #include <asm/apicdef.h>
+#include <asm/delay.h>
 #include <linux/atomic.h>
 #include <linux/jump_label.h>
 #include "kvm_cache_regs.h"
@@ -1073,6 +1074,7 @@ static void apic_timer_expired(struct kv
 {
 	struct kvm_vcpu *vcpu = apic->vcpu;
 	wait_queue_head_t *q = &vcpu->wq;
+	struct kvm_timer *ktimer = &apic->lapic_timer;
 
 	/*
 	 * Note: KVM_REQ_PENDING_TIMER is implicitly checked in
@@ -1087,11 +1089,64 @@ static void apic_timer_expired(struct kv
 
 	if (waitqueue_active(q))
 		wake_up_interruptible(q);
+
+	if (apic_lvtt_tscdeadline(apic))
+		ktimer->expired_tscdeadline = ktimer->tscdeadline;
+}
+
+/*
+ * On APICv, this test will cause a busy wait
+ * during a higher-priority task.
+ */
+
+static bool lapic_timer_int_injected(struct kvm_vcpu *vcpu)
+{
+	struct kvm_lapic *apic = vcpu->arch.apic;
+	u32 reg = kvm_apic_get_reg(apic, APIC_LVTT);
+
+	if (kvm_apic_hw_enabled(apic)) {
+		int vec = reg & APIC_VECTOR_MASK;
+
+		if (kvm_x86_ops->test_posted_interrupt)
+			return kvm_x86_ops->test_posted_interrupt(vcpu, vec);
+		else {
+			if (apic_test_vector(vec, apic->regs + APIC_ISR))
+				return true;
+		}
+	}
+	return false;
+}
+
+void wait_lapic_expire(struct kvm_vcpu *vcpu)
+{
+	struct kvm_lapic *apic = vcpu->arch.apic;
+	u64 guest_tsc, tsc_deadline;
+
+	if (!kvm_vcpu_has_lapic(vcpu))
+		return;
+
+	if (apic->lapic_timer.expired_tscdeadline == 0)
+		return;
+
+	if (!lapic_timer_int_injected(vcpu))
+		return;
+
+	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
+	apic->lapic_timer.expired_tscdeadline = 0;
+	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
+
+	while (guest_tsc < tsc_deadline) {
+		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
+
+		__delay(delay);
+		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
+	}
 }
 
 static void start_apic_timer(struct kvm_lapic *apic)
 {
 	ktime_t now;
+
 	atomic_set(&apic->lapic_timer.pending, 0);
 
 	if (apic_lvtt_period(apic) || apic_lvtt_oneshot(apic)) {
@@ -1137,6 +1192,7 @@ static void start_apic_timer(struct kvm_
 		/* lapic timer in tsc deadline mode */
 		u64 guest_tsc, tscdeadline = apic->lapic_timer.tscdeadline;
 		u64 ns = 0;
+		ktime_t expire;
 		struct kvm_vcpu *vcpu = apic->vcpu;
 		unsigned long this_tsc_khz = vcpu->arch.virtual_tsc_khz;
 		unsigned long flags;
@@ -1151,8 +1207,10 @@ static void start_apic_timer(struct kvm_
 		if (likely(tscdeadline > guest_tsc)) {
 			ns = (tscdeadline - guest_tsc) * 1000000ULL;
 			do_div(ns, this_tsc_khz);
+			expire = ktime_add_ns(now, ns);
+			expire = ktime_sub_ns(expire, lapic_timer_advance_ns);
 			hrtimer_start(&apic->lapic_timer.timer,
-				ktime_add_ns(now, ns), HRTIMER_MODE_ABS);
+				      expire, HRTIMER_MODE_ABS);
 		} else
 			apic_timer_expired(apic);
 
Index: kvm/arch/x86/kvm/lapic.h
===================================================================
--- kvm.orig/arch/x86/kvm/lapic.h
+++ kvm/arch/x86/kvm/lapic.h
@@ -14,6 +14,7 @@ struct kvm_timer {
 	u32 timer_mode;
 	u32 timer_mode_mask;
 	u64 tscdeadline;
+	u64 expired_tscdeadline;
 	atomic_t pending;			/* accumulated triggered timers */
 };
 
@@ -170,4 +171,6 @@ static inline bool kvm_apic_has_events(s
 
 bool kvm_apic_pending_eoi(struct kvm_vcpu *vcpu, int vector);
 
+void wait_lapic_expire(struct kvm_vcpu *vcpu);
+
 #endif
Index: kvm/arch/x86/kvm/x86.c
===================================================================
--- kvm.orig/arch/x86/kvm/x86.c
+++ kvm/arch/x86/kvm/x86.c
@@ -108,6 +108,10 @@ EXPORT_SYMBOL_GPL(kvm_max_guest_tsc_khz)
 static u32 tsc_tolerance_ppm = 250;
 module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR);
 
+/* lapic timer advance (tscdeadline mode only) in nanoseconds */
+unsigned int lapic_timer_advance_ns = 0;
+module_param(lapic_timer_advance_ns, uint, S_IRUGO | S_IWUSR);
+
 static bool backwards_tsc_observed = false;
 
 #define KVM_NR_SHARED_MSRS 16
@@ -6311,6 +6315,7 @@ static int vcpu_enter_guest(struct kvm_v
 	}
 
 	trace_kvm_entry(vcpu->vcpu_id);
+	wait_lapic_expire(vcpu);
 	kvm_x86_ops->run(vcpu);
 
 	/*
Index: kvm/arch/x86/kvm/x86.h
===================================================================
--- kvm.orig/arch/x86/kvm/x86.h
+++ kvm/arch/x86/kvm/x86.h
@@ -170,5 +170,7 @@ extern u64 kvm_supported_xcr0(void);
 
 extern unsigned int min_timer_period_us;
 
+extern unsigned int lapic_timer_advance_ns;
+
 extern struct static_key kvm_no_apic_vcpu;
 #endif

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Radim Krcmar]

2014-12-16 09:08-0500, Marcelo Tosatti:
> For the hrtimer which emulates the tscdeadline timer in the guest,
> add an option to advance expiration, and busy spin on VM-entry waiting
> for the actual expiration time to elapse.
> 
> This allows achieving low latencies in cyclictest (or any scenario 
> which requires strict timing regarding timer expiration).
> 
> Reduces average cyclictest latency from 12us to 8us
> on Core i5 desktop.
> 
> Note: this option requires tuning to find the appropriate value 
> for a particular hardware/guest combination. One method is to measure the 
> average delay between apic_timer_fn and VM-entry. 
> Another method is to start with 1000ns, and increase the value
> in say 500ns increments until avg cyclictest numbers stop decreasing.
> 
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>

> +++ kvm/arch/x86/kvm/lapic.c
> @@ -1087,11 +1089,64 @@ static void apic_timer_expired(struct kv
[...]
> +/*
> + * On APICv, this test will cause a busy wait
> + * during a higher-priority task.
> + */

(A bit confusing ... this test doesn't busy wait.)

> +
> +static bool lapic_timer_int_injected(struct kvm_vcpu *vcpu)
[...]
> +void wait_lapic_expire(struct kvm_vcpu *vcpu)
> +{
[...]
> +	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
> +	apic->lapic_timer.expired_tscdeadline = 0;
> +	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> +
> +	while (guest_tsc < tsc_deadline) {
> +		int delay = min(tsc_deadline - guest_tsc, 1000ULL);

Why break the __delay() loop into smaller parts?

> +
> +		__delay(delay);

(Does not have to call delay_tsc, but I guess it won't change.)

> +		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> +	}
>  }
>  

Btw. simple automatic delta tuning had worse results?

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Marcelo Tosatti]

On Wed, Dec 17, 2014 at 03:58:13PM +0100, Radim Krcmar wrote:
> 2014-12-16 09:08-0500, Marcelo Tosatti:
> > For the hrtimer which emulates the tscdeadline timer in the guest,
> > add an option to advance expiration, and busy spin on VM-entry waiting
> > for the actual expiration time to elapse.
> > 
> > This allows achieving low latencies in cyclictest (or any scenario 
> > which requires strict timing regarding timer expiration).
> > 
> > Reduces average cyclictest latency from 12us to 8us
> > on Core i5 desktop.
> > 
> > Note: this option requires tuning to find the appropriate value 
> > for a particular hardware/guest combination. One method is to measure the 
> > average delay between apic_timer_fn and VM-entry. 
> > Another method is to start with 1000ns, and increase the value
> > in say 500ns increments until avg cyclictest numbers stop decreasing.
> > 
> > Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
> 
> Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>
> 
> > +++ kvm/arch/x86/kvm/lapic.c
> > @@ -1087,11 +1089,64 @@ static void apic_timer_expired(struct kv
> [...]
> > +/*
> > + * On APICv, this test will cause a busy wait
> > + * during a higher-priority task.
> > + */
> 
> (A bit confusing ... this test doesn't busy wait.)
> 
> > +
> > +static bool lapic_timer_int_injected(struct kvm_vcpu *vcpu)
> [...]
> > +void wait_lapic_expire(struct kvm_vcpu *vcpu)
> > +{
> [...]
> > +	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
> > +	apic->lapic_timer.expired_tscdeadline = 0;
> > +	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > +
> > +	while (guest_tsc < tsc_deadline) {
> > +		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
> 
> Why break the __delay() loop into smaller parts?

So that you can handle interrupts, in case this code ever moves
outside IRQ protected region.

> > +		__delay(delay);
> 
> (Does not have to call delay_tsc, but I guess it won't change.)
> 
> > +		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > +	}
> >  }
> >  
> 
> Btw. simple automatic delta tuning had worse results?

Haven't tried automatic tuning.

So what happens on a realtime environment is this: you execute the fixed
number of instructions from interrupt handling all the way to VM-entry.

Well, almost fixed. Fixed is the number of apic_timer_fn plus KVM
instructions. You can also execute host scheduler and timekeeping 
processing.

In practice, the length to execute that instruction sequence is a bell
shaped normal distribution around the average (the right side is
slightly higher due to host scheduler and timekeeping processing).

You want to advance the timer by the rightmost bucket, that way you
guarantee lower possible latencies (which is the interest here).

That said, i don't see advantage in automatic tuning for the usecase 
which this targets.

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Paolo Bonzini]

On 17/12/2014 18:41, Marcelo Tosatti wrote:
>> > Btw. simple automatic delta tuning had worse results?
> Haven't tried automatic tuning.
> 
> So what happens on a realtime environment is this: you execute the fixed
> number of instructions from interrupt handling all the way to VM-entry.
> 
> Well, almost fixed. Fixed is the number of apic_timer_fn plus KVM
> instructions. You can also execute host scheduler and timekeeping 
> processing.
> 
> In practice, the length to execute that instruction sequence is a bell
> shaped normal distribution around the average (the right side is
> slightly higher due to host scheduler and timekeeping processing).
> 
> You want to advance the timer by the rightmost bucket, that way you
> guarantee lower possible latencies (which is the interest here).
> 
> That said, i don't see advantage in automatic tuning for the usecase 
> which this targets.

Yeah, the value of the parameter can be computed pretty easily from
either the tracepoint or the kvm-unit-tests testcase.

Paolo

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Radim Krcmar]

2014-12-17 15:41-0200, Marcelo Tosatti:
> On Wed, Dec 17, 2014 at 03:58:13PM +0100, Radim Krcmar wrote:
> > 2014-12-16 09:08-0500, Marcelo Tosatti:
> > > +	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
> > > +	apic->lapic_timer.expired_tscdeadline = 0;
> > > +	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > > +
> > > +	while (guest_tsc < tsc_deadline) {
> > > +		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
> > 
> > Why break the __delay() loop into smaller parts?
> 
> So that you can handle interrupts, in case this code ever moves
> outside IRQ protected region.

__delay() works only if it is delay_tsc(), which has this handled ...
(It even considers rescheduling with unsynchronized TSC.)

delay_tsc(delay) translates roughly to

  end = read_tsc() + delay;
  while (read_tsc() < end);

so the code of our while loop has a structure like

  while ((guest_tsc = read_tsc()) < tsc_deadline) {
    end = read_tsc() + min(tsc_deadline - guest_tsc, 1000);
    while (read_tsc() < end);
  }

which complicates our original idea of

  while (read_tsc() < tsc_deadline);

(but I'm completely fine with it.)

> > > +		__delay(delay);
> > 
> > (Does not have to call delay_tsc, but I guess it won't change.)
> > 
> > > +		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > > +	}
> > >  }
> > >  
> > 
> > Btw. simple automatic delta tuning had worse results?
> 
> Haven't tried automatic tuning.
> 
> So what happens on a realtime environment is this: you execute the fixed
> number of instructions from interrupt handling all the way to VM-entry.
> 
> Well, almost fixed. Fixed is the number of apic_timer_fn plus KVM
> instructions. You can also execute host scheduler and timekeeping 
> processing.
> 
> In practice, the length to execute that instruction sequence is a bell
> shaped normal distribution around the average (the right side is
> slightly higher due to host scheduler and timekeeping processing).
> 
> You want to advance the timer by the rightmost bucket, that way you
> guarantee lower possible latencies (which is the interest here).

(Lower latencies would likely be achieved by having a timer that issues
 posted interrupts from another CPU, and the guest set to busy idle.)

> That said, i don't see advantage in automatic tuning for the usecase 
> which this targets.

Thanks, it doesn't make much difference in the long RT setup checklist.


---
I was asking just because I consider programming to equal automation ...
If we know that we will always set this to the rightmost bucket anyway,
it could be done like this

  if ((s64)(delta = guest_tsc - tsc_deadline) > 0)
    tsc_deadline_delta += delta;
  ...
  advance_ns = kvm_tsc_to_ns(tsc_deadline_delta);

instead of a script that runs a test and sets the variable.
(On the other hand, it would probably have to be more complicated to
 reach the same level of flexibility.)

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Marcelo Tosatti]

On Wed, Dec 17, 2014 at 08:36:27PM +0100, Radim Krcmar wrote:
> 2014-12-17 15:41-0200, Marcelo Tosatti:
> > On Wed, Dec 17, 2014 at 03:58:13PM +0100, Radim Krcmar wrote:
> > > 2014-12-16 09:08-0500, Marcelo Tosatti:
> > > > +	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
> > > > +	apic->lapic_timer.expired_tscdeadline = 0;
> > > > +	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > > > +
> > > > +	while (guest_tsc < tsc_deadline) {
> > > > +		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
> > > 
> > > Why break the __delay() loop into smaller parts?
> > 
> > So that you can handle interrupts, in case this code ever moves
> > outside IRQ protected region.
> 
> __delay() works only if it is delay_tsc(), which has this handled ...
> (It even considers rescheduling with unsynchronized TSC.)
> 
> delay_tsc(delay) translates roughly to
> 
>   end = read_tsc() + delay;
>   while (read_tsc() < end);
> 
> so the code of our while loop has a structure like
> 
>   while ((guest_tsc = read_tsc()) < tsc_deadline) {
>     end = read_tsc() + min(tsc_deadline - guest_tsc, 1000);
>     while (read_tsc() < end);
>   }
> 
> which complicates our original idea of
> 
>   while (read_tsc() < tsc_deadline);
> 
> (but I'm completely fine with it.)

True. I can change to a direct wait if that is preferred.

> > > > +		__delay(delay);
> > > 
> > > (Does not have to call delay_tsc, but I guess it won't change.)
> > > 
> > > > +		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> > > > +	}
> > > >  }
> > > >  
> > > 
> > > Btw. simple automatic delta tuning had worse results?
> > 
> > Haven't tried automatic tuning.
> > 
> > So what happens on a realtime environment is this: you execute the fixed
> > number of instructions from interrupt handling all the way to VM-entry.
> > 
> > Well, almost fixed. Fixed is the number of apic_timer_fn plus KVM
> > instructions. You can also execute host scheduler and timekeeping 
> > processing.
> > 
> > In practice, the length to execute that instruction sequence is a bell
> > shaped normal distribution around the average (the right side is
> > slightly higher due to host scheduler and timekeeping processing).
> > 
> > You want to advance the timer by the rightmost bucket, that way you
> > guarantee lower possible latencies (which is the interest here).
> 
> (Lower latencies would likely be achieved by having a timer that issues
>  posted interrupts from another CPU, and the guest set to busy idle.)

Yes.

> > That said, i don't see advantage in automatic tuning for the usecase 
> > which this targets.
> 
> Thanks, it doesn't make much difference in the long RT setup checklist.

Exactly.

> ---
> I was asking just because I consider programming to equal automation ...
> If we know that we will always set this to the rightmost bucket anyway,
> it could be done like this
> 
>   if ((s64)(delta = guest_tsc - tsc_deadline) > 0)
>     tsc_deadline_delta += delta;
>   ...
>   advance_ns = kvm_tsc_to_ns(tsc_deadline_delta);
> 
> instead of a script that runs a test and sets the variable.
> (On the other hand, it would probably have to be more complicated to
>  reach the same level of flexibility.)

You'd have to guarantee the vcpus are never interrupted by other work,
such as processing host interrupts, otherwise you could get high
increments for tsc_deadline_delta.

So to tune that value you do:

1) Boot guest.
2) Setup certain vCPUs as realtime (large checklist), which includes
pinning and host interrupt routing.
3) Measure with cyclictest on those vCPUs with the realtime conditions.

So its also a matter of configuration.

But yes the code above would set advance_ns to the rightmost bucket.

Re: [patch 2/3] KVM: x86: add option to advance tscdeadline hrtimer expiration

[Paolo Bonzini]

On 18/12/2014 13:24, Marcelo Tosatti wrote:
> True. I can change to a direct wait if that is preferred.

What about this:

 	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
 	trace_kvm_wait_lapic_expire(vcpu->vcpu_id, guest_tsc - tsc_deadline);
 
	/* We know that __delay is delay_tsc, see kvm_timer_init.  */
 	if (guest_tsc < tsc_deadline)
 		__delay(tsc_deadline - guest_tsc);

and a check in kvm_timer_init:

	if (check_tsc_unstable() && lapic_timer_advance_ns) {
		pr_info("kvm: unstable TSC, disabling lapic_timer_advance_ns\n");
		lapic_timer_advance_ns = 0;
	}

Paolo

[patch 3/3] KVM: x86: add tracepoint to wait_lapic_expire

[Marcelo Tosatti]

[-- Attachment #1: lapic-wait-tracepoint --]
[-- Type: text/plain, Size: 1918 bytes --]

Add tracepoint to wait_lapic_expire.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Index: kvm/arch/x86/kvm/lapic.c
===================================================================
--- kvm.orig/arch/x86/kvm/lapic.c
+++ kvm/arch/x86/kvm/lapic.c
@@ -1120,7 +1120,7 @@ static bool lapic_timer_int_injected(str
 void wait_lapic_expire(struct kvm_vcpu *vcpu)
 {
 	struct kvm_lapic *apic = vcpu->arch.apic;
-	u64 guest_tsc, tsc_deadline;
+	u64 orig_guest_tsc, guest_tsc, tsc_deadline;
 
 	if (!kvm_vcpu_has_lapic(vcpu))
 		return;
@@ -1133,7 +1133,7 @@ void wait_lapic_expire(struct kvm_vcpu *
 
 	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
 	apic->lapic_timer.expired_tscdeadline = 0;
-	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
+	orig_guest_tsc = guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
 
 	while (guest_tsc < tsc_deadline) {
 		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
@@ -1141,6 +1141,8 @@ void wait_lapic_expire(struct kvm_vcpu *
 		__delay(delay);
 		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
 	}
+
+	trace_kvm_wait_lapic_expire(vcpu->vcpu_id, orig_guest_tsc - tsc_deadline);
 }
 
 static void start_apic_timer(struct kvm_lapic *apic)
Index: kvm/arch/x86/kvm/trace.h
===================================================================
--- kvm.orig/arch/x86/kvm/trace.h
+++ kvm/arch/x86/kvm/trace.h
@@ -914,6 +914,25 @@ TRACE_EVENT(kvm_pvclock_update,
 		  __entry->flags)
 );
 
+TRACE_EVENT(kvm_wait_lapic_expire,
+	TP_PROTO(unsigned int vcpu_id, s64 delta),
+	TP_ARGS(vcpu_id, delta),
+
+	TP_STRUCT__entry(
+		__field(	unsigned int,	vcpu_id		)
+		__field(	s64,		delta		)
+	),
+
+	TP_fast_assign(
+		__entry->vcpu_id	   = vcpu_id;
+		__entry->delta             = delta;
+	),
+
+	TP_printk("vcpu %u: delta %lld",
+		  __entry->vcpu_id,
+		  __entry->delta)
+);
+
 #endif /* _TRACE_KVM_H */
 
 #undef TRACE_INCLUDE_PATH

Re: [patch 3/3] KVM: x86: add tracepoint to wait_lapic_expire

[Radim Krcmar]

2014-12-16 09:08-0500, Marcelo Tosatti:
> Add tracepoint to wait_lapic_expire.
> 
> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>

> +++ kvm/arch/x86/kvm/lapic.c
> @@ -1120,7 +1120,7 @@ static bool lapic_timer_int_injected(str
>  void wait_lapic_expire(struct kvm_vcpu *vcpu)
>  {
>  	struct kvm_lapic *apic = vcpu->arch.apic;
> -	u64 guest_tsc, tsc_deadline;
> +	u64 orig_guest_tsc, guest_tsc, tsc_deadline;
>  
>  	if (!kvm_vcpu_has_lapic(vcpu))
>  		return;
> @@ -1133,7 +1133,7 @@ void wait_lapic_expire(struct kvm_vcpu *
>  
>  	tsc_deadline = apic->lapic_timer.expired_tscdeadline;
>  	apic->lapic_timer.expired_tscdeadline = 0;
> -	guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
> +	orig_guest_tsc = guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());

(We could move the tracepoint here and save orig_guest_tsc.)

>  	while (guest_tsc < tsc_deadline) {
>  		int delay = min(tsc_deadline - guest_tsc, 1000ULL);
> @@ -1141,6 +1141,8 @@ void wait_lapic_expire(struct kvm_vcpu *
>  		__delay(delay);
>  		guest_tsc = kvm_x86_ops->read_l1_tsc(vcpu, native_read_tsc());
>  	}
> +
> +	trace_kvm_wait_lapic_expire(vcpu->vcpu_id, orig_guest_tsc - tsc_deadline);
>  }
>  

[patch 0/3] KVM: add option to advance tscdeadline hrtimer expiration (v6)

[Marcelo Tosatti]

See patches for details.

v2:
- fix email address.

v3:
- use module parameter for configuration of value (Paolo/Radim)

v4:
- fix check for tscdeadline mode while waiting for expiration (Paolo)
- use proper delay function (Radim)
- fix LVTT tscdeadline mode check in hrtimer interrupt handler (Radim)
- add comment regarding PPR and APICv (Paolo)

v5:
- use tscdeadline expiration and guest tsc difference in
  tracepoint (Paolo)

v6:
- Remove unnecessary delay loop (Radim)
- Disable feature if __delay is not TSC based (Paolo)

[patch 1/3] KVM: x86: add method to test PIR bitmap vector

[Marcelo Tosatti]

[-- Attachment #1: add-test-pir --]
[-- Type: text/plain, Size: 2342 bytes --]

kvm_x86_ops->test_posted_interrupt() returns true/false depending
whether 'vector' is set.

Next patch makes use of this interface.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/arch/x86/include/asm/kvm_host.h
@@ -743,6 +743,7 @@ struct kvm_x86_ops {
 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
+	bool (*test_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
 	int (*get_tdp_level)(void);
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -435,6 +435,11 @@ static int pi_test_and_set_pir(int vecto
 	return test_and_set_bit(vector, (unsigned long *)pi_desc->pir);
 }
 
+static int pi_test_pir(int vector, struct pi_desc *pi_desc)
+{
+	return test_bit(vector, (unsigned long *)pi_desc->pir);
+}
+
 struct vcpu_vmx {
 	struct kvm_vcpu       vcpu;
 	unsigned long         host_rsp;
@@ -5939,6 +5944,7 @@ static __init int hardware_setup(void)
 	else {
 		kvm_x86_ops->hwapic_irr_update = NULL;
 		kvm_x86_ops->deliver_posted_interrupt = NULL;
+		kvm_x86_ops->test_posted_interrupt = NULL;
 		kvm_x86_ops->sync_pir_to_irr = vmx_sync_pir_to_irr_dummy;
 	}
 
@@ -6960,6 +6966,13 @@ static int handle_invvpid(struct kvm_vcp
 	return 1;
 }
 
+static bool vmx_test_pir(struct kvm_vcpu *vcpu, int vector)
+{
+	struct vcpu_vmx *vmx = to_vmx(vcpu);
+
+	return pi_test_pir(vector, &vmx->pi_desc);
+}
+
 /*
  * 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
@@ -9374,6 +9387,7 @@ static struct kvm_x86_ops vmx_x86_ops =
 	.hwapic_isr_update = vmx_hwapic_isr_update,
 	.sync_pir_to_irr = vmx_sync_pir_to_irr,
 	.deliver_posted_interrupt = vmx_deliver_posted_interrupt,
+	.test_posted_interrupt = vmx_test_pir,
 
 	.set_tss_addr = vmx_set_tss_addr,
 	.get_tdp_level = get_ept_level,

[patch 0/3] KVM: add option to advance tscdeadline hrtimer expiration (v4)

[Marcelo Tosatti]

See patches for details.

v2:
- fix email address.

v3:
- use module parameter for configuration of value (Paolo/Radim)

v4:
- fix check for tscdeadline mode while waiting for expiration (Paolo)
- use proper delay function (Radim)
- fix LVTT tscdeadline mode check in hrtimer interrupt handler (Radim)
- add comment regarding PPR and APICv (Paolo)

[patch 1/3] KVM: x86: add method to test PIR bitmap vector

[Marcelo Tosatti]

[-- Attachment #1: add-test-pir --]
[-- Type: text/plain, Size: 2342 bytes --]

kvm_x86_ops->test_posted_interrupt() returns true/false depending
whether 'vector' is set.

Next patch makes use of this interface.

Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>

Index: kvm/arch/x86/include/asm/kvm_host.h
===================================================================
--- kvm.orig/arch/x86/include/asm/kvm_host.h
+++ kvm/arch/x86/include/asm/kvm_host.h
@@ -743,6 +743,7 @@ struct kvm_x86_ops {
 	void (*set_virtual_x2apic_mode)(struct kvm_vcpu *vcpu, bool set);
 	void (*set_apic_access_page_addr)(struct kvm_vcpu *vcpu, hpa_t hpa);
 	void (*deliver_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
+	bool (*test_posted_interrupt)(struct kvm_vcpu *vcpu, int vector);
 	void (*sync_pir_to_irr)(struct kvm_vcpu *vcpu);
 	int (*set_tss_addr)(struct kvm *kvm, unsigned int addr);
 	int (*get_tdp_level)(void);
Index: kvm/arch/x86/kvm/vmx.c
===================================================================
--- kvm.orig/arch/x86/kvm/vmx.c
+++ kvm/arch/x86/kvm/vmx.c
@@ -435,6 +435,11 @@ static int pi_test_and_set_pir(int vecto
 	return test_and_set_bit(vector, (unsigned long *)pi_desc->pir);
 }
 
+static int pi_test_pir(int vector, struct pi_desc *pi_desc)
+{
+	return test_bit(vector, (unsigned long *)pi_desc->pir);
+}
+
 struct vcpu_vmx {
 	struct kvm_vcpu       vcpu;
 	unsigned long         host_rsp;
@@ -5939,6 +5944,7 @@ static __init int hardware_setup(void)
 	else {
 		kvm_x86_ops->hwapic_irr_update = NULL;
 		kvm_x86_ops->deliver_posted_interrupt = NULL;
+		kvm_x86_ops->test_posted_interrupt = NULL;
 		kvm_x86_ops->sync_pir_to_irr = vmx_sync_pir_to_irr_dummy;
 	}
 
@@ -6960,6 +6966,13 @@ static int handle_invvpid(struct kvm_vcp
 	return 1;
 }
 
+static bool vmx_test_pir(struct kvm_vcpu *vcpu, int vector)
+{
+	struct vcpu_vmx *vmx = to_vmx(vcpu);
+
+	return pi_test_pir(vector, &vmx->pi_desc);
+}
+
 /*
  * 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
@@ -9374,6 +9387,7 @@ static struct kvm_x86_ops vmx_x86_ops =
 	.hwapic_isr_update = vmx_hwapic_isr_update,
 	.sync_pir_to_irr = vmx_sync_pir_to_irr,
 	.deliver_posted_interrupt = vmx_deliver_posted_interrupt,
+	.test_posted_interrupt = vmx_test_pir,
 
 	.set_tss_addr = vmx_set_tss_addr,
 	.get_tdp_level = get_ept_level,