Commit 108b249c authored by Paolo Bonzini's avatar Paolo Bonzini
Browse files

KVM: x86: introduce get_kvmclock_ns 



Introduce a function that reads the exact nanoseconds value that is
provided to the guest in kvmclock.  This crystallizes the notion of
kvmclock as a thin veneer over a stable TSC, that the guest will
(hopefully) convert with NTP.  In other words, kvmclock is *not* a
paravirtualized host-to-guest NTP.

Drop the get_kernel_ns() function, that was used both to get the base
value of the master clock and to get the current value of kvmclock.
The former use is replaced by ktime_get_boot_ns(), the latter is
the purpose of get_kernel_ns().

This also allows KVM to provide a Hyper-V time reference counter that
is synchronized with the time that is computed from the TSC page.

Reviewed-by: default avatarRoman Kagan <rkagan@virtuozzo.com>
Signed-off-by: default avatarPaolo Bonzini <pbonzini@redhat.com>
parent 67198ac3

arch/x86/entry/vdso/vclock_gettime.c

+1 −1
Original line number Original line Diff line number Diff line
@@ -129,7 +129,7 @@ static notrace cycle_t vread_pvclock(int *mode) 
			return 0;

			return 0;

		}

		}





		ret = __pvclock_read_cycles(pvti);

		ret = __pvclock_read_cycles(pvti, rdtsc_ordered());

	} while (pvclock_read_retry(pvti, version));

	} while (pvclock_read_retry(pvti, version));





	/* refer to vread_tsc() comment for rationale */

	/* refer to vread_tsc() comment for rationale */

arch/x86/include/asm/pvclock.h

+3 −2
Original line number Original line Diff line number Diff line
@@ -87,9 +87,10 @@ static inline u64 pvclock_scale_delta(u64 delta, u32 mul_frac, int shift) 
}

}





static __always_inline

static __always_inline

cycle_t __pvclock_read_cycles(const struct pvclock_vcpu_time_info *src)

cycle_t __pvclock_read_cycles(const struct pvclock_vcpu_time_info *src,

			      u64 tsc)

{

{

	u64 delta = rdtsc_ordered() - src->tsc_timestamp;

	u64 delta = tsc - src->tsc_timestamp;

	cycle_t offset = pvclock_scale_delta(delta, src->tsc_to_system_mul,

	cycle_t offset = pvclock_scale_delta(delta, src->tsc_to_system_mul,

					     src->tsc_shift);

					     src->tsc_shift);

	return src->system_time + offset;

	return src->system_time + offset;

arch/x86/kernel/pvclock.c

+1 −1
Original line number Original line Diff line number Diff line
@@ -80,7 +80,7 @@ cycle_t pvclock_clocksource_read(struct pvclock_vcpu_time_info *src) 




	do {

	do {

		version = pvclock_read_begin(src);

		version = pvclock_read_begin(src);

		ret = __pvclock_read_cycles(src);

		ret = __pvclock_read_cycles(src, rdtsc_ordered());

		flags = src->flags;

		flags = src->flags;

	} while (pvclock_read_retry(src, version));

	} while (pvclock_read_retry(src, version));

arch/x86/kvm/hyperv.c

+1 −1
Original line number Original line Diff line number Diff line
@@ -386,7 +386,7 @@ static void synic_init(struct kvm_vcpu_hv_synic *synic) 




static u64 get_time_ref_counter(struct kvm *kvm)

static u64 get_time_ref_counter(struct kvm *kvm)

{

{

	return div_u64(get_kernel_ns() + kvm->arch.kvmclock_offset, 100);

	return div_u64(get_kvmclock_ns(kvm), 100);

}

}





static void stimer_mark_pending(struct kvm_vcpu_hv_stimer *stimer,

static void stimer_mark_pending(struct kvm_vcpu_hv_stimer *stimer,

arch/x86/kvm/x86.c

+36 −12
Original line number Original line Diff line number Diff line
@@ -1431,7 +1431,7 @@ void kvm_write_tsc(struct kvm_vcpu *vcpu, struct msr_data *msr) 




	raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags);

	raw_spin_lock_irqsave(&kvm->arch.tsc_write_lock, flags);

	offset = kvm_compute_tsc_offset(vcpu, data);

	offset = kvm_compute_tsc_offset(vcpu, data);

	ns = get_kernel_ns();

	ns = ktime_get_boot_ns();

	elapsed = ns - kvm->arch.last_tsc_nsec;

	elapsed = ns - kvm->arch.last_tsc_nsec;





	if (vcpu->arch.virtual_tsc_khz) {

	if (vcpu->arch.virtual_tsc_khz) {
@@ -1722,6 +1722,34 @@ static void kvm_gen_update_masterclock(struct kvm *kvm) 
#endif

#endif

}

}





static u64 __get_kvmclock_ns(struct kvm *kvm)

{

	struct kvm_vcpu *vcpu = kvm_get_vcpu(kvm, 0);

	struct kvm_arch *ka = &kvm->arch;

	s64 ns;



	if (vcpu->arch.hv_clock.flags & PVCLOCK_TSC_STABLE_BIT) {

		u64 tsc = kvm_read_l1_tsc(vcpu, rdtsc());

		ns = __pvclock_read_cycles(&vcpu->arch.hv_clock, tsc);

	} else {

		ns = ktime_get_boot_ns() + ka->kvmclock_offset;

	}



	return ns;

}



u64 get_kvmclock_ns(struct kvm *kvm)

{

	unsigned long flags;

	s64 ns;



	local_irq_save(flags);

	ns = __get_kvmclock_ns(kvm);

	local_irq_restore(flags);



	return ns;

}



static void kvm_setup_pvclock_page(struct kvm_vcpu *v)

static void kvm_setup_pvclock_page(struct kvm_vcpu *v)

{

{

	struct kvm_vcpu_arch *vcpu = &v->arch;

	struct kvm_vcpu_arch *vcpu = &v->arch;
@@ -1811,7 +1839,7 @@ static int kvm_guest_time_update(struct kvm_vcpu *v) 
	}

	}

	if (!use_master_clock) {

	if (!use_master_clock) {

		host_tsc = rdtsc();

		host_tsc = rdtsc();

		kernel_ns = get_kernel_ns();

		kernel_ns = ktime_get_boot_ns();

	}

	}





	tsc_timestamp = kvm_read_l1_tsc(v, host_tsc);

	tsc_timestamp = kvm_read_l1_tsc(v, host_tsc);
@@ -4054,7 +4082,6 @@ long kvm_arch_vm_ioctl(struct file *filp, 
	case KVM_SET_CLOCK: {

	case KVM_SET_CLOCK: {

		struct kvm_clock_data user_ns;

		struct kvm_clock_data user_ns;

		u64 now_ns;

		u64 now_ns;

		s64 delta;





		r = -EFAULT;

		r = -EFAULT;

		if (copy_from_user(&user_ns, argp, sizeof(user_ns)))

		if (copy_from_user(&user_ns, argp, sizeof(user_ns)))
@@ -4066,10 +4093,9 @@ long kvm_arch_vm_ioctl(struct file *filp, 




		r = 0;

		r = 0;

		local_irq_disable();

		local_irq_disable();

		now_ns = get_kernel_ns();

		now_ns = __get_kvmclock_ns(kvm);

		delta = user_ns.clock - now_ns;

		kvm->arch.kvmclock_offset += user_ns.clock - now_ns;

		local_irq_enable();

		local_irq_enable();

		kvm->arch.kvmclock_offset = delta;

		kvm_gen_update_masterclock(kvm);

		kvm_gen_update_masterclock(kvm);

		break;

		break;

	}

	}
@@ -4077,10 +4103,8 @@ long kvm_arch_vm_ioctl(struct file *filp, 
		struct kvm_clock_data user_ns;

		struct kvm_clock_data user_ns;

		u64 now_ns;

		u64 now_ns;





		local_irq_disable();

		now_ns = get_kvmclock_ns(kvm);

		now_ns = get_kernel_ns();

		user_ns.clock = now_ns;

		user_ns.clock = kvm->arch.kvmclock_offset + now_ns;

		local_irq_enable();

		user_ns.flags = 0;

		user_ns.flags = 0;

		memset(&user_ns.pad, 0, sizeof(user_ns.pad));

		memset(&user_ns.pad, 0, sizeof(user_ns.pad));
@@ -7544,7 +7568,7 @@ int kvm_arch_hardware_enable(void) 
	 * before any KVM threads can be running.  Unfortunately, we can't

	 * before any KVM threads can be running.  Unfortunately, we can't

	 * bring the TSCs fully up to date with real time, as we aren't yet far

	 * bring the TSCs fully up to date with real time, as we aren't yet far

	 * enough into CPU bringup that we know how much real time has actually

	 * enough into CPU bringup that we know how much real time has actually

	 * elapsed; our helper function, get_kernel_ns() will be using boot

	 * elapsed; our helper function, ktime_get_boot_ns() will be using boot

	 * variables that haven't been updated yet.

	 * variables that haven't been updated yet.

	 *

	 *

	 * So we simply find the maximum observed TSC above, then record the

	 * So we simply find the maximum observed TSC above, then record the
@@ -7779,7 +7803,7 @@ int kvm_arch_init_vm(struct kvm *kvm, unsigned long type) 
	mutex_init(&kvm->arch.apic_map_lock);

	mutex_init(&kvm->arch.apic_map_lock);

	spin_lock_init(&kvm->arch.pvclock_gtod_sync_lock);

	spin_lock_init(&kvm->arch.pvclock_gtod_sync_lock);





	kvm->arch.kvmclock_offset = -get_kernel_ns();

	kvm->arch.kvmclock_offset = -ktime_get_boot_ns();

	pvclock_update_vm_gtod_copy(kvm);

	pvclock_update_vm_gtod_copy(kvm);





	INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn);

	INIT_DELAYED_WORK(&kvm->arch.kvmclock_update_work, kvmclock_update_fn);

CRA Git | Maintained and supported by SUSTech CRA and CCSE