KVM: x86: add module parameter to disable periodic kvmclock sync

The periodic kvmclock sync can be an undesired source of latencies.

When running cyclictest on a guest, a latency spike is visible.
With kvmclock periodic sync disabled, the spike is gone.

Guests should use ntp which means the propagations of ntp corrections
from the host clock are unnecessary.

v2:
-> Make parameter read-only (Radim)
-> Return early on kvmclock_sync_fn (Andrew)

Reported-and-tested-by: Luiz Capitulino <lcapitulino@redhat.com>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index ba7b0cc52fedad6d2bc2fa594a432dfaa88d90b1..2211213a84e7a4184d297842ecc354b0760a6573 100644 (file)
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -99,6 +99,9 @@ module_param(ignore_msrs, bool, S_IRUGO | S_IWUSR);
 unsigned int min_timer_period_us = 500;
 module_param(min_timer_period_us, uint, S_IRUGO | S_IWUSR);
 
+static bool __read_mostly kvmclock_periodic_sync = true;
+module_param(kvmclock_periodic_sync, bool, S_IRUGO);
+
 bool kvm_has_tsc_control;
 EXPORT_SYMBOL_GPL(kvm_has_tsc_control);
 u32  kvm_max_guest_tsc_khz;
@@ -1767,6 +1770,9 @@ static void kvmclock_sync_fn(struct work_struct *work)
                                           kvmclock_sync_work);
        struct kvm *kvm = container_of(ka, struct kvm, arch);
 
+       if (!kvmclock_periodic_sync)
+               return;
+
        schedule_delayed_work(&kvm->arch.kvmclock_update_work, 0);
        schedule_delayed_work(&kvm->arch.kvmclock_sync_work,
                                        KVMCLOCK_SYNC_PERIOD);
@@ -7221,6 +7227,9 @@ void kvm_arch_vcpu_postcreate(struct kvm_vcpu *vcpu)
        kvm_write_tsc(vcpu, &msr);
        vcpu_put(vcpu);
 
+       if (!kvmclock_periodic_sync)
+               return;
+
        schedule_delayed_work(&kvm->arch.kvmclock_sync_work,
                                        KVMCLOCK_SYNC_PERIOD);
 }