Re: [PATCH v3 1/2] KVM: LAPIC: Optimize timer latency consider world switch time

From: Radim KrÄmÃÅ
Date: Wed Jun 12 2019 - 15:32:07 EST


2019-06-12 21:22+0200, Radim KrÄmÃÅ:
> 2019-06-12 08:14-0700, Sean Christopherson:
> > On Wed, Jun 12, 2019 at 05:40:18PM +0800, Wanpeng Li wrote:
> > > diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> > > @@ -145,6 +145,12 @@ module_param(tsc_tolerance_ppm, uint, S_IRUGO | S_IWUSR);
> > > static int __read_mostly lapic_timer_advance_ns = -1;
> > > module_param(lapic_timer_advance_ns, int, S_IRUGO | S_IWUSR);
> > >
> > > +/*
> > > + * lapic timer vmentry advance (tscdeadline mode only) in nanoseconds.
> > > + */
> > > +u32 __read_mostly vmentry_advance_ns = 300;
> >
> > Enabling this by default makes me nervous, e.g. nothing guarantees that
> > future versions of KVM and/or CPUs will continue to have 300ns of overhead
> > between wait_lapic_expire() and VM-Enter.
> >
> > If we want it enabled by default so that it gets tested, the default
> > value should be extremely conservative, e.g. set the default to a small
> > percentage (25%?) of the latency of VM-Enter itself on modern CPUs,
> > VM-Enter latency being the min between VMLAUNCH and VMLOAD+VMRUN+VMSAVE.
>
> I share the sentiment. We definitely must not enter the guest before
> the deadline has expired and CPUs are approaching 5 GHz (in turbo), so
> 300 ns would be too much even today.
>
> I wrote a simple testcase for rough timing and there are 267 cycles
> (111 ns @ 2.4 GHz) between doing rdtsc() right after
> kvm_wait_lapic_expire() [1] and doing rdtsc() in the guest as soon as
> possible (see the attached kvm-unit-test).

I forgot to attach it, pasting here as a patch for kvm-unit-tests.

---
diff --git a/x86/Makefile.common b/x86/Makefile.common
index e612dbe..ceed648 100644
--- a/x86/Makefile.common
+++ b/x86/Makefile.common
@@ -58,7 +58,7 @@ tests-common = $(TEST_DIR)/vmexit.flat $(TEST_DIR)/tsc.flat \
$(TEST_DIR)/init.flat $(TEST_DIR)/smap.flat \
$(TEST_DIR)/hyperv_synic.flat $(TEST_DIR)/hyperv_stimer.flat \
$(TEST_DIR)/hyperv_connections.flat \
- $(TEST_DIR)/umip.flat
+ $(TEST_DIR)/umip.flat $(TEST_DIR)/vmentry_latency.flat

ifdef API
tests-api = api/api-sample api/dirty-log api/dirty-log-perf
diff --git a/x86/vmentry_latency.c b/x86/vmentry_latency.c
new file mode 100644
index 0000000..3859f09
--- /dev/null
+++ b/x86/vmentry_latency.c
@@ -0,0 +1,45 @@
+#include "x86/vm.h"
+
+static u64 get_last_hypervisor_tsc_delta(void)
+{
+ u64 a = 0, b, c, d;
+ u64 tsc;
+
+ /*
+ * The first vmcall is there to force a vm exit just before measuring.
+ */
+ asm volatile ("vmcall" : "+a"(a), "=b"(b), "=c"(c), "=d"(d));
+
+ tsc = rdtsc();
+
+ /*
+ * The second hypercall recovers the value that was stored when vm
+ * entering to execute the rdtsc()
+ */
+ a = 11;
+ asm volatile ("vmcall" : "+a"(a), "=b"(b), "=c"(c), "=d"(d));
+
+ return tsc - a;
+}
+
+static void vmentry_latency(void)
+{
+ unsigned i = 1000000;
+ u64 min = -1;
+
+ while (i--) {
+ u64 latency = get_last_hypervisor_tsc_delta();
+ if (latency < min)
+ min = latency;
+ }
+
+ printf("vm entry latency is %"PRIu64" TSC cycles\n", min);
+}
+
+int main(void)
+{
+ setup_vm();
+ vmentry_latency();
+
+ return 0;
+}