[Devel] [PATCH RHEL7 COMMIT] ms/kvm: dynamic halt-polling

[Konstantin Khorenko]

The commit is pushed to "branch-rh7-3.10.0-327.10.1.vz7.12.x-ovz" and will appear at https://src.openvz.org/scm/ovz/vzkernel.git
after rh7-3.10.0-327.10.1.vz7.12.11
------>
commit 98050d9d5de51020fb5418531b4b1958d8c4fe6d
Author: Wanpeng Li <wanpeng.li at hotmail.com>
Date:   Tue Apr 19 14:22:48 2016 +0400

    ms/kvm: dynamic halt-polling
    
    There is a downside of always-poll since poll is still happened for idle
    vCPUs which can waste cpu usage. This patchset add the ability to adjust
    halt_poll_ns dynamically, to grow halt_poll_ns when shot halt is detected,
    and to shrink halt_poll_ns when long halt is detected.
    
    There are two new kernel parameters for changing the halt_poll_ns:
    halt_poll_ns_grow and halt_poll_ns_shrink.
    
                            no-poll      always-poll    dynamic-poll
    "Idle (X) vCPU %c0" is the percent of time the physical cpu spent in
    c0 over 60 seconds (each vCPU is pinned to a pCPU). (nohz) means the
    guest was tickless. (250HZ) means the guest was ticking at 250HZ.
    
    The big win is with ticking operating systems. Running the linux guest
    with nohz=off (and HZ=250), we save 3.4%~12.8% CPUs/second and get close
    to no-polling overhead levels by using the dynamic-poll. The savings
    should be even higher for higher frequency ticks.
    
    Suggested-by: David Matlack <dmatlack at google.com>
    Signed-off-by: Wanpeng Li <wanpeng.li at hotmail.com>
    [Simplify the patch. - Paolo]
    Signed-off-by: Paolo Bonzini <pbonzini at redhat.com>
    (cherry picked from commit aca6ff29c4063a8d467cdee241e6b3bf7dc4a171)
        #PSBM-43100:fixed
    
    Conflicts:
    	virt/kvm/kvm_main.c
    
    https://jira.sw.ru/browse/PSBM-43100
    
    Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov at virtuozzo.com>
---
 virt/kvm/kvm_main.c | 53 +++++++++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 51 insertions(+), 2 deletions(-)

diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index a320bf1..ea234fc 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -66,9 +66,18 @@
 MODULE_AUTHOR("Qumranet");
 MODULE_LICENSE("GPL");
 
-unsigned int halt_poll_ns = 0;
+/* halt polling only reduces halt latency by 5-7 us, 500us is enough */
+static unsigned int halt_poll_ns = 500000;
 module_param(halt_poll_ns, uint, S_IRUGO | S_IWUSR);
 
+/* Default doubles per-vcpu halt_poll_ns. */
+static unsigned int halt_poll_ns_grow = 2;
+module_param(halt_poll_ns_grow, int, S_IRUGO);
+
+/* Default resets per-vcpu halt_poll_ns . */
+static unsigned int halt_poll_ns_shrink;
+module_param(halt_poll_ns_shrink, int, S_IRUGO);
+
 /*
  * Ordering of locks:
  *
@@ -1934,6 +1943,31 @@ void kvm_vcpu_mark_page_dirty(struct kvm_vcpu *vcpu, gfn_t gfn)
 }
 EXPORT_SYMBOL_GPL(kvm_vcpu_mark_page_dirty);
 
+static void grow_halt_poll_ns(struct kvm_vcpu *vcpu)
+{
+	int val = vcpu->halt_poll_ns;
+
+	/* 10us base */
+	if (val == 0 && halt_poll_ns_grow)
+		val = 10000;
+	else
+		val *= halt_poll_ns_grow;
+
+	vcpu->halt_poll_ns = val;
+}
+
+static void shrink_halt_poll_ns(struct kvm_vcpu *vcpu)
+{
+	int val = vcpu->halt_poll_ns;
+
+	if (halt_poll_ns_shrink == 0)
+		val = 0;
+	else
+		val /= halt_poll_ns_shrink;
+
+	vcpu->halt_poll_ns = val;
+}
+
 static int kvm_vcpu_check_block(struct kvm_vcpu *vcpu)
 {
 	if (kvm_arch_vcpu_runnable(vcpu)) {
@@ -1956,6 +1990,7 @@ void kvm_vcpu_block(struct kvm_vcpu *vcpu)
 	ktime_t start, cur;
 	DEFINE_WAIT(wait);
 	bool waited = false;
+	u64 block_ns;
 
 	start = cur = ktime_get();
 	if (vcpu->halt_poll_ns) {
@@ -1988,7 +2023,21 @@ void kvm_vcpu_block(struct kvm_vcpu *vcpu)
 	cur = ktime_get();
 
 out:
-	trace_kvm_vcpu_wakeup(ktime_to_ns(cur) - ktime_to_ns(start), waited);
+	block_ns = ktime_to_ns(cur) - ktime_to_ns(start);
+
+	if (halt_poll_ns) {
+		if (block_ns <= vcpu->halt_poll_ns)
+			;
+		/* we had a long block, shrink polling */
+		else if (vcpu->halt_poll_ns && block_ns > halt_poll_ns)
+			shrink_halt_poll_ns(vcpu);
+		/* we had a short halt and our poll time is too small */
+		else if (vcpu->halt_poll_ns < halt_poll_ns &&
+			block_ns < halt_poll_ns)
+			grow_halt_poll_ns(vcpu);
+	}
+
+	trace_kvm_vcpu_wakeup(block_ns, waited);
 }
 EXPORT_SYMBOL_GPL(kvm_vcpu_block);