Re: [kvm-devel] [PATCH 0/2][KVM] guest time accounting

[Laurent Vivier]

Avi Kivity wrote:
> Laurent Vivier wrote:
>>> - perhaps the new fields should be guarded by a #ifdef CONFIG_HYPERVISOR
>>> (selected by CONFIG_KVM)?  that way the (minor) additional overhead is
>>> only incurred if it can possibly be used.  I imagine that our canine
>>> cousin will want to use this as well.
>>>     
>>
>> There is also a CONFIG_VIRTUALIZATION and a CONFIG_VIRT_CPU_ACCOUNTING
>> (from
>> s390 and powerpc) Which one to use ?
>>   
> 
> Are these options for using the kernel as a guest or host?  I'd guess
> the former.

I didn't find CONFIG_HYPERVISOR.
The good one seems to be CONFIG_VIRTUALIZATION that is used to activate CONFIG_KVM.

>> I'm wondering if we can have a more accurate accounting:
>>
>> - For the moment we add all system time since the previous entering to
>> the VCPU
>> to the guest time (and I guess there is some real system time in it ???)
>>
>> - Perhaps we can sum nanoseconds spent in the VCPU and add it to
>> cpustat when
>> these ns are greater than 1 ms ? (I'm trying to make something in this
>> way)
>>   

Ingo (or other guru), could you have a look to the attached patch, it is what I
was thinking about when I wrote this.

Laurent
-- 
------------- Laurent.Vivier@xxxxxxxx  --------------
          "Software is hard" - Donald Knuth
Index: kvm/drivers/kvm/vmx.c
===================================================================
--- kvm.orig/drivers/kvm/vmx.c	2007-08-13 14:11:38.000000000 +0200
+++ kvm/drivers/kvm/vmx.c	2007-08-13 14:29:44.000000000 +0200
@@ -2052,6 +2052,7 @@
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
 	u8 fail;
 	int r;
+	ktime_t now, delta;
 
 preempted:
 	if (vcpu->guest_debug.enabled)
@@ -2078,6 +2079,7 @@
 	local_irq_disable();
 
 	vcpu->guest_mode = 1;
+	now = ktime_get();
 	if (vcpu->requests)
 		if (test_and_clear_bit(KVM_TLB_FLUSH, &vcpu->requests))
 		    vmx_flush_tlb(vcpu);
@@ -2198,6 +2200,8 @@
 		[cr2]"i"(offsetof(struct kvm_vcpu, cr2))
 	      : "cc", "memory" );
 
+	delta = ktime_sub(ktime_get(), now);
+	current->vtime = ktime_add(current->vtime, delta);
 	vcpu->guest_mode = 0;
 	local_irq_enable();
 
Index: kvm/include/linux/sched.h
===================================================================
--- kvm.orig/include/linux/sched.h	2007-08-13 14:25:58.000000000 +0200
+++ kvm/include/linux/sched.h	2007-08-13 14:29:44.000000000 +0200
@@ -1192,6 +1192,9 @@
 #ifdef CONFIG_FAULT_INJECTION
 	int make_it_fail;
 #endif
+#ifdef CONFIG_VIRTUALIZATION
+	ktime_t vtime;
+#endif
 };
 
 /*
Index: kvm/kernel/sched.c
===================================================================
--- kvm.orig/kernel/sched.c	2007-08-13 14:11:38.000000000 +0200
+++ kvm/kernel/sched.c	2007-08-13 14:34:47.000000000 +0200
@@ -3212,6 +3212,29 @@
 	return ns;
 }
 
+#ifdef CONFIG_VIRTUALIZATION
+static cputime_t account_guest_time(struct task_struct *p, cputime_t cputime)
+{
+	struct cpu_usage_stat *cpustat = &kstat_this_cpu.cpustat;
+	ktime_t kmsec = ktime_set(0, NSEC_PER_MSEC);
+	cputime_t cmsec = msecs_to_cputime(1);
+
+	while ((ktime_to_ns(p->vtime) >= NSEC_PER_MSEC) &&
+	       (cputime_to_msecs(cputime) >= 1)) {
+		p->vtime = ktime_sub(p->vtime, kmsec);
+		p->utime = cputime_add(p->utime, cmsec);
+
+		cputime = cputime_sub(cputime, cmsec);
+
+		cpustat->guest = cputime64_add(cpustat->guest,
+				cputime_to_cputime64(cmsec));
+		cpustat->user = cputime64_add(cpustat->user,
+				cputime_to_cputime64(cmsec));
+	}
+	return cputime;
+}
+#endif
+
 /*
  * Account user cpu time to a process.
  * @p: the process that the cpu time gets accounted to
@@ -3246,6 +3269,10 @@
 	struct rq *rq = this_rq();
 	cputime64_t tmp;
 
+#ifdef CONFIG_VIRTUALIZATION
+	cputime = account_guest_time(p, cputime);
+#endif
+
 	p->stime = cputime_add(p->stime, cputime);
 
 	/* Add system time to cpustat. */
Index: kvm/drivers/kvm/svm.c
===================================================================
--- kvm.orig/drivers/kvm/svm.c	2007-08-13 14:31:16.000000000 +0200
+++ kvm/drivers/kvm/svm.c	2007-08-13 14:33:10.000000000 +0200
@@ -1392,6 +1392,7 @@
 	u16 gs_selector;
 	u16 ldt_selector;
 	int r;
+	ktime_t now, delta;
 
 again:
 	r = kvm_mmu_reload(vcpu);
@@ -1404,6 +1405,7 @@
 	clgi();
 
 	vcpu->guest_mode = 1;
+	now = ktime_get();
 	if (vcpu->requests)
 		if (test_and_clear_bit(KVM_TLB_FLUSH, &vcpu->requests))
 		    svm_flush_tlb(vcpu);
@@ -1536,6 +1538,8 @@
 #endif
 		: "cc", "memory" );
 
+	delta = ktime_sub(ktime_get(), now);
+	current->vtime = ktime_add(current->vtime, delta);
 	vcpu->guest_mode = 0;
 
 	if (vcpu->fpu_active) {

Attachment: signature.asc
Description: OpenPGP digital signature