On Wed, Jun 15, 2011 at 12:09:31AM -0300, Glauber Costa wrote:
> On 06/14/2011 04:45 AM, Gleb Natapov wrote:
> >On Mon, Jun 13, 2011 at 07:31:33PM -0400, Glauber Costa wrote:
> >>To implement steal time, we need the hypervisor to pass the guest information
> >>about how much time was spent running other processes outside the VM.
> >>This is per-vcpu, and using the kvmclock structure for that is an abuse
> >>we decided not to make.
> >>
> >>In this patchset, I am introducing a new msr, KVM_MSR_STEAL_TIME, that
> >>holds the memory area address containing information about steal time
> >>
> >>This patch contains the hypervisor part for it. I am keeping it separate from
> >>the headers to facilitate backports to people who wants to backport the kernel
> >>part but not the hypervisor, or the other way around.
> >>
> >>Signed-off-by: Glauber Costa<glommer@xxxxxxxxxx>
> >>CC: Rik van Riel<riel@xxxxxxxxxx>
> >>CC: Jeremy Fitzhardinge<jeremy.fitzhardinge@xxxxxxxxxx>
> >>CC: Peter Zijlstra<peterz@xxxxxxxxxxxxx>
> >>CC: Avi Kivity<avi@xxxxxxxxxx>
> >>CC: Anthony Liguori<aliguori@xxxxxxxxxx>
> >>CC: Eric B Munson<emunson@xxxxxxxxx>
> >>---
> >> arch/x86/include/asm/kvm_host.h | 8 +++++
> >> arch/x86/include/asm/kvm_para.h | 4 ++
> >> arch/x86/kvm/x86.c | 60 +++++++++++++++++++++++++++++++++++++--
> >> 3 files changed, 69 insertions(+), 3 deletions(-)
> >>
> >>diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
> >>index fc38eca..5dce014 100644
> >>--- a/arch/x86/include/asm/kvm_host.h
> >>+++ b/arch/x86/include/asm/kvm_host.h
> >>@@ -388,6 +388,14 @@ struct kvm_vcpu_arch {
> >> unsigned int hw_tsc_khz;
> >> unsigned int time_offset;
> >> struct page *time_page;
> >>+
> >>+ struct {
> >>+ u64 msr_val;
> >>+ gpa_t stime;
> >>+ struct kvm_steal_time steal;
> >>+ u64 this_time_out;
> >>+ } st;
> >>+
> >> u64 last_guest_tsc;
> >> u64 last_kernel_ns;
> >> u64 last_tsc_nsec;
> >>diff --git a/arch/x86/include/asm/kvm_para.h b/arch/x86/include/asm/kvm_para.h
> >>index ac306c4..0341e61 100644
> >>--- a/arch/x86/include/asm/kvm_para.h
> >>+++ b/arch/x86/include/asm/kvm_para.h
> >>@@ -45,6 +45,10 @@ struct kvm_steal_time {
> >> __u32 pad[6];
> >> };
> >>
> >>+#define KVM_STEAL_ALIGNMENT_BITS 5
> >>+#define KVM_STEAL_VALID_BITS ((-1ULL<< (KVM_STEAL_ALIGNMENT_BITS + 1)))
> >>+#define KVM_STEAL_RESERVED_MASK (((1<< KVM_STEAL_ALIGNMENT_BITS) - 1 )<< 1)
> >>+
> >> #define KVM_MAX_MMU_OP_BATCH 32
> >>
> >> #define KVM_ASYNC_PF_ENABLED (1<< 0)
> >>diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> >>index 6645634..10fe028 100644
> >>--- a/arch/x86/kvm/x86.c
> >>+++ b/arch/x86/kvm/x86.c
> >>@@ -797,12 +797,12 @@ EXPORT_SYMBOL_GPL(kvm_get_dr);
> >> * kvm-specific. Those are put in the beginning of the list.
> >> */
> >>
> >>-#define KVM_SAVE_MSRS_BEGIN 8
> >>+#define KVM_SAVE_MSRS_BEGIN 9
> >> static u32 msrs_to_save[] = {
> >> MSR_KVM_SYSTEM_TIME, MSR_KVM_WALL_CLOCK,
> >> MSR_KVM_SYSTEM_TIME_NEW, MSR_KVM_WALL_CLOCK_NEW,
> >> HV_X64_MSR_GUEST_OS_ID, HV_X64_MSR_HYPERCALL,
> >>- HV_X64_MSR_APIC_ASSIST_PAGE, MSR_KVM_ASYNC_PF_EN,
> >>+ HV_X64_MSR_APIC_ASSIST_PAGE, MSR_KVM_ASYNC_PF_EN, MSR_KVM_STEAL_TIME,
> >> MSR_IA32_SYSENTER_CS, MSR_IA32_SYSENTER_ESP, MSR_IA32_SYSENTER_EIP,
> >> MSR_STAR,
> >> #ifdef CONFIG_X86_64
> >>@@ -1480,6 +1480,34 @@ static void kvmclock_reset(struct kvm_vcpu *vcpu)
> >> }
> >> }
> >>
> >>+static void record_steal_time(struct kvm_vcpu *vcpu)
> >>+{
> >>+ u64 delta;
> >>+
> >>+ if (vcpu->arch.st.stime&& vcpu->arch.st.this_time_out) {
> >>+
> >>+ if (unlikely(kvm_read_guest(vcpu->kvm, vcpu->arch.st.stime,
> >>+ &vcpu->arch.st.steal, sizeof(struct kvm_steal_time)))) {
> >>+
> >>+ vcpu->arch.st.stime = 0;
> >>+ return;
> >>+ }
> >>+
> >>+ delta = (get_kernel_ns() - vcpu->arch.st.this_time_out);
> >>+
> >>+ vcpu->arch.st.steal.steal += delta;
> >>+ vcpu->arch.st.steal.version += 2;
> >>+
> >>+ if (unlikely(kvm_write_guest(vcpu->kvm, vcpu->arch.st.stime,
> >Why not use kvm_write_guest_cached() here and introduce kvm_read_guest_cached()
> >for the read above?
>
> Actually, I'd expect most read/writes to benefit from caching, no?
> So why don't we just rename kvm_write_guest_cached() to
> kvm_write_guest(), and the few places - if any - that need to force
> transversing of the gfn mappings, get renamed to
> kvm_write_guest_uncached ?
>
Good idea. I do not see any places where kvm_write_guest_uncached is
needed from a brief look. Avi?
> >>+ &vcpu->arch.st.steal, sizeof(struct kvm_steal_time)))) {
> >>+
> >>+ vcpu->arch.st.stime = 0;
> >>+ return;
> >>+ }
> >>+ }
> >>+
> >>+}
> >>+
> >> int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data)
> >> {
> >> switch (msr) {
> >>@@ -1562,6 +1590,23 @@ int kvm_set_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 data)
> >> if (kvm_pv_enable_async_pf(vcpu, data))
> >> return 1;
> >> break;
> >>+ case MSR_KVM_STEAL_TIME:
> >>+ vcpu->arch.st.msr_val = data;
> >>+
> >>+ if (!(data& KVM_MSR_ENABLED)) {
> >>+ vcpu->arch.st.stime = 0;
> >>+ break;
> >>+ }
> >>+
> >>+ if (data& KVM_STEAL_RESERVED_MASK)
> >>+ return 1;
> >>+
> >>+ vcpu->arch.st.this_time_out = get_kernel_ns();
> >>+ vcpu->arch.st.stime = data& KVM_STEAL_VALID_BITS;
> >>+ record_steal_time(vcpu);
> >>+
> >>+ break;
> >>+
> >> case MSR_IA32_MCG_CTL:
> >> case MSR_IA32_MCG_STATUS:
> >> case MSR_IA32_MC0_CTL ... MSR_IA32_MC0_CTL + 4 * KVM_MAX_MCE_BANKS - 1:
> >>@@ -1847,6 +1892,9 @@ int kvm_get_msr_common(struct kvm_vcpu *vcpu, u32 msr, u64 *pdata)
> >> case MSR_KVM_ASYNC_PF_EN:
> >> data = vcpu->arch.apf.msr_val;
> >> break;
> >>+ case MSR_KVM_STEAL_TIME:
> >>+ data = vcpu->arch.st.msr_val;
> >>+ break;
> >> case MSR_IA32_P5_MC_ADDR:
> >> case MSR_IA32_P5_MC_TYPE:
> >> case MSR_IA32_MCG_CAP:
> >>@@ -2158,6 +2206,8 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
> >> kvm_migrate_timers(vcpu);
> >> vcpu->cpu = cpu;
> >> }
> >>+
> >>+ record_steal_time(vcpu);
> >> }
> >>
> >> void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
> >>@@ -2165,6 +2215,7 @@ void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
> >> kvm_x86_ops->vcpu_put(vcpu);
> >> kvm_put_guest_fpu(vcpu);
> >> kvm_get_msr(vcpu, MSR_IA32_TSC,&vcpu->arch.last_guest_tsc);
> >>+ vcpu->arch.st.this_time_out = get_kernel_ns();
> >> }
> >>
> >Shouldn't we call record_steal_time(vcpu)/vcpu->arch.st.this_time_out = get_kernel_ns();
> >just before/after entering/exiting a guest? vcpu_(put|get) are called
> >for each vcpu ioctl, not only VCPU_RUN.
> Sorry, missed that the first time I've read your e-mail.
>
> If done like you said, time spent on the hypervisor is accounted as
> steal time. I don't think it is.
I thought that this is the point of a steal time. Running other
tasks/guests is a hypervisor overhead too after all :) Also what about
time spend serving host interrupts between put/get? It will not be
accounted as steal time, correct?
>
> Steal time is time spent running someone else's job instead of
> yours. The name for the time spent in the hypervisor doing something
> for *you* is just overhead.
OK. That is the question of a definition I guess. If you define it like
that the code is correct.
>
> >
> >> static int is_efer_nx(void)
> >>@@ -2477,7 +2528,8 @@ static void do_cpuid_ent(struct kvm_cpuid_entry2 *entry, u32 function,
> >> (1<< KVM_FEATURE_NOP_IO_DELAY) |
> >> (1<< KVM_FEATURE_CLOCKSOURCE2) |
> >> (1<< KVM_FEATURE_ASYNC_PF) |
> >>- (1<< KVM_FEATURE_CLOCKSOURCE_STABLE_BIT);
> >>+ (1<< KVM_FEATURE_CLOCKSOURCE_STABLE_BIT) |
> >>+ (1<< KVM_FEATURE_STEAL_TIME);
> >> entry->ebx = 0;
> >> entry->ecx = 0;
> >> entry->edx = 0;
> >>@@ -6200,6 +6252,8 @@ int kvm_arch_vcpu_reset(struct kvm_vcpu *vcpu)
> >>
> >> kvmclock_reset(vcpu);
> >>
> >>+ vcpu->arch.st.stime = 0;
> >>+
> >> kvm_clear_async_pf_completion_queue(vcpu);
> >> kvm_async_pf_hash_reset(vcpu);
> >> vcpu->arch.apf.halted = false;
> >>--
> >>1.7.3.4
> >>
> >>--
> >>To unsubscribe from this list: send the line "unsubscribe kvm" in
> >>the body of a message to majordomo@xxxxxxxxxxxxxxx
> >>More majordomo info at http://vger.kernel.org/majordomo-info.html
> >
> >--
> > Gleb.
--
Gleb.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
