Re: [PATCH v2 2/4] KVM: Add paravirt remote TLB flush

[Wanpeng Li]

2017-11-10 16:24 GMT+08:00 Paolo Bonzini <pbonzini@xxxxxxxxxx>:
> On 10/11/2017 08:04, Wanpeng Li wrote:
>> From: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>>
>> Remote flushing api's does a busy wait which is fine in bare-metal
>> scenario. But with-in the guest, the vcpus might have been pre-empted
>> or blocked. In this scenario, the initator vcpu would end up
>> busy-waiting for a long amount of time.
>>
>> This patch set implements para-virt flush tlbs making sure that it
>> does not wait for vcpus that are sleeping. And all the sleeping vcpus
>> flush the tlb on guest enter.
>>
>> The best result is achieved when we're overcommiting the host by running
>> multiple vCPUs on each pCPU. In this case PV tlb flush avoids touching
>> vCPUs which are not scheduled and avoid the wait on the main CPU.
>>
>> Test on a Haswell i7 desktop 4 cores (2HT), so 8 pCPUs, running ebizzy in
>> one linux guest.
>>
>> ebizzy -M
>>               vanilla    optimized     boost
>>  8 vCPUs       10152       10083       -0.68%
>> 16 vCPUs        1224        4866       297.5%
>> 24 vCPUs        1109        3871       249%
>> 32 vCPUs        1025        3375       229.3%
>>
>> Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
>> Cc: Radim KrÄmÃÅ <rkrcmar@xxxxxxxxxx>
>> Signed-off-by: Wanpeng Li <wanpeng.li@xxxxxxxxxxx>
>> ---
>>  Documentation/virtual/kvm/cpuid.txt  |  4 ++++
>>  arch/x86/include/uapi/asm/kvm_para.h |  2 ++
>>  arch/x86/kernel/kvm.c                | 31 +++++++++++++++++++++++++++++++
>>  3 files changed, 37 insertions(+)
>>
>> diff --git a/Documentation/virtual/kvm/cpuid.txt b/Documentation/virtual/kvm/cpuid.txt
>> index 117066a..9693fcc 100644
>> --- a/Documentation/virtual/kvm/cpuid.txt
>> +++ b/Documentation/virtual/kvm/cpuid.txt
>> @@ -60,6 +60,10 @@ KVM_FEATURE_PV_DEDICATED           ||     8 || guest checks this feature bit
>>                                     ||       || mizations such as usage of
>>                                     ||       || qspinlocks.
>>  ------------------------------------------------------------------------------
>> +KVM_FEATURE_PV_TLB_FLUSH           ||     9 || guest checks this feature bit
>> +                                   ||       || before enabling paravirtualized
>> +                                   ||       || tlb flush.
>> +------------------------------------------------------------------------------
>>  KVM_FEATURE_CLOCKSOURCE_STABLE_BIT ||    24 || host will warn if no guest-side
>>                                     ||       || per-cpu warps are expected in
>>                                     ||       || kvmclock.
>> diff --git a/arch/x86/include/uapi/asm/kvm_para.h b/arch/x86/include/uapi/asm/kvm_para.h
>> index 9ead1ed..a028479 100644
>> --- a/arch/x86/include/uapi/asm/kvm_para.h
>> +++ b/arch/x86/include/uapi/asm/kvm_para.h
>> @@ -25,6 +25,7 @@
>>  #define KVM_FEATURE_PV_EOI           6
>>  #define KVM_FEATURE_PV_UNHALT                7
>>  #define KVM_FEATURE_PV_DEDICATED     8
>> +#define KVM_FEATURE_PV_TLB_FLUSH     9
>>
>>  /* The last 8 bits are used to indicate how to interpret the flags field
>>   * in pvclock structure. If no bits are set, all flags are ignored.
>> @@ -53,6 +54,7 @@ struct kvm_steal_time {
>>
>>  #define KVM_VCPU_NOT_PREEMPTED      (0 << 0)
>>  #define KVM_VCPU_PREEMPTED          (1 << 0)
>> +#define KVM_VCPU_SHOULD_FLUSH       (1 << 1)
>>
>>  #define KVM_CLOCK_PAIRING_WALLCLOCK 0
>>  struct kvm_clock_pairing {
>> diff --git a/arch/x86/kernel/kvm.c b/arch/x86/kernel/kvm.c
>> index 66ed3bc..50f4b6a 100644
>> --- a/arch/x86/kernel/kvm.c
>> +++ b/arch/x86/kernel/kvm.c
>> @@ -465,6 +465,33 @@ static void __init kvm_apf_trap_init(void)
>>       update_intr_gate(X86_TRAP_PF, async_page_fault);
>>  }
>>
>> +static cpumask_t flushmask;
>
> Hi Wanpeng,
>
> are you going to send v3 with a percpu variable?

Hi Peterz,

I found big performance difference as I discuss with you several days ago.

ebizzy -M
                        vanilla    static/local cpumask     per-cpu cpumask
 8 vCPUs       10152            10083                          10117
16 vCPUs        1224              4866                          10008
24 vCPUs        1109              3871                            9928
32 vCPUs        1025              3375                            9811

In addition, I can observe ~50% perf top time is occupied by
smp_call_function_many(), ~30% perf top time is occupied by
call_function_interrupt() in the guest when running ebizzy for
static/local cpumask variable. However, I almost can't observe these
IPI stuffs after changing to per-cpu variable. Any opinions?

Regards,
Wanpeng Li

>
> Paolo
>
>> +static void kvm_flush_tlb_others(const struct cpumask *cpumask,
>> +                     const struct flush_tlb_info *info)
>> +{
>> +     u8 state;
>> +     int cpu;
>> +     struct kvm_steal_time *src;
>> +
>> +     cpumask_copy(&flushmask, cpumask);
>> +     /*
>> +      * We have to call flush only on online vCPUs. And
>> +      * queue flush_on_enter for pre-empted vCPUs
>> +      */
>> +     for_each_cpu(cpu, cpumask) {
>> +             src = &per_cpu(steal_time, cpu);
>> +             state = src->preempted;
>> +             if ((state & KVM_VCPU_PREEMPTED)) {
>> +                     if (cmpxchg(&src->preempted, state, state |
>> +                             KVM_VCPU_SHOULD_FLUSH) == state)
>> +                             __cpumask_clear_cpu(cpu, &flushmask);
>> +             }
>> +     }
>> +
>> +     native_flush_tlb_others(&flushmask, info);
>> +}
>> +
>>  void __init kvm_guest_init(void)
>>  {
>>       int i;
>> @@ -484,6 +511,10 @@ void __init kvm_guest_init(void)
>>               pv_time_ops.steal_clock = kvm_steal_clock;
>>       }
>>
>> +     if (kvm_para_has_feature(KVM_FEATURE_PV_TLB_FLUSH) &&
>> +             !kvm_para_has_feature(KVM_FEATURE_PV_DEDICATED))
>> +             pv_mmu_ops.flush_tlb_others = kvm_flush_tlb_others;
>> +
>>       if (kvm_para_has_feature(KVM_FEATURE_PV_EOI))
>>               apic_set_eoi_write(kvm_guest_apic_eoi_write);
>>
>>
>