From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christoffer Dall Subject: Re: [PATCH] KVM: arm64: Count guest exit due to various reasons Date: Thu, 22 Oct 2015 23:16:01 +0200 Message-ID: <20151022211601.GA25602@cbox> References: <1445155482-12482-1-git-send-email-amittomer25@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Received: from localhost (localhost [127.0.0.1]) by mm01.cs.columbia.edu (Postfix) with ESMTP id 750DD40F90 for ; Thu, 22 Oct 2015 17:12:50 -0400 (EDT) Received: from mm01.cs.columbia.edu ([127.0.0.1]) by localhost (mm01.cs.columbia.edu [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 6GC0erGU4ylo for ; Thu, 22 Oct 2015 17:12:49 -0400 (EDT) Received: from mail-lf0-f53.google.com (mail-lf0-f53.google.com [209.85.215.53]) by mm01.cs.columbia.edu (Postfix) with ESMTPS id 41B384053F for ; Thu, 22 Oct 2015 17:12:48 -0400 (EDT) Received: by lfbn126 with SMTP id n126so28585028lfb.2 for ; Thu, 22 Oct 2015 14:15:26 -0700 (PDT) Content-Disposition: inline In-Reply-To: <1445155482-12482-1-git-send-email-amittomer25@gmail.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: kvmarm-bounces@lists.cs.columbia.edu Sender: kvmarm-bounces@lists.cs.columbia.edu To: Amit Cc: marc.zyngier@arm.com, andre.przywara@arm.com, kvmarm@lists.cs.columbia.edu List-Id: kvmarm@lists.cs.columbia.edu On Sun, Oct 18, 2015 at 01:34:42PM +0530, Amit wrote: > From: Amit Singh Tomar > > This patch adds guest exit statistics to debugfs, This can be helpful > while measuring KVM performance. > > Signed-off-by: Amit Singh Tomar > --- > arch/arm/include/asm/kvm_host.h | 2 ++ > arch/arm/kvm/arm.c | 1 + > arch/arm/kvm/guest.c | 2 ++ > arch/arm/kvm/mmio.c | 1 + > arch/arm64/include/asm/kvm_host.h | 5 +++++ > arch/arm64/kvm/guest.c | 8 ++++++++ > arch/arm64/kvm/handle_exit.c | 3 +++ > 7 files changed, 22 insertions(+) > > diff --git a/arch/arm/include/asm/kvm_host.h b/arch/arm/include/asm/kvm_host.h > index c4072d9..d1f3159 100644 > --- a/arch/arm/include/asm/kvm_host.h > +++ b/arch/arm/include/asm/kvm_host.h > @@ -147,6 +147,8 @@ struct kvm_vcpu_stat { > u32 halt_successful_poll; > u32 halt_attempted_poll; > u32 halt_wakeup; > + u32 mmio_exit_stat; I think it would be even more useful to distinguish between mmio exits handled in the kernel vs. those going to userspace. > + u32 exits; > }; > > int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init); > diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c > index 78b2869..5d0a2c5 100644 > --- a/arch/arm/kvm/arm.c > +++ b/arch/arm/kvm/arm.c > @@ -569,6 +569,7 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run) > ret = kvm_call_hyp(__kvm_vcpu_run, vcpu); > > vcpu->mode = OUTSIDE_GUEST_MODE; > + vcpu->stat.exits++; > /* > * Back from guest > *************************************************************/ > diff --git a/arch/arm/kvm/guest.c b/arch/arm/kvm/guest.c > index 96e935b..0eea5fc 100644 > --- a/arch/arm/kvm/guest.c > +++ b/arch/arm/kvm/guest.c > @@ -33,6 +33,8 @@ > #define VCPU_STAT(x) { #x, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU } > > struct kvm_stats_debugfs_item debugfs_entries[] = { > + VCPU_STAT(mmio_exit_stat), > + VCPU_STAT(exits), > { NULL } > }; > > diff --git a/arch/arm/kvm/mmio.c b/arch/arm/kvm/mmio.c > index 974b1c6..51d518e 100644 > --- a/arch/arm/kvm/mmio.c > +++ b/arch/arm/kvm/mmio.c > @@ -184,6 +184,7 @@ int io_mem_abort(struct kvm_vcpu *vcpu, struct kvm_run *run, > } > > rt = vcpu->arch.mmio_decode.rt; > + vcpu->stat.mmio_exit_stat++; > > if (is_write) { > data = vcpu_data_guest_to_host(vcpu, *vcpu_reg(vcpu, rt), len); > diff --git a/arch/arm64/include/asm/kvm_host.h b/arch/arm64/include/asm/kvm_host.h > index ed03968..b32c938 100644 > --- a/arch/arm64/include/asm/kvm_host.h > +++ b/arch/arm64/include/asm/kvm_host.h > @@ -194,6 +194,11 @@ struct kvm_vcpu_stat { > u32 halt_successful_poll; > u32 halt_attempted_poll; > u32 halt_wakeup; > + u32 hvc_exit_stat; > + u32 wfx_exit_stat; > + u32 wfi_exit_stat; > + u32 mmio_exit_stat; > + u32 exits; why more exit stats on arm64 than on arm? why u32? Would it not make sense to make these u64 or just unsigned long? > }; > > int kvm_vcpu_preferred_target(struct kvm_vcpu_init *init); > diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c > index d250160..b1943cd 100644 > --- a/arch/arm64/kvm/guest.c > +++ b/arch/arm64/kvm/guest.c > @@ -34,7 +34,15 @@ > > #include "trace.h" > > +#define VM_STAT(x) { #x, offsetof(struct kvm, stat.x), KVM_STAT_VM } > +#define VCPU_STAT(x) { #x, offsetof(struct kvm_vcpu, stat.x), KVM_STAT_VCPU } > + > struct kvm_stats_debugfs_item debugfs_entries[] = { > + VCPU_STAT(hvc_exit_stat), > + VCPU_STAT(wfx_exit_stat), > + VCPU_STAT(wfi_exit_stat), > + VCPU_STAT(mmio_exit_stat), > + VCPU_STAT(exits), > { NULL } > }; > > diff --git a/arch/arm64/kvm/handle_exit.c b/arch/arm64/kvm/handle_exit.c > index 68a0759..6e38c50 100644 > --- a/arch/arm64/kvm/handle_exit.c > +++ b/arch/arm64/kvm/handle_exit.c > @@ -39,6 +39,7 @@ static int handle_hvc(struct kvm_vcpu *vcpu, struct kvm_run *run) > > trace_kvm_hvc_arm64(*vcpu_pc(vcpu), *vcpu_reg(vcpu, 0), > kvm_vcpu_hvc_get_imm(vcpu)); > + vcpu->stat.hvc_exit_stat++; > > ret = kvm_psci_call(vcpu); > if (ret < 0) { > @@ -71,9 +72,11 @@ static int kvm_handle_wfx(struct kvm_vcpu *vcpu, struct kvm_run *run) > { > if (kvm_vcpu_get_hsr(vcpu) & ESR_ELx_WFx_ISS_WFE) { > trace_kvm_wfx_arm64(*vcpu_pc(vcpu), true); > + vcpu->stat.wfx_exit_stat++; either you should use one called wfx_exit_stat or you should distinguish between wfi and wfe; this is just a weird way of doing it. > kvm_vcpu_on_spin(vcpu); > } else { > trace_kvm_wfx_arm64(*vcpu_pc(vcpu), false); > + vcpu->stat.wfi_exit_stat++; > kvm_vcpu_block(vcpu); > } > > -- > 1.9.1 > Thanks, -Christoffer