linux-kernel - Re: [PATCH] KVM: X86: correct trace_kvm_pv_tlb

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <YnL0gUcUq5MbWvdH@google.com>
Date:   Wed, 4 May 2022 21:47:45 +0000
From:   Sean Christopherson <seanjc@...gle.com>
To:     Jon Kohler <jon@...anix.com>
Cc:     Paolo Bonzini <pbonzini@...hat.com>,
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        Wanpeng Li <wanpengli@...cent.com>,
        Jim Mattson <jmattson@...gle.com>,
        Joerg Roedel <joro@...tes.org>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>, Borislav Petkov <bp@...en8.de>,
        Dave Hansen <dave.hansen@...ux.intel.com>, x86@...nel.org,
        "H. Peter Anvin" <hpa@...or.com>, kvm@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] KVM: X86: correct trace_kvm_pv_tlb_flush stats

On Wed, May 04, 2022, Jon Kohler wrote:
> The trace point in record_steal_time() is above the conditional
> that fires kvm_vcpu_flush_tlb_guest(), so even when we might
> not be flushing tlb, we still record that we are.

No, it records whether not a TLB flush is needed.

> Fix by nestling trace_kvm_pv_tlb_flush() under appropriate
> conditional. This results in the stats for kvm:kvm_pv_tlb_flush,
> as trivially observable by perf stat -e "kvm:*" -a sleep Xs, in
> reporting the amount of times we actually do a pv tlb flush,
> instead of just the amount of times we happen to call
> record_steal_time().
> 
> Signed-off-by: Jon Kohler <jon@...anix.com>
> ---
>  arch/x86/kvm/x86.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 4790f0d7d40b..8d4e0e58ec34 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -3410,9 +3410,9 @@ static void record_steal_time(struct kvm_vcpu *vcpu)
>  
>  		vcpu->arch.st.preempted = 0;
>  
> -		trace_kvm_pv_tlb_flush(vcpu->vcpu_id,
> -				       st_preempted & KVM_VCPU_FLUSH_TLB);
>  		if (st_preempted & KVM_VCPU_FLUSH_TLB)
> +			trace_kvm_pv_tlb_flush(vcpu->vcpu_id,
> +				st_preempted & KVM_VCPU_FLUSH_TLB);

If you're going to trace only when a flush is needed, this should simply be:

			trace_kvm_pv_tlb_flush(vcpu->vcpu_id);

I haven't used this tracepoint often (at all?) so I don't have a strong preference,
but I can see the "no TLB flush needed" information being extremely valuable when
debugging a supsected TLB flushing bug.

>  			kvm_vcpu_flush_tlb_guest(vcpu);
>  
>  		if (!user_access_begin(st, sizeof(*st)))
> -- 
> 2.30.1 (Apple Git-130)
>