From: Alexander Graf <graf@amazon.com>
To: Sean Christopherson <sean.j.christopherson@intel.com>,
Paolo Bonzini <pbonzini@redhat.com>
Cc: Vitaly Kuznetsov <vkuznets@redhat.com>,
Wanpeng Li <wanpengli@tencent.com>,
Jim Mattson <jmattson@google.com>, Joerg Roedel <joro@8bytes.org>,
<kvm@vger.kernel.org>, <linux-kernel@vger.kernel.org>,
KarimAllah Raslan <karahmed@amazon.de>
Subject: Re: [PATCH] KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02
Date: Thu, 30 Apr 2020 23:22:20 +0200 [thread overview]
Message-ID: <853bea8c-41b8-ba3e-0a7c-c5df3b5dac9e@amazon.com> (raw)
In-Reply-To: <20200430204123.2608-1-sean.j.christopherson@intel.com>
On 30.04.20 22:41, Sean Christopherson wrote:
>
> Skip the Indirect Branch Prediction Barrier that is triggered on a VMCS
> switch when running with spectre_v2_user=on/auto if the switch is
> between two VMCSes in the same guest, i.e. between vmcs01 and vmcs02.
> The IBPB is intended to prevent one guest from attacking another, which
> is unnecessary in the nested case as it's the same guest from KVM's
> perspective.
>
> This all but eliminates the overhead observed for nested VMX transitions
> when running with CONFIG_RETPOLINE=y and spectre_v2_user=on/auto, which
> can be significant, e.g. roughly 3x on current systems.
>
> Reported-by: Alexander Graf <graf@amazon.com>
> Cc: KarimAllah Raslan <karahmed@amazon.de>
> Cc: stable@vger.kernel.org
> Fixes: 15d45071523d ("KVM/x86: Add IBPB support")
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
I can confirm that with kvm-unit-test's vmcall benchmark, the patch does
make a big difference:
BEFORE: vmcall 33488
AFTER: vmcall 14898
So we're at least getting a good chunk of performance back :)
> ---
> arch/x86/kvm/vmx/nested.c | 2 +-
> arch/x86/kvm/vmx/vmx.c | 12 ++++++++----
> arch/x86/kvm/vmx/vmx.h | 3 ++-
> 3 files changed, 11 insertions(+), 6 deletions(-)
>
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 2c36f3f53108..1a02bdfeeb2b 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -303,7 +303,7 @@ static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
> cpu = get_cpu();
> prev = vmx->loaded_vmcs;
> vmx->loaded_vmcs = vmcs;
> - vmx_vcpu_load_vmcs(vcpu, cpu);
> + vmx_vcpu_load_vmcs(vcpu, cpu, prev);
> vmx_sync_vmcs_host_state(vmx, prev);
> put_cpu();
>
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index 3ab6ca6062ce..818dd8ba5e9f 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -1311,10 +1311,12 @@ static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)
> pi_set_on(pi_desc);
> }
>
> -void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu)
> +void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu,
> + struct loaded_vmcs *buddy)
> {
> struct vcpu_vmx *vmx = to_vmx(vcpu);
> bool already_loaded = vmx->loaded_vmcs->cpu == cpu;
> + struct vmcs *prev;
>
> if (!already_loaded) {
> loaded_vmcs_clear(vmx->loaded_vmcs);
> @@ -1333,10 +1335,12 @@ void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu)
> local_irq_enable();
> }
>
> - if (per_cpu(current_vmcs, cpu) != vmx->loaded_vmcs->vmcs) {
> + prev = per_cpu(current_vmcs, cpu);
> + if (prev != vmx->loaded_vmcs->vmcs) {
> per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs;
> vmcs_load(vmx->loaded_vmcs->vmcs);
> - indirect_branch_prediction_barrier();
> + if (!buddy || buddy->vmcs != prev)
> + indirect_branch_prediction_barrier();
I fail to understand the logic here though. What exactly are you trying
to catch? We only do the barrier when the current_vmcs as loaded by
vmx_vcpu_load_vmcs is different from the vmcs of the context that was
issuing the vmcs load.
Isn't this a really complicated way to say "Don't flush for nested"? Why
not just make it explicit and pass in a bool that says "nested = true"
from vmx_switch_vmcs()? Is there any case I'm missing where that would
be unsafe?
Thanks,
Alex
Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879
next prev parent reply other threads:[~2020-04-30 21:22 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-04-30 20:41 [PATCH] KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02 Sean Christopherson
2020-04-30 21:22 ` Alexander Graf [this message]
2020-05-01 13:58 ` Sean Christopherson
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=853bea8c-41b8-ba3e-0a7c-c5df3b5dac9e@amazon.com \
--to=graf@amazon.com \
--cc=jmattson@google.com \
--cc=joro@8bytes.org \
--cc=karahmed@amazon.de \
--cc=kvm@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=pbonzini@redhat.com \
--cc=sean.j.christopherson@intel.com \
--cc=vkuznets@redhat.com \
--cc=wanpengli@tencent.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).