linux-kernel - [PATCH v2 0/5] RFC: KVM: SVM: Allow L1's AVIC to co-exist with nesting

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-next>] [day] [month] [year] [list]

Message-Id: <20211213104634.199141-1-mlevitsk@redhat.com>
Date:   Mon, 13 Dec 2021 12:46:29 +0200
From:   Maxim Levitsky <mlevitsk@...hat.com>
To:     kvm@...r.kernel.org
Cc:     Jim Mattson <jmattson@...gle.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Joerg Roedel <joro@...tes.org>,
        x86@...nel.org (maintainer:X86 ARCHITECTURE (32-BIT AND 64-BIT)),
        Vitaly Kuznetsov <vkuznets@...hat.com>,
        Borislav Petkov <bp@...en8.de>, linux-kernel@...r.kernel.org,
        Paolo Bonzini <pbonzini@...hat.com>,
        Dave Hansen <dave.hansen@...ux.intel.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Sean Christopherson <seanjc@...gle.com>,
        Wanpeng Li <wanpengli@...cent.com>,
        Ingo Molnar <mingo@...hat.com>,
        Maxim Levitsky <mlevitsk@...hat.com>
Subject: [PATCH v2 0/5] RFC: KVM: SVM: Allow L1's AVIC to co-exist with nesting

This patch series aims to lift long standing restriction of
using AVIC only when nested virtualization is not exposed
to the guest.

Notes about specific patches:

Patch 1 - this is an unrelated fix to KVM for a corner case I found
while writing a unit test for the feature.

Patch 2 addresses the fact that AVIC appears to be disabled
in CPUID on several Milan systems I am tesing on.

This adds a workaround (with a big warning and a kernel taint) to enable
it anyway if you really know what you are doing.

Patch 3 is from Paolo, which was done after our long discussion about the
various races that AVIC inhibition is subject to. Thanks Paolo!
It replaces two patches I sent in the previous version which attempted
to fix the same issue but weren't quite right.

Patch 4 is the more or less the same patch 5 in V1, but with proper
justification.

Patch 6 is the patch that adds the AVIC co-existance itself and
in this version I fixed (and partially tested) it in regard to AVIC inhibition
due to interrupt window.

Everything was tested on a Zen3 (Milan) machine.

On Zen2 machines, the errata #1235 makes my tests fail quite fast.
For general use though, most of the time this errata doesn't cause
long hangs.

Best regards,
	Maxim Levitsky

Maxim Levitsky (5):
  KVM: nSVM: deal with L1 hypervisor that intercepts interrupts but lets
    L2 control EFLAGS.IF
  KVM: SVM: allow to force AVIC to be enabled
  KVM: SVM: fix race between interrupt delivery and AVIC inhibition
  KVM: x86: don't touch irr_pending in kvm_apic_update_apicv when
    inhibiting it
  KVM: SVM: allow AVIC to co-exist with a nested guest running

 arch/x86/include/asm/kvm-x86-ops.h |  1 +
 arch/x86/include/asm/kvm_host.h    |  7 ++-
 arch/x86/kvm/lapic.c               |  5 +-
 arch/x86/kvm/svm/avic.c            | 91 +++++++++++++++++++-----------
 arch/x86/kvm/svm/nested.c          | 11 ++--
 arch/x86/kvm/svm/svm.c             | 51 +++++++++++------
 arch/x86/kvm/svm/svm.h             |  1 +
 arch/x86/kvm/x86.c                 | 17 +++++-
 8 files changed, 125 insertions(+), 59 deletions(-)

-- 
2.26.3