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: <20230201194604.11135-1-minipli@grsecurity.net>
Date:   Wed,  1 Feb 2023 20:45:58 +0100
From:   Mathias Krause <minipli@...ecurity.net>
To:     kvm@...r.kernel.org
Cc:     linux-kernel@...r.kernel.org,
        Sean Christopherson <seanjc@...gle.com>,
        Paolo Bonzini <pbonzini@...hat.com>,
        Mathias Krause <minipli@...ecurity.net>
Subject: [PATCH v3 0/6] KVM: MMU: performance tweaks for heavy CR0.WP users

v2: https://lore.kernel.org/kvm/20230118145030.40845-1-minipli@grsecurity.net/

This series is a resurrection of the missing pieces of Paolo's previous
attempt[1] to avoid needless MMU roots unloading. The performance gap
between TDP and legacy MMU is still existent, especially noticeable under
grsecurity which implements kernel W^X by toggling CR0.WP, which happens
very frequently.

Patches 1-13 and 17 of the old series had been merged, but, unfortunately,
the remaining parts never saw a v3. I therefore took care of these, took
Sean's feedback into account[2] and simplified the whole approach to just
handle the case we care most about explicitly.

Patch 1 is a v3 of [3], addressing Sean's feedback.

Patch 2 is specifically useful for grsecurity, as handle_cr() is by far
*the* top vmexit reason.

Patch 3 is the most important one, as it skips unloading the MMU roots for
CR0.WP toggling.

Sean was suggesting another change on top of v2 of this series, to skip
intercepting CR0.WP writes completely for VMX[4]. That turned out to be
yet another performance boost and is implemenmted in patch 6.

While patches 1 and 2 bring small performance improvements already, the
big gains come from patches 3 and 6.

I used 'ssdd 10 50000' from rt-tests[5] as a micro-benchmark, running on a
grsecurity L1 VM. Below table shows the results (runtime in seconds, lower
is better):

                         legacy     TDP    shadow
    kvm.git/queue        11.55s   13.91s    75.2s
    + patches 1-3         7.32s    7.31s    74.6s
    + patches 4-6         4.89s    4.89s    73.4s

This series builds on top of kvm.git/queue, namely commit de60733246ff
("Merge branch 'kvm-hw-enable-refactor' into HEAD").

Patches 1-3 didn't change from v2, beside minor changlog mangling.

Patches 4-6 are new to v3.

Thanks,
Mathias

[1] https://lore.kernel.org/kvm/20220217210340.312449-1-pbonzini@redhat.com/
[2] https://lore.kernel.org/kvm/YhATewkkO%2Fl4P9UN@google.com/
[3] https://lore.kernel.org/kvm/YhAB1d1%2FnQbx6yvk@google.com/
[4] https://lore.kernel.org/kvm/Y8cTMnyBzNdO5dY3@google.com/
[5] https://git.kernel.org/pub/scm/utils/rt-tests/rt-tests.git

Mathias Krause (5):
  KVM: VMX: Avoid retpoline call for control register caused exits
  KVM: x86: Do not unload MMU roots when only toggling CR0.WP
  KVM: x86: Make use of kvm_read_cr*_bits() when testing bits
  KVM: x86/mmu: Fix comment typo
  KVM: VMX: Make CR0.WP a guest owned bit

Paolo Bonzini (1):
  KVM: x86/mmu: Avoid indirect call for get_cr3

 arch/x86/kvm/kvm_cache_regs.h   |  3 ++-
 arch/x86/kvm/mmu/mmu.c          | 31 ++++++++++++++++++++-----------
 arch/x86/kvm/mmu/paging_tmpl.h  |  2 +-
 arch/x86/kvm/mmu/spte.c         |  2 +-
 arch/x86/kvm/pmu.c              |  4 ++--
 arch/x86/kvm/vmx/capabilities.h |  1 +
 arch/x86/kvm/vmx/nested.c       |  4 ++--
 arch/x86/kvm/vmx/vmx.c          | 15 ++++++++++++---
 arch/x86/kvm/vmx/vmx.h          |  8 ++++++++
 arch/x86/kvm/x86.c              |  9 +++++++++
 10 files changed, 58 insertions(+), 21 deletions(-)

-- 
2.39.1

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ