Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 21 Dec 2022 18:34:48 -0800
Mime-Version: 1.0
Message-ID: <20221222023457.1764-1-vipinsh@google.com>
Subject: [Patch v3 0/9] NUMA aware page table's pages allocation
From:   Vipin Sharma <vipinsh@google.com>
To:     seanjc@google.com, pbonzini@redhat.com, bgardon@google.com,
        dmatlack@google.com
Cc:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
        Vipin Sharma <vipinsh@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi,

This series has expanded from v2 based on the feedbacks. Main items happening in
this series are:

1. KVM MMU shrinker now shrinks KVM caches.
   MMU shrinker will free shadow page caches and split caches whenever shrinker
   is invoked.

2. Page table's pages are allocated on NUMA node during fault and split.
   Pages of page tables will be allocated based on the underlying physical page
   a page table entry is point to. Got performance improvement up to 150% in
   416 vCPUs VM during live migrations.

3. Cache size is reduced from 40 to 5.
   40 is current cache size for KVM memory caches. Reduced them to 5. I didn't
   see any negative performance impact while running perf and dirty_log_perf_test.
   I also saw lesser number of calls to get a free page.

Thanks
Vipin

v3:
- Split patches into smaller ones.
- Repurposed KVM MMU shrinker to free cache pages instead of oldest page table
  pages
- Reduced cache size from 40 to 5
- Removed __weak function and initializing node value in all architectures.
- Some name changes.

v2: https://lore.kernel.org/lkml/20221201195718.1409782-1-vipinsh@google.com/
- All page table pages will be allocated on underlying physical page's
  NUMA node.
- Introduced module parameter, numa_aware_pagetable, to disable this
  feature.
- Using kvm_pfn_to_refcounted_page to get page from a pfn.

v1: https://lore.kernel.org/all/20220801151928.270380-1-vipinsh@google.com/

Vipin Sharma (9):
  KVM: x86/mmu: Repurpose KVM MMU shrinker to purge shadow page caches
  KVM: x86/mmu: Remove zapped_obsolete_pages from struct kvm_arch{}
  KVM: x86/mmu: Shrink split_shadow_page_cache via KVM MMU shrinker
  KVM: Add module param to make page tables NUMA aware
  KVM: x86/mmu: Allocate TDP page table's page on correct NUMA node on
    split
  KVM: Provide NUMA node support to kvm_mmu_memory_cache{}
  KVM: x86/mmu: Allocate page table's pages on NUMA node of the
    underlying pages
  KVM: x86/mmu: Make split_shadow_page_cache NUMA aware
  KVM: x86/mmu: Reduce default cache size in KVM from 40 to
    PT64_ROOT_MAX_LEVEL

 arch/arm64/kvm/arm.c             |   2 +-
 arch/arm64/kvm/mmu.c             |   4 +-
 arch/mips/kvm/mips.c             |   2 +
 arch/riscv/kvm/mmu.c             |   2 +-
 arch/riscv/kvm/vcpu.c            |   2 +-
 arch/x86/include/asm/kvm_host.h  |  15 +-
 arch/x86/include/asm/kvm_types.h |   2 +-
 arch/x86/kvm/mmu/mmu.c           | 282 +++++++++++++++++++------------
 arch/x86/kvm/mmu/mmu_internal.h  |   2 +
 arch/x86/kvm/mmu/paging_tmpl.h   |   4 +-
 arch/x86/kvm/mmu/tdp_mmu.c       |  24 ++-
 include/linux/kvm_host.h         |  27 +++
 include/linux/kvm_types.h        |   2 +
 virt/kvm/kvm_main.c              |  35 +++-
 14 files changed, 277 insertions(+), 128 deletions(-)

-- 
2.39.0.314.g84b9a713c41-goog