Linux KVM/arm64 development list
 help / color / mirror / Atom feed
From: Oliver Upton <oliver.upton@linux.dev>
To: kvmarm@lists.linux.dev
Cc: Marc Zyngier <maz@kernel.org>, Joey Gouly <joey.gouly@arm.com>,
	Suzuki K Poulose <suzuki.poulose@arm.com>,
	Zenghui Yu <yuzenghui@huawei.com>,
	Oliver Upton <oliver.upton@linux.dev>
Subject: [PATCH 0/3] KVM: arm64: nv: Fixes for stage-2 MMU recycling
Date: Tue,  1 Oct 2024 00:17:06 +0000	[thread overview]
Message-ID: <20241001001709.1303668-1-oliver.upton@linux.dev> (raw)

I recently found that the path for reclaiming stage-2 MMUs is a bit
dodgy. We call kvm_stage2_unmap_range() from vcpu_load() when
repurposing a valid MMU, which itself is allowed to block. We repeat
this silliness on the next vcpu_load(), eventually leading to a stack
overflow.

It also looks like the nested hooks into the MMU notifiers are similarly
screwed up, ignoring the @may_block parameter and allowing the page
table walk to block unconditionally.

In addition to that, we are not handling vcpu->arch.hw_mmu safely in
preemptible code, i.e. we could be using a stale pointer on the loaded
hardware MMU.

Set of fixes to address all of the above, by:

1) Allowing vCPUs to 'pin' the loaded stage-2 MMU, guaranteeing that
   vcpu->arch.hw_mmu is stable when used in preemptible code

2) Preventing unmap walks from blocking when disallowed by the calling
   context.

   This is the correct fix for the MMU notifiers, and a "fix" for the
   reclaim case. Unmapping a whole MMU behind the write lock w/o periodically
   releasing the lock hurts vCPU scheduling, especially on larger VMs.

3) Push the cleanup of reclaimed MMUs into a vCPU request, allowing the
   walk to release the lock + CPU as needed.

Applies to 6.12-rc1, tested on M2. An easy way to force the reclaim case
is to constantly rerun a KVM selftest inside the L1 guest, which is how
I found this in the first place.

Oliver Upton (3):
  KVM: arm64: Treat stage-2 MMUs as refcounted generally
  KVM: arm64: nv: Do not block when unmapping stage-2 if disallowed
  KVM: arm64: nv: Punt stage-2 recycling to a vCPU request

 arch/arm64/include/asm/kvm_host.h   | 13 ++++++++--
 arch/arm64/include/asm/kvm_mmu.h    | 40 ++++++++++++++++++++++++++++-
 arch/arm64/include/asm/kvm_nested.h |  4 ++-
 arch/arm64/kvm/arm.c                | 12 +++++----
 arch/arm64/kvm/hyp/nvhe/hyp-main.c  |  2 +-
 arch/arm64/kvm/hyp/nvhe/pkvm.c      |  2 +-
 arch/arm64/kvm/hyp/nvhe/switch.c    |  2 +-
 arch/arm64/kvm/hyp/nvhe/tlb.c       |  4 +--
 arch/arm64/kvm/hyp/vhe/switch.c     |  2 +-
 arch/arm64/kvm/hyp/vhe/tlb.c        |  4 +--
 arch/arm64/kvm/mmu.c                | 34 +++++++++++++-----------
 arch/arm64/kvm/nested.c             | 39 ++++++++++++++++++++--------
 arch/arm64/kvm/sys_regs.c           |  6 ++---
 13 files changed, 119 insertions(+), 45 deletions(-)


base-commit: 9852d85ec9d492ebef56dc5f229416c925758edc
-- 
2.46.1.824.gd892dcdcdd-goog


             reply	other threads:[~2024-10-01  0:17 UTC|newest]

Thread overview: 19+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-01  0:17 Oliver Upton [this message]
2024-10-01  0:17 ` [PATCH 1/3] KVM: arm64: Treat stage-2 MMUs as refcounted generally Oliver Upton
2024-10-01  0:17 ` [PATCH 2/3] KVM: arm64: nv: Do not block when unmapping stage-2 if disallowed Oliver Upton
2024-10-01  0:17 ` [PATCH 3/3] KVM: arm64: nv: Punt stage-2 recycling to a vCPU request Oliver Upton
2024-10-01 19:05   ` Sean Christopherson
2024-10-01 20:41     ` Oliver Upton
2024-10-01 23:28       ` Sean Christopherson
2024-10-01 23:49         ` Marc Zyngier
2024-10-02  0:06           ` Oliver Upton
2024-10-02  0:23             ` Sean Christopherson
2024-10-02 23:31               ` Marc Zyngier
2024-10-03  0:04                 ` Oliver Upton
2024-10-03  0:12                   ` Oliver Upton
2024-10-03 16:45                     ` Sean Christopherson
2024-10-03 17:52                       ` Oliver Upton
2024-10-03 18:23                         ` Sean Christopherson
2024-10-03 22:03                           ` Oliver Upton
2024-10-01 23:23   ` Marc Zyngier
2024-10-02  0:06     ` Oliver Upton

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=20241001001709.1303668-1-oliver.upton@linux.dev \
    --to=oliver.upton@linux.dev \
    --cc=joey.gouly@arm.com \
    --cc=kvmarm@lists.linux.dev \
    --cc=maz@kernel.org \
    --cc=suzuki.poulose@arm.com \
    --cc=yuzenghui@huawei.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