[PATCH v2 0/4] KVM: SVM: Fix IRQ window inhibit handling

[Sean Christopherson <seanjc@google.com>]

Patch 1 fixes a bugs where KVM will keep AVIC inhibit for too long when
running a nested guest and AVIC is inhibited to open an IRQ window.

Patch 2 refcounts IRQ window inhibits so that "closing" an IRQ window on
one vCPU doesn't clobber other vCPUs' windows when AVIC is enabled.

Patch 3 optimizes IRQ window inhibits by avoiding contention on
apicv_update_lock when KVM wants to inhibit AVIC to open an IRQ windows
because AVIC is _already_ inhibited.

Patch 4 further optimizes IRQ window inhibits by isolating the refcount
and the lock in their own cacheline, e.g. to avoid false sharing and cache
line contention with things like apicv_inhibit_reasons, which is read on
every VM-Enter.

v2:
 - Formalize the SoB chains (thanks Naveen!)
 - Explicitly isolate the write-mostly fields with ____cacheline_aligned.
 - Add Naveen's Tested-by tags.

v1 / RFC: https://lore.kernel.org/all/cover.1752819570.git.naveen@kernel.org

Original: https://lore.kernel.org/all/cover.1738595289.git.naveen@kernel.org

Sean Christopherson (4):
  KVM: SVM: Fix clearing IRQ window inhibit with nested guests
  KVM: SVM: Fix IRQ window inhibit handling across multiple vCPUs
  KVM: SVM: Optimize IRQ window inhibit handling
  KVM: Isolate apicv_update_lock and apicv_nr_irq_window_req in a
    cacheline

 arch/x86/include/asm/kvm_host.h | 29 +++++++++++++++-
 arch/x86/kvm/svm/svm.c          | 60 ++++++++++++++++++++-------------
 arch/x86/kvm/svm/svm.h          |  1 +
 arch/x86/kvm/x86.c              | 45 ++++++++++++++++++++++++-
 4 files changed, 110 insertions(+), 25 deletions(-)


base-commit: e81f7c908e1664233974b9f20beead78cde6343a
-- 
2.52.0.457.g6b5491de43-goog