[PATCH v2 0/2] KVM: arm64: Stop stage-2 teardown scaling with the IPA size

[Fuad Tabba <fuad.tabba@linux.dev>]

Hi folks,

Changes since v1 [1]:
  - Run all teardown from stage2_destroy_range() (Oliver, Suzuki)
  - Dropped the Fixes tags (Marc)
  - Dropped the patch moving the teardown into kvm_vm_s2_ops
  - Use gfn_to_gpa() for a mapping's address. (Vincent)

Two patches that make stage-2 teardown take time proportional to what
is mapped, rather than to the size of the IPA space as it has since
4ddfab5436b6.

Under QEMU TCG ('-cpu max', 16K pages with LPA2), destroying even an
empty VM takes about 7 seconds in pKVM. Without pKVM it takes about 19
seconds if the VM requests a 52-bit IPA (crosvm requests the host's
limit). With the series, both take well under a second. I tested only
under QEMU, since I don't have hardware with LPA2.

The first has the page-table walk return where it stopped, so that the
teardown steps over an invalid entry larger than a block-sized chunk in
one go. The second has pKVM return where its next mapping starts.

pKVM gives every VM the host's IPA limit. My series honoring
KVM_VM_TYPE_ARM_IPA_SIZE in pKVM [2] doesn't change that for crosvm,
which requests the limit.

Based on Linux 7.4-rc1 (6628e7d4c6b25).

Cheers,
/fuad

[1] https://lore.kernel.org/all/20261103084742.1302597-1-fuad.tabba@linux.dev/
[2] https://lore.kernel.org/all/20260917091826.2639205-1-fuad.tabba@linux.dev/

Fuad Tabba (2):
  KVM: arm64: Resume stage-2 teardown where the walk stopped
  KVM: arm64: Resume a pKVM stage-2 teardown at the next mapping

 arch/arm64/include/asm/kvm_pgtable.h |  8 ++++++--
 arch/arm64/include/asm/kvm_pkvm.h    |  4 ++--
 arch/arm64/kvm/hyp/pgtable.c         | 15 ++++++++++++---
 arch/arm64/kvm/mmu.c                 |  4 ++--
 arch/arm64/kvm/pkvm.c                | 12 +++++++++---
 5 files changed, 31 insertions(+), 12 deletions(-)


base-commit: 6628e7d4c6b25c8d18f1345e3a20d21829aae7e4
-- 
2.47.3