Patch 1 fixes a leak of root-1 shadow pages, patch 2 is a minor
optimization to the zap all flow that avoids re-reading and re-checking
the root-1 SPTEs after they've been zapped by "zap all" flows.
I'm still somewhat on the fence for patch 2, feel free to drop it.
v2:
- Replaced magic number silliness with Paolo's much more clever suggestion.
- Elaborated on the benefits of the optimization.
- Add Ben's somewhat reluctant review for the optimization.
v1: https://lkml.kernel.org/r/20210812050717.3176478-1-seanjc@xxxxxxxxxx
Sean Christopherson (2):
KVM: x86/mmu: Don't leak non-leaf SPTEs when zapping all SPTEs
KVM: x86/mmu: Don't step down in the TDP iterator when zapping all
SPTEs
arch/x86/kvm/mmu/tdp_mmu.c | 35 ++++++++++++++++++++++++-----------
1 file changed, 24 insertions(+), 11 deletions(-)