I've been playing with a hacked-up kvmtool that places the memory much higher than usual, in an effort to emulate what some arm64 platforms do (typically AMD Seattle). Unexpectedly (hey, what could possibly go wrong...), I've discovered a couple of mm related issues, ranging from the host exploding to the guest being stuck on a page fault. Additionally, I noticed a minor documentation issue. With these fixes in, I'm able to run a guest that has its memory located just under the 40bit limit. * From v1 [1]: - Reworked patch #1 to use split_page() instead. This leads to much refactoring, but is slightly nicer (more common code). - Dropped Mark's Tested-by (too many changes) [1]: https://lists.cs.columbia.edu/pipermail/kvmarm/2015-February/013672.html Marc Zyngier (3): arm64: KVM: Fix stage-2 PGD allocation to have per-page refcounting arm64: KVM: Do not use pgd_index to index stage-2 pgd arm64: KVM: Fix outdated comment about VTCR_EL2.PS arch/arm/include/asm/kvm_mmu.h | 12 ++---- arch/arm/kvm/mmu.c | 85 ++++++++++++++++++++++++++++++---------- arch/arm64/include/asm/kvm_arm.h | 5 ++- arch/arm64/include/asm/kvm_mmu.h | 48 +++-------------------- 4 files changed, 77 insertions(+), 73 deletions(-) -- 2.1.4 _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
