On 15/02/2017 19:28, Cao, Lei wrote:
> + spin_lock(&kvm->mmu_lock);
> + kvm_arch_mmu_enable_log_dirty_pt_masked(kvm, memslot, offset, mask);
> + spin_unlock(&kvm->mmu_lock);
> +
> + while (mask) {
> + clear_bit_le(offset + __ffs(mask), memslot->dirty_bitmap);
> + mask &= mask - 1;
> + }
These two steps should be done in the opposite order. So far nothing I
cannot fix on commit though (and this is going to be material for 4.12
anyway).
Paolo
