This paper proposes HyperSafe, which provides lifetime control flow integrity for the hypervisor through two techniques: 1) non-bypassable memory lockdown of hypervisor code and control data, and 2) restricted pointer indexing to allow only valid control transfers. It implements these techniques in a type-1 hypervisor using LLVM for pointer analysis and enforcement with restricted pointer indexing. This aims to prevent attacks against the hypervisor, such as VM escapes or installation of a hypervisor rootkit.