Performance Evaluation of In-kernel System Calls

Modern operating systems use mechanisms and instructions implemented in instruction set architectures to perform system calls, which typically involve a context switch from user space (unprivileged mode) to kernel space (privileged mode). By leveraging these mechanisms, we can implement in-kernel system calls, which are system calls invoked from kernel space. We present a performance evaluation of executing in-kernel system calls on ARMv7 and x86 64 architectures. The results we obtained indicate that implementing in-kernel system calls in operating systems is feasible, as long as the incurred overhead is taken into consideration.