mlockall

mlockall is a Unix-like operating system facility that requests the kernel to lock a process’s virtual address space into RAM, preventing paging to swap. By keeping memory resident, it can reduce paging latency and improve predictability for real-time tasks or sensitive operations.

The function accepts flags such as MCL_CURRENT and MCL_FUTURE. MCL_CURRENT locks all pages currently mapped into the process address space, while MCL_FUTURE locks pages that will be mapped in the future. Using both flags provides comprehensive protection for the process’s memory against paging for the duration of its execution.

Attempting to lock memory typically requires elevated privileges (for example, a capability such as CAP_IPC_LOCK or root) and sufficient lockable memory as governed by the process’s RLIMIT_MEMLOCK limit. The call returns 0 on success and -1 on failure, with errno indicating the cause (common errors include EAGAIN when memory cannot be locked due to limits or contention, EINVAL for invalid flags, EACCES or EPERM for insufficient privileges, and ENOMEM).

Applications that rely on mlockall include real-time systems, cryptographic software, and database servers that demand low latency or deterministic memory access. Locking memory increases RAM usage and can impose memory pressure on the host, potentially affecting other processes or system stability under heavy load.

To release the lock, munlockall can be used; mlock and munlock provide finer-grained control over individual memory regions. While supported on many Unix-like systems, the exact behavior and available flags can vary by platform.