发布于 2026年9月16日 · 我们于 2026年9月16日 确认该职位仍然有效
这是您的公司吗?US$ 15 – US$ 25 (每个项目)
I need a clean, maintainable way to throttle overall memory usage directly inside the Linux kernel running on a small embedded board. My goal is to guarantee that critical processes never starve for RAM while background tasks are gently reined in before the OOM-killer steps in. The exact board can still change—at the moment I am evaluating Raspberry Pi, Arduino-style MCU boards with Linux, and BeagleBone—but whichever one we settle on, I want a single patch or loadable module that can be ported with minimal tweaks. Upstream code quality matters more than quick hacks. What I am after: a kernel-space mechanism (cgroup tweaks, custom LRU hooks, or another elegant approach you prefer) that detects when overall usage crosses a threshold I define and then slows further allocations for selected processes until memory returns to a safe level. User-space wrappers are fine as helpers, yet the core logic must live in the kernel so it reacts fast and works even if userland is stuck. Deliverables • Patch or .ko module with clear comments • Simple config file or sysfs/proc interface for threshold tuning • Step-by-step build instructions and brief validation demo showing memory pressure before/after the throttle kicks in I will test on my side and merge it into our Yocto build once it compiles cleanly with no warnings. If you have prior work on cgroups, mm subsystem, or similar embedded memory governors, that will help us hit the finish line faster.