tasklets

Tasklets are a lightweight mechanism in the Linux kernel for deferring work from interrupt context to a later point without creating a full kernel thread. They are a form of bottom-half handling, allowing interrupt handlers to perform short, deferred processing while keeping interrupt latency low.

A tasklet consists of a function to execute and a data pointer, packaged in a per-CPU tasklet structure. When code schedules a tasklet using tasklet_schedule, the tasklet is enqueued on the local CPU’s tasklet list. The kernel processes non-empty tasklet lists during softirq handling, so tasklets run in softirq context on the originating CPU. They cannot sleep and typically run to completion before control returns to user code. If a tasklet is needed on another CPU, it must be scheduled on that CPU as well.

Tasklets are one of several bottom-half mechanisms in the Linux kernel, alongside softirqs and workqueues. They offer low-overhead deferral suitable for short, non-blocking tasks such as quick data processing or signaling work for later stages. However, due to their restrictions—especially the inability to sleep and per-CPU execution—longer or potentially blocking work is better handled by workqueues or threaded interrupt handlers.

History and evolution: Tasklets appeared as a simple way to defer work in the kernel and are still present for compatibility, but modern Linux kernel development often favors workqueues for deferrable work that may sleep or require more complex synchronization. Tasklets thus remain a historical reference point for bottom-half design in the kernel, illustrating the spectrum of interrupt handling strategies. See also bottom halves, softirq, and workqueue.