📄 中文摘要
Unix 系统中可执行的原子操作能力，对并发编程和系统可靠性至关重要。文件系统操作、进程管理、内存访问等多个层面的原子性保证，是构建稳定系统的基石。尽管许多操作表面单一，底层实现可能涉及多步，并非天然原子。例如，`link(2)`、`rename(2)`、`mkdir(2)` 等系统调用在特定条件下展现原子性行为，开发者可利用这些特性构建健壮应用。然而，原子操作存在局限性，跨文件系统、网络文件系统或涉及多个独立操作时，原子性难以保证。`flock(2)` 和 `lockf(3)` 等文件锁机制，在实现更复杂原子性保证中发挥作用。理解这些基本原语，有助于开发者设计和实现高并发、高可靠的 Unix 应用程序，有效避免竞态条件和数据不一致性问题。
📄 English Summary
Things Unix can do atomically (2010)
This article delves into the atomic capabilities within Unix systems, highlighting the critical role of atomicity in concurrent programming and system reliability. The author meticulously analyzes atomic guarantees across various layers, including file system operations, process management, and memory access. It emphasizes that while many operations appear singular on the surface, their underlying implementations often involve multiple steps and are not inherently atomic. Through concrete examples like `link(2)`, `rename(2)`, and `mkdir(2)` system calls, the article elucidates their atomic behavior under specific conditions and how these properties can be leveraged to build robust applications. Concurrently, the article discusses the limitations of atomic operations, such as the challenges in ensuring atomicity across file systems, network file systems, or when involving multiple independent operations. It also touches upon file locking mechanisms like `flock(2)` and `lockf(3)` and their role in achieving more complex atomic guarantees. By understanding these fundamental primitives, developers can better design and implement highly concurrent and reliable Unix applications, effectively avoiding race conditions and data inconsistency issues. The article aims to provide readers with a comprehensive perspective on Unix atomic operations, helping them grasp the underlying semantics of system calls to write safer and more efficient code.