Lockless MPSC FIFO queues for io_uring
Summary
LWN reports on a new lockless MPSC FIFO queue for io_uring in the Linux 7.2 kernel, replacing the previous lockless singly linked list. The article explains how producers push to the tail using an atomic exchange, how the consumer maintains a separate head to avoid cache contention, and how the queue handles empty and concurrent states to achieve higher throughput with fewer locks. It highlights performance benefits, simplified code paths, and notes community discussion on lock-free semantics and real-time considerations.