Loading feed…
Phoronix0 Hot0 comments2 min read
Linux's IO_uring Sees Some "Crazy" Patches That Deliver Very Nice Improvements
Jens Axboe, the io_uring lead developer, posted an RFC patch series that changes how io_uring handles blocking operations like fsync, statx, openat, and xattr calls. Instead of always punting these to io-wq worker threads (which costs a thread wakeup, context switch, and task_work round trip), the new approach issues them inline in blocking mode and only offloads when the request actually blocks. When blocking does occur, an idle io-wq worker takes over the submitting task's identity (tid, signal state, credentials, scheduling attributes, cgroup, register state) to complete the syscall transparently, while the original task finishes the work as an io-wq worker. Axboe notes similar approaches were attempted roughly 20 years ago. Early benchmarks shared on X show significant performance gains, though it remains unclear if this will land in time for the Linux v7.4 cycle.
Source: https://www.phoronix.com/news/Linux-IO-uring-Crazy-Friday. 8 Sync News only summarizes and links out; content copyright belongs to the authors and original sources.
Recommended for you
Bình luận
Đang tải bình luận…
