Many IO_uring Improvements Submitted For Linux 5.19
There is a lot for Linux 5.19 particularly on the IO_uring side with improvements and new features abound.
On the IO_uring core side there is multi-shot accept, support for more types of cancellation types, support for cooperative task_work signaling, speed-ups for registered files, speed-ups for registered resource handling, support for sparse buffers and file maps, and other IO_uring improvements. See the full list here.
A separate pull meanwhile adds fsetxattr/setxattr and fgetxattr/getxattr support to IO_uring. Another pull is also adding socket(2) support to IO_uring with Linux 5.19. This socket(2) support is intended for use when using direct/registered file descriptors with IO_uring.
As part of enhancing IO_uring for networking use, another pull adds a hint to IO_uring for indicating whether more data is still available after receive or if the socket is empty.
More exciting IO_uring work for Linux 5.19 includes IO_uring passthrough support. This IO_uring passthrough support is initially wired up for NVMe passthrough support.
Meanwhile when it comes to the block subsystem changes for Linux 5.19 there is a lot of the usual churn. There is a fix for blk-throttle accounting, BFQ fixes, and a lot of ongoing clean-up work. Lastly the block driver changes include various updates/improvements to the NVMe code, MD RAID code improvements, and other fixes.
And a lot of other code continues pouring in for this first day of the Linux 5.19 merge window.
2 Comments
