Here's v6 of the io_uring interface. Various tweaks and fixes to the internals, especially in terms of the application facing API. See changelog below for those. In terms of features, the ring_fd (returned from io_uring_setup(2)) is now pollable, so you can use poll(2)/epoll(2) and friends to get notified of completions. In the same ballpark, I stole the IOCB_CMD_POLL feature from aio, and Christoph massaged it to better fit the new API. This means we now have two new commands: - IORING_OP_POLL_ADD. Adds an fd to poll for. A cqe will be posted to the CQ ring when the desired events are available. - IORING_OP_POLL_REMOVE. Remove a previous queued POLL_ADD. The io_uring instance is now accounted fully, and will fail if we cannot get enough memory under the RLIMIT_MEMLOCK limit. The liburing library has grown a few test cases, and now also the beginnings of man pages for the project. So far we have io_uring_register(2) documented, io_uring_setup(2) and io_uring_enter(2) should be coming shortly, as well as a io_uring(7) man page. Thanks to Jeff Moyer for leading this effort! Clone the liburing repo here: git://git.kernel.dk/liburing As usual, patches are against 5.0-rc2, and can also be found in my io_uring branch here: git://git.kernel.dk/linux-block io_uring Changes since v5: - Limit fixed buffers to UIO_MAXIOV (was USHRT_MAX) - Limit fixed buffers to non-file backed memory (Dave Chinner) - Fail file/buffer registration if already setup (Jeff Moyer) - Return -ENXIO for attempt to enter dead ring - Reorder two patches - Cleanup some patch dependencies - Add support for polling the ring_fd (poll(2)/epoll(2)) - Add IORING_OP_POLL_ADD/IORING_OP_POLL_REMOVE support - Limit io_uring SQ size to 4k entries (IORING_MAX_ENTRIES) - Check sqe->flags validity in one location, not for each opcode - Account ring memory in RLIMIT_MEMLOCK - Re-init completion event after io_uring_register() - Fix SQ thread missing submission - Fix silly use-after-free in io_free_req() Documentation/filesystems/vfs.txt | 3 + arch/x86/entry/syscalls/syscall_32.tbl | 3 + arch/x86/entry/syscalls/syscall_64.tbl | 3 + block/bio.c | 59 +- fs/Makefile | 1 + fs/block_dev.c | 19 +- fs/file.c | 15 +- fs/file_table.c | 9 +- fs/gfs2/file.c | 2 + fs/io_uring.c | 2376 ++++++++++++++++++++++++ fs/iomap.c | 48 +- fs/xfs/xfs_file.c | 1 + include/linux/bio.h | 14 + include/linux/blk_types.h | 1 + include/linux/file.h | 2 + include/linux/fs.h | 6 +- include/linux/iomap.h | 1 + include/linux/sched/user.h | 2 +- include/linux/syscalls.h | 7 + include/uapi/linux/io_uring.h | 141 ++ init/Kconfig | 9 + kernel/sys_ni.c | 4 + 22 files changed, 2686 insertions(+), 40 deletions(-) -- Jens Axboe
