/
githubmirror
/
liburing
Обзор
Документация
Войти
/
githubmirror
/
liburing
Код
Запросы
0
Пакеты
0
Релизы
0
Аналитика
Безопасность
master
man/io_uring_cancelation.7
324 строки
8 KB
Jens Axboe
man/io_uring_cancelation.7: various cleanups
04 фев 2026, 19:04
04 фев 2026, 19:04
e3d35ea
Код
Авторство
О чём код?
.\" Copyright (C) 2025 Jens Axboe <axboe@kernel.dk> .\" SPDX-License-Identifier: LGPL-2.0-or-later .\" .TH io_uring_cancelation 7 "January 18, 2025" "Linux" "Linux Programmer's Manual" .SH NAME io_uring_cancelation \- io_uring request cancelation overview .SH DESCRIPTION io_uring provides mechanisms to cancel in-flight requests before they complete naturally. This is useful for implementing timeouts, handling connection drops, closing connections that go away, or when a request is no longer needed. .SS Why cancel requests? Common scenarios requiring cancelation: .IP \(bu 2 .B Timeouts: Cancel a read or accept that has been pending too long .IP \(bu .B Connection management: Cancel pending operations when a connection is closed .IP \(bu .B Resource cleanup: Cancel operations on a file descriptor being closed .IP \(bu .B Multishot termination: Stop a multishot operation that is no longer needed .SS Basic cancelation The primary cancelation mechanism is .B IORING_OP_ASYNC_CANCEL (set up with .BR io_uring_prep_cancel (3) or related functions). By default, it cancels a request matching a specific .IR user_data : .PP .in +4n .EX /* Submit a read with user_data = 1234 */ sqe = io_uring_get_sqe(ring); io_uring_prep_read(sqe, fd, buf, len, 0); io_uring_sqe_set_data64(sqe, 1234); io_uring_submit(ring); /* Later, cancel it */ sqe = io_uring_get_sqe(ring); io_uring_prep_cancel64(sqe, 1234, 0); io_uring_submit(ring); .EE .in .SS Cancelation results When a cancelation is submitted, two CQEs are generated: .PP .B The canceled request's CQE: .RS 4 .IP \(bu 2 .I res is set to .B -ECANCELED (or occasionally .B -EINTR if the operation was already in progress) .IP \(bu The .I user_data identifies which request was canceled .RE .PP .B The cancel request's CQE: .RS 4 .IP \(bu 2 .I res is 0 on success (request was found and canceled) .IP \(bu .I res is .B -ENOENT if no matching request was found .IP \(bu .I res is .B -EALREADY if the request was found but already completing .RE .PP The order of these CQEs is not guaranteed. The application may receive the cancel CQE before or after the canceled request's CQE. .SS Cancelation flags Various flags modify cancelation behavior: .PP .B IORING_ASYNC_CANCEL_ALL .RS 4 Cancel all matching requests, not just the first one found. The cancel CQE's .I res indicates how many requests were canceled. .RE .PP .B IORING_ASYNC_CANCEL_FD .RS 4 Match requests by file descriptor instead of .IR user_data . Cancels requests operating on the specified fd: .PP .in +4n .EX io_uring_prep_cancel_fd(sqe, fd, IORING_ASYNC_CANCEL_FD); .EE .in .RE .PP .B IORING_ASYNC_CANCEL_ANY .RS 4 Cancel any single request, ignoring .I user_data matching. Useful for draining a ring of all pending requests when combined with .BR IORING_ASYNC_CANCEL_ALL . .RE .PP .B IORING_ASYNC_CANCEL_FD_FIXED .RS 4 The file descriptor is a fixed file (registered file index) rather than a regular fd. .RE .PP Flags can be combined: .PP .in +4n .EX /* Cancel all requests on a specific fd */ io_uring_prep_cancel_fd(sqe, fd, IORING_ASYNC_CANCEL_FD | IORING_ASYNC_CANCEL_ALL); /* Cancel all pending requests in the ring */ io_uring_prep_cancel(sqe, NULL, IORING_ASYNC_CANCEL_ANY | IORING_ASYNC_CANCEL_ALL); .EE .in .SS Race conditions Cancelation is inherently racy. Between submitting the cancel request and the kernel processing it: .IP \(bu 2 The target request may complete successfully .IP \(bu The target request may fail for another reason .IP \(bu The target request may already be in a non-cancelable state .PP Applications must handle these cases: .PP .in +4n .EX io_uring_wait_cqe(ring, &cqe); if (cqe->user_data == cancel_user_data) { /* This is the cancel operation's result */ if (cqe->res == -ENOENT) { /* Request already completed or not found */ } else if (cqe->res == -EALREADY) { /* Request was found but completing */ } else if (cqe->res >= 0) { /* Successfully canceled res requests */ } } else { /* This is the original request's result */ if (cqe->res == -ECANCELED) { /* Request was canceled */ } else { /* Request completed normally (or with error) */ } } .EE .in .SS Link timeouts For timing out a single operation, link timeouts are often simpler than explicit cancelation. See .BR io_uring_linked_requests (7) for details: .PP .in +4n .EX sqe = io_uring_get_sqe(ring); io_uring_prep_read(sqe, fd, buf, len, 0); sqe->flags |= IOSQE_IO_LINK; sqe = io_uring_get_sqe(ring); io_uring_prep_link_timeout(sqe, &timeout, 0); .EE .in .PP The kernel handles the cancelation automatically if the timeout expires. .SS Canceling multishot requests Multishot requests (see .BR io_uring_multishot (7)) continue generating completions until canceled or an error occurs. To stop a multishot request: .PP .in +4n .EX /* Cancel a multishot accept */ io_uring_prep_cancel64(sqe, accept_user_data, 0); .EE .in .PP After cancelation: .IP \(bu 2 The multishot generates a final CQE with .B -ECANCELED .IP \(bu The .B IORING_CQE_F_MORE flag is not set on this final CQE .IP \(bu The cancel CQE indicates success .SS Canceling by file descriptor When a file descriptor is closed (either via .BR close (2) or .BR IORING_OP_CLOSE ), pending requests operating on that fd are .B not automatically canceled. This differs from synchronous I/O behavior and is a common source of confusion. In synchronous I/O, closing a file descriptor is typically the last reference to the underlying file, so the close completes any pending operations. However, io_uring holds its own reference to the file for each pending request. Closing the application's fd does not release these references \(em the pending read, recv, or other operation continues to hold a reference and will not automatically complete or fail. If an application expects a pending read on an fd to post a completion when the fd is closed, that will not happen. The request must be explicitly canceled: .PP .in +4n .EX /* Cancel all operations on fd before closing */ sqe = io_uring_get_sqe(ring); io_uring_prep_cancel_fd(sqe, fd, IORING_ASYNC_CANCEL_FD | IORING_ASYNC_CANCEL_ALL); io_uring_submit(ring); /* Wait for cancelations, then close */ .EE .in .SS Shutdown cancelation When an io_uring instance is closed (via .BR io_uring_queue_exit (3) or closing the ring file descriptor), all pending requests are automatically canceled. Manual cancelation before shutdown is not required. However, if the application needs to ensure all requests are completed before proceeding (e.g., to process their results or free associated resources), explicit cancelation can be used: .PP .in +4n .EX /* Cancel everything */ sqe = io_uring_get_sqe(ring); io_uring_prep_cancel(sqe, NULL, IORING_ASYNC_CANCEL_ANY | IORING_ASYNC_CANCEL_ALL); io_uring_submit(ring); /* Wait for all CQEs */ while (pending_count > 0) { io_uring_wait_cqe(ring, &cqe); pending_count--; io_uring_cqe_seen(ring, cqe); } .EE .in .SS Synchronous cancelation For cases where the application needs to cancel requests and wait for the cancelation to complete in a single blocking call, .BR io_uring_register_sync_cancel (3) provides a synchronous interface: .PP .in +4n .EX struct io_uring_sync_cancel_reg reg = { .addr = user_data, .timeout.tv_sec = 5, }; ret = io_uring_register_sync_cancel(ring, ®); .EE .in .PP This blocks until the matching request is canceled or the timeout expires. It is useful when the application cannot easily integrate asynchronous cancelation into its event loop. .SH NOTES .IP \(bu 2 Not all operations are cancelable. Operations that have already been submitted to hardware (e.g., disk I/O in progress) typically cannot be canceled. .IP \(bu Cancelation is asynchronous. The cancel request itself may take time to complete. .IP \(bu When using .BR IORING_ASYNC_CANCEL_ALL , the cancel CQE's .I res field contains the count of canceled requests. .IP \(bu Fixed files can be canceled using .B IORING_ASYNC_CANCEL_FD_FIXED with the file index instead of a regular fd. .IP \(bu Poll operations and multishot requests are generally good candidates for cancelation. Completed disk I/O is not. .SH SEE ALSO .BR io_uring (7), .BR io_uring_linked_requests (7), .BR io_uring_multishot (7), .BR io_uring_prep_cancel (3), .BR io_uring_prep_cancel64 (3), .BR io_uring_prep_cancel_fd (3), .BR io_uring_prep_link_timeout (3), .BR io_uring_register_sync_cancel (3)