/
githubmirror
/
liburing
Обзор
Документация
Войти
/
githubmirror
/
liburing
Код
Запросы
0
Пакеты
0
Релизы
0
Аналитика
Безопасность
master
man/io_uring_multishot.7
246 строк
7 KB
Jens Axboe
man: add io_uring_multishot.7 overview page
19 янв 2026, 06:24
19 янв 2026, 06:24
9b7740b
Код
Авторство
О чём код?
.\" Copyright (C) 2025 Jens Axboe <axboe@kernel.dk> .\" SPDX-License-Identifier: LGPL-2.0-or-later .\" .TH io_uring_multishot 7 "January 18, 2025" "Linux" "Linux Programmer's Manual" .SH NAME io_uring_multishot \- io_uring multishot requests overview .SH DESCRIPTION Multishot requests are a class of io_uring operations where a single submission queue entry (SQE) can generate multiple completion queue entries (CQEs). This is in contrast to normal "oneshot" operations where each SQE produces exactly one CQE. .SS Why use multishot requests? Traditional I/O operations require submitting a new request for each operation. For high-frequency operations like accepting connections or receiving data, this creates overhead: .IP \(bu 2 CPU cycles spent preparing and submitting SQEs .IP \(bu Memory bandwidth for SQE/CQE processing .IP \(bu Potential for gaps between completions and new submissions .PP Multishot requests eliminate this overhead by keeping the operation active after each completion. The kernel automatically re-arms the operation, generating a new CQE when the next event occurs. Additionally, the internal poll mechanism remains persistent for the request, avoiding the need to manipulate poll state for each operation. Multishot operations are most beneficial for: .IP \(bu 2 Network servers accepting many connections .IP \(bu Applications receiving data on long-lived connections .IP \(bu Event monitoring with poll .IP \(bu Any scenario with repeated identical operations .SS How multishot works When a multishot operation completes, the CQE has the .B IORING_CQE_F_MORE flag set in .IR cqe->flags . This indicates that the operation remains active and more completions will follow. The operation continues until: .IP \(bu 2 An error occurs (the final CQE will not have .B IORING_CQE_F_MORE set) .IP \(bu The operation is explicitly canceled .IP \(bu A termination condition specific to the operation is met (e.g., buffer exhaustion for receives) .PP The final CQE for a multishot operation will not have .B IORING_CQE_F_MORE set, indicating the operation has terminated. .SS Multishot accept .BR io_uring_prep_multishot_accept (3) and .BR io_uring_prep_multishot_accept_direct (3) set up a multishot accept operation. Each incoming connection generates a CQE with the new file descriptor in .IR cqe->res . .PP .in +4n .EX struct io_uring_sqe *sqe = io_uring_get_sqe(ring); io_uring_prep_multishot_accept(sqe, listen_fd, NULL, NULL, 0); .EE .in .PP The operation continues accepting connections until an error occurs or it is canceled. Using the direct variant with .B IORING_FILE_INDEX_ALLOC allows accepted sockets to be placed directly into the fixed file table. .SS Multishot receive .BR io_uring_prep_recv_multishot (3) sets up a multishot receive operation. Each time data arrives on the socket, a CQE is generated. This is typically used with provided buffers (see .BR io_uring_provided_buffers (7)): .PP .in +4n .EX struct io_uring_sqe *sqe = io_uring_get_sqe(ring); io_uring_prep_recv_multishot(sqe, sockfd, NULL, 0, 0); sqe->buf_group = bgid; sqe->flags |= IOSQE_BUFFER_SELECT; .EE .in .PP Each completion includes: .IP \(bu 2 .B IORING_CQE_F_MORE if more completions will follow .IP \(bu .B IORING_CQE_F_BUFFER indicating a buffer was selected .IP \(bu The buffer ID in the upper bits of .I cqe->flags .IP \(bu The number of bytes received in .I cqe->res .PP The multishot receive terminates when an error occurs, the connection closes, or the buffer ring is exhausted. .SS Multishot recvmsg .BR io_uring_prep_recvmsg_multishot (3) is similar to multishot receive but uses the .I msghdr structure for scatter/gather I/O and ancillary data. A provided buffer is used for each message, with the kernel writing a .I struct io_uring_recvmsg_out header at the start of the buffer containing the actual message parameters. .SS Multishot read .BR io_uring_prep_read_multishot (3) sets up a multishot read operation, typically used with pipes or other stream-oriented file descriptors. Like multishot receive, this is used with provided buffers: .PP .in +4n .EX struct io_uring_sqe *sqe = io_uring_get_sqe(ring); io_uring_prep_read_multishot(sqe, fd, 0, 0, bgid); .EE .in .PP The operation generates a CQE each time data becomes available to read. .SS Multishot poll .BR io_uring_prep_poll_multishot (3) sets up a multishot poll operation, or it can be done manually by setting the .B IORING_POLL_ADD_MULTI flag: .PP .in +4n .EX struct io_uring_sqe *sqe = io_uring_get_sqe(ring); io_uring_prep_poll_multishot(sqe, fd, POLLIN); /* or equivalently: */ io_uring_prep_poll_add(sqe, fd, POLLIN); sqe->len |= IORING_POLL_ADD_MULTI; .EE .in .PP Each time the polled condition becomes true, a CQE is generated with the triggered events in .IR cqe->res . Unlike oneshot poll which is automatically removed after triggering, multishot poll remains active. For level-triggered events, the application should be careful to handle the event (e.g., read all available data) before the next poll completion, or spurious wakeups may occur. .SS Multishot waitid .BR io_uring_prep_waitid (3) can operate in multishot mode by setting .B IORING_ACCEPT_MULTISHOT in the flags. This allows waiting for multiple child process state changes with a single SQE. .SS Handling multishot completions Applications must check for .B IORING_CQE_F_MORE to determine if the operation is still active: .PP .in +4n .EX struct io_uring_cqe *cqe; while (io_uring_peek_cqe(ring, &cqe) == 0) { if (cqe->res < 0) { /* Error occurred, operation terminated */ handle_error(cqe->res); } else { process_completion(cqe); } if (!(cqe->flags & IORING_CQE_F_MORE)) { /* Operation terminated, may need to resubmit */ rearm_if_needed(); } io_uring_cqe_seen(ring, cqe); } .EE .in .SS Canceling multishot operations Multishot operations can be canceled using .BR io_uring_prep_cancel (3) or related functions. The cancellation request generates its own CQE, and the multishot operation generates a final CQE (typically with .BR -ECANCELED ) without .B IORING_CQE_F_MORE set. .PP .in +4n .EX struct io_uring_sqe *sqe = io_uring_get_sqe(ring); io_uring_prep_cancel64(sqe, user_data, 0); .EE .in .SS Integration with provided buffers Multishot receive and read operations are designed to work with provided buffer rings (see .BR io_uring_provided_buffers (7)). Each completion consumes a buffer from the ring, and the application must return buffers to the ring to keep the operation running. If the buffer ring becomes empty, the multishot operation terminates with .BR -ENOBUFS . Applications should ensure adequate buffers are available and promptly return used buffers to the ring. .SH NOTES .IP \(bu 2 Always check .B IORING_CQE_F_MORE to know if a multishot operation is still active. .IP \(bu Multishot operations may generate many CQEs quickly. Ensure the CQ ring is large enough to avoid overflow. .IP \(bu When using provided buffers with multishot receives, monitor buffer availability to prevent premature termination. .IP \(bu Multishot operations are edge-triggered conceptually \(em they generate completions when events occur, not continuously while conditions are true. .IP \(bu Error completions from multishot operations do not have .B IORING_CQE_F_MORE set, indicating termination. .SH SEE ALSO .BR io_uring (7), .BR io_uring_provided_buffers (7), .BR io_uring_prep_multishot_accept (3), .BR io_uring_prep_recv_multishot (3), .BR io_uring_prep_recvmsg_multishot (3), .BR io_uring_prep_read_multishot (3), .BR io_uring_prep_poll_add (3), .BR io_uring_prep_poll_multishot (3), .BR io_uring_prep_cancel (3)