man-pages/man2/io_submit.2

287 lines
6.6 KiB
Groff

.\" Copyright (C) 2003 Free Software Foundation, Inc.
.\" and Copyright (C) 2017 Goldwyn Rodrigues <rgoldwyn@suse.de>
.\"
.\" %%%LICENSE_START(GPL_NOVERSION_ONELINE)
.\" This file is distributed according to the GNU General Public License.
.\" %%%LICENSE_END
.\"
.TH IO_SUBMIT 2 2021-03-22 "Linux" "Linux Programmer's Manual"
.SH NAME
io_submit \- submit asynchronous I/O blocks for processing
.SH SYNOPSIS
.nf
.BR "#include <linux/aio_abi.h>" " /* Defines needed types */"
.PP
.BI "int io_submit(aio_context_t " ctx_id ", long " nr \
", struct iocb **" iocbpp );
.fi
.PP
.IR Note :
There is no glibc wrapper for this system call; see NOTES.
.SH DESCRIPTION
.IR Note :
this page describes the raw Linux system call interface.
The wrapper function provided by
.I libaio
uses a different type for the
.I ctx_id
argument.
See NOTES.
.PP
The
.BR io_submit ()
system call
queues \fInr\fP I/O request blocks for processing in
the AIO context \fIctx_id\fP.
The
.I iocbpp
argument should be an array of \fInr\fP AIO control blocks,
which will be submitted to context \fIctx_id\fP.
.PP
The
.I iocb
(I/O control block) structure defined in
.IR linux/aio_abi.h
defines the parameters that control the I/O operation.
.PP
.in +4n
.EX
#include <linux/aio_abi.h>
struct iocb {
__u64 aio_data;
__u32 PADDED(aio_key, aio_rw_flags);
__u16 aio_lio_opcode;
__s16 aio_reqprio;
__u32 aio_fildes;
__u64 aio_buf;
__u64 aio_nbytes;
__s64 aio_offset;
__u64 aio_reserved2;
__u32 aio_flags;
__u32 aio_resfd;
};
.EE
.in
.PP
The fields of this structure are as follows:
.TP
.I aio_data
This data is copied into the
.I data
field of the
.I io_event
structure upon I/O completion (see
.BR io_getevents (2)).
.TP
.I aio_key
This is an internal field used by the kernel.
Do not modify this field after an
.BR io_submit ()
call.
.TP
.I aio_rw_flags
This defines the R/W flags passed with structure.
The valid values are:
.RS
.TP
.BR RWF_APPEND " (since Linux 4.16)"
.\" commit e1fc742e14e01d84d9693c4aca4ab23da65811fb
Append data to the end of the file.
See the description of the flag of the same name in
.BR pwritev2 (2)
as well as the description of
.B O_APPEND
in
.BR open (2).
The
.I aio_offset
field is ignored.
The file offset is not changed.
.TP
.BR RWF_DSYNC " (since Linux 4.13)"
Write operation complete according to requirement of
synchronized I/O data integrity.
See the description of the flag of the same name in
.BR pwritev2 (2)
as well the description of
.B O_DSYNC
in
.BR open (2).
.TP
.BR RWF_HIPRI " (since Linux 4.13)"
High priority request, poll if possible
.TP
.BR RWF_NOWAIT " (since Linux 4.14)"
Don't wait if the I/O will block for operations such as
file block allocations, dirty page flush, mutex locks,
or a congested block device inside the kernel.
If any of these conditions are met, the control block is returned
immediately with a return value of
.B \-EAGAIN
in the
.I res
field of the
.I io_event
structure (see
.BR io_getevents (2)).
.TP
.BR RWF_SYNC " (since Linux 4.13)"
Write operation complete according to requirement of
synchronized I/O file integrity.
See the description of the flag of the same name in
.BR pwritev2 (2)
as well the description of
.B O_SYNC
in
.BR open (2).
.RE
.TP
.I aio_lio_opcode
This defines the type of I/O to be performed by the
.I iocb
structure.
The
valid values are defined by the enum defined in
.IR linux/aio_abi.h :
.IP
.in +4n
.EX
enum {
IOCB_CMD_PREAD = 0,
IOCB_CMD_PWRITE = 1,
IOCB_CMD_FSYNC = 2,
IOCB_CMD_FDSYNC = 3,
IOCB_CMD_POLL = 5,
IOCB_CMD_NOOP = 6,
IOCB_CMD_PREADV = 7,
IOCB_CMD_PWRITEV = 8,
};
.EE
.in
.TP
.I aio_reqprio
This defines the requests priority.
.TP
.I aio_fildes
The file descriptor on which the I/O operation is to be performed.
.TP
.I aio_buf
This is the buffer used to transfer data for a read or write operation.
.TP
.I aio_nbytes
This is the size of the buffer pointed to by
.IR aio_buf .
.TP
.I aio_offset
This is the file offset at which the I/O operation is to be performed.
.TP
.I aio_flags
This is the set of flags associated with the
.I iocb
structure.
The valid values are:
.RS
.TP
.BR IOCB_FLAG_RESFD
Asynchronous I/O control must signal the file
descriptor mentioned in
.I aio_resfd
upon completion.
.TP
.BR IOCB_FLAG_IOPRIO " (since Linux 4.18)"
.\" commit d9a08a9e616beeccdbd0e7262b7225ffdfa49e92
Interpret the
.I aio_reqprio
field as an
.B IOPRIO_VALUE
as defined by
.IR linux/ioprio.h .
.RE
.TP
.I aio_resfd
The file descriptor to signal in the event of asynchronous I/O completion.
.SH RETURN VALUE
On success,
.BR io_submit ()
returns the number of \fIiocb\fPs submitted (which may be
less than \fInr\fP, or 0 if \fInr\fP is zero).
For the failure return, see NOTES.
.SH ERRORS
.TP
.B EAGAIN
Insufficient resources are available to queue any \fIiocb\fPs.
.TP
.B EBADF
The file descriptor specified in the first \fIiocb\fP is invalid.
.TP
.B EFAULT
One of the data structures points to invalid data.
.TP
.B EINVAL
The AIO context specified by \fIctx_id\fP is invalid.
\fInr\fP is less than 0.
The \fIiocb\fP at
.I *iocbpp[0]
is not properly initialized, the operation specified is invalid for the file
descriptor in the \fIiocb\fP, or the value in the
.I aio_reqprio
field is invalid.
.TP
.B ENOSYS
.BR io_submit ()
is not implemented on this architecture.
.TP
.B EPERM
The
.I aio_reqprio
field is set with the class
.BR IOPRIO_CLASS_RT ,
but the submitting context does not have the
.B CAP_SYS_ADMIN
capability.
.SH VERSIONS
The asynchronous I/O system calls first appeared in Linux 2.5.
.SH CONFORMING TO
.BR io_submit ()
is Linux-specific and should not be used in
programs that are intended to be portable.
.SH NOTES
Glibc does not provide a wrapper for this system call.
You could invoke it using
.BR syscall (2).
But instead, you probably want to use the
.BR io_submit ()
wrapper function provided by
.\" http://git.fedorahosted.org/git/?p=libaio.git
.IR libaio .
.PP
Note that the
.I libaio
wrapper function uses a different type
.RI ( io_context_t )
.\" But glibc is confused, since <libaio.h> uses 'io_context_t' to declare
.\" the system call.
for the
.I ctx_id
argument.
Note also that the
.I libaio
wrapper does not follow the usual C library conventions for indicating errors:
on error it returns a negated error number
(the negative of one of the values listed in ERRORS).
If the system call is invoked via
.BR syscall (2),
then the return value follows the usual conventions for
indicating an error: \-1, with
.I errno
set to a (positive) value that indicates the error.
.SH SEE ALSO
.BR io_cancel (2),
.BR io_destroy (2),
.BR io_getevents (2),
.BR io_setup (2),
.BR aio (7)
.\" .SH AUTHOR
.\" Kent Yoder.