path: root/man/io_uring_enter.2

.\" Copyright (C) 2019 Jens Axboe <axboe@kernel.dk>
.\" Copyright (C) 2019 Red Hat, Inc.
.\"
.\" %%%LICENSE_START(LGPL_V2.1)
.\" This file is distributed according to the GNU Lesser General Public License.
.\" %%%LICENSE_END
.\"
.TH IO_URING_ENTER 2 2019-01-22 "Linux" "Linux Programmer's Manual"
.SH NAME
io_uring_enter \- initiate and/or complete asynchronous I/O
.SH SYNOPSIS
.nf
.BR "#include <linux/io_uring.h>"
.PP
.BI "int io_uring_enter(unsigned int " fd ", unsigned int " to_submit ,
.BI "                   unsigned int " min_complete ", unsigned int " flags ,
.BI "                   sigset_t *" sig );
.fi
.PP
.SH DESCRIPTION
.PP
.BR io_uring_enter ()
is used to initiate and complete I/O using the shared submission and
completion queues setup by a call to
.BR io_uring_setup (2).
A single call can both submit new I/O and wait for completions of I/O
initiated by this call or previous calls to
.BR io_uring_enter ().

.I fd
is the file descriptor returned by
.BR io_uring_setup (2).
.I to_submit
specifies the number of I/Os to submit from the submission queue.  If
the
.B IORING_ENTER_GETEVENTS
bit is set in
.IR flags ,
then the system call will attempt to wait for
.I min_complete
event completions before returning.  If the io_uring instance was
configured for polling, by specifying
.B IORING_SETUP_IOPOLL
in the call to
.BR io_uring_setup (2),
then min_complete has a slightly different meaning.  Passing a value
of 0 instructs the kernel to return any events which are already complete,
without blocking.  If
.I min_complete
is a non-zero value, the kernel will still return immediately if any
completion events are available.  If no event completions are
available, then the call will poll either until one or more
completions become available, or until the process has exceeded its
scheduler time slice.

Note that, for interrupt driven I/O (where
.B IORING_SETUP_IOPOLL
was not specified in the call to
.BR io_uring_setup (2)),
an application may check the completion queue for event completions
without entering the kernel at all.
.PP
When the system call returns that a certain amount of SQEs have been
consumed and submitted, it's safe to reuse SQE entries in the ring. This is
true even if the actual IO submission had to be punted to async context,
which means that the SQE may in fact not have been submitted yet. If the
kernel requires later use of a particular SQE entry, it will have made a
private copy of it.

.I sig
is a pointer to a signal mask (see
.BR sigprocmask (2));
if
.I sig
is not NULL,
.BR io_uring_enter ()
first replaces the current signal mask by the one pointed to by
.IR sig ,
then waits for events to become available in the completion queue, and
then restores the original signal mask.  The following
.BR io_uring_enter ()
call:
.PP
.in +4n
.EX
ret = io_uring_enter(fd, 0, 1, IORING_ENTER_GETEVENTS, &sig);
.EE
.in
.PP
is equivalent to
.I atomically
executing the following calls:
.PP
.in +4n
.EX
pthread_sigmask(SIG_SETMASK, &sig, &orig);
ret = io_uring_enter(fd, 0, 1, IORING_ENTER_GETEVENTS, NULL);
pthread_sigmask(SIG_SETMASK, &orig, NULL);
.EE
.in
.PP
See the description of
.BR pselect (2)
for an explanation of why the
.I sig
parameter is necessary.

Submission queue entries are represented using the following data
structure:
.PP
.in +4n
.EX
/*
 * IO submission data structure (Submission Queue Entry)
 */
struct io_uring_sqe {
    __u8    opcode;         /* type of operation for this sqe */
    __u8    flags;          /* IOSQE_ flags */
    __u16   ioprio;         /* ioprio for the request */
    __s32   fd;             /* file descriptor to do IO on */
    union {
        __u64   off;            /* offset into file */
        __u64   addr2;
    };
    __u64   addr;           /* pointer to buffer or iovecs */
    __u32   len;            /* buffer size or number of iovecs */
    union {
        __kernel_rwf_t  rw_flags;
        __u32    fsync_flags;
        __u16    poll_events;
        __u32    sync_range_flags;
        __u32    msg_flags;
        __u32    timeout_flags;
        __u32    accept_flags;
        __u32    cancel_flags;
    };
    __u64    user_data;     /* data to be passed back at completion time */
    union {
        __u16    buf_index; /* index into fixed buffers, if used */
        __u64    __pad2[3];
    };
};
.EE
.in
.PP
The
.I opcode
describes the operation to be performed.  It can be one of:
.TP
.B IORING_OP_NOP
Do not perform any I/O.  This is useful for testing the performance of
the io_uring implementation itself.
.TP
.B IORING_OP_READV
.TP
.B IORING_OP_WRITEV
Vectored read and write operations, similar to
.BR preadv2 (2)
and
.BR pwritev2 (2).

.TP
.B IORING_OP_READ_FIXED
.TP
.B IORING_OP_WRITE_FIXED
Read from or write to pre-mapped buffers.  See
.BR io_uring_register (2)
for details on how to setup a context for fixed reads and writes.

.TP
.B IORING_OP_FSYNC
File sync.  See also
.BR fsync (2).
Note that, while I/O is initiated in the order in which it appears in
the submission queue, completions are unordered.  For example, an
application which places a write I/O followed by an fsync in the
submission queue cannot expect the fsync to apply to the write.  The
two operations execute in parallel, so the fsync may complete before
the write is issued to the storage.  The same is also true for
previously issued writes that have not completed prior to the fsync.

.TP
.B IORING_OP_POLL_ADD
Poll the
.I fd
specified in the submission queue entry for the events
specified in the
.I poll_events
field.  Unlike poll or epoll without
.BR EPOLLONESHOT ,
this interface always works in one shot mode.  That is, once the poll
operation is completed, it will have to be resubmitted.

.TP
.B IORING_OP_POLL_REMOVE
Remove an existing poll request.  If found, the
.I res
field of the
.I "struct io_uring_cqe"
will contain 0.  If not found,
.I res
will contain
.B -ENOENT.

.TP
.B IORING_OP_SYNC_FILE_RANGE
Issue the equivalent of a \fBsync_file_range\fR (2) on the file descriptor. The
.I fd
field is the file descriptor to sync, the
.I off
field holds the offset in bytes, the
.I len
field holds the length in bytes, and the
.I flags
field holds the flags for the command. See also
.BR sync_file_range (2).
for the general description of the related system call.

.TP
.B IORING_OP_SENDMSG
Issue the equivalent of a
.BR sendmsg(2)
system call.
.I fd
must be set to the socket file descriptor,
.I addr
must contains a pointer to the msghdr structure, and
.I flags
holds the flags associated with the system call. See also
.BR sendmsg (2).
for the general description of the related system call.

.TP
.B IORING_OP_RECVMSG
Works just like IORING_OP_SENDMSG, except for
.BR recvmsg(2)
instead. See the description of IORING_OP_SENDMSG.

.TP
.B IORING_OP_TIMEOUT
This command will register a timeout operation. The
.I addr
field must contain a pointer to a struct timespec64 structure,
.I len
must contain 1 to signify one timespec64 structure,
.I timeout_flags
may contain IORING_TIMEOUT_ABS
for an absolutel timeout value, or 0 for a relative timeout.
.I off
may contain a completion event count. If not set, this defaults to 1. A timeout
will trigger a wakeup event on the completion ring for anyone waiting for
events. A timeout condition is met when either the specified timeout expires,
or the specified number of events have completed. Either condition will
trigger the event. io_uring timeouts use the
.B CLOCK_MONOTONIC
clock source. The request will complete with
.I -ETIME
if the timeout got completed through expiration of the timer, or
.I 0
if the timeout got completed through requests completing on their own. If
the timeout was cancelled before it expired, the request will complete with
.I -ECANCELED.

.TP
.B IORING_OP_TIMEOUT_REMOVE
Attempt to remove an existing timeout operation.
.I addr
must contain the
.I user_data
field of the previously issued timeout operation. If the specified timeout
request is found and cancelled successfully, this request will terminate
with a result value of
.I 0
If the timeout request was found but expiration was already in progress,
this request will terminate with a result value of
.I -EBUSY
If the timeout request wasn't found, the request will terminate with a result
value of
.I -ENOENT

.TP
.B IORING_OP_ACCEPT
Issue the equivalent of an
.BR accept4(2)
system call.
.I fd
must be set to the socket file descriptor,
.I addr
must contains the pointer to the sockaddr structure, and
.I addr2
must contain a pointer to the socklen_t addrlen field. See also
.BR accept4(2)
for the general description of the related system call.

.TP
.B IORING_OP_ASYNC_CANCEL
Attempt to cancel an already issued request.
.I addr
must contain the
.I user_data
field of the request that should be cancelled. The cancellation request will
complete with one of the following results codes. I found, the
.I res
field of the cqe will contain 0. If not found,
.I res
will contain -ENOENT. If found and attempted cancelled, the
.I res
field will contain -EALREADY. In this case, the request may or may not
terminate. In general, requests that are interruptible (like socket IO) will
get cancelled, while disk IO requests cannot be cancelled if already started.

.TP
.B IORING_OP_LINK_TIMEOUT
This request must be linked with another request through
.I IOSQE_IO_LINK
which is described below. Unlike
.I IORING_OP_TIMEOUT,
.I IORING_OP_LINK_TIMEOUT
acts on the linked request, not the completion queue. The format of the command
is otherwise like
.I IORING_OP_TIMEOUT,
except there's no completion event count as it's tied to a specific request.
If used, the timeout specified in the command will cancel the linked command,
unless the linked command completes before the timeout. The timeout will
complete with
.I -ETIME
if the timer expired and the linked request was attempted cancelled, or
.I -ECANCELED
if the timer got cancelled because of completion of the linked request. Like
.B IORING_OP_TIMEOUT
the clock source used is
.B CLOCK_MONOTONIC


.TP
.B IORING_OP_CONNECT
Issue the equivalent of a
.BR connect(2)
system call.
.I fd
must be set to the socket file descriptor,
.I addr
must contains the pointer to the sockaddr structure, and
.I off
must contain the socklen_t addrlen field. See also
.BR connect(2)
for the general description of the related system call.

.TP
.B IORING_OP_FALLOCATE
Issue the equivalent of a
.BR fallocate(2)
system call.
.I fd
must be set to the file descriptor,
.I off
must contain the offset on which to operate, and
.I len
must the length. See also
.BR fallocate(2)
for the general description of the related system call.

.TP
.B IORING_OP_OPENAT
Issue the equivalent of a
.BR openat(2)
system call.
.I fd
is the
.I dirfd
argument,
.I addr
must contain a pointer to the
.I *pathname
argument,
.I open_flags
should contain any flags passed in, and
.I mode
is access mode of the file. See also
.BR openat(2)
for the general description of the related system call.

.TP
.B IORING_OP_CLOSE
Issue the equivalent of a
.BR close(2)
system call.
.I fd
is the file descriptor to be closed. See also
.BR close(2)
for the general description of the related system call.

.TP
.B IORING_OP_STATX
Issue the equivalent of a
.BR statx(2)
system call.
.I fd
is the
.I dirfd
argument,
.I addr
must contain a pointer to the
.I *pathname
string,
.I statx_flags
is the
.I flags
argument,
.I len
should be the
.I mask
argument, and
.I addr
must contain a pointer to the
.I statxbuf
to be filled in. See also
.BR statx(2)
for the general description of the related system call.

.TP
.B IORING_OP_READ IORING_OP_WRITE
Issue the equivalent of a
.BR read(2)
or
.BR write(2)
system call.
.I fd
is the file descriptor to be operated on,
.I addr
contains the buffer in question, and
.I len
contains the length of the IO operation. These are non-vectored versions of the
.B IORING_OP_READV
and
.B IORING_OP_WRITEV
opcodes. See also
.BR read(2)
and
.BR write(2)
for the general description of the related system call.

.TP
.B IORING_OP_FILES_UPDATE
This command is an alternative to using
.B IORING_REGISTER_FILES_UPDATE
which then works in an async fashion, like the rest of the io_uring commands.
The arguments passed in are the same.
.I addr
must contain a pointer to the array of file descriptors,
.I len
must contain the length of the array, and
.I off
must contain the offset at which to operate. Note that the array of file
descriptors pointed to in
.I addr
must remain valid until this operation has completed.

.PP
The
.I flags
field is a bit mask. The supported flags are:
.TP
.B IOSQE_FIXED_FILE
When this flag is specified,
.I fd
is an index into the files array registered with the io_uring instance (see the
.B IORING_REGISTER_FILES
section of the
.BR io_uring_register (2)
man page).
.TP
.B IOSQE_IO_DRAIN
When this flag is specified, the SQE will not be started before previously
submitted SQEs have completed, and new SQEs will not be started before this
one completes.
.TP
.B IOSQE_IO_LINK
When this flag is specified, it forms a link with the next SQE in the
submission ring. That next SQE will not be started before this one completes.
This, in effect, forms a chain of SQEs, which can be arbitrarily long. The tail
of the chain is denoted by the first SQE that does not have this flag set.
This flag has no effect on previous SQE submissions, nor does it impact SQEs
that are outside of the chain tail. This means that multiple chains can be
executing in parallel, or chains and individual SQEs. Only members inside the
chain are serialized.
.TP
.B IOSQE_IO_HARDLINK
Like IOSQE_IO_LINK, but it doesn't sever regardless of the completion result.
Note that the link will still sever if we fail submitting the parent request,
hard links are only resilient in the presence of completion results for
requests that did submit correctly. IOSQE_IO_HARDLINK implies IOSQE_IO_LINK.

.PP
.I ioprio
specifies the I/O priority.  See
.BR ioprio_get (2)
for a description of Linux I/O priorities.

.I fd
specifies the file descriptor against which the operation will be
performed, with the exception noted above.

If the operation is one of
.B IORING_OP_READ_FIXED
or
.BR IORING_OP_WRITE_FIXED ,
.I addr
and
.I len
must fall within the buffer located at
.I buf_index
in the fixed buffer array.  If the operation is either
.B IORING_OP_READV
or
.BR IORING_OP_WRITEV ,
then
.I addr
points to an iovec array of
.I len
entries.

.IR rw_flags ,
specified for read and write operations, contains a bitwise OR of
per-I/O flags, as described in the
.BR preadv2 (2)
man page.

The
.I fsync_flags
bit mask may contain either 0, for a normal file integrity sync, or
.B IORING_FSYNC_DATASYNC
to provide data sync only semantics.  See the descriptions of
.B O_SYNC
and
.B O_DSYNC
in the
.BR open (2)
manual page for more information.

The bits that may be set in
.I poll_events
are defined in \fI<poll.h>\fP, and documented in
.BR poll (2).

.I user_data
is an application-supplied value that will be copied into
the completion queue entry (see below).
.I buf_index
is an index into an array of fixed buffers, and is only valid if fixed
buffers were registered
.PP
Once the submission queue entry is initialized, I/O is submitted by
placing the index of the submission queue entry into the tail of the
submission queue.  After one or more indexes are added to the queue,
and the queue tail is advanced, the
.BR io_uring_enter (2)
system call can be invoked to initiate the I/O.

Completions use the following data structure:
.PP
.in +4n
.EX
/*
 * IO completion data structure (Completion Queue Entry)
 */
struct io_uring_cqe {
    __u64    user_data; /* sqe->data submission passed back */
    __s32    res;       /* result code for this event */
    __u32    flags;
};
.EE
.in
.PP
.I user_data
is copied from the field of the same name in the submission queue
entry.  The primary use case is to store data that the application
will need to access upon completion of this particular I/O.  The
.I flags
is reserved for future use.
.I res
is the operation-specific result.
.PP
For read and write opcodes, the
return values match those documented in the
.BR preadv2 (2)
and
.BR pwritev2 (2)
man pages.
Return codes for the io_uring-specific opcodes are documented in the
description of the opcodes above.
.PP
.SH RETURN VALUE
.BR io_uring_enter ()
returns the number of I/Os successfully consumed.  This can be zero
if
.I to_submit
was zero or if the submission queue was empty. The errors below that refer to
an error in a submission queue entry will be returned though a completion queue
entry, rather than through the system call itself.

Errors that occur not on behalf of a submission queue entry are returned via the
system call directly. On such an error, -1 is returned and
.I errno
is set appropriately.
.PP
.SH ERRORS
.TP
.B EAGAIN
The kernel was unable to allocate memory for the request, or otherwise ran out
of resources to handle it. The application should wait for some completions and
try again.
.TP
.B EBUSY
The application is attempting to overcommit the number of requests it can have
pending. The application should wait for some completions and try again. May
occur if the application tries to queue more requests than we have room for in
the CQ ring.
.TP
.B EBADF
The
.I fd
field in the submission queue entry is invalid, or the
.B IOSQE_FIXED_FILE
flag was set in the submission queue entry, but no files were registered
with the io_uring instance.
.TP
.B EFAULT
buffer is outside of the process' accessible address space
.TP
.B EFAULT
.B IORING_OP_READ_FIXED
or
.B IORING_OP_WRITE_FIXED
was specified in the
.I opcode
field of the submission queue entry, but either buffers were not
registered for this io_uring instance, or the address range described
by
.I addr
and
.I len
does not fit within the buffer registered at
.IR buf_index .
.TP
.B EINVAL
The
.I index
member of the submission queue entry is invalid.
.TP
.B EINVAL
The
.I flags
field or
.I opcode
in a submission queue entry is invalid.
.TP
.B EINVAL
.B IORING_OP_NOP
was specified in the submission queue entry, but the io_uring context
was setup for polling
.RB ( IORING_SETUP_IOPOLL
was specified in the call to io_uring_setup).
.TP
.B EINVAL
.B IORING_OP_READV
or
.B IORING_OP_WRITEV
was specified in the submission queue entry, but the io_uring instance
has fixed buffers registered.
.TP
.B EINVAL
.B IORING_OP_READ_FIXED
or
.B IORING_OP_WRITE_FIXED
was specified in the submission queue entry, and the
.I buf_index
is invalid.
.TP
.B EINVAL
.BR IORING_OP_READV ,
.BR IORING_OP_WRITEV ,
.BR IORING_OP_READ_FIXED ,
.B IORING_OP_WRITE_FIXED
or
.B IORING_OP_FSYNC
was specified in the submission queue entry, but the io_uring instance
was configured for IOPOLLing, or any of
.IR addr ,
.IR ioprio ,
.IR off ,
.IR len ,
or
.I buf_index
was set in the submission queue entry.
.TP
.B EINVAL
.B IORING_OP_POLL_ADD
or
.B IORING_OP_POLL_REMOVE
was specified in the
.I opcode
field of the submission queue entry, but the io_uring instance was
configured for busy-wait polling
.RB ( IORING_SETUP_IOPOLL ),
or any of
.IR ioprio ,
.IR off ,
.IR len ,
or
.I buf_index
was non-zero in the submission queue entry.
.TP
.B EINVAL
.B IORING_OP_POLL_ADD
was specified in the
.I opcode
field of the submission queue entry, and the
.I addr
field was non-zero.
.TP
.B ENXIO
The io_uring instance is in the process of being torn down.
.TP
.B EOPNOTSUPP
.I fd
does not refer to an io_uring instance.
.TP
.B EOPNOTSUPP
.I opcode
is valid, but not supported by this kernel.