man2/vmsplice.2

   1 .\" This manpage is Copyright (C) 2006 Jens Axboe
   2 .\" and Copyright (C) 2006 Michael Kerrisk <mtk.manpages@gmail.com>
   3 .\"
   4 .\" %%%LICENSE_START(VERBATIM)
   5 .\" Permission is granted to make and distribute verbatim copies of this
   6 .\" manual provided the copyright notice and this permission notice are
   7 .\" preserved on all copies.
   8 .\"
   9 .\" Permission is granted to copy and distribute modified versions of this
  10 .\" manual under the conditions for verbatim copying, provided that the
  11 .\" entire resulting derived work is distributed under the terms of a
  12 .\" permission notice identical to this one.
  13 .\"
  14 .\" Since the Linux kernel and libraries are constantly changing, this
  15 .\" manual page may be incorrect or out-of-date.  The author(s) assume no
  16 .\" responsibility for errors or omissions, or for damages resulting from
  17 .\" the use of the information contained herein.  The author(s) may not
  18 .\" have taken the same level of care in the production of this manual,
  19 .\" which is licensed free of charge, as they might when working
  20 .\" professionally.
  21 .\"
  22 .\" Formatted or processed versions of this manual, if unaccompanied by
  23 .\" the source, must acknowledge the copyright and authors of this work.
  24 .\" %%%LICENSE_END
  25 .\"
  26 .TH VMSPLICE 2 2014-10-02 "Linux" "Linux Programmer's Manual"
  27 .SH NAME
  28 vmsplice \- splice user pages into a pipe
  29 .SH SYNOPSIS
  30 .nf
  31 .BR "#define _GNU_SOURCE" "         /* See feature_test_macros(7) */"
  32 .B #include <fcntl.h>
  33 .B #include <sys/uio.h>
  34
  35 .BI "ssize_t vmsplice(int " fd ", const struct iovec *" iov ,
  36 .BI "                 unsigned long " nr_segs ", unsigned int " flags );
  37 .fi
  38 .\" Return type was long before glibc 2.7
  39 .SH DESCRIPTION
  40 .\" Linus: vmsplice() system call to basically do a "write to
  41 .\" the buffer", but using the reference counting and VM traversal
  42 .\" to actually fill the buffer. This means that the user needs to
  43 .\" be careful not to reuse the user-space buffer it spliced into
  44 .\" the kernel-space one (contrast this to "write()", which copies
  45 .\" the actual data, and you can thus reuse the buffer immediately
  46 .\" after a successful write), but that is often easy to do.
  47 The
  48 .BR vmsplice ()
  49 system call maps
  50 .I nr_segs
  51 ranges of user memory described by
  52 .I iov
  53 into a pipe.
  54 The file descriptor
  55 .I fd
  56 must refer to a pipe.
  57
  58 The pointer
  59 .I iov
  60 points to an array of
  61 .I iovec
  62 structures as defined in
  63 .IR <sys/uio.h> :
  64
  65 .in +4n
  66 .nf
  67 struct iovec {
  68     void  *iov_base;        /* Starting address */
  69     size_t iov_len;         /* Number of bytes */
  70 };
  71 .in
  72 .fi
  73
  74 The
  75 .I flags
  76 argument is a bit mask that is composed by ORing together
  77 zero or more of the following values:
  78 .RS
  79 .TP 1.9i
  80 .B SPLICE_F_MOVE
  81 Unused for
  82 .BR vmsplice ();
  83 see
  84 .BR splice (2).
  85 .TP
  86 .B SPLICE_F_NONBLOCK
  87 .\" Not used for vmsplice
  88 .\" May be in the future -- therefore EAGAIN
  89 Do not block on I/O; see
  90 .BR splice (2)
  91 for further details.
  92 .TP
  93 .B SPLICE_F_MORE
  94 Currently has no effect for
  95 .BR vmsplice (),
  96 but may be implemented in the future; see
  97 .BR splice (2).
  98 .TP
  99 .B SPLICE_F_GIFT
 100 The user pages are a gift to the kernel.
 101 The application may not modify this memory ever,
 102 .\" FIXME . Explain the following line in a little more detail:
 103 otherwise the page cache and on-disk data may differ.
 104 Gifting pages to the kernel means that a subsequent
 105 .BR splice (2)
 106 .B SPLICE_F_MOVE
 107 can successfully move the pages;
 108 if this flag is not specified, then a subsequent
 109 .BR splice (2)
 110 .B SPLICE_F_MOVE
 111 must copy the pages.
 112 Data must also be properly page aligned, both in memory and length.
 113 .RE
 114 .\" FIXME
 115 .\" It looks like the page-alignment requirement went away with
 116 .\" commit bd1a68b59c8e3bce45fb76632c64e1e063c3962d
 117 .\"
 118 .\" .... if we expect to later SPLICE_F_MOVE to the cache.
 119 .SH RETURN VALUE
 120 Upon successful completion,
 121 .BR vmsplice ()
 122 returns the number of bytes transferred to the pipe.
 123 On error,
 124 .BR vmsplice ()
 125 returns \-1 and
 126 .I errno
 127 is set to indicate the error.
 128 .SH ERRORS
 129 .TP
 130 .B EAGAIN
 131 .B SPLICE_F_NONBLOCK
 132 was specified in
 133 .IR flags ,
 134 and the operation would block.
 135 .TP
 136 .B EBADF
 137 .I fd
 138 either not valid, or doesn't refer to a pipe.
 139 .TP
 140 .B EINVAL
 141 .I nr_segs
 142 is greater than
 143 .BR IOV_MAX ;
 144 or memory not aligned if
 145 .B SPLICE_F_GIFT
 146 set.
 147 .TP
 148 .B ENOMEM
 149 Out of memory.
 150 .SH VERSIONS
 151 The
 152 .BR vmsplice ()
 153 system call first appeared in Linux 2.6.17;
 154 library support was added to glibc in version 2.5.
 155 .SH CONFORMING TO
 156 This system call is Linux-specific.
 157 .SH NOTES
 158 .BR vmsplice ()
 159 follows the other vectorized read/write type functions when it comes to
 160 limitations on the number of segments being passed in.
 161 This limit is
 162 .B IOV_MAX
 163 as defined in
 164 .IR <limits.h> .
 165 Currently,
 166 .\" UIO_MAXIOV in kernel source
 167 this limit is 1024.
 168 .SH SEE ALSO
 169 .BR splice (2),
 170 .BR tee (2)