Internal change
PiperOrigin-RevId: 517447696
Change-Id: Ib010c584202dcc3bcf71404e4ed6cb4c05adac27
diff --git a/ChangeLog b/ChangeLog
new file mode 100644
index 0000000..b1536b4
--- /dev/null
+++ b/ChangeLog
@@ -0,0 +1,54 @@
+0.3.107
+ - Make tests compile again on modern systems (warnings + -Werror)
+ - Add 'make partcheck' and don't require manual setup for testing.
+ - Change test harness to compile against this dir, not global install
+ - Fix 5.t for archs where PROT_WRITE mappings are readable.
+ - Allow sending of SIGXFSZ on aio over limits
+ - Explicitly specify bash for runtests.sh
+ - Put deprecating comments on never-merged io_prep_poll
+ - Add io_prep_preadv and io_prep_pwritev
+ - Add eventfd support (io_set_eventfd).
+
+0.4.0
+ - remove libredhat-kernel
+ - add rough outline for man pages
+ - make the compiled io_getevents() add the extra parameter and
+ pass the timeout for updating as per 2.5
+ - fixes for ia64, now works
+ - fixes for x86-64
+ - powerpc support from Gianni Tedesco <gianni@ecsc.co.uk>
+ - disable the NULL check in harness/cases/4.t on ia64: ia64
+ maps the 0 page and causes this check to fail.
+
+0.3.15
+ - use real syscall interface, but don't break source compatibility
+ yet (that will happen with 0.4.0)
+
+0.3.13
+ - add test cases
+
+0.3.11
+ - use library versioning of libredhat-kernel to always provide a
+ fallback
+
+0.3.9
+ - add io_queue_release function
+
+0.3.8
+ - make clean deletes libredhat-kernel.so.1
+ - const struct timespec *
+ - add make srpm target
+
+0.3.7
+ - fix assembly function .types
+ - export io_getevents
+ - fix io_submit function prototype to match the kernel
+ - provide /usr/lib/libredhat-kernel.so link for compilation
+ (do NOT link against libredhat-kernel.so directly)
+ - fix soname to libaio.so.1
+ - fix dummy libredhat-kernel's soname
+ - work around nfs bug
+ - provide and install libredhat-kernel.so.1 stub
+ - Makefile improvements
+ - make sure dummy libredhat-kernel.so only returns -ENOSYS
+
diff --git a/INSTALL b/INSTALL
new file mode 100644
index 0000000..29b9077
--- /dev/null
+++ b/INSTALL
@@ -0,0 +1,18 @@
+To install the library, execute the command:
+
+ make prefix=`pwd`/usr install
+
+which will install the binaries and header files into the directory
+usr. Set prefix=/usr to get them installed into the main system.
+
+Please note: Do not attempt to install on the system the
+"libredhat-kernel.so" file. It is a dummy shared library
+provided only for the purpose of being able to bootstrap
+this facility while running on systems without the correct
+libredhat-kernel.so built. The contents of the included
+libredhat-kernel.so are only stubs; this library is NOT
+functional for anything except the internal purpose of
+linking libaio.so against the provided stubs. At runtime,
+libaio.so requires a real libredhat-kernel.so library; this
+is provided by the Red Hat kernel RPM packages with async
+I/O functionality.
diff --git a/LICENSE b/LICENSE
new file mode 100644
index 0000000..c4792dd
--- /dev/null
+++ b/LICENSE
@@ -0,0 +1,515 @@
+
+ GNU LESSER GENERAL PUBLIC LICENSE
+ Version 2.1, February 1999
+
+ Copyright (C) 1991, 1999 Free Software Foundation, Inc.
+ 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ Everyone is permitted to copy and distribute verbatim copies
+ of this license document, but changing it is not allowed.
+
+[This is the first released version of the Lesser GPL. It also counts
+ as the successor of the GNU Library Public License, version 2, hence
+ the version number 2.1.]
+
+ Preamble
+
+ The licenses for most software are designed to take away your
+freedom to share and change it. By contrast, the GNU General Public
+Licenses are intended to guarantee your freedom to share and change
+free software--to make sure the software is free for all its users.
+
+ This license, the Lesser General Public License, applies to some
+specially designated software packages--typically libraries--of the
+Free Software Foundation and other authors who decide to use it. You
+can use it too, but we suggest you first think carefully about whether
+this license or the ordinary General Public License is the better
+strategy to use in any particular case, based on the explanations
+below.
+
+ When we speak of free software, we are referring to freedom of use,
+not price. Our General Public Licenses are designed to make sure that
+you have the freedom to distribute copies of free software (and charge
+for this service if you wish); that you receive source code or can get
+it if you want it; that you can change the software and use pieces of
+it in new free programs; and that you are informed that you can do
+these things.
+
+ To protect your rights, we need to make restrictions that forbid
+distributors to deny you these rights or to ask you to surrender these
+rights. These restrictions translate to certain responsibilities for
+you if you distribute copies of the library or if you modify it.
+
+ For example, if you distribute copies of the library, whether gratis
+or for a fee, you must give the recipients all the rights that we gave
+you. You must make sure that they, too, receive or can get the source
+code. If you link other code with the library, you must provide
+complete object files to the recipients, so that they can relink them
+with the library after making changes to the library and recompiling
+it. And you must show them these terms so they know their rights.
+
+ We protect your rights with a two-step method: (1) we copyright the
+library, and (2) we offer you this license, which gives you legal
+permission to copy, distribute and/or modify the library.
+
+ To protect each distributor, we want to make it very clear that
+there is no warranty for the free library. Also, if the library is
+modified by someone else and passed on, the recipients should know
+that what they have is not the original version, so that the original
+author's reputation will not be affected by problems that might be
+introduced by others.
+^L
+ Finally, software patents pose a constant threat to the existence of
+any free program. We wish to make sure that a company cannot
+effectively restrict the users of a free program by obtaining a
+restrictive license from a patent holder. Therefore, we insist that
+any patent license obtained for a version of the library must be
+consistent with the full freedom of use specified in this license.
+
+ Most GNU software, including some libraries, is covered by the
+ordinary GNU General Public License. This license, the GNU Lesser
+General Public License, applies to certain designated libraries, and
+is quite different from the ordinary General Public License. We use
+this license for certain libraries in order to permit linking those
+libraries into non-free programs.
+
+ When a program is linked with a library, whether statically or using
+a shared library, the combination of the two is legally speaking a
+combined work, a derivative of the original library. The ordinary
+General Public License therefore permits such linking only if the
+entire combination fits its criteria of freedom. The Lesser General
+Public License permits more lax criteria for linking other code with
+the library.
+
+ We call this license the "Lesser" General Public License because it
+does Less to protect the user's freedom than the ordinary General
+Public License. It also provides other free software developers Less
+of an advantage over competing non-free programs. These disadvantages
+are the reason we use the ordinary General Public License for many
+libraries. However, the Lesser license provides advantages in certain
+special circumstances.
+
+ For example, on rare occasions, there may be a special need to
+encourage the widest possible use of a certain library, so that it
+becomes
+a de-facto standard. To achieve this, non-free programs must be
+allowed to use the library. A more frequent case is that a free
+library does the same job as widely used non-free libraries. In this
+case, there is little to gain by limiting the free library to free
+software only, so we use the Lesser General Public License.
+
+ In other cases, permission to use a particular library in non-free
+programs enables a greater number of people to use a large body of
+free software. For example, permission to use the GNU C Library in
+non-free programs enables many more people to use the whole GNU
+operating system, as well as its variant, the GNU/Linux operating
+system.
+
+ Although the Lesser General Public License is Less protective of the
+users' freedom, it does ensure that the user of a program that is
+linked with the Library has the freedom and the wherewithal to run
+that program using a modified version of the Library.
+
+ The precise terms and conditions for copying, distribution and
+modification follow. Pay close attention to the difference between a
+"work based on the library" and a "work that uses the library". The
+former contains code derived from the library, whereas the latter must
+be combined with the library in order to run.
+^L
+ GNU LESSER GENERAL PUBLIC LICENSE
+ TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
+
+ 0. This License Agreement applies to any software library or other
+program which contains a notice placed by the copyright holder or
+other authorized party saying it may be distributed under the terms of
+this Lesser General Public License (also called "this License").
+Each licensee is addressed as "you".
+
+ A "library" means a collection of software functions and/or data
+prepared so as to be conveniently linked with application programs
+(which use some of those functions and data) to form executables.
+
+ The "Library", below, refers to any such software library or work
+which has been distributed under these terms. A "work based on the
+Library" means either the Library or any derivative work under
+copyright law: that is to say, a work containing the Library or a
+portion of it, either verbatim or with modifications and/or translated
+straightforwardly into another language. (Hereinafter, translation is
+included without limitation in the term "modification".)
+
+ "Source code" for a work means the preferred form of the work for
+making modifications to it. For a library, complete source code means
+all the source code for all modules it contains, plus any associated
+interface definition files, plus the scripts used to control
+compilation
+and installation of the library.
+
+ Activities other than copying, distribution and modification are not
+covered by this License; they are outside its scope. The act of
+running a program using the Library is not restricted, and output from
+such a program is covered only if its contents constitute a work based
+on the Library (independent of the use of the Library in a tool for
+writing it). Whether that is true depends on what the Library does
+and what the program that uses the Library does.
+
+ 1. You may copy and distribute verbatim copies of the Library's
+complete source code as you receive it, in any medium, provided that
+you conspicuously and appropriately publish on each copy an
+appropriate copyright notice and disclaimer of warranty; keep intact
+all the notices that refer to this License and to the absence of any
+warranty; and distribute a copy of this License along with the
+Library.
+
+ You may charge a fee for the physical act of transferring a copy,
+and you may at your option offer warranty protection in exchange for a
+fee.
+
+ 2. You may modify your copy or copies of the Library or any portion
+of it, thus forming a work based on the Library, and copy and
+distribute such modifications or work under the terms of Section 1
+above, provided that you also meet all of these conditions:
+
+ a) The modified work must itself be a software library.
+
+ b) You must cause the files modified to carry prominent notices
+ stating that you changed the files and the date of any change.
+
+ c) You must cause the whole of the work to be licensed at no
+ charge to all third parties under the terms of this License.
+
+ d) If a facility in the modified Library refers to a function or a
+ table of data to be supplied by an application program that uses
+ the facility, other than as an argument passed when the facility
+ is invoked, then you must make a good faith effort to ensure that,
+ in the event an application does not supply such function or
+ table, the facility still operates, and performs whatever part of
+ its purpose remains meaningful.
+
+ (For example, a function in a library to compute square roots has
+ a purpose that is entirely well-defined independent of the
+ application. Therefore, Subsection 2d requires that any
+ application-supplied function or table used by this function must
+ be optional: if the application does not supply it, the square
+ root function must still compute square roots.)
+
+These requirements apply to the modified work as a whole. If
+identifiable sections of that work are not derived from the Library,
+and can be reasonably considered independent and separate works in
+themselves, then this License, and its terms, do not apply to those
+sections when you distribute them as separate works. But when you
+distribute the same sections as part of a whole which is a work based
+on the Library, the distribution of the whole must be on the terms of
+this License, whose permissions for other licensees extend to the
+entire whole, and thus to each and every part regardless of who wrote
+it.
+
+Thus, it is not the intent of this section to claim rights or contest
+your rights to work written entirely by you; rather, the intent is to
+exercise the right to control the distribution of derivative or
+collective works based on the Library.
+
+In addition, mere aggregation of another work not based on the Library
+with the Library (or with a work based on the Library) on a volume of
+a storage or distribution medium does not bring the other work under
+the scope of this License.
+
+ 3. You may opt to apply the terms of the ordinary GNU General Public
+License instead of this License to a given copy of the Library. To do
+this, you must alter all the notices that refer to this License, so
+that they refer to the ordinary GNU General Public License, version 2,
+instead of to this License. (If a newer version than version 2 of the
+ordinary GNU General Public License has appeared, then you can specify
+that version instead if you wish.) Do not make any other change in
+these notices.
+^L
+ Once this change is made in a given copy, it is irreversible for
+that copy, so the ordinary GNU General Public License applies to all
+subsequent copies and derivative works made from that copy.
+
+ This option is useful when you wish to copy part of the code of
+the Library into a program that is not a library.
+
+ 4. You may copy and distribute the Library (or a portion or
+derivative of it, under Section 2) in object code or executable form
+under the terms of Sections 1 and 2 above provided that you accompany
+it with the complete corresponding machine-readable source code, which
+must be distributed under the terms of Sections 1 and 2 above on a
+medium customarily used for software interchange.
+
+ If distribution of object code is made by offering access to copy
+from a designated place, then offering equivalent access to copy the
+source code from the same place satisfies the requirement to
+distribute the source code, even though third parties are not
+compelled to copy the source along with the object code.
+
+ 5. A program that contains no derivative of any portion of the
+Library, but is designed to work with the Library by being compiled or
+linked with it, is called a "work that uses the Library". Such a
+work, in isolation, is not a derivative work of the Library, and
+therefore falls outside the scope of this License.
+
+ However, linking a "work that uses the Library" with the Library
+creates an executable that is a derivative of the Library (because it
+contains portions of the Library), rather than a "work that uses the
+library". The executable is therefore covered by this License.
+Section 6 states terms for distribution of such executables.
+
+ When a "work that uses the Library" uses material from a header file
+that is part of the Library, the object code for the work may be a
+derivative work of the Library even though the source code is not.
+Whether this is true is especially significant if the work can be
+linked without the Library, or if the work is itself a library. The
+threshold for this to be true is not precisely defined by law.
+
+ If such an object file uses only numerical parameters, data
+structure layouts and accessors, and small macros and small inline
+functions (ten lines or less in length), then the use of the object
+file is unrestricted, regardless of whether it is legally a derivative
+work. (Executables containing this object code plus portions of the
+Library will still fall under Section 6.)
+
+ Otherwise, if the work is a derivative of the Library, you may
+distribute the object code for the work under the terms of Section 6.
+Any executables containing that work also fall under Section 6,
+whether or not they are linked directly with the Library itself.
+^L
+ 6. As an exception to the Sections above, you may also combine or
+link a "work that uses the Library" with the Library to produce a
+work containing portions of the Library, and distribute that work
+under terms of your choice, provided that the terms permit
+modification of the work for the customer's own use and reverse
+engineering for debugging such modifications.
+
+ You must give prominent notice with each copy of the work that the
+Library is used in it and that the Library and its use are covered by
+this License. You must supply a copy of this License. If the work
+during execution displays copyright notices, you must include the
+copyright notice for the Library among them, as well as a reference
+directing the user to the copy of this License. Also, you must do one
+of these things:
+
+ a) Accompany the work with the complete corresponding
+ machine-readable source code for the Library including whatever
+ changes were used in the work (which must be distributed under
+ Sections 1 and 2 above); and, if the work is an executable linked
+ with the Library, with the complete machine-readable "work that
+ uses the Library", as object code and/or source code, so that the
+ user can modify the Library and then relink to produce a modified
+ executable containing the modified Library. (It is understood
+ that the user who changes the contents of definitions files in the
+ Library will not necessarily be able to recompile the application
+ to use the modified definitions.)
+
+ b) Use a suitable shared library mechanism for linking with the
+ Library. A suitable mechanism is one that (1) uses at run time a
+ copy of the library already present on the user's computer system,
+ rather than copying library functions into the executable, and (2)
+ will operate properly with a modified version of the library, if
+ the user installs one, as long as the modified version is
+ interface-compatible with the version that the work was made with.
+
+ c) Accompany the work with a written offer, valid for at
+ least three years, to give the same user the materials
+ specified in Subsection 6a, above, for a charge no more
+ than the cost of performing this distribution.
+
+ d) If distribution of the work is made by offering access to copy
+ from a designated place, offer equivalent access to copy the above
+ specified materials from the same place.
+
+ e) Verify that the user has already received a copy of these
+ materials or that you have already sent this user a copy.
+
+ For an executable, the required form of the "work that uses the
+Library" must include any data and utility programs needed for
+reproducing the executable from it. However, as a special exception,
+the materials to be distributed need not include anything that is
+normally distributed (in either source or binary form) with the major
+components (compiler, kernel, and so on) of the operating system on
+which the executable runs, unless that component itself accompanies
+the executable.
+
+ It may happen that this requirement contradicts the license
+restrictions of other proprietary libraries that do not normally
+accompany the operating system. Such a contradiction means you cannot
+use both them and the Library together in an executable that you
+distribute.
+^L
+ 7. You may place library facilities that are a work based on the
+Library side-by-side in a single library together with other library
+facilities not covered by this License, and distribute such a combined
+library, provided that the separate distribution of the work based on
+the Library and of the other library facilities is otherwise
+permitted, and provided that you do these two things:
+
+ a) Accompany the combined library with a copy of the same work
+ based on the Library, uncombined with any other library
+ facilities. This must be distributed under the terms of the
+ Sections above.
+
+ b) Give prominent notice with the combined library of the fact
+ that part of it is a work based on the Library, and explaining
+ where to find the accompanying uncombined form of the same work.
+
+ 8. You may not copy, modify, sublicense, link with, or distribute
+the Library except as expressly provided under this License. Any
+attempt otherwise to copy, modify, sublicense, link with, or
+distribute the Library is void, and will automatically terminate your
+rights under this License. However, parties who have received copies,
+or rights, from you under this License will not have their licenses
+terminated so long as such parties remain in full compliance.
+
+ 9. You are not required to accept this License, since you have not
+signed it. However, nothing else grants you permission to modify or
+distribute the Library or its derivative works. These actions are
+prohibited by law if you do not accept this License. Therefore, by
+modifying or distributing the Library (or any work based on the
+Library), you indicate your acceptance of this License to do so, and
+all its terms and conditions for copying, distributing or modifying
+the Library or works based on it.
+
+ 10. Each time you redistribute the Library (or any work based on the
+Library), the recipient automatically receives a license from the
+original licensor to copy, distribute, link with or modify the Library
+subject to these terms and conditions. You may not impose any further
+restrictions on the recipients' exercise of the rights granted herein.
+You are not responsible for enforcing compliance by third parties with
+this License.
+^L
+ 11. If, as a consequence of a court judgment or allegation of patent
+infringement or for any other reason (not limited to patent issues),
+conditions are imposed on you (whether by court order, agreement or
+otherwise) that contradict the conditions of this License, they do not
+excuse you from the conditions of this License. If you cannot
+distribute so as to satisfy simultaneously your obligations under this
+License and any other pertinent obligations, then as a consequence you
+may not distribute the Library at all. For example, if a patent
+license would not permit royalty-free redistribution of the Library by
+all those who receive copies directly or indirectly through you, then
+the only way you could satisfy both it and this License would be to
+refrain entirely from distribution of the Library.
+
+If any portion of this section is held invalid or unenforceable under
+any particular circumstance, the balance of the section is intended to
+apply, and the section as a whole is intended to apply in other
+circumstances.
+
+It is not the purpose of this section to induce you to infringe any
+patents or other property right claims or to contest validity of any
+such claims; this section has the sole purpose of protecting the
+integrity of the free software distribution system which is
+implemented by public license practices. Many people have made
+generous contributions to the wide range of software distributed
+through that system in reliance on consistent application of that
+system; it is up to the author/donor to decide if he or she is willing
+to distribute software through any other system and a licensee cannot
+impose that choice.
+
+This section is intended to make thoroughly clear what is believed to
+be a consequence of the rest of this License.
+
+ 12. If the distribution and/or use of the Library is restricted in
+certain countries either by patents or by copyrighted interfaces, the
+original copyright holder who places the Library under this License
+may add an explicit geographical distribution limitation excluding those
+countries, so that distribution is permitted only in or among
+countries not thus excluded. In such case, this License incorporates
+the limitation as if written in the body of this License.
+
+ 13. The Free Software Foundation may publish revised and/or new
+versions of the Lesser General Public License from time to time.
+Such new versions will be similar in spirit to the present version,
+but may differ in detail to address new problems or concerns.
+
+Each version is given a distinguishing version number. If the Library
+specifies a version number of this License which applies to it and
+"any later version", you have the option of following the terms and
+conditions either of that version or of any later version published by
+the Free Software Foundation. If the Library does not specify a
+license version number, you may choose any version ever published by
+the Free Software Foundation.
+^L
+ 14. If you wish to incorporate parts of the Library into other free
+programs whose distribution conditions are incompatible with these,
+write to the author to ask for permission. For software which is
+copyrighted by the Free Software Foundation, write to the Free
+Software Foundation; we sometimes make exceptions for this. Our
+decision will be guided by the two goals of preserving the free status
+of all derivatives of our free software and of promoting the sharing
+and reuse of software generally.
+
+ NO WARRANTY
+
+ 15. BECAUSE THE LIBRARY IS LICENSED FREE OF CHARGE, THERE IS NO
+WARRANTY FOR THE LIBRARY, TO THE EXTENT PERMITTED BY APPLICABLE LAW.
+EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR
+OTHER PARTIES PROVIDE THE LIBRARY "AS IS" WITHOUT WARRANTY OF ANY
+KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE
+IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE
+LIBRARY IS WITH YOU. SHOULD THE LIBRARY PROVE DEFECTIVE, YOU ASSUME
+THE COST OF ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
+
+ 16. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN
+WRITING WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY
+AND/OR REDISTRIBUTE THE LIBRARY AS PERMITTED ABOVE, BE LIABLE TO YOU
+FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR
+CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE
+LIBRARY (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
+RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
+FAILURE OF THE LIBRARY TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
+SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
+DAMAGES.
+
+ END OF TERMS AND CONDITIONS
+^L
+ How to Apply These Terms to Your New Libraries
+
+ If you develop a new library, and you want it to be of the greatest
+possible use to the public, we recommend making it free software that
+everyone can redistribute and change. You can do so by permitting
+redistribution under these terms (or, alternatively, under the terms
+of the ordinary General Public License).
+
+ To apply these terms, attach the following notices to the library.
+It is safest to attach them to the start of each source file to most
+effectively convey the exclusion of warranty; and each file should
+have at least the "copyright" line and a pointer to where the full
+notice is found.
+
+
+ <one line to give the library's name and a brief idea of what it
+does.>
+ Copyright (C) <year> <name of author>
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+
+Also add information on how to contact you by electronic and paper
+mail.
+
+You should also get your employer (if you work as a programmer) or
+your
+school, if any, to sign a "copyright disclaimer" for the library, if
+necessary. Here is a sample; alter the names:
+
+ Yoyodyne, Inc., hereby disclaims all copyright interest in the
+ library `Frob' (a library for tweaking knobs) written by James
+Random Hacker.
+
+ <signature of Ty Coon>, 1 April 1990
+ Ty Coon, President of Vice
+
+That's all there is to it!
+
+
diff --git a/Makefile b/Makefile
new file mode 100644
index 0000000..8d134cc
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,64 @@
+prefix=/usr
+includedir=$(prefix)/include
+libdir=$(prefix)/lib
+
+ARCH := $(shell uname -m | sed -e s/i.86/i386/)
+CFLAGS := -nostdlib -nostartfiles -Wall -I. -g -fomit-frame-pointer -O2 -fPIC
+SO_CFLAGS=-shared $(CFLAGS)
+L_CFLAGS=$(CFLAGS)
+LINK_FLAGS=
+
+soname=libaio.so.1
+minor=0
+micro=1
+libname=$(soname).$(minor).$(micro)
+all_targets += libaio.a $(libname)
+
+all: $(all_targets)
+
+# libaio provided functions
+libaio_srcs := io_queue_init.c io_queue_release.c
+libaio_srcs += io_queue_wait.c io_queue_run.c
+
+# real syscalls
+libaio_srcs += io_getevents.c io_submit.c io_cancel.c
+libaio_srcs += io_setup.c io_destroy.c
+
+# internal functions
+libaio_srcs += raw_syscall.c
+
+# old symbols
+libaio_srcs += compat-0_1.c
+
+libaio_objs := $(patsubst %.c,%.ol,$(libaio_srcs))
+libaio_sobjs := $(patsubst %.c,%.os,$(libaio_srcs))
+
+$(libaio_objs) $(libaio_sobjs): libaio.h vsys_def.h
+
+%.os: %.c
+ $(CC) $(SO_CFLAGS) -c -o $@ $<
+
+%.ol: %.c
+ $(CC) $(L_CFLAGS) -c -o $@ $<
+
+
+libaio.a: $(libaio_objs)
+ rm -f libaio.a
+ ar r libaio.a $^
+ ranlib libaio.a
+
+$(libname): $(libaio_sobjs) libaio.map
+ $(CC) $(SO_CFLAGS) -Wl,--version-script=libaio.map -Wl,-soname=$(soname) -o $@ $(libaio_sobjs) $(LINK_FLAGS)
+
+install: $(all_targets)
+ install -D -m 644 libaio.h $(includedir)/libaio.h
+ install -D -m 644 libaio.a $(libdir)/libaio.a
+ install -D -m 755 $(libname) $(libdir)/$(libname)
+ ln -sf $(libname) $(libdir)/$(soname)
+ ln -sf $(libname) $(libdir)/libaio.so
+
+$(libaio_objs): libaio.h
+
+clean:
+ rm -f $(all_targets) $(libaio_objs) $(libaio_sobjs) $(soname).new
+ rm -f *.so* *.a *.o
diff --git a/TODO b/TODO
new file mode 100644
index 0000000..0a9ac15
--- /dev/null
+++ b/TODO
@@ -0,0 +1,4 @@
+- Write man pages.
+- Make -static links against libaio work.
+- Fallback on userspace if the kernel calls return -ENOSYS.
+
diff --git a/compat-0_1.c b/compat-0_1.c
new file mode 100644
index 0000000..136396f
--- /dev/null
+++ b/compat-0_1.c
@@ -0,0 +1,62 @@
+/* libaio Linux async I/O interface
+
+ compat-0_1.c : compatibility symbols for libaio 0.1.x-0.3.x
+
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <stdlib.h>
+#include <asm/errno.h>
+
+#include "libaio.h"
+#include "vsys_def.h"
+
+#include "syscall.h"
+
+
+/* ABI change. Provide partial compatibility on this one for now. */
+SYMVER(compat0_1_io_cancel, io_cancel, 0.1);
+int compat0_1_io_cancel(io_context_t ctx, struct iocb *iocb)
+{
+ struct io_event event;
+
+ /* FIXME: the old ABI would return the event on the completion queue */
+ return io_cancel(ctx, iocb, &event);
+}
+
+SYMVER(compat0_1_io_queue_wait, io_queue_wait, 0.1);
+int compat0_1_io_queue_wait(io_context_t ctx, struct timespec *when)
+{
+ struct timespec timeout;
+ if (when)
+ timeout = *when;
+ return io_getevents(ctx, 0, 0, NULL, when ? &timeout : NULL);
+}
+
+
+/* ABI change. Provide backwards compatibility for this one. */
+SYMVER(compat0_1_io_getevents, io_getevents, 0.1);
+int compat0_1_io_getevents(io_context_t ctx_id, long nr,
+ struct io_event *events,
+ const struct timespec *const_timeout)
+{
+ struct timespec timeout;
+ if (const_timeout)
+ timeout = *const_timeout;
+ return io_getevents(ctx_id, 1, nr, events,
+ const_timeout ? &timeout : NULL);
+}
+
diff --git a/harness/Makefile b/harness/Makefile
new file mode 100644
index 0000000..0bb8e3d
--- /dev/null
+++ b/harness/Makefile
@@ -0,0 +1,56 @@
+# foo.
+TEST_SRCS:=$(shell find cases/ -name \*.t | sort -n -t/ -k2)
+EXTRAPROGS:=cases/8.p cases/10.p
+PARTPROGS:=$(filter-out $(EXTRAPROGS), $(patsubst %.t,%.p,$(TEST_SRCS)))
+PROGS:=$(PARTPROGS) $(EXTRAPROGS)
+HARNESS_SRCS:=main.c
+# io_queue.c
+
+CFLAGS=-Wall -Werror -I../src -g -O
+#-lpthread -lrt
+
+all: $(PROGS)
+
+$(PROGS): %.p: %.t $(HARNESS_SRCS)
+ $(CC) $(CFLAGS) -DTEST_NAME=\"$<\" -o $@ main.c ../src/libaio.a
+
+clean:
+ rm -f $(PROGS) *.o runtests.out rofile wofile rwfile
+
+.PHONY:
+
+testdir/rofile: testdir .PHONY
+ rm -f $@
+ echo "test" >$@
+ chmod 400 $@
+
+testdir/wofile: testdir .PHONY
+ rm -f $@
+ echo "test" >$@
+ chmod 200 $@
+
+testdir/rwfile: testdir .PHONY
+ rm -f $@
+ echo "test" >$@
+ chmod 600 $@
+
+testdir testdir.enospc testdir.ext2:
+ mkdir $@
+
+root: .PHONY
+ @if [ `id -u` -ne 0 ]; then echo Need root for check, try partcheck >&2; exit 1; fi
+
+partcheck: $(PARTPROGS) testdir/rofile testdir/rwfile testdir/wofile
+ ./runtests.sh $(PARTPROGS)
+
+ext2.img:
+ dd if=/dev/zero bs=1M count=10 of=$@
+ mke2fs -F -b 4096 $@
+
+extracheck: $(EXTRAPROGS) root testdir.ext2 testdir.enospc ext2.img
+ mount -o loop -t ext2 ext2-enospc.img testdir.enospc
+ ./runtests.sh cases/10.p; ret=$$?; umount testdir.enospc; exit $$ret
+ mount -o loop -t ext2 ext2.img testdir.ext2
+ ./runtests.sh cases/8.p; ret=$$?; umount testdir.ext2; exit $$ret
+
+check: partcheck extracheck
diff --git a/harness/README b/harness/README
new file mode 100644
index 0000000..5557370
--- /dev/null
+++ b/harness/README
@@ -0,0 +1,19 @@
+Notes on running this test suite:
+
+To run the test suite, run "make check". All test cases should pass
+and there should be 0 fails.
+
+Several of the test cases require a directory on the filesystem under
+test for the creation of test files, as well as the generation of
+error conditions. The test cases assume the directories (or symlinks
+to directories) are as follows:
+
+ testdir/
+ - used for general read/write test cases. Must have at
+ least as much free space as the machine has RAM (up
+ to 768MB).
+ testdir.enospc/
+ - a filesystem that has space for writing 8KB out, but
+ fails with -ENOSPC beyond 8KB.
+ testdir.ext2/
+ - must be an ext2 filesystem.
diff --git a/harness/attic/0.t b/harness/attic/0.t
new file mode 100644
index 0000000..033e62c
--- /dev/null
+++ b/harness/attic/0.t
@@ -0,0 +1,9 @@
+/* 0.t
+ Test harness check: okay.
+*/
+int test_main(void)
+{
+ printf("test_main: okay\n");
+ return 0;
+}
+
diff --git a/harness/attic/1.t b/harness/attic/1.t
new file mode 100644
index 0000000..799ffd1
--- /dev/null
+++ b/harness/attic/1.t
@@ -0,0 +1,9 @@
+/* 1.t
+ Test harness check: fail.
+*/
+int test_main(void)
+{
+ printf("test_main: fail\n");
+ return 1;
+}
+
diff --git a/harness/cases/10.t b/harness/cases/10.t
new file mode 100644
index 0000000..9d3beb2
--- /dev/null
+++ b/harness/cases/10.t
@@ -0,0 +1,53 @@
+/* 10.t - uses testdir.enospc/rwfile
+- Check results on out-of-space and out-of-quota. (10.t)
+ - write that fills filesystem but does not go over should succeed
+ - write that fills filesystem and goes over should be partial
+ - write to full filesystem should return -ENOSPC
+ - read beyond end of file after ENOSPC should return 0
+*/
+#include "aio_setup.h"
+
+#include <sys/time.h>
+#include <sys/resource.h>
+#include <unistd.h>
+
+int test_main(void)
+{
+/* Note: changing either of these requires updating the ext2-enospc.img
+ * filesystem image. Also, if SIZE is less than PAGE_SIZE, problems
+ * crop up due to ext2's preallocation.
+ */
+#define LIMIT 65536
+#define SIZE 65536
+ char *buf;
+ int rwfd;
+ int status = 0, res;
+
+ rwfd = open("testdir.enospc/rwfile", O_RDWR|O_CREAT|O_TRUNC, 0600);
+ assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ buf = malloc(SIZE); assert(buf != NULL);
+ memset(buf, 0, SIZE);
+
+
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT-SIZE, WRITE, SIZE);
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT-SIZE, READ, SIZE);
+
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT, WRITE, -ENOSPC);
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT, READ, 0);
+
+ res = ftruncate(rwfd, 0); assert(res == 0);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 1+LIMIT-SIZE, WRITE, SIZE-1);
+ status |= attempt_rw(rwfd, buf, SIZE, 1+LIMIT-SIZE, READ, SIZE-1);
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT, READ, 0);
+
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT, WRITE, -ENOSPC);
+ status |= attempt_rw(rwfd, buf, SIZE, LIMIT, READ, 0);
+ status |= attempt_rw(rwfd, buf, 0, LIMIT, WRITE, 0);
+
+ res = close(rwfd); assert(res == 0);
+ res = unlink("testdir.enospc/rwfile"); assert(res == 0);
+ return status;
+}
+
diff --git a/harness/cases/11.t b/harness/cases/11.t
new file mode 100644
index 0000000..efcf6d4
--- /dev/null
+++ b/harness/cases/11.t
@@ -0,0 +1,39 @@
+/* 11.t - uses testdir/rwfile
+- repeated read / write of same page (to check accounting) (11.t)
+*/
+#include "aio_setup.h"
+
+#include <sys/time.h>
+#include <sys/resource.h>
+#include <unistd.h>
+
+int test_main(void)
+{
+#define COUNT 1000000
+#define SIZE 256
+ char *buf;
+ int rwfd;
+ int status = 0;
+ int i;
+
+ rwfd = open("testdir/rwfile", O_RDWR|O_CREAT|O_TRUNC, 0600);
+ assert(rwfd != -1);
+ buf = malloc(SIZE); assert(buf != NULL);
+ memset(buf, 0, SIZE);
+
+ for (i=0; i<COUNT; i++) {
+ status |= attempt_rw(rwfd, buf, SIZE, 0, WRITE_SILENT, SIZE);
+ if (status)
+ break;
+ }
+ printf("completed %d out of %d writes\n", i, COUNT);
+ for (i=0; i<COUNT; i++) {
+ status |= attempt_rw(rwfd, buf, SIZE, 0, READ_SILENT, SIZE);
+ if (status)
+ break;
+ }
+ printf("completed %d out of %d reads\n", i, COUNT);
+
+ return status;
+}
+
diff --git a/harness/cases/12.t b/harness/cases/12.t
new file mode 100644
index 0000000..e87d1dc
--- /dev/null
+++ b/harness/cases/12.t
@@ -0,0 +1,53 @@
+/* 12.t
+- ioctx access across fork() (12.t)
+ */
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+#include <signal.h>
+
+#include "aio_setup.h"
+
+void test_child(void)
+{
+ int res;
+ res = attempt_io_submit(io_ctx, 0, NULL, -EINVAL);
+ fflush(stdout);
+ _exit(res);
+}
+
+int test_main(void)
+{
+ int res, status;
+ pid_t pid;
+ sigset_t set;
+
+ if (attempt_io_submit(io_ctx, 0, NULL, 0))
+ return 1;
+
+ sigemptyset(&set);
+ sigaddset(&set, SIGCHLD);
+ sigprocmask(SIG_BLOCK, &set, NULL);
+
+ fflush(NULL);
+ pid = fork(); assert(pid != -1);
+
+ if (pid == 0)
+ test_child();
+
+ res = waitpid(pid, &status, 0);
+
+ if (WIFEXITED(status)) {
+ int failed = (WEXITSTATUS(status) != 0);
+ printf("child exited with status %d%s\n", WEXITSTATUS(status),
+ failed ? " -- FAILED" : "");
+ return failed;
+ }
+
+ /* anything else: failed */
+ if (WIFSIGNALED(status))
+ printf("child killed by signal %d -- FAILED.\n",
+ WTERMSIG(status));
+
+ return 1;
+}
diff --git a/harness/cases/13.t b/harness/cases/13.t
new file mode 100644
index 0000000..5f18005
--- /dev/null
+++ b/harness/cases/13.t
@@ -0,0 +1,66 @@
+/* 13.t - uses testdir/rwfile
+- Submit multiple writes larger than aio-max-size (deadlocks on older
+ aio code)
+*/
+#include "aio_setup.h"
+
+#include <sys/time.h>
+#include <sys/resource.h>
+#include <unistd.h>
+
+int test_main(void)
+{
+#define SIZE (1024 * 1024)
+#define IOS 8
+ struct iocb iocbs[IOS];
+ struct iocb *iocb_list[IOS];
+ char *bufs[IOS];
+ int rwfd;
+ int status = 0, res;
+ int i;
+
+ rwfd = open("testdir/rwfile", O_RDWR|O_CREAT|O_TRUNC, 0600);
+ assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+
+ for (i=0; i<IOS; i++) {
+ bufs[i] = malloc(SIZE);
+ assert(bufs[i] != NULL);
+ memset(bufs[i], 0, SIZE);
+
+ io_prep_pwrite(&iocbs[i], rwfd, bufs[i], SIZE, i * SIZE);
+ iocb_list[i] = &iocbs[i];
+ }
+
+ status |= attempt_io_submit(io_ctx, IOS, iocb_list, IOS);
+
+ for (i=0; i<IOS; i++) {
+ struct timespec ts = { tv_sec: 30, tv_nsec: 0 };
+ struct io_event event;
+ struct iocb *iocb;
+
+ res = io_getevents(io_ctx, 0, 1, &event, &ts);
+ if (res != 1) {
+ status |= 1;
+ printf("io_getevents failed [%d] with res=%d [%s]\n",
+ i, res, (res < 0) ? strerror(-res) : "okay");
+ break;
+ }
+
+ if (event.res != SIZE)
+ status |= 1;
+
+ iocb = (void *)event.obj;
+ printf("event[%d]: write[%d] %s, returned: %ld [%s]\n",
+ i, (int)(iocb - &iocbs[0]),
+ (event.res != SIZE) ? "failed" : "okay",
+ (long)event.res,
+ (event.res < 0) ? strerror(-event.res) : "okay"
+ );
+ }
+
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ res = close(rwfd); assert(res == 0);
+ return status;
+}
+
diff --git a/harness/cases/14.t b/harness/cases/14.t
new file mode 100644
index 0000000..87773e3
--- /dev/null
+++ b/harness/cases/14.t
@@ -0,0 +1,93 @@
+#include <sys/types.h>
+#include <sys/wait.h>
+#include <unistd.h>
+#include <signal.h>
+
+#include "aio_setup.h"
+#include <sys/mman.h>
+
+#define SIZE 768*1024*1024
+
+//just submit an I/O
+
+int test_child(void)
+{
+ char *buf;
+ int rwfd;
+ int res;
+ long size;
+ struct iocb iocb;
+ struct iocb *iocbs[] = { &iocb };
+ int loop = 10;
+ int i;
+
+ aio_setup(1024);
+
+ size = SIZE;
+
+ printf("size = %ld\n", size);
+
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd !=
+-1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ buf = malloc(size); assert(buf !=
+NULL);
+
+ for(i=0;i<loop;i++) {
+
+ switch(i%2) {
+ case 0:
+ io_prep_pwrite(&iocb, rwfd, buf, size, 0);
+ break;
+ case 1:
+ io_prep_pread(&iocb, rwfd, buf, size, 0);
+ }
+
+ res = io_submit(io_ctx, 1, iocbs);
+ if (res != 1) {
+ printf("child: submit: io_submit res=%d [%s]\n", res,
+strerror(-res));
+ _exit(1);
+ }
+ }
+
+ res = ftruncate(rwfd, 0); assert(res == 0);
+
+ _exit(0);
+}
+
+/* from 12.t */
+int test_main(void)
+{
+ int res, status;
+ pid_t pid;
+ sigset_t set;
+
+ if (attempt_io_submit(io_ctx, 0, NULL, 0))
+ return 1;
+
+ sigemptyset(&set);
+ sigaddset(&set, SIGCHLD);
+ sigprocmask(SIG_BLOCK, &set, NULL);
+ fflush(NULL);
+ pid = fork(); assert(pid != -1);
+
+ if (pid == 0)
+ test_child();
+
+ res = waitpid(pid, &status, 0);
+
+ if (WIFEXITED(status)) {
+ int failed = (WEXITSTATUS(status) != 0);
+ printf("child exited with status %d%s\n", WEXITSTATUS(status),
+ failed ? " -- FAILED" : "");
+ return failed;
+ }
+
+ /* anything else: failed */
+ if (WIFSIGNALED(status))
+ printf("child killed by signal %d -- FAILED.\n",
+ WTERMSIG(status));
+
+ return 1;
+}
diff --git a/harness/cases/15.t b/harness/cases/15.t
new file mode 100644
index 0000000..933d008
--- /dev/null
+++ b/harness/cases/15.t
@@ -0,0 +1,94 @@
+/* 15.t
+- pwritev and preadv tests.
+*/
+#include "aio_setup.h"
+#include <sys/mman.h>
+#include <sys/uio.h>
+#include <errno.h>
+
+int test_main(void)
+{
+#define SIZE 512
+#define NUM_IOV 10
+ char buf[SIZE*NUM_IOV];
+ struct iovec iov[NUM_IOV];
+ int rwfd;
+ int status = 0, res, i;
+
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd != -1);
+ res = ftruncate(rwfd, sizeof(buf)); assert(res == 0);
+
+ for (i = 0; i < NUM_IOV; i++) {
+ iov[i].iov_base = buf + i*SIZE;
+ iov[i].iov_len = SIZE;
+ memset(iov[i].iov_base, i, SIZE);
+ }
+ status |= attempt_rw(rwfd, iov, NUM_IOV, 0, WRITEV, SIZE*NUM_IOV);
+ res = pread(rwfd, buf, sizeof(buf), 0); assert(res == sizeof(buf));
+ for (i = 0; i < NUM_IOV; i++) {
+ unsigned int j;
+ for (j = 0; j < SIZE; j++) {
+ if (buf[i*SIZE + j] != i) {
+ printf("Unexpected value after writev at %i\n",
+ i*SIZE + j);
+ status |= 1;
+ break;
+ }
+ }
+ }
+ if (!status)
+ printf("Checking memory: [Success]\n");
+
+ memset(buf, 0, sizeof(buf));
+ status |= attempt_rw(rwfd, iov, NUM_IOV, 0, READV, SIZE*NUM_IOV);
+ for (i = 0; i < NUM_IOV; i++) {
+ unsigned int j;
+ for (j = 0; j < SIZE; j++) {
+ if (buf[i*SIZE + j] != i) {
+ printf("Unexpected value after readv at %i\n",
+ i*SIZE + j);
+ status |= 1;
+ break;
+ }
+ }
+ }
+
+ /* Check that offset works. */
+ status |= attempt_rw(rwfd, iov+1, NUM_IOV-1, SIZE, WRITEV,
+ SIZE*(NUM_IOV-1));
+ memset(buf, 0, sizeof(buf));
+ res = pread(rwfd, buf, sizeof(buf), 0); assert(res == sizeof(buf));
+ for (i = 1; i < NUM_IOV; i++) {
+ unsigned int j;
+ for (j = 0; j < SIZE; j++) {
+ if (buf[i*SIZE + j] != i) {
+ printf("Unexpected value after offset writev at %i\n",
+ i*SIZE + j);
+ status |= 1;
+ break;
+ }
+ }
+ }
+ if (!status)
+ printf("Checking memory: [Success]\n");
+
+ memset(buf, 0, sizeof(buf));
+ status |= attempt_rw(rwfd, iov+1, NUM_IOV-1, SIZE, READV,
+ SIZE*(NUM_IOV-1));
+ for (i = 1; i < NUM_IOV; i++) {
+ unsigned int j;
+ for (j = 0; j < SIZE; j++) {
+ if (buf[i*SIZE + j] != i) {
+ printf("Unexpected value after offset readv at %i\n",
+ i*SIZE + j);
+ status |= 1;
+ break;
+ }
+ }
+ }
+ if (!status)
+ printf("Checking memory: [Success]\n");
+
+ return status;
+}
+
diff --git a/harness/cases/16.t b/harness/cases/16.t
new file mode 100644
index 0000000..c3157cc
--- /dev/null
+++ b/harness/cases/16.t
@@ -0,0 +1,94 @@
+/* 16.t
+- eventfd tests.
+*/
+#include <stdint.h>
+#include <err.h>
+#include <sys/syscall.h> /* For SYS_xxx definitions */
+
+#ifndef SYS_eventfd
+#if defined(__i386__)
+#define SYS_eventfd 323
+#elif defined(__x86_64__)
+#define SYS_eventfd 284
+#elif defined(__ia64__)
+#define SYS_eventfd 1309
+#elif defined(__PPC__)
+#define SYS_eventfd 307
+#elif defined(__s390__)
+#define SYS_eventfd 318
+#elif defined(__alpha__)
+#define SYS_eventfd 478
+#else
+#error define SYS_eventfd for your arch!
+#endif
+#endif
+
+int test_main(void)
+{
+ /* 10 MB takes long enough that we would fail if eventfd
+ * returned immediately. */
+#define SIZE 10000000
+ char *buf;
+ struct io_event io_event;
+ struct iocb iocb;
+ struct iocb *iocbs[] = { &iocb };
+ int rwfd, efd;
+ int res;
+ io_context_t io_ctx;
+ uint64_t event;
+ struct timespec notime = { .tv_sec = 0, .tv_nsec = 0 };
+
+ buf = malloc(SIZE); assert(buf);
+ efd = syscall(SYS_eventfd, 0);
+ if (efd < 0) {
+ if (errno == ENOSYS) {
+ printf("No eventfd support. [SKIPPING]\n");
+ exit(0);
+ }
+ err(1, "Failed to get eventfd");
+ }
+
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ memset(buf, 0x42, SIZE);
+
+ /* Write test. */
+ res = io_queue_init(1024, &io_ctx); assert(res == 0);
+ io_prep_pwrite(&iocb, rwfd, buf, SIZE, 0);
+ io_set_eventfd(&iocb, efd);
+ res = io_submit(io_ctx, 1, iocbs); assert(res == 1);
+
+ alarm(30);
+ res = read(efd, &event, sizeof(event)); assert(res == sizeof(event));
+ assert(event == 1);
+
+ /* This should now be ready. */
+ res = io_getevents(io_ctx, 0, 1, &io_event, ¬ime);
+ if (res != 1)
+ err(1, "io_getevents did not return 1 event after eventfd");
+ assert(io_event.res == SIZE);
+ printf("eventfd write test [SUCCESS]\n");
+
+ /* Read test. */
+ memset(buf, 0, SIZE);
+ io_prep_pread(&iocb, rwfd, buf, SIZE, 0);
+ io_set_eventfd(&iocb, efd);
+ res = io_submit(io_ctx, 1, iocbs); assert(res == 1);
+
+ alarm(30);
+ res = read(efd, &event, sizeof(event)); assert(res == sizeof(event));
+ assert(event == 1);
+
+ /* This should now be ready. */
+ res = io_getevents(io_ctx, 0, 1, &io_event, ¬ime);
+ if (res != 1)
+ err(1, "io_getevents did not return 1 event after eventfd");
+ assert(io_event.res == SIZE);
+
+ for (res = 0; res < SIZE; res++)
+ assert(buf[res] == 0x42);
+ printf("eventfd read test [SUCCESS]\n");
+
+ return 0;
+}
+
diff --git a/harness/cases/2.t b/harness/cases/2.t
new file mode 100644
index 0000000..3a0212d
--- /dev/null
+++ b/harness/cases/2.t
@@ -0,0 +1,41 @@
+/* 2.t
+- io_setup (#2)
+ - with invalid context pointer
+ - with maxevents <= 0
+ - with an already initialized ctxp
+*/
+
+int attempt(int n, io_context_t *ctxp, int expect)
+{
+ int res;
+
+ printf("expect %3d: io_setup(%5d, %p) = ", expect, n, ctxp);
+ fflush(stdout);
+ res = io_setup(n, ctxp);
+ printf("%3d [%s]%s\n", res, strerror(-res),
+ (res != expect) ? " -- FAILED" : "");
+ if (res != expect)
+ return 1;
+
+ return 0;
+}
+
+int test_main(void)
+{
+ io_context_t ctx;
+ int status = 0;
+
+ ctx = NULL;
+ status |= attempt(-1000, KERNEL_RW_POINTER, -EFAULT);
+ status |= attempt( 1000, KERNEL_RW_POINTER, -EFAULT);
+ status |= attempt( 0, KERNEL_RW_POINTER, -EFAULT);
+ status |= attempt(-1000, &ctx, -EINVAL);
+ status |= attempt( -1, &ctx, -EINVAL);
+ status |= attempt( 0, &ctx, -EINVAL);
+ assert(ctx == NULL);
+ status |= attempt( 1, &ctx, 0);
+ status |= attempt( 1, &ctx, -EINVAL);
+
+ return status;
+}
+
diff --git a/harness/cases/3.t b/harness/cases/3.t
new file mode 100644
index 0000000..7773d80
--- /dev/null
+++ b/harness/cases/3.t
@@ -0,0 +1,25 @@
+/* 3.t
+- io_submit/io_getevents with invalid addresses (3.t)
+
+*/
+#include "aio_setup.h"
+
+int test_main(void)
+{
+ struct iocb a, b;
+ struct iocb *good_ios[] = { &a, &b };
+ struct iocb *bad1_ios[] = { NULL, &b };
+ struct iocb *bad2_ios[] = { KERNEL_RW_POINTER, &a };
+ int status = 0;
+
+ status |= attempt_io_submit(BAD_CTX, 1, good_ios, -EINVAL);
+ status |= attempt_io_submit( io_ctx, 0, good_ios, 0);
+ status |= attempt_io_submit( io_ctx, 1, NULL, -EFAULT);
+ status |= attempt_io_submit( io_ctx, 1, (void *)-1, -EFAULT);
+ status |= attempt_io_submit( io_ctx, 2, bad1_ios, -EFAULT);
+ status |= attempt_io_submit( io_ctx, 2, bad2_ios, -EFAULT);
+ status |= attempt_io_submit( io_ctx, -1, good_ios, -EINVAL);
+
+ return status;
+}
+
diff --git a/harness/cases/4.t b/harness/cases/4.t
new file mode 100644
index 0000000..972b4f2
--- /dev/null
+++ b/harness/cases/4.t
@@ -0,0 +1,72 @@
+/* 4.t
+- read of descriptor without read permission (4.t)
+- write to descriptor without write permission (4.t)
+- check that O_APPEND writes actually append
+
+*/
+#include "aio_setup.h"
+
+#define SIZE 512
+#define READ 'r'
+#define WRITE 'w'
+int attempt(int fd, void *buf, int count, long long pos, int rw, int expect)
+{
+ struct iocb iocb;
+ int res;
+
+ switch(rw) {
+ case READ: io_prep_pread (&iocb, fd, buf, count, pos); break;
+ case WRITE: io_prep_pwrite(&iocb, fd, buf, count, pos); break;
+ }
+
+ printf("expect %3d: (%c), res = ", expect, rw);
+ fflush(stdout);
+ res = sync_submit(&iocb);
+ printf("%3d [%s]%s\n", res, (res <= 0) ? strerror(-res) : "Success",
+ (res != expect) ? " -- FAILED" : "");
+ if (res != expect)
+ return 1;
+
+ return 0;
+}
+
+int test_main(void)
+{
+ char buf[SIZE];
+ int rofd, wofd, rwfd;
+ int status = 0, res;
+
+ memset(buf, 0, SIZE);
+
+ rofd = open("testdir/rofile", O_RDONLY); assert(rofd != -1);
+ wofd = open("testdir/wofile", O_WRONLY); assert(wofd != -1);
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd != -1);
+
+ status |= attempt(rofd, buf, SIZE, 0, WRITE, -EBADF);
+ status |= attempt(wofd, buf, SIZE, 0, READ, -EBADF);
+ status |= attempt(rwfd, buf, SIZE, 0, WRITE, SIZE);
+ status |= attempt(rwfd, buf, SIZE, 0, READ, SIZE);
+ status |= attempt(rwfd, buf, SIZE, -1, READ, -EINVAL);
+ status |= attempt(rwfd, buf, SIZE, -1, WRITE, -EINVAL);
+
+ rwfd = open("testdir/rwfile", O_RDWR|O_APPEND); assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ status |= attempt(rwfd, buf, SIZE, 0, READ, 0);
+ status |= attempt(rwfd, "1234", 4, 0, WRITE, 4);
+ status |= attempt(rwfd, "5678", 4, 0, WRITE, 4);
+ memset(buf, 0, SIZE);
+ status |= attempt(rwfd, buf, SIZE, 0, READ, 8);
+ printf("read after append: [%s]\n", buf);
+ assert(memcmp(buf, "12345678", 8) == 0);
+
+ status |= attempt(rwfd, KERNEL_RW_POINTER, SIZE, 0, READ, -EFAULT);
+ status |= attempt(rwfd, KERNEL_RW_POINTER, SIZE, 0, WRITE, -EFAULT);
+
+ /* Some architectures map the 0 page. Ugh. */
+#if !defined(__ia64__)
+ status |= attempt(rwfd, NULL, SIZE, 0, WRITE, -EFAULT);
+#endif
+
+ return status;
+}
+
diff --git a/harness/cases/5.t b/harness/cases/5.t
new file mode 100644
index 0000000..2b4b4bb
--- /dev/null
+++ b/harness/cases/5.t
@@ -0,0 +1,54 @@
+/* 5.t
+- Write from a mmap() of the same file. (5.t)
+*/
+#include "aio_setup.h"
+#include <sys/mman.h>
+#include <errno.h>
+
+int test_main(void)
+{
+ int page_size = getpagesize();
+#define SIZE 512
+ char *buf;
+ int rwfd;
+ int status = 0, res;
+
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd != -1);
+ res = ftruncate(rwfd, 512); assert(res == 0);
+
+ buf = mmap(0, page_size, PROT_READ|PROT_WRITE, MAP_SHARED, rwfd, 0);
+ assert(buf != (char *)-1);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 0, WRITE, SIZE);
+ status |= attempt_rw(rwfd, buf, SIZE, 0, READ, SIZE);
+
+ res = munmap(buf, page_size); assert(res == 0);
+ buf = mmap(0, page_size, PROT_READ|PROT_WRITE, MAP_SHARED, rwfd, 0);
+ assert(buf != (char *)-1);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 0, READ, SIZE);
+ status |= attempt_rw(rwfd, buf, SIZE, 0, WRITE, SIZE);
+
+ res = munmap(buf, page_size); assert(res == 0);
+ buf = mmap(0, page_size, PROT_READ, MAP_SHARED, rwfd, 0);
+ assert(buf != (char *)-1);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 0, WRITE, SIZE);
+ status |= attempt_rw(rwfd, buf, SIZE, 0, READ, -EFAULT);
+
+ res = munmap(buf, page_size); assert(res == 0);
+ buf = mmap(0, page_size, PROT_WRITE, MAP_SHARED, rwfd, 0);
+ assert(buf != (char *)-1);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 0, READ, SIZE);
+
+ /* Whether PROT_WRITE is readable is arch-dependent. So compare
+ * against read result. */
+ res = read(rwfd, buf, SIZE);
+ if (res < 0)
+ res = -errno;
+ status |= attempt_rw(rwfd, buf, SIZE, 0, WRITE, res);
+
+ return status;
+}
+
diff --git a/harness/cases/6.t b/harness/cases/6.t
new file mode 100644
index 0000000..cea4b01
--- /dev/null
+++ b/harness/cases/6.t
@@ -0,0 +1,57 @@
+/* 6.t
+- huge reads (pinned pages) (6.t)
+- huge writes (6.t)
+*/
+#include "aio_setup.h"
+#include <sys/mman.h>
+
+long getmemsize(void)
+{
+ FILE *f = fopen("/proc/meminfo", "r");
+ long size;
+ int gotit = 0;
+ char str[256];
+
+ assert(f != NULL);
+ while (NULL != fgets(str, 255, f)) {
+ str[255] = 0;
+ if (0 == memcmp(str, "MemTotal:", 9)) {
+ if (1 == sscanf(str + 9, "%ld", &size)) {
+ gotit = 1;
+ break;
+ }
+ }
+ }
+ fclose(f);
+
+ assert(gotit != 0);
+ return size;
+}
+
+int test_main(void)
+{
+ char *buf;
+ int rwfd;
+ int status = 0, res;
+ long size;
+
+ size = getmemsize();
+ printf("size = %ld\n", size);
+ assert(size >= (16 * 1024));
+ if (size > (768 * 1024))
+ size = 768 * 1024;
+ size *= 1024;
+
+ rwfd = open("testdir/rwfile", O_RDWR); assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ buf = malloc(size); assert(buf != NULL);
+
+ //memset(buf, 0, size);
+ status |= attempt_rw(rwfd, buf, size, 0, WRITE, size);
+ status |= attempt_rw(rwfd, buf, size, 0, READ, size);
+
+ //res = ftruncate(rwfd, 0); assert(res == 0);
+
+ return status;
+}
+
diff --git a/harness/cases/7.t b/harness/cases/7.t
new file mode 100644
index 0000000..f877d8a
--- /dev/null
+++ b/harness/cases/7.t
@@ -0,0 +1,30 @@
+/* 7.t
+- Write overlapping the file size rlimit boundary: should be a short
+ write. (7.t)
+- Write at the file size rlimit boundary: should give EFBIG. (I think
+ the spec requires that you do NOT deliver SIGXFSZ in this case, where
+ you would do so for sync IO.) (7.t)
+- Special case: a write of zero bytes at or beyond the file size rlimit
+ boundary must return success. (7.t)
+*/
+
+#include <sys/resource.h>
+#include <signal.h>
+
+void SET_RLIMIT(long long limit)
+{
+ struct rlimit rlim;
+ int res;
+
+ /* Seems that we do send SIGXFSZ, but hard to fix... */
+ signal(SIGXFSZ, SIG_IGN);
+ rlim.rlim_cur = limit; assert(rlim.rlim_cur == limit);
+ rlim.rlim_max = limit; assert(rlim.rlim_max == limit);
+
+ res = setrlimit(RLIMIT_FSIZE, &rlim); assert(res == 0);
+}
+
+#define LIMIT 8192
+#define FILENAME "testdir/rwfile"
+
+#include "common-7-8.h"
diff --git a/harness/cases/8.t b/harness/cases/8.t
new file mode 100644
index 0000000..e59199f
--- /dev/null
+++ b/harness/cases/8.t
@@ -0,0 +1,28 @@
+/* 8.t
+- Ditto for the above three tests at the offset maximum (largest
+ possible ext2/3 file size.) (8.t)
+ */
+#include <sys/types.h>
+#include <unistd.h>
+
+long long get_fs_limit(int fd)
+{
+ long long min = 0, max = 9223372036854775807LL;
+ char c = 0;
+
+ while (max - min > 1) {
+ if (pwrite64(fd, &c, 1, (min + max) / 2) == -1)
+ max = (min + max) / 2;
+ else {
+ ftruncate(fd, 0);
+ min = (min + max) / 2;
+ }
+ }
+ return max;
+}
+
+#define SET_RLIMIT(x) do ; while (0)
+#define LIMIT get_fs_limit(rwfd)
+#define FILENAME "testdir.ext2/rwfile"
+
+#include "common-7-8.h"
diff --git a/harness/cases/aio_setup.h b/harness/cases/aio_setup.h
new file mode 100644
index 0000000..1914915
--- /dev/null
+++ b/harness/cases/aio_setup.h
@@ -0,0 +1,108 @@
+#include <time.h>
+io_context_t io_ctx;
+#define BAD_CTX ((io_context_t)-1)
+
+void aio_setup(int n)
+{
+ int res = io_queue_init(n, &io_ctx);
+ if (res != 0) {
+ printf("io_queue_setup(%d) returned %d (%s)\n",
+ n, res, strerror(-res));
+ exit(3);
+ }
+}
+
+int attempt_io_submit(io_context_t ctx, long nr, struct iocb *ios[], int expect)
+{
+ int res;
+
+ printf("expect %3d: io_submit(%10p, %3ld, %10p) = ", expect, ctx, nr, ios);
+ fflush(stdout);
+ res = io_submit(ctx, nr, ios);
+ printf("%3d [%s]%s\n", res, (res <= 0) ? strerror(-res) : "",
+ (res != expect) ? " -- FAILED" : "");
+ if (res != expect)
+ return 1;
+
+ return 0;
+}
+
+int sync_submit(struct iocb *iocb)
+{
+ struct io_event event;
+ struct iocb *iocbs[] = { iocb };
+ int res;
+
+ /* 30 second timeout should be enough */
+ struct timespec ts;
+ ts.tv_sec = 30;
+ ts.tv_nsec = 0;
+
+ res = io_submit(io_ctx, 1, iocbs);
+ if (res != 1) {
+ printf("sync_submit: io_submit res=%d [%s]\n", res, strerror(-res));
+ return res;
+ }
+
+ res = io_getevents(io_ctx, 0, 1, &event, &ts);
+ if (res != 1) {
+ printf("sync_submit: io_getevents res=%d [%s]\n", res, strerror(-res));
+ return res;
+ }
+ return event.res;
+}
+
+#define SETUP aio_setup(1024)
+
+
+#define READ 'r'
+#define WRITE 'w'
+#define READ_SILENT 'R'
+#define WRITE_SILENT 'W'
+#define READV '<'
+#define WRITEV '>'
+
+int attempt_rw(int fd, void *buf, int count, long long pos, int rw, int expect)
+{
+ struct iocb iocb;
+ int res;
+ int silent = 0;
+
+ switch(rw) {
+ case READ_SILENT:
+ silent = 1;
+ case READ:
+ io_prep_pread (&iocb, fd, buf, count, pos);
+ break;
+ case WRITE_SILENT:
+ silent = 1;
+ case WRITE:
+ io_prep_pwrite(&iocb, fd, buf, count, pos);
+ break;
+ case WRITEV:
+ io_prep_pwritev(&iocb, fd, buf, count, pos);
+ break;
+ case READV:
+ io_prep_preadv(&iocb, fd, buf, count, pos);
+ break;
+ }
+
+ if (!silent) {
+ printf("expect %5d: (%c), res = ", expect, rw);
+ fflush(stdout);
+ }
+ res = sync_submit(&iocb);
+ if (!silent || res != expect) {
+ if (silent)
+ printf("expect %5d: (%c), res = ", expect, rw);
+ printf("%5d [%s]%s\n", res,
+ (res <= 0) ? strerror(-res) : "Success",
+ (res != expect) ? " -- FAILED" : "");
+ }
+
+ if (res != expect)
+ return 1;
+
+ return 0;
+}
+
diff --git a/harness/cases/common-7-8.h b/harness/cases/common-7-8.h
new file mode 100644
index 0000000..fc54bbf
--- /dev/null
+++ b/harness/cases/common-7-8.h
@@ -0,0 +1,38 @@
+/* common-7-8.h
+*/
+#include "aio_setup.h"
+
+#include <errno.h>
+#include <unistd.h>
+
+#define SIZE 512
+
+int test_main(void)
+{
+ char *buf;
+ int rwfd;
+ int status = 0, res;
+ long long limit;
+
+ rwfd = open(FILENAME, O_RDWR|O_CREAT, 0600); assert(rwfd != -1);
+ res = ftruncate(rwfd, 0); assert(res == 0);
+ buf = malloc(SIZE); assert(buf != NULL);
+ memset(buf, 0, SIZE);
+
+ limit = LIMIT;
+
+ SET_RLIMIT(limit);
+
+ status |= attempt_rw(rwfd, buf, SIZE, limit-SIZE, WRITE, SIZE);
+ status |= attempt_rw(rwfd, buf, SIZE, limit-SIZE, READ, SIZE);
+
+ status |= attempt_rw(rwfd, buf, SIZE, 1+limit-SIZE, WRITE, SIZE-1);
+ status |= attempt_rw(rwfd, buf, SIZE, 1+limit-SIZE, READ, SIZE-1);
+
+ status |= attempt_rw(rwfd, buf, SIZE, limit, WRITE, -EFBIG);
+ status |= attempt_rw(rwfd, buf, SIZE, limit, READ, 0);
+ status |= attempt_rw(rwfd, buf, 0, limit, WRITE, 0);
+
+ return status;
+}
+
diff --git a/harness/ext2-enospc.img b/harness/ext2-enospc.img
new file mode 100644
index 0000000..fa56745
--- /dev/null
+++ b/harness/ext2-enospc.img
Binary files differ
diff --git a/harness/main.c b/harness/main.c
new file mode 100644
index 0000000..9ecd5da
--- /dev/null
+++ b/harness/main.c
@@ -0,0 +1,40 @@
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <errno.h>
+#include <assert.h>
+#include <stdlib.h>
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+
+#include <libaio.h>
+
+#if __LP64__ == 0
+#define KERNEL_RW_POINTER ((void *)0xc0010000)
+#else
+//#warning Not really sure where kernel memory is. Guessing.
+#define KERNEL_RW_POINTER ((void *)0xffffffff81000000)
+#endif
+
+
+char test_name[] = TEST_NAME;
+
+#include TEST_NAME
+
+int main(void)
+{
+ int res;
+
+#if defined(SETUP)
+ SETUP;
+#endif
+
+ res = test_main();
+ printf("test %s completed %s.\n", test_name,
+ res ? "FAILED" : "PASSED"
+ );
+ fflush(stdout);
+ return res ? 1 : 0;
+}
diff --git a/harness/runtests.sh b/harness/runtests.sh
new file mode 100755
index 0000000..717c72a
--- /dev/null
+++ b/harness/runtests.sh
@@ -0,0 +1,19 @@
+#!/bin/bash
+
+passes=0
+fails=0
+
+echo "Test run starting at" `date`
+
+while [ $# -ge 1 ] ; do
+ this_test=$1
+ shift
+ echo "Starting $this_test"
+ $this_test 2>&1
+ res=$?
+ if [ $res -eq 0 ] ; then str="" ; passes=$[passes + 1] ; else str=" -- FAILED" ; fails=$[fails + 1] ; fi
+ echo "Completed $this_test with $res$str".
+done
+
+echo "Pass: $passes Fail: $fails"
+echo "Test run complete at" `date`
diff --git a/io_cancel.c b/io_cancel.c
new file mode 100644
index 0000000..2f0f5f4
--- /dev/null
+++ b/io_cancel.c
@@ -0,0 +1,23 @@
+/* io_cancel.c
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <libaio.h>
+#include "syscall.h"
+
+io_syscall3(int, io_cancel_0_4, io_cancel, io_context_t, ctx, struct iocb *, iocb, struct io_event *, event)
+DEFSYMVER(io_cancel_0_4, io_cancel, 0.4)
diff --git a/io_destroy.c b/io_destroy.c
new file mode 100644
index 0000000..0ab6bd1
--- /dev/null
+++ b/io_destroy.c
@@ -0,0 +1,23 @@
+/* io_destroy
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <errno.h>
+#include <libaio.h>
+#include "syscall.h"
+
+io_syscall1(int, io_destroy, io_destroy, io_context_t, ctx)
diff --git a/io_getevents.c b/io_getevents.c
new file mode 100644
index 0000000..5a05174
--- /dev/null
+++ b/io_getevents.c
@@ -0,0 +1,57 @@
+/* io_getevents.c
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <libaio.h>
+#include <errno.h>
+#include <stdlib.h>
+#include <time.h>
+#include "syscall.h"
+
+io_syscall5(int, __io_getevents_0_4, io_getevents, io_context_t, ctx, long, min_nr, long, nr, struct io_event *, events, struct timespec *, timeout)
+
+#define AIO_RING_MAGIC 0xa10a10a1
+
+/* Ben will hate me for this */
+struct aio_ring {
+ unsigned id; /* kernel internal index number */
+ unsigned nr; /* number of io_events */
+ unsigned head;
+ unsigned tail;
+
+ unsigned magic;
+ unsigned compat_features;
+ unsigned incompat_features;
+ unsigned header_length; /* size of aio_ring */
+};
+
+int io_getevents_0_4(io_context_t ctx, long min_nr, long nr, struct io_event * events, struct timespec * timeout)
+{
+ struct aio_ring *ring;
+ ring = (struct aio_ring*)ctx;
+ if (ring==NULL || ring->magic != AIO_RING_MAGIC)
+ goto do_syscall;
+ if (timeout!=NULL && timeout->tv_sec == 0 && timeout->tv_nsec == 0) {
+ if (ring->head == ring->tail)
+ return 0;
+ }
+
+do_syscall:
+ return __io_getevents_0_4(ctx, min_nr, nr, events, timeout);
+}
+
+DEFSYMVER(io_getevents_0_4, io_getevents, 0.4)
diff --git a/io_queue_init.c b/io_queue_init.c
new file mode 100644
index 0000000..563d137
--- /dev/null
+++ b/io_queue_init.c
@@ -0,0 +1,33 @@
+/* io_queue_init.c
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <libaio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <errno.h>
+
+#include "syscall.h"
+
+int io_queue_init(int maxevents, io_context_t *ctxp)
+{
+ if (maxevents > 0) {
+ *ctxp = NULL;
+ return io_setup(maxevents, ctxp);
+ }
+ return -EINVAL;
+}
diff --git a/io_queue_release.c b/io_queue_release.c
new file mode 100644
index 0000000..94bbb86
--- /dev/null
+++ b/io_queue_release.c
@@ -0,0 +1,27 @@
+/* io_queue_release.c
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <libaio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <errno.h>
+
+int io_queue_release(io_context_t ctx)
+{
+ return io_destroy(ctx);
+}
diff --git a/io_queue_run.c b/io_queue_run.c
new file mode 100644
index 0000000..e0132f4
--- /dev/null
+++ b/io_queue_run.c
@@ -0,0 +1,39 @@
+/* io_submit
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <libaio.h>
+#include <errno.h>
+#include <stdlib.h>
+#include <time.h>
+
+int io_queue_run(io_context_t ctx)
+{
+ static struct timespec timeout = { 0, 0 };
+ struct io_event event;
+ int ret;
+
+ /* FIXME: batch requests? */
+ while (1 == (ret = io_getevents(ctx, 0, 1, &event, &timeout))) {
+ io_callback_t cb = (io_callback_t)event.data;
+ struct iocb *iocb = event.obj;
+
+ cb(ctx, iocb, event.res, event.res2);
+ }
+
+ return ret;
+}
diff --git a/io_queue_wait.c b/io_queue_wait.c
new file mode 100644
index 0000000..538d2f3
--- /dev/null
+++ b/io_queue_wait.c
@@ -0,0 +1,31 @@
+/* io_submit
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#define NO_SYSCALL_ERRNO
+#include <sys/types.h>
+#include <libaio.h>
+#include <errno.h>
+#include "syscall.h"
+
+struct timespec;
+
+int io_queue_wait_0_4(io_context_t ctx, struct timespec *timeout)
+{
+ return io_getevents(ctx, 0, 0, NULL, timeout);
+}
+DEFSYMVER(io_queue_wait_0_4, io_queue_wait, 0.4)
diff --git a/io_setup.c b/io_setup.c
new file mode 100644
index 0000000..4ba1afc
--- /dev/null
+++ b/io_setup.c
@@ -0,0 +1,23 @@
+/* io_setup
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <errno.h>
+#include <libaio.h>
+#include "syscall.h"
+
+io_syscall2(int, io_setup, io_setup, int, maxevents, io_context_t *, ctxp)
diff --git a/io_submit.c b/io_submit.c
new file mode 100644
index 0000000..e22ba54
--- /dev/null
+++ b/io_submit.c
@@ -0,0 +1,23 @@
+/* io_submit
+ libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#include <errno.h>
+#include <libaio.h>
+#include "syscall.h"
+
+io_syscall3(int, io_submit, io_submit, io_context_t, ctx, long, nr, struct iocb **, iocbs)
diff --git a/libaio.h b/libaio.h
new file mode 100644
index 0000000..b0d5ab9
--- /dev/null
+++ b/libaio.h
@@ -0,0 +1,265 @@
+/* /usr/include/libaio.h
+ *
+ * Copyright 2000,2001,2002 Red Hat, Inc.
+ *
+ * Written by Benjamin LaHaise <bcrl@redhat.com>
+ *
+ * libaio Linux async I/O interface
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+#ifndef __LIBAIO_H
+#define __LIBAIO_H
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <sys/types.h>
+#include <string.h>
+
+struct timespec;
+struct sockaddr;
+struct iovec;
+
+typedef struct io_context *io_context_t;
+
+typedef enum io_iocb_cmd {
+ IO_CMD_PREAD = 0,
+ IO_CMD_PWRITE = 1,
+
+ IO_CMD_FSYNC = 2,
+ IO_CMD_FDSYNC = 3,
+
+ IO_CMD_POLL = 5, /* Never implemented in mainline, see io_prep_poll */
+ IO_CMD_NOOP = 6,
+ IO_CMD_PREADV = 7,
+ IO_CMD_PWRITEV = 8,
+} io_iocb_cmd_t;
+
+/* little endian, 32 bits */
+#if defined(__i386__) || (defined(__arm__) && !defined(__ARMEB__)) || \
+ defined(__sh__) || defined(__bfin__) || defined(__MIPSEL__) || \
+ defined(__cris__)
+#define PADDED(x, y) x; unsigned y
+#define PADDEDptr(x, y) x; unsigned y
+#define PADDEDul(x, y) unsigned long x; unsigned y
+
+/* little endian, 64 bits */
+#elif defined(__ia64__) || defined(__x86_64__) || defined(__alpha__) || \
+ (defined(__aarch64__) && defined(__AARCH64EL__)) || \
+ (defined(__powerpc64__) && defined(_LITTLE_ENDIAN))
+#define PADDED(x, y) x, y
+#define PADDEDptr(x, y) x
+#define PADDEDul(x, y) unsigned long x
+
+/* big endian, 64 bits */
+#elif (defined(__powerpc64__) && defined(_BIG_ENDIAN)) || \
+ defined(__s390x__) || \
+ (defined(__sparc__) && defined(__arch64__)) || \
+ (defined(__aarch64__) && defined(__AARCH64EB__))
+#define PADDED(x, y) unsigned y; x
+#define PADDEDptr(x,y) x
+#define PADDEDul(x, y) unsigned long x
+
+/* big endian, 32 bits */
+#elif defined(__PPC__) || defined(__s390__) || \
+ (defined(__arm__) && defined(__ARMEB__)) || \
+ defined(__sparc__) || defined(__MIPSEB__) || defined(__m68k__) || \
+ defined(__hppa__) || defined(__frv__) || defined(__avr32__)
+#define PADDED(x, y) unsigned y; x
+#define PADDEDptr(x, y) unsigned y; x
+#define PADDEDul(x, y) unsigned y; unsigned long x
+
+#else
+#error endian?
+#endif
+
+struct io_iocb_poll {
+ PADDED(int events, __pad1);
+}; /* result code is the set of result flags or -'ve errno */
+
+struct io_iocb_sockaddr {
+ struct sockaddr *addr;
+ int len;
+}; /* result code is the length of the sockaddr, or -'ve errno */
+
+struct io_iocb_common {
+ PADDEDptr(void *buf, __pad1);
+ PADDEDul(nbytes, __pad2);
+ long long offset;
+ long long __pad3;
+ unsigned flags;
+ unsigned resfd;
+}; /* result code is the amount read or -'ve errno */
+
+struct io_iocb_vector {
+ const struct iovec *vec;
+ int nr;
+ long long offset;
+}; /* result code is the amount read or -'ve errno */
+
+struct iocb {
+ PADDEDptr(void *data, __pad1); /* Return in the io completion event */
+ PADDED(unsigned key, __pad2); /* For use in identifying io requests */
+
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+
+ union {
+ struct io_iocb_common c;
+ struct io_iocb_vector v;
+ struct io_iocb_poll poll;
+ struct io_iocb_sockaddr saddr;
+ } u;
+};
+
+struct io_event {
+ PADDEDptr(void *data, __pad1);
+ PADDEDptr(struct iocb *obj, __pad2);
+ PADDEDul(res, __pad3);
+ PADDEDul(res2, __pad4);
+};
+
+#undef PADDED
+#undef PADDEDptr
+#undef PADDEDul
+
+typedef void (*io_callback_t)(io_context_t ctx, struct iocb *iocb, long res, long res2);
+
+/* library wrappers */
+extern int io_queue_init(int maxevents, io_context_t *ctxp);
+/*extern int io_queue_grow(io_context_t ctx, int new_maxevents);*/
+extern int io_queue_release(io_context_t ctx);
+/*extern int io_queue_wait(io_context_t ctx, struct timespec *timeout);*/
+extern int io_queue_run(io_context_t ctx);
+
+/* Actual syscalls */
+extern int io_setup(int maxevents, io_context_t *ctxp);
+extern int io_destroy(io_context_t ctx);
+extern int io_submit(io_context_t ctx, long nr, struct iocb *ios[]);
+extern int io_cancel(io_context_t ctx, struct iocb *iocb, struct io_event *evt);
+extern int io_getevents(io_context_t ctx_id, long min_nr, long nr, struct io_event *events, struct timespec *timeout);
+
+
+static inline void io_set_callback(struct iocb *iocb, io_callback_t cb)
+{
+ iocb->data = (void *)cb;
+}
+
+static inline void io_prep_pread(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_PREAD;
+ iocb->aio_reqprio = 0;
+ iocb->u.c.buf = buf;
+ iocb->u.c.nbytes = count;
+ iocb->u.c.offset = offset;
+}
+
+static inline void io_prep_pwrite(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_PWRITE;
+ iocb->aio_reqprio = 0;
+ iocb->u.c.buf = buf;
+ iocb->u.c.nbytes = count;
+ iocb->u.c.offset = offset;
+}
+
+static inline void io_prep_preadv(struct iocb *iocb, int fd, const struct iovec *iov, int iovcnt, long long offset)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_PREADV;
+ iocb->aio_reqprio = 0;
+ iocb->u.c.buf = (void *)iov;
+ iocb->u.c.nbytes = iovcnt;
+ iocb->u.c.offset = offset;
+}
+
+static inline void io_prep_pwritev(struct iocb *iocb, int fd, const struct iovec *iov, int iovcnt, long long offset)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_PWRITEV;
+ iocb->aio_reqprio = 0;
+ iocb->u.c.buf = (void *)iov;
+ iocb->u.c.nbytes = iovcnt;
+ iocb->u.c.offset = offset;
+}
+
+/* Jeff Moyer says this was implemented in Red Hat AS2.1 and RHEL3.
+ * AFAICT, it was never in mainline, and should not be used. --RR */
+static inline void io_prep_poll(struct iocb *iocb, int fd, int events)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_POLL;
+ iocb->aio_reqprio = 0;
+ iocb->u.poll.events = events;
+}
+
+static inline int io_poll(io_context_t ctx, struct iocb *iocb, io_callback_t cb, int fd, int events)
+{
+ io_prep_poll(iocb, fd, events);
+ io_set_callback(iocb, cb);
+ return io_submit(ctx, 1, &iocb);
+}
+
+static inline void io_prep_fsync(struct iocb *iocb, int fd)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_FSYNC;
+ iocb->aio_reqprio = 0;
+}
+
+static inline int io_fsync(io_context_t ctx, struct iocb *iocb, io_callback_t cb, int fd)
+{
+ io_prep_fsync(iocb, fd);
+ io_set_callback(iocb, cb);
+ return io_submit(ctx, 1, &iocb);
+}
+
+static inline void io_prep_fdsync(struct iocb *iocb, int fd)
+{
+ memset(iocb, 0, sizeof(*iocb));
+ iocb->aio_fildes = fd;
+ iocb->aio_lio_opcode = IO_CMD_FDSYNC;
+ iocb->aio_reqprio = 0;
+}
+
+static inline int io_fdsync(io_context_t ctx, struct iocb *iocb, io_callback_t cb, int fd)
+{
+ io_prep_fdsync(iocb, fd);
+ io_set_callback(iocb, cb);
+ return io_submit(ctx, 1, &iocb);
+}
+
+static inline void io_set_eventfd(struct iocb *iocb, int eventfd)
+{
+ iocb->u.c.flags |= (1 << 0) /* IOCB_FLAG_RESFD */;
+ iocb->u.c.resfd = eventfd;
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __LIBAIO_H */
diff --git a/libaio.lds b/libaio.lds
new file mode 100644
index 0000000..dc37725
--- /dev/null
+++ b/libaio.lds
@@ -0,0 +1,22 @@
+LIBAIO_0.1 {
+ global:
+ io_queue_init;
+ io_queue_run;
+ io_queue_wait;
+ io_queue_release;
+ io_cancel;
+ io_submit;
+ io_getevents;
+ local:
+ *;
+
+};
+
+LIBAIO_0.4 {
+ global:
+ io_setup;
+ io_destroy;
+ io_cancel;
+ io_getevents;
+ io_queue_wait;
+} LIBAIO_0.1;
diff --git a/libaio.map b/libaio.map
new file mode 100644
index 0000000..dc37725
--- /dev/null
+++ b/libaio.map
@@ -0,0 +1,22 @@
+LIBAIO_0.1 {
+ global:
+ io_queue_init;
+ io_queue_run;
+ io_queue_wait;
+ io_queue_release;
+ io_cancel;
+ io_submit;
+ io_getevents;
+ local:
+ *;
+
+};
+
+LIBAIO_0.4 {
+ global:
+ io_setup;
+ io_destroy;
+ io_cancel;
+ io_getevents;
+ io_queue_wait;
+} LIBAIO_0.1;
diff --git a/libaio.spec b/libaio.spec
new file mode 100644
index 0000000..32b0d42
--- /dev/null
+++ b/libaio.spec
@@ -0,0 +1,196 @@
+Name: libaio
+Version: 0.3.109
+Release: 1
+Summary: Linux-native asynchronous I/O access library
+License: LGPL
+Group: System Environment/Libraries
+Source: %{name}-%{version}.tar.gz
+BuildRoot: %{_tmppath}/%{name}-root
+# Fix ExclusiveArch as we implement this functionality on more architectures
+ExclusiveArch: i386 x86_64 ia64 s390 s390x ppc ppc64 ppc64pseries ppc64iseries alpha alphaev6 %{arm}
+
+%description
+The Linux-native asynchronous I/O facility ("async I/O", or "aio") has a
+richer API and capability set than the simple POSIX async I/O facility.
+This library, libaio, provides the Linux-native API for async I/O.
+The POSIX async I/O facility requires this library in order to provide
+kernel-accelerated async I/O capabilities, as do applications which
+require the Linux-native async I/O API.
+
+%package devel
+Summary: Development files for Linux-native asynchronous I/O access
+Group: Development/System
+Requires: libaio
+Provides: libaio.so.1
+
+%description devel
+This package provides header files to include and libraries to link with
+for the Linux-native asynchronous I/O facility ("async I/O", or "aio").
+
+%prep
+%setup
+
+%build
+make
+
+%install
+[ "$RPM_BUILD_ROOT" != "/" ] && rm -rf $RPM_BUILD_ROOT
+
+make install prefix=$RPM_BUILD_ROOT/usr \
+ libdir=$RPM_BUILD_ROOT/%{_libdir} \
+ root=$RPM_BUILD_ROOT
+
+%clean
+[ "$RPM_BUILD_ROOT" != "/" ] && rm -rf $RPM_BUILD_ROOT
+
+%post -p /sbin/ldconfig
+
+%postun -p /sbin/ldconfig
+
+%files
+%defattr(-,root,root)
+%attr(0755,root,root) %{_libdir}/libaio.so.*
+%doc COPYING TODO
+
+%files devel
+%defattr(-,root,root)
+%attr(0644,root,root) %{_includedir}/*
+%attr(0755,root,root) %{_libdir}/libaio.so
+%attr(0644,root,root) %{_libdir}/libaio.a
+
+%changelog
+* Tue Jun 9 2009 Jeff Moyer <jmoyer@redhat.com> - 0.3.108-1
+- add ARM architecture support (grabbed from Debian arches tree)
+- replace check of __i386__ with __LP64__ in test harness (Jeff Moyer)
+- change the .spec's copyright tag to license
+
+* Wed Jan 9 2008 Jeff Moyer <jmoyer@redhat.com> - 0.3.107-1
+- Fix the test harness (Rusty Russell)
+- Add eventfd support (Rusty Russell)
+
+* Tue Jan 3 2006 Jeff Moyer <jmoyer@redhat.com> - 0.3.106-1
+- Add a .proc directive for the ia64_aio_raw_syscall macro. This sounds a lot
+ like the previous entry, but that one fixed the __ia64_raw_syscall macro,
+ located in syscall-ia64.h. This macro is in raw_syscall.c, which pretty much
+ only exists for ia64. This bug prevented the package from building with
+ newer version of gcc.
+
+* Mon Aug 1 2005 Jeff Moyer <jmoyer@redhat.com> - 0.3.105-1
+- Add a .proc directive for the ia64 raw syscall macro.
+
+* Fri Apr 1 2005 Jeff Moyer <jmoyer@redhat.com> - 0.3.104-1
+- Add Alpha architecture support. (Sergey Tikhonov <tsv@solvo.ru>)
+
+* Tue Jan 25 2005 Jeff Moyer <jmoyer@redhat.com> - 0.3.103-1
+- Fix SONAME breakage. In changing file names around, I also changed the
+ SONAME, which is a no no.
+
+* Thu Oct 14 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.102-1
+- S390 asm had a bug; I forgot to update the clobber list. Lucky for me,
+ newer compilers complain about such things.
+- Also update the s390 asm to look more like the new kernel variants.
+
+* Wed Oct 13 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.101-1
+- Revert syscall return values to be -ERRNO. This was an inadvertant bug
+ introduced when clobber lists changed.
+- add ppc64pseries and ppc64iseries to exclusivearch
+
+* Tue Sep 14 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.100-1
+- Switch around the tests for _PPC_ and _powerpc64_ so that the ppc64
+ platforms get the right padding.
+
+* Wed Jul 14 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.99-4
+- Ok, there was a race in moving the cvs module. Someone rebuild from
+ the old cvs into fc3. *sigh* bumping rev.
+
+* Wed Jul 14 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.99-3
+- Actually provide libaio.so.1.
+
+* Tue Mar 30 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.99-2
+- Apparently the 0.3.93 patch was not meant for 0.3.96. Backed it out.
+
+* Tue Mar 30 2004 Jeff Moyer <jmoyer@redhat.com> - 0.3.99-1
+- Fix compat calls.
+- make library .so.1.0.0 and make symlinks properly.
+- Fix header file for inclusion in c++ code.
+
+* Thu Feb 26 2004 Jeff Moyer <jmoyer@redhat.com> 0.3.98-2
+- bah. fix version nr in changelog.
+
+* Thu Feb 26 2004 Jeff Moyer <jmoyer@redhat.com> 0.3.98-1
+- fix compiler warnings.
+
+* Thu Feb 26 2004 Jeff Moyer <jmoyer@redhat.com> 0.3.97-2
+- make srpm was using rpm to do a build. changed that to use rpmbuild if
+ it exists, and fallback to rpm if it doesn't.
+
+* Tue Feb 24 2004 Jeff Moyer <jmoyer@redhat.com> 0.3.97-1
+- Use libc syscall(2) instead of rolling our own calling mechanism. This
+ change is inspired due to a failure to build with newer gcc, since clobber
+ lists were wrong.
+- Add -fpic to the CFLAGS for all architectures. Should address bz #109457.
+- change a #include from <linux/types.h> to <sys/types.h>. Fixes a build
+ issue on s390.
+
+* Wed Jul 7 2003 Bill Nottingham <notting@redhat.com> 0.3.96-3
+- fix paths on lib64 arches
+
+* Wed Jun 18 2003 Michael K. Johnson <johnsonm@redhat.com> 0.3.96-2
+- optimization in io_getevents from Arjan van de Ven in 0.3.96-1
+- deal with ia64 in 0.3.96-2
+
+* Wed May 28 2003 Michael K. Johnson <johnsonm@redhat.com> 0.3.95-1
+- ppc bugfix from Julie DeWandel
+
+* Tue May 20 2003 Michael K. Johnson <johnsonm@redhat.com> 0.3.94-1
+- symbol versioning fix from Ulrich Drepper
+
+* Mon Jan 27 2003 Benjamin LaHaise <bcrl@redhat.com>
+- bump to 0.3.93-3 for rebuild.
+
+* Mon Dec 16 2002 Benjamin LaHaise <bcrl@redhat.com>
+- libaio 0.3.93 test release
+- add powerpc support from Gianni Tedesco <gianni@ecsc.co.uk>
+- add s/390 support from Arnd Bergmann <arnd@bergmann-dalldorf.de>
+
+* Fri Sep 12 2002 Benjamin LaHaise <bcrl@redhat.com>
+- libaio 0.3.92 test release
+- build on x86-64
+
+* Thu Sep 12 2002 Benjamin LaHaise <bcrl@redhat.com>
+- libaio 0.3.91 test release
+- build on ia64
+- remove libredhat-kernel from the .spec file
+
+* Thu Sep 5 2002 Benjamin LaHaise <bcrl@redhat.com>
+- libaio 0.3.90 test release
+
+* Mon Apr 29 2002 Benjamin LaHaise <bcrl@redhat.com>
+- add requires initscripts >= 6.47-1 to get boot time libredhat-kernel
+ linkage correct.
+- typo fix
+
+* Thu Apr 25 2002 Benjamin LaHaise <bcrl@redhat.com>
+- make /usr/lib/libredhat-kernel.so point to /lib/libredhat-kernel.so.1.0.0
+
+* Mon Apr 15 2002 Tim Powers <timp@redhat.com>
+- make the post scriptlet not use /bin/sh
+
+* Sat Apr 12 2002 Benjamin LaHaise <bcrl@redhat.com>
+- add /lib/libredhat-kernel* to %files.
+
+* Fri Apr 12 2002 Benjamin LaHaise <bcrl@redhat.com>
+- make the dummy install as /lib/libredhat-kernel.so.1.0.0 so
+ that ldconfig will link against it if no other is installed.
+
+* Tue Jan 22 2002 Benjamin LaHaise <bcrl@redhat.com>
+- add io_getevents
+
+* Tue Jan 22 2002 Michael K. Johnson <johnsonm@redhat.com>
+- Make linker happy with /usr/lib symlink for libredhat-kernel.so
+
+* Mon Jan 21 2002 Michael K. Johnson <johnsonm@redhat.com>
+- Added stub library
+
+* Sun Jan 20 2002 Michael K. Johnson <johnsonm@redhat.com>
+- Initial packaging
diff --git a/man/aio.3 b/man/aio.3
new file mode 100644
index 0000000..6dc3c63
--- /dev/null
+++ b/man/aio.3
@@ -0,0 +1,315 @@
+.TH aio 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio \- Asynchronous IO
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.fi
+.SH DESCRIPTION
+The POSIX.1b standard defines a new set of I/O operations which can
+significantly reduce the time an application spends waiting at I/O. The
+new functions allow a program to initiate one or more I/O operations and
+then immediately resume normal work while the I/O operations are
+executed in parallel. This functionality is available if the
+.IR "unistd.h"
+file defines the symbol
+.B "_POSIX_ASYNCHRONOUS_IO"
+.
+
+These functions are part of the library with realtime functions named
+.IR "librt"
+. They are not actually part of the
+.IR "libc"
+binary.
+The implementation of these functions can be done using support in the
+kernel (if available) or using an implementation based on threads at
+userlevel. In the latter case it might be necessary to link applications
+with the thread library
+.IR "libpthread"
+in addition to
+.IR "librt"
+and
+.IR "libaio"
+.
+
+All AIO operations operate on files which were opened previously. There
+might be arbitrarily many operations running for one file. The
+asynchronous I/O operations are controlled using a data structure named
+.IR "struct aiocb"
+It is defined in
+.IR "aio.h"
+ as follows.
+
+.nf
+struct aiocb
+{
+ int aio_fildes; /* File desriptor. */
+ int aio_lio_opcode; /* Operation to be performed. */
+ int aio_reqprio; /* Request priority offset. */
+ volatile void *aio_buf; /* Location of buffer. */
+ size_t aio_nbytes; /* Length of transfer. */
+ struct sigevent aio_sigevent; /* Signal number and value. */
+
+ /* Internal members. */
+ struct aiocb *__next_prio;
+ int __abs_prio;
+ int __policy;
+ int __error_code;
+ __ssize_t __return_value;
+
+#ifndef __USE_FILE_OFFSET64
+ __off_t aio_offset; /* File offset. */
+ char __pad[sizeof (__off64_t) - sizeof (__off_t)];
+#else
+ __off64_t aio_offset; /* File offset. */
+#endif
+ char __unused[32];
+};
+
+.fi
+The POSIX.1b standard mandates that the
+.IR "struct aiocb"
+structure
+contains at least the members described in the following table. There
+might be more elements which are used by the implementation, but
+depending upon these elements is not portable and is highly deprecated.
+
+.TP
+.IR "int aio_fildes"
+This element specifies the file descriptor to be used for the
+operation. It must be a legal descriptor, otherwise the operation will
+fail.
+
+The device on which the file is opened must allow the seek operation.
+I.e., it is not possible to use any of the AIO operations on devices
+like terminals where an
+.IR "lseek"
+ call would lead to an error.
+.TP
+.IR "off_t aio_offset"
+This element specifies the offset in the file at which the operation (input
+or output) is performed. Since the operations are carried out in arbitrary
+order and more than one operation for one file descriptor can be
+started, one cannot expect a current read/write position of the file
+descriptor.
+.TP
+.IR "volatile void *aio_buf"
+This is a pointer to the buffer with the data to be written or the place
+where the read data is stored.
+.TP
+.IR "size_t aio_nbytes"
+This element specifies the length of the buffer pointed to by
+.IR "aio_buf"
+.
+.TP
+.IR "int aio_reqprio"
+If the platform has defined
+.B "_POSIX_PRIORITIZED_IO"
+and
+.B "_POSIX_PRIORITY_SCHEDULING"
+, the AIO requests are
+processed based on the current scheduling priority. The
+.IR "aio_reqprio"
+element can then be used to lower the priority of the
+AIO operation.
+.TP
+.IR "struct sigevent aio_sigevent"
+This element specifies how the calling process is notified once the
+operation terminates. If the
+.IR "sigev_notify"
+element is
+.B "SIGEV_NONE"
+, no notification is sent. If it is
+.B "SIGEV_SIGNAL"
+,
+the signal determined by
+.IR "sigev_signo"
+is sent. Otherwise,
+.IR "sigev_notify"
+must be
+.B "SIGEV_THREAD"
+. In this case, a thread
+is created which starts executing the function pointed to by
+.IR "sigev_notify_function"
+.
+.TP
+.IR "int aio_lio_opcode"
+This element is only used by the
+.IR "lio_listio"
+ and
+.IR "lio_listio64"
+ functions. Since these functions allow an
+arbitrary number of operations to start at once, and each operation can be
+input or output (or nothing), the information must be stored in the
+control block. The possible values are:
+.TP
+.B "LIO_READ"
+Start a read operation. Read from the file at position
+.IR "aio_offset"
+ and store the next
+.IR "aio_nbytes"
+ bytes in the
+buffer pointed to by
+.IR "aio_buf"
+.
+.TP
+.B "LIO_WRITE"
+Start a write operation. Write
+.IR "aio_nbytes"
+bytes starting at
+.IR "aio_buf"
+into the file starting at position
+.IR "aio_offset"
+.
+.TP
+.B "LIO_NOP"
+Do nothing for this control block. This value is useful sometimes when
+an array of
+.IR "struct aiocb"
+values contains holes, i.e., some of the
+values must not be handled although the whole array is presented to the
+.IR "lio_listio"
+function.
+
+When the sources are compiled using
+.B "_FILE_OFFSET_BITS == 64"
+on a
+32 bit machine, this type is in fact
+.IR "struct aiocb64"
+, since the LFS
+interface transparently replaces the
+.IR "struct aiocb"
+definition.
+.PP
+For use with the AIO functions defined in the LFS, there is a similar type
+defined which replaces the types of the appropriate members with larger
+types but otherwise is equivalent to
+.IR "struct aiocb"
+. Particularly,
+all member names are the same.
+
+.nf
+/* The same for the 64bit offsets. Please note that the members aio_fildes
+ to __return_value have to be the same in aiocb and aiocb64. */
+#ifdef __USE_LARGEFILE64
+struct aiocb64
+{
+ int aio_fildes; /* File desriptor. */
+ int aio_lio_opcode; /* Operation to be performed. */
+ int aio_reqprio; /* Request priority offset. */
+ volatile void *aio_buf; /* Location of buffer. */
+ size_t aio_nbytes; /* Length of transfer. */
+ struct sigevent aio_sigevent; /* Signal number and value. */
+
+ /* Internal members. */
+ struct aiocb *__next_prio;
+ int __abs_prio;
+ int __policy;
+ int __error_code;
+ __ssize_t __return_value;
+
+ __off64_t aio_offset; /* File offset. */
+ char __unused[32];
+};
+
+.fi
+.TP
+.IR "int aio_fildes"
+This element specifies the file descriptor which is used for the
+operation. It must be a legal descriptor since otherwise the operation
+fails for obvious reasons.
+The device on which the file is opened must allow the seek operation.
+I.e., it is not possible to use any of the AIO operations on devices
+like terminals where an
+.IR "lseek"
+ call would lead to an error.
+.TP
+.IR "off64_t aio_offset"
+This element specifies at which offset in the file the operation (input
+or output) is performed. Since the operation are carried in arbitrary
+order and more than one operation for one file descriptor can be
+started, one cannot expect a current read/write position of the file
+descriptor.
+.TP
+.IR "volatile void *aio_buf"
+This is a pointer to the buffer with the data to be written or the place
+where the read data is stored.
+.TP
+.IR "size_t aio_nbytes"
+This element specifies the length of the buffer pointed to by
+.IR "aio_buf"
+.
+.TP
+.IR "int aio_reqprio"
+If for the platform
+.B "_POSIX_PRIORITIZED_IO"
+and
+.B "_POSIX_PRIORITY_SCHEDULING"
+are defined the AIO requests are
+processed based on the current scheduling priority. The
+.IR "aio_reqprio"
+element can then be used to lower the priority of the
+AIO operation.
+.TP
+.IR "struct sigevent aio_sigevent"
+This element specifies how the calling process is notified once the
+operation terminates. If the
+.IR "sigev_notify"
+, element is
+.B "SIGEV_NONE"
+no notification is sent. If it is
+.B "SIGEV_SIGNAL"
+,
+the signal determined by
+.IR "sigev_signo"
+is sent. Otherwise,
+.IR "sigev_notify"
+ must be
+.B "SIGEV_THREAD"
+in which case a thread
+which starts executing the function pointed to by
+.IR "sigev_notify_function"
+.
+.TP
+.IR "int aio_lio_opcode"
+This element is only used by the
+.IR "lio_listio"
+and
+.IR "lio_listio64"
+functions. Since these functions allow an
+arbitrary number of operations to start at once, and since each operation can be
+input or output (or nothing), the information must be stored in the
+control block. See the description of
+.IR "struct aiocb"
+for a description
+of the possible values.
+.PP
+When the sources are compiled using
+.B "_FILE_OFFSET_BITS == 64"
+on a
+32 bit machine, this type is available under the name
+.IR "struct aiocb64"
+, since the LFS transparently replaces the old interface.
+.SH "RETURN VALUES"
+.SH ERRORS
+.SH "SEE ALSO"
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_cancel.3 b/man/aio_cancel.3
new file mode 100644
index 0000000..502c83c
--- /dev/null
+++ b/man/aio_cancel.3
@@ -0,0 +1,137 @@
+.TH aio_cancel 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_cancel - Cancel asynchronous I/O requests
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_cancel (int fildes " , struct aiocb *aiocbp " )"
+.fi
+.SH DESCRIPTION
+When one or more requests are asynchronously processed, it might be
+useful in some situations to cancel a selected operation, e.g., if it
+becomes obvious that the written data is no longer accurate and would
+have to be overwritten soon. As an example, assume an application, which
+writes data in files in a situation where new incoming data would have
+to be written in a file which will be updated by an enqueued request.
+The POSIX AIO implementation provides such a function, but this function
+is not capable of forcing the cancellation of the request. It is up to the
+implementation to decide whether it is possible to cancel the operation
+or not. Therefore using this function is merely a hint.
+.B "The libaio implementation does not implement the cancel operation in the"
+.B "POSIX libraries".
+.PP
+The
+.IR aio_cancel
+function can be used to cancel one or more
+outstanding requests. If the
+.IR aiocbp
+parameter is
+.IR NULL
+, the
+function tries to cancel all of the outstanding requests which would process
+the file descriptor
+.IR fildes
+(i.e., whose
+.IR aio_fildes
+member
+is
+.IR fildes
+). If
+.IR aiocbp is not
+.IR NULL
+,
+.IR aio_cancel
+attempts to cancel the specific request pointed to by
+.IR aiocbp.
+
+For requests which were successfully canceled, the normal notification
+about the termination of the request should take place. I.e., depending
+on the
+.IR "struct sigevent"
+object which controls this, nothing
+happens, a signal is sent or a thread is started. If the request cannot
+be canceled, it terminates the usual way after performing the operation.
+After a request is successfully canceled, a call to
+.IR aio_error
+with
+a reference to this request as the parameter will return
+.B ECANCELED
+and a call to
+.IR aio_return
+will return
+.IR -1.
+If the request wasn't canceled and is still running the error status is
+still
+.B EINPROGRESS.
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+, this
+function is in fact
+.IR aio_cancel64
+since the LFS interface
+transparently replaces the normal implementation.
+
+.SH "RETURN VALUES"
+.TP
+.B AIO_CANCELED
+If there were
+requests which haven't terminated and which were successfully canceled.
+.TP
+.B AIO_NOTCANCELED
+If there is one or more requests left which couldn't be canceled,
+. In this case
+.IR aio_error
+must be used to find out which of the, perhaps multiple, requests (in
+.IR aiocbp
+is
+.IR NULL
+) weren't successfully canceled.
+.TP
+.B AIO_ALLDONE
+If all
+requests already terminated at the time
+.IR aio_cancel
+is called the
+return value is
+.
+.SH ERRORS
+If an error occurred during the execution of
+.IR aio_cancel
+the
+function returns
+.IR -1
+and sets
+.IR errno
+to one of the following
+values.
+.TP
+.B EBADF
+The file descriptor
+.IR fildes
+is not valid.
+.TP
+.B ENOSYS
+.IR aio_cancel
+is not implemented.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_cancel64.3 b/man/aio_cancel64.3
new file mode 100644
index 0000000..ede775b
--- /dev/null
+++ b/man/aio_cancel64.3
@@ -0,0 +1,50 @@
+.TH aio_cancel64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_cancel64 \- Cancel asynchronous I/O requests
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_cancel64 (int fildes, struct aiocb64 *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to
+.IR aio_cancel
+with the only difference
+that the argument is a reference to a variable of type
+.IR struct aiocb64
+.
+
+When the sources are compiled with
+.IR _FILE_OFFSET_BITS == 64
+, this
+function is available under the name
+.IR aio_cancel
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+See aio_cancel(3).
+.SH ERRORS
+See aio_cancel(3).
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_error.3 b/man/aio_error.3
new file mode 100644
index 0000000..12b82cf
--- /dev/null
+++ b/man/aio_error.3
@@ -0,0 +1,81 @@
+.TH aio_error 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_error \- Getting the Status of AIO Operations
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_error (const struct aiocb *aiocbp)"
+.fi
+.SH DESCRIPTION
+The function
+.IR aio_error
+determines the error state of the request described by the
+.IR "struct aiocb"
+variable pointed to by
+.I aiocbp
+.
+
+When the operation is performed truly asynchronously (as with
+.IR "aio_read"
+and
+.IR "aio_write"
+and with
+.IR "lio_listio"
+when the mode is
+.IR "LIO_NOWAIT"
+), one sometimes needs to know whether a
+specific request already terminated and if so, what the result was.
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this function is in fact
+.IR "aio_error64"
+since the LFS interface transparently replaces the normal implementation.
+.SH "RETURN VALUES"
+If the request has not yet terminated the value returned is always
+.IR "EINPROGRESS"
+. Once the request has terminated the value
+.IR "aio_error"
+returns is either
+.I 0
+if the request completed successfully or it returns the value which would be stored in the
+.IR "errno"
+variable if the request would have been done using
+.IR "read"
+,
+.IR "write"
+, or
+.IR "fsync"
+.
+.SH ERRORS
+.TP
+.IR "ENOSYS"
+if it is not implemented. It
+could also return
+.TP
+.IR "EINVAL"
+if the
+.I aiocbp
+parameter does not
+refer to an asynchronous operation whose return status is not yet known.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_error64.3 b/man/aio_error64.3
new file mode 100644
index 0000000..3333161
--- /dev/null
+++ b/man/aio_error64.3
@@ -0,0 +1,64 @@
+.TH aio_error64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_error64 \- Return errors
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_error64 (const struct aiocb64 *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to
+.IR aio_error
+with the only difference
+that the argument is a reference to a variable of type
+.IR "struct aiocb64".
+.PP
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is available under the name
+.IR aio_error
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+If the request has not yet terminated the value returned is always
+.IR "EINPROGRESS"
+. Once the request has terminated the value
+.IR "aio_error"
+returns is either
+.I 0
+if the request completed successfully or it returns the value which would be stored in the
+.IR "errno"
+variable if the request would have been done using
+.IR "read"
+,
+.IR "write"
+, or
+.IR "fsync"
+.
+.SH ERRORS
+See
+.IR aio_error(3).
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_fsync.3 b/man/aio_fsync.3
new file mode 100644
index 0000000..637f0f6
--- /dev/null
+++ b/man/aio_fsync.3
@@ -0,0 +1,139 @@
+.TH aio_fsync 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_fsync \- Synchronize a file's complete in-core state with that on disk
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_fsync (int op, struct aiocb aiocbp)"
+.fi
+.SH DESCRIPTION
+.PP
+When dealing with asynchronous operations it is sometimes necessary to
+get into a consistent state. This would mean for AIO that one wants to
+know whether a certain request or a group of request were processed.
+This could be done by waiting for the notification sent by the system
+after the operation terminated, but this sometimes would mean wasting
+resources (mainly computation time). Instead POSIX.1b defines two
+functions which will help with most kinds of consistency.
+.PP
+The
+.IR aio_fsync
+and
+.IR "aio_fsync64"
+functions are only available
+if the symbol
+.IR "_POSIX_SYNCHRONIZED_IO"
+is defined in
+.I unistd.h
+.
+
+Calling this function forces all I/O operations operating queued at the
+time of the function call operating on the file descriptor
+.IR "aiocbp->aio_fildes"
+into the synchronized I/O completion state . The
+.IR "aio_fsync"
+function returns
+immediately but the notification through the method described in
+.IR "aiocbp->aio_sigevent"
+will happen only after all requests for this
+file descriptor have terminated and the file is synchronized. This also
+means that requests for this very same file descriptor which are queued
+after the synchronization request are not affected.
+
+If
+.IR "op"
+is
+.IR "O_DSYNC"
+the synchronization happens as with a call
+to
+.IR "fdatasync"
+. Otherwise
+.IR "op"
+should be
+.IR "O_SYNC"
+and
+the synchronization happens as with
+.IR "fsync"
+.
+
+As long as the synchronization has not happened, a call to
+.IR "aio_error"
+with the reference to the object pointed to by
+.IR "aiocbp"
+returns
+.IR "EINPROGRESS"
+. Once the synchronization is
+done
+.IR "aio_error"
+return
+.IR 0
+if the synchronization was not
+successful. Otherwise the value returned is the value to which the
+.IR "fsync"
+or
+.IR "fdatasync"
+function would have set the
+.IR "errno"
+variable. In this case nothing can be assumed about the
+consistency for the data written to this file descriptor.
+
+.SH "RETURN VALUES"
+The return value of this function is
+.IR 0
+if the request was
+successfully enqueued. Otherwise the return value is
+.IR -1
+and
+.IR "errno".
+.SH ERRORS
+.TP
+.B EAGAIN
+The request could not be enqueued due to temporary lack of resources.
+.TP
+.B EBADF
+The file descriptor
+.IR "aiocbp->aio_fildes"
+is not valid or not open
+for writing.
+.TP
+.B EINVAL
+The implementation does not support I/O synchronization or the
+.IR "op"
+parameter is other than
+.IR "O_DSYNC"
+and
+.IR "O_SYNC"
+.
+.TP
+.B ENOSYS
+This function is not implemented.
+.PP
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+ this
+function is in fact
+.IR "aio_return64"
+since the LFS interface
+transparently replaces the normal implementation.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_fsync64.3 b/man/aio_fsync64.3
new file mode 100644
index 0000000..5dce22d
--- /dev/null
+++ b/man/aio_fsync64.3
@@ -0,0 +1,51 @@
+.TH aio_fsync64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_fsync64 \- Synchronize a file's complete in-core state with that on disk
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_fsync64 (int op, struct aiocb64 *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to
+.IR aio_fsync
+with the only difference
+that the argument is a reference to a variable of type
+.IR "struct aiocb64".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is available under the name
+.IR aio_fsync
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+See
+.IR aio_fsync.
+.SH ERRORS
+See
+.IR aio_fsync.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_init.3 b/man/aio_init.3
new file mode 100644
index 0000000..3b0ec95
--- /dev/null
+++ b/man/aio_init.3
@@ -0,0 +1,96 @@
+.TH aio_init 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_init \- How to optimize the AIO implementation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "void aio_init (const struct aioinit *init)"
+.fi
+.SH DESCRIPTION
+
+The POSIX standard does not specify how the AIO functions are
+implemented. They could be system calls, but it is also possible to
+emulate them at userlevel.
+
+At the point of this writing, the available implementation is a userlevel
+implementation which uses threads for handling the enqueued requests.
+While this implementation requires making some decisions about
+limitations, hard limitations are something which is best avoided
+in the GNU C library. Therefore, the GNU C library provides a means
+for tuning the AIO implementation according to the individual use.
+
+.BI "struct aioinit"
+.PP
+This data type is used to pass the configuration or tunable parameters
+to the implementation. The program has to initialize the members of
+this struct and pass it to the implementation using the
+.IR aio_init
+function.
+.TP
+.B "int aio_threads"
+This member specifies the maximal number of threads which may be used
+at any one time.
+.TP
+.B "int aio_num"
+This number provides an estimate on the maximal number of simultaneously
+enqueued requests.
+.TP
+.B "int aio_locks"
+Unused.
+.TP
+.B "int aio_usedba"
+Unused.
+.TP
+.B "int aio_debug"
+Unused.
+.TP
+.B "int aio_numusers"
+Unused.
+.TP
+.B "int aio_reserved[2]"
+Unused.
+.PP
+This function must be called before any other AIO function. Calling it
+is completely voluntary, as it is only meant to help the AIO
+implementation perform better.
+
+Before calling the
+.IR aio_init
+, function the members of a variable of
+type
+.IR "struct aioinit"
+must be initialized. Then a reference to
+this variable is passed as the parameter to
+.IR aio_init
+which itself
+may or may not pay attention to the hints.
+
+It is a extension which follows a proposal from the SGI implementation in
+.IR Irix 6
+. It is not covered by POSIX.1b or Unix98.
+.SH "RETURN VALUES"
+The function has no return value.
+.SH ERRORS
+The function has no error cases defined.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_read.3 b/man/aio_read.3
new file mode 100644
index 0000000..5bcb6c8
--- /dev/null
+++ b/man/aio_read.3
@@ -0,0 +1,146 @@
+.TH aio_read 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_read \- Initiate an asynchronous read operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_read (struct aiocb *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function initiates an asynchronous read operation. It
+immediately returns after the operation was enqueued or when an
+error was encountered.
+
+The first
+.IR "aiocbp->aio_nbytes"
+bytes of the file for which
+.IR "aiocbp->aio_fildes"
+is a descriptor are written to the buffer
+starting at
+.IR "aiocbp->aio_buf"
+. Reading starts at the absolute
+position
+.IR "aiocbp->aio_offset"
+in the file.
+
+If prioritized I/O is supported by the platform the
+.IR "aiocbp->aio_reqprio"
+value is used to adjust the priority before
+the request is actually enqueued.
+
+The calling process is notified about the termination of the read
+request according to the
+.IR "aiocbp->aio_sigevent"
+value.
+
+.SH "RETURN VALUES"
+When
+.IR "aio_read"
+returns, the return value is zero if no error
+occurred that can be found before the process is enqueued. If such an
+early error is found, the function returns
+.IR -1
+and sets
+.IR "errno".
+
+.PP
+If
+.IR "aio_read"
+returns zero, the current status of the request
+can be queried using
+.IR "aio_error"
+and
+.IR "aio_return"
+functions.
+As long as the value returned by
+.IR "aio_error"
+is
+.IR "EINPROGRESS"
+the operation has not yet completed. If
+.IR "aio_error"
+returns zero,
+the operation successfully terminated, otherwise the value is to be
+interpreted as an error code. If the function terminated, the result of
+the operation can be obtained using a call to
+.IR "aio_return"
+. The
+returned value is the same as an equivalent call to
+.IR "read"
+would
+have returned.
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is in fact
+.IR "aio_read64"
+since the LFS interface transparently
+replaces the normal implementation.
+
+.SH ERRORS
+In the case of an early error:
+.TP
+.B EAGAIN
+The request was not enqueued due to (temporarily) exceeded resource
+limitations.
+.TP
+.B ENOSYS
+The
+.IR "aio_read"
+function is not implemented.
+.TP
+.B EBADF
+The
+.IR "aiocbp->aio_fildes"
+descriptor is not valid. This condition
+need not be recognized before enqueueing the request and so this error
+might also be signaled asynchronously.
+.TP
+.B EINVAL
+The
+.IR "aiocbp->aio_offset"
+or
+.IR "aiocbp->aio_reqpiro"
+value is
+invalid. This condition need not be recognized before enqueueing the
+request and so this error might also be signaled asynchronously.
+
+.PP
+In the case of a normal return, possible error codes returned by
+.IR "aio_error"
+are:
+.TP
+.B EBADF
+The
+.IR "aiocbp->aio_fildes"
+descriptor is not valid.
+.TP
+.B ECANCELED
+The operation was canceled before the operation was finished
+.TP
+.B EINVAL
+The
+.IR "aiocbp->aio_offset"
+value is invalid.
+.PP
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_read64.3 b/man/aio_read64.3
new file mode 100644
index 0000000..8e407a5
--- /dev/null
+++ b/man/aio_read64.3
@@ -0,0 +1,60 @@
+.TH aio_read64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_read64 \- Initiate an asynchronous read operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_read64 (struct aiocb *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to the
+.IR "aio_read"
+function. The only
+difference is that on
+.IR "32 bit"
+machines, the file descriptor should
+be opened in the large file mode. Internally,
+.IR "aio_read64"
+uses
+functionality equivalent to
+.IR "lseek64"
+to position the file descriptor correctly for the reading,
+as opposed to
+.IR "lseek"
+functionality used in
+.IR "aio_read".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+, this
+function is available under the name
+.IR "aio_read"
+and so transparently
+replaces the interface for small files on 32 bit machines.
+.SH "RETURN VALUES"
+See
+.IR aio_read.
+.SH ERRORS
+See
+.IR aio_read.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_return.3 b/man/aio_return.3
new file mode 100644
index 0000000..1e3335f
--- /dev/null
+++ b/man/aio_return.3
@@ -0,0 +1,71 @@
+.TH aio_return 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_return \- Retrieve status of asynchronous I/O operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "ssize_t aio_return (const struct aiocb *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function can be used to retrieve the return status of the operation
+carried out by the request described in the variable pointed to by
+.IR aiocbp
+. As long as the error status of this request as returned
+by
+.IR aio_error
+is
+.IR EINPROGRESS
+the return of this function is
+undefined.
+
+Once the request is finished this function can be used exactly once to
+retrieve the return value. Following calls might lead to undefined
+behavior.
+When the sources are compiled with
+.B "_FILE_OFFSET_BITS == 64"
+this function is in fact
+.IR aio_return64
+since the LFS interface
+transparently replaces the normal implementation.
+.SH "RETURN VALUES"
+The return value itself is the value which would have been
+returned by the
+.IR read
+,
+.IR write
+, or
+.IR fsync
+call.
+.SH ERRORS
+The function can return
+.TP
+.B ENOSYS
+if it is not implemented.
+.TP
+.B EINVAL
+if the
+.IR aiocbp
+parameter does not
+refer to an asynchronous operation whose return status is not yet known.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_return64.3 b/man/aio_return64.3
new file mode 100644
index 0000000..7e78362
--- /dev/null
+++ b/man/aio_return64.3
@@ -0,0 +1,51 @@
+.TH aio_read64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_read64 \- Retrieve status of asynchronous I/O operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_return64 (const struct aiocb64 *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to
+.IR "aio_return"
+with the only difference
+that the argument is a reference to a variable of type
+.IR "struct aiocb64".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is available under the name
+.IR "aio_return"
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+See
+.IR aio_return.
+.SH ERRORS
+See
+.IR aio_return.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_suspend.3 b/man/aio_suspend.3
new file mode 100644
index 0000000..cae1b65
--- /dev/null
+++ b/man/aio_suspend.3
@@ -0,0 +1,123 @@
+.TH aio_suspend 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_suspend \- Wait until one or more requests of a specific set terminates.
+.SH SYNOPSYS
+.nf
+.B "#include <errno.h>"
+.sp
+.br
+.B "#include <aio.h>"
+.sp
+.br
+.BI "int aio_suspend (const struct aiocb *const list[], int nent, const struct timespec *timeout)"
+.fi
+.SH DESCRIPTION
+Another method of synchronization is to wait until one or more requests of a
+specific set terminated. This could be achieved by the
+.IR "aio_*"
+functions to notify the initiating process about the termination but in
+some situations this is not the ideal solution. In a program which
+constantly updates clients somehow connected to the server it is not
+always the best solution to go round robin since some connections might
+be slow. On the other hand letting the
+.IR "aio_*"
+function notify the
+caller might also be not the best solution since whenever the process
+works on preparing data for on client it makes no sense to be
+interrupted by a notification since the new client will not be handled
+before the current client is served. For situations like this
+.IR "aio_suspend"
+should be used.
+.PP
+When calling this function, the calling thread is suspended until at
+least one of the requests pointed to by the
+.IR "nent"
+elements of the
+array
+.IR "list"
+has completed. If any of the requests has already
+completed at the time
+.IR "aio_suspend"
+is called, the function returns
+immediately. Whether a request has terminated or not is determined by
+comparing the error status of the request with
+.IR "EINPROGRESS"
+. If
+an element of
+.IR "list"
+is
+.IR "NULL"
+, the entry is simply ignored.
+
+If no request has finished, the calling process is suspended. If
+.IR "timeout"
+is
+.IR "NULL"
+, the process is not woken until a request
+has finished. If
+.IR "timeout"
+is not
+.IR "NULL"
+, the process remains
+suspended at least as long as specified in
+.IR "timeout"
+. In this case,
+.IR "aio_suspend"
+returns with an error.
+.PP
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is in fact
+.IR "aio_suspend64"
+since the LFS interface
+transparently replaces the normal implementation.
+.SH "RETURN VALUES"
+The return value of the function is
+.IR 0
+if one or more requests
+from the
+.IR "list"
+have terminated. Otherwise the function returns
+.IR -1
+and
+.IR "errno"
+is set.
+.SH ERRORS
+.TP
+.B EAGAIN
+None of the requests from the
+.IR "list"
+completed in the time specified
+by
+.IR "timeout"
+.
+.TP
+.B EINTR
+A signal interrupted the
+.IR "aio_suspend"
+function. This signal might
+also be sent by the AIO implementation while signalling the termination
+of one of the requests.
+.TP
+.B ENOSYS
+The
+.IR "aio_suspend"
+function is not implemented.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_suspend64.3 b/man/aio_suspend64.3
new file mode 100644
index 0000000..2f289ec
--- /dev/null
+++ b/man/aio_suspend64.3
@@ -0,0 +1,51 @@
+.TH aio_suspend64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_suspend64 \- Wait until one or more requests of a specific set terminates
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_suspend64 (const struct aiocb64 *const list[], int nent, const struct timespec *timeout)"
+.fi
+.SH DESCRIPTION
+This function is similar to
+.IR "aio_suspend"
+with the only difference
+that the argument is a reference to a variable of type
+.IR "struct aiocb64".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+this
+function is available under the name
+.IR "aio_suspend"
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+See
+.IR aio_suspend.
+.SH ERRORS
+See
+.IR aio_suspend.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_write(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_write.3 b/man/aio_write.3
new file mode 100644
index 0000000..7c0cfd0
--- /dev/null
+++ b/man/aio_write.3
@@ -0,0 +1,176 @@
+.TH aio_write 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_write \- Initiate an asynchronous write operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_write (struct aiocb * aiocbp);"
+.fi
+.SH DESCRIPTION
+This function initiates an asynchronous write operation. The function
+call immediately returns after the operation was enqueued or if before
+this happens an error was encountered.
+
+The first
+.IR "aiocbp->aio_nbytes"
+bytes from the buffer starting at
+.IR "aiocbp->aio_buf"
+are written to the file for which
+.IR "aiocbp->aio_fildes"
+is an descriptor, starting at the absolute
+position
+.IR "aiocbp->aio_offset"
+in the file.
+
+If prioritized I/O is supported by the platform, the
+.IR "aiocbp->aio_reqprio "
+value is used to adjust the priority before
+the request is actually enqueued.
+
+The calling process is notified about the termination of the read
+request according to the
+.IR "aiocbp->aio_sigevent"
+value.
+
+When
+.IR "aio_write"
+returns, the return value is zero if no error
+occurred that can be found before the process is enqueued. If such an
+early error is found the function returns
+.IR -1
+and sets
+.IR "errno"
+to one of the following values.
+
+.TP
+.B EAGAIN
+The request was not enqueued due to (temporarily) exceeded resource
+limitations.
+.TP
+.B ENOSYS
+The
+.IR "aio_write"
+function is not implemented.
+.TP
+.B EBADF
+The
+.IR "aiocbp->aio_fildes"
+descriptor is not valid. This condition
+may not be recognized before enqueueing the request, and so this error
+might also be signaled asynchronously.
+.TP
+.B EINVAL
+The
+.IR "aiocbp->aio_offset"
+or
+.IR "aiocbp->aio_reqprio"
+value is
+invalid. This condition may not be recognized before enqueueing the
+request and so this error might also be signaled asynchronously.
+.PP
+
+In the case
+.IR "aio_write"
+returns zero, the current status of the
+request can be queried using
+.IR "aio_error"
+and
+.IR "aio_return"
+functions. As long as the value returned by
+.IR "aio_error"
+is
+.IR "EINPROGRESS"
+the operation has not yet completed. If
+.IR "aio_error"
+returns zero, the operation successfully terminated,
+otherwise the value is to be interpreted as an error code. If the
+function terminated, the result of the operation can be get using a call
+to
+.IR "aio_return"
+. The returned value is the same as an equivalent
+call to
+.IR "read"
+would have returned. Possible error codes returned
+by
+.IR "aio_error"
+are:
+
+.TP
+.B EBADF
+The
+.IR "aiocbp->aio_fildes"
+descriptor is not valid.
+.TP
+.B ECANCELED
+The operation was canceled before the operation was finished.
+.TP
+.B EINVAL
+The
+.IR "aiocbp->aio_offset"
+value is invalid.
+.PP
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+, this
+function is in fact
+.IR "aio_write64"
+since the LFS interface transparently
+replaces the normal implementation.
+.SH "RETURN VALUES"
+When
+.IR "aio_write"
+returns, the return value is zero if no error
+occurred that can be found before the process is enqueued. If such an
+early error is found the function returns
+.IR -1
+and sets
+.IR "errno"
+to one of the following values.
+.SH ERRORS
+.TP
+.B EAGAIN
+The request was not enqueued due to (temporarily) exceeded resource
+limitations.
+.TP
+.B ENOSYS
+The
+.IR "aio_write"
+function is not implemented.
+.TP
+.B EBADF
+The
+.IR "aiocbp->aio_fildes"
+descriptor is not valid. This condition
+may not be recognized before enqueueing the request, and so this error
+might also be signaled asynchronously.
+.TP
+.B EINVAL
+The
+.IR "aiocbp->aio_offset"
+or
+.IR "aiocbp->aio_reqprio"
+value is
+invalid. This condition may not be recognized before enqueueing the
+request and so this error might also be signaled asynchronously.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write64(3),
+.BR errno(3),
diff --git a/man/aio_write64.3 b/man/aio_write64.3
new file mode 100644
index 0000000..1080903
--- /dev/null
+++ b/man/aio_write64.3
@@ -0,0 +1,61 @@
+.TH aio_write64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+aio_write64 \- Initiate an asynchronous write operation
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <aio.h>
+.sp
+.br
+.BI "int aio_write64 (struct aiocb *aiocbp)"
+.fi
+.SH DESCRIPTION
+This function is similar to the
+.IR "aio_write"
+function. The only
+difference is that on
+.IR "32 bit"
+machines the file descriptor should
+be opened in the large file mode. Internally
+.IR "aio_write64"
+uses
+functionality equivalent to
+.IR "lseek64"
+to position the file descriptor correctly for the writing,
+as opposed to
+.IR "lseek"
+functionality used in
+.IR "aio_write".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+, this
+function is available under the name
+.IR "aio_write"
+and so transparently
+replaces the interface for small files on 32 bit machines.
+.SH "RETURN VALUES"
+See
+.IR aio_write.
+.SH ERRORS
+See
+.IR aio_write.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR errno(3),
diff --git a/man/io.3 b/man/io.3
new file mode 100644
index 0000000..d910a68
--- /dev/null
+++ b/man/io.3
@@ -0,0 +1,351 @@
+.TH io 3 2002-09-12 "Linux 2.4" Linux IO"
+.SH NAME
+io \- Asynchronous IO
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <libio.h>
+.sp
+.fi
+.SH DESCRIPTION
+The libaio library defines a new set of I/O operations which can
+significantly reduce the time an application spends waiting at I/O. The
+new functions allow a program to initiate one or more I/O operations and
+then immediately resume normal work while the I/O operations are
+executed in parallel.
+
+These functions are part of the library with realtime functions named
+.IR "libaio"
+. They are not actually part of the
+.IR "libc"
+binary.
+The implementation of these functions can be done using support in the
+kernel.
+
+All IO operations operate on files which were opened previously. There
+might be arbitrarily many operations running for one file. The
+asynchronous I/O operations are controlled using a data structure named
+.IR "struct iocb"
+It is defined in
+.IR "libio.h"
+as follows.
+
+.nf
+
+typedef struct io_context *io_context_t;
+
+typedef enum io_iocb_cmd {
+ IO_CMD_PREAD = 0,
+ IO_CMD_PWRITE = 1,
+
+ IO_CMD_FSYNC = 2,
+ IO_CMD_FDSYNC = 3,
+
+ IO_CMD_POLL = 5,
+ IO_CMD_NOOP = 6,
+} io_iocb_cmd_t;
+
+struct io_iocb_common {
+ void *buf;
+ unsigned __pad1;
+ long nbytes;
+ unsigned __pad2;
+ long long offset;
+ long long __pad3, __pad4;
+}; /* result code is the amount read or -'ve errno */
+
+
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+ union {
+ struct io_iocb_common c;
+ struct io_iocb_vector v;
+ struct io_iocb_poll poll;
+ struct io_iocb_sockaddr saddr;
+ } u;
+};
+
+
+.fi
+.TP
+.IR "int aio_fildes"
+This element specifies the file descriptor to be used for the
+operation. It must be a legal descriptor, otherwise the operation will
+fail.
+
+The device on which the file is opened must allow the seek operation.
+I.e., it is not possible to use any of the IO operations on devices
+like terminals where an
+.IR "lseek"
+call would lead to an error.
+.TP
+.IR "long u.c.offset"
+This element specifies the offset in the file at which the operation (input
+or output) is performed. Since the operations are carried out in arbitrary
+order and more than one operation for one file descriptor can be
+started, one cannot expect a current read/write position of the file
+descriptor.
+.TP
+.IR "void *buf"
+This is a pointer to the buffer with the data to be written or the place
+where the read data is stored.
+.TP
+.IR "long u.c.nbytes"
+This element specifies the length of the buffer pointed to by
+.IR "io_buf"
+.
+.TP
+.IR "int aio_reqprio"
+Is not currently used.
+.TP
+.B "IO_CMD_PREAD"
+Start a read operation. Read from the file at position
+.IR "u.c.offset"
+and store the next
+.IR "u.c.nbytes"
+bytes in the
+buffer pointed to by
+.IR "buf"
+.
+.TP
+.B "IO_CMD_PWRITE"
+Start a write operation. Write
+.IR "u.c.nbytes"
+bytes starting at
+.IR "buf"
+into the file starting at position
+.IR "u.c.offset"
+.
+.TP
+.B "IO_CMD_NOP"
+Do nothing for this control block. This value is useful sometimes when
+an array of
+.IR "struct iocb"
+values contains holes, i.e., some of the
+values must not be handled although the whole array is presented to the
+.IR "io_submit"
+function.
+.TP
+.B "IO_CMD_FSYNC"
+.TP
+.B "IO_CMD_POLL"
+This is experimental.
+.SH EXAMPLE
+.nf
+/*
+ * Simplistic version of copy command using async i/o
+ *
+ * From: Stephen Hemminger <shemminger@osdl.org>
+ * Copy file by using a async I/O state machine.
+ * 1. Start read request
+ * 2. When read completes turn it into a write request
+ * 3. When write completes decrement counter and free resources
+ *
+ *
+ * Usage: aiocp file(s) desination
+ */
+
+#include <unistd.h>
+#include <stdio.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <sys/param.h>
+#include <fcntl.h>
+#include <errno.h>
+
+#include <libaio.h>
+
+#define AIO_BLKSIZE (64*1024)
+#define AIO_MAXIO 32
+
+static int busy = 0; // # of I/O's in flight
+static int tocopy = 0; // # of blocks left to copy
+static int dstfd = -1; // destination file descriptor
+static const char *dstname = NULL;
+static const char *srcname = NULL;
+
+
+/* Fatal error handler */
+static void io_error(const char *func, int rc)
+{
+ if (rc == -ENOSYS)
+ fprintf(stderr, "AIO not in this kernel\n");
+ else if (rc < 0 && -rc < sys_nerr)
+ fprintf(stderr, "%s: %s\n", func, sys_errlist[-rc]);
+ else
+ fprintf(stderr, "%s: error %d\n", func, rc);
+
+ if (dstfd > 0)
+ close(dstfd);
+ if (dstname)
+ unlink(dstname);
+ exit(1);
+}
+
+/*
+ * Write complete callback.
+ * Adjust counts and free resources
+ */
+static void wr_done(io_context_t ctx, struct iocb *iocb, long res, long res2)
+{
+ if (res2 != 0) {
+ io_error("aio write", res2);
+ }
+ if (res != iocb->u.c.nbytes) {
+ fprintf(stderr, "write missed bytes expect %d got %d\n", iocb->u.c.nbytes, res2);
+ exit(1);
+ }
+ --tocopy;
+ --busy;
+ free(iocb->u.c.buf);
+
+ memset(iocb, 0xff, sizeof(iocb)); // paranoia
+ free(iocb);
+ write(2, "w", 1);
+}
+
+/*
+ * Read complete callback.
+ * Change read iocb into a write iocb and start it.
+ */
+static void rd_done(io_context_t ctx, struct iocb *iocb, long res, long res2)
+{
+ /* library needs accessors to look at iocb? */
+ int iosize = iocb->u.c.nbytes;
+ char *buf = iocb->u.c.buf;
+ off_t offset = iocb->u.c.offset;
+
+ if (res2 != 0)
+ io_error("aio read", res2);
+ if (res != iosize) {
+ fprintf(stderr, "read missing bytes expect %d got %d\n", iocb->u.c.nbytes, res);
+ exit(1);
+ }
+
+
+ /* turn read into write */
+ io_prep_pwrite(iocb, dstfd, buf, iosize, offset);
+ io_set_callback(iocb, wr_done);
+ if (1 != (res = io_submit(ctx, 1, &iocb)))
+ io_error("io_submit write", res);
+ write(2, "r", 1);
+}
+
+
+int main(int argc, char *const *argv)
+{
+ int srcfd;
+ struct stat st;
+ off_t length = 0, offset = 0;
+ io_context_t myctx;
+
+ if (argc != 3 || argv[1][0] == '-') {
+ fprintf(stderr, "Usage: aiocp SOURCE DEST");
+ exit(1);
+ }
+ if ((srcfd = open(srcname = argv[1], O_RDONLY)) < 0) {
+ perror(srcname);
+ exit(1);
+ }
+ if (fstat(srcfd, &st) < 0) {
+ perror("fstat");
+ exit(1);
+ }
+ length = st.st_size;
+
+ if ((dstfd = open(dstname = argv[2], O_WRONLY | O_CREAT, 0666)) < 0) {
+ close(srcfd);
+ perror(dstname);
+ exit(1);
+ }
+
+ /* initialize state machine */
+ memset(&myctx, 0, sizeof(myctx));
+ io_queue_init(AIO_MAXIO, &myctx);
+ tocopy = howmany(length, AIO_BLKSIZE);
+
+ while (tocopy > 0) {
+ int i, rc;
+ /* Submit as many reads as once as possible upto AIO_MAXIO */
+ int n = MIN(MIN(AIO_MAXIO - busy, AIO_MAXIO / 2),
+ howmany(length - offset, AIO_BLKSIZE));
+ if (n > 0) {
+ struct iocb *ioq[n];
+
+ for (i = 0; i < n; i++) {
+ struct iocb *io = (struct iocb *) malloc(sizeof(struct iocb));
+ int iosize = MIN(length - offset, AIO_BLKSIZE);
+ char *buf = (char *) malloc(iosize);
+
+ if (NULL == buf || NULL == io) {
+ fprintf(stderr, "out of memory\n");
+ exit(1);
+ }
+
+ io_prep_pread(io, srcfd, buf, iosize, offset);
+ io_set_callback(io, rd_done);
+ ioq[i] = io;
+ offset += iosize;
+ }
+
+ rc = io_submit(myctx, n, ioq);
+ if (rc < 0)
+ io_error("io_submit", rc);
+
+ busy += n;
+ }
+
+ // Handle IO's that have completed
+ rc = io_queue_run(myctx);
+ if (rc < 0)
+ io_error("io_queue_run", rc);
+
+ // if we have maximum number of i/o's in flight
+ // then wait for one to complete
+ if (busy == AIO_MAXIO) {
+ rc = io_queue_wait(myctx, NULL);
+ if (rc < 0)
+ io_error("io_queue_wait", rc);
+ }
+
+ }
+
+ close(srcfd);
+ close(dstfd);
+ exit(0);
+}
+
+/*
+ * Results look like:
+ * [alanm@toolbox ~/MOT3]$ ../taio kernel-source-2.4.8-0.4g.ppc.rpm abc
+ * rrrrrrrrrrrrrrrwwwrwrrwwrrwrwwrrwrwrwwrrwrwrrrrwwrwwwrrwrrrwwwwwwwwwwwwwwwww
+ * rrrrrrrrrrrrrrwwwrrwrwrwrwrrwwwwwwwwwwwwwwrrrrrrrrrrrrrrrrrrwwwwrwrwwrwrwrwr
+ * wrrrrrrrwwwwwwwwwwwwwrrrwrrrwrrwrwwwwwwwwwwrrrrwwrwrrrrrrrrrrrwwwwwwwwwwwrww
+ * wwwrrrrrrrrwwrrrwwrwrwrwwwrrrrrrrwwwrrwwwrrwrwwwwwwwwrrrrrrrwwwrrrrrrrwwwwww
+ * wwwwwwwrwrrrrrrrrwrrwrrwrrwrwrrrwrrrwrrrwrwwwwwwwwwwwwwwwwwwrrrwwwrrrrrrrrrr
+ * rrwrrrrrrwrrwwwwwwwwwwwwwwwwrwwwrrwrwwrrrrrrrrrrrrrrrrrrrwwwwwwwwwwwwwwwwwww
+ * rrrrrwrrwrwrwrrwrrrwwwwwwwwrrrrwrrrwrwwrwrrrwrrwrrrrwwwwwwwrwrwwwwrwwrrrwrrr
+ * rrrwwwwwwwrrrrwwrrrrrrrrrrrrwrwrrrrwwwwwwwwwwwwwwrwrrrrwwwwrwrrrrwrwwwrrrwww
+ * rwwrrrrrrrwrrrrrrrrrrrrwwwwrrrwwwrwrrwwwwwwwwwwwwwwwwwwwwwrrrrrrrwwwwwwwrw
+ */
+.fi
+.SH "SEE ALSO"
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_cancel.1 b/man/io_cancel.1
new file mode 100644
index 0000000..16e898a
--- /dev/null
+++ b/man/io_cancel.1
@@ -0,0 +1,21 @@
+.\"/* sys_io_cancel:
+.\" * Attempts to cancel an iocb previously passed to io_submit. If
+.\" * the operation is successfully cancelled, the resulting event is
+.\" * copied into the memory pointed to by result without being placed
+.\" * into the completion queue and 0 is returned. May fail with
+.\" * -EFAULT if any of the data structures pointed to are invalid.
+.\" * May fail with -EINVAL if aio_context specified by ctx_id is
+.\" * invalid. May fail with -EAGAIN if the iocb specified was not
+.\" * cancelled. Will fail with -ENOSYS if not implemented.
+.\" */
+.\"
+.TH io_cancel 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_cancel \- cancel io requests
+.SH SYNOPSIS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.LP
+.BI "int io_submit(io_context_t " ctx ", struct iocb *" iocb ", struct io_event *" result ");"
+
diff --git a/man/io_cancel.3 b/man/io_cancel.3
new file mode 100644
index 0000000..9a16084
--- /dev/null
+++ b/man/io_cancel.3
@@ -0,0 +1,65 @@
+.TH io_cancel 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_cancel \- Cancel io requests
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <libaio.h>
+.sp
+.br
+.BI "int io_cancel(io_context_t ctx, struct iocb *iocb)"
+.br
+.sp
+struct iocb {
+ void *data; /* Return in the io completion event */
+ unsigned key; /* For use in identifying io requests */
+ short aio_lio_opcode;
+ short aio_reqprio; /* Not used */
+ int aio_fildes;
+};
+.fi
+.SH DESCRIPTION
+Attempts to cancel an iocb previously passed to io_submit. If
+the operation is successfully cancelled, the resulting event is
+copied into the memory pointed to by result without being placed
+into the completion queue.
+.PP
+When one or more requests are asynchronously processed, it might be
+useful in some situations to cancel a selected operation, e.g., if it
+becomes obvious that the written data is no longer accurate and would
+have to be overwritten soon. As an example, assume an application, which
+writes data in files in a situation where new incoming data would have
+to be written in a file which will be updated by an enqueued request.
+.SH "RETURN VALUES"
+0 is returned on success , otherwise returns Errno.
+.SH ERRORS
+.TP
+.B EFAULT
+If any of the data structures pointed to are invalid.
+.TP
+.B EINVAL
+If aio_context specified by ctx_id is
+invalid.
+.TP
+.B EAGAIN
+If the iocb specified was not
+cancelled.
+.TP
+.B ENOSYS
+if not implemented.
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_destroy.1 b/man/io_destroy.1
new file mode 100644
index 0000000..177683b
--- /dev/null
+++ b/man/io_destroy.1
@@ -0,0 +1,17 @@
+.\"/* sys_io_destroy:
+.\" * Destroy the aio_context specified. May cancel any outstanding
+.\" * AIOs and block on completion. Will fail with -ENOSYS if not
+.\" * implemented. May fail with -EFAULT if the context pointed to
+.\" * is invalid.
+.\" */
+.\" libaio provides this as io_queue_release.
+.TH io_destroy 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_destroy \- destroy an io context
+.SH SYNOPSIS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.LP
+.BI "int io_destroy(io_context_t " ctx ");"
+
diff --git a/man/io_fsync.3 b/man/io_fsync.3
new file mode 100644
index 0000000..53eb63d
--- /dev/null
+++ b/man/io_fsync.3
@@ -0,0 +1,82 @@
+./" static inline int io_fsync(io_context_t ctx, struct iocb *iocb, io_callback_t cb, int fd)
+./" {
+./" io_prep_fsync(iocb, fd);
+./" io_set_callback(iocb, cb);
+./" return io_submit(ctx, 1, &iocb);
+./" }
+.TH io_fsync 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+io_fsync \- Synchronize a file's complete in-core state with that on disk
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <libaio.h>
+.sp
+.br
+.BI "int io_fsync(io_context_t ctx, struct iocb *iocb, io_callback_t cb, int fd)"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.sp
+typedef void (*io_callback_t)(io_context_t ctx, struct iocb *iocb, long res, long res2);
+.sp
+.fi
+.SH DESCRIPTION
+When dealing with asynchronous operations it is sometimes necessary to
+get into a consistent state. This would mean for AIO that one wants to
+know whether a certain request or a group of request were processed.
+This could be done by waiting for the notification sent by the system
+after the operation terminated, but this sometimes would mean wasting
+resources (mainly computation time).
+.PP
+Calling this function forces all I/O operations operating queued at the
+time of the function call operating on the file descriptor
+.IR "iocb->io_fildes"
+into the synchronized I/O completion state . The
+.IR "io_fsync"
+function returns
+immediately but the notification through the method described in
+.IR "io_callback"
+will happen only after all requests for this
+file descriptor have terminated and the file is synchronized. This also
+means that requests for this very same file descriptor which are queued
+after the synchronization request are not affected.
+.SH "RETURN VALUES"
+Returns 0, otherwise returns errno.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I ctx
+refers to an unitialized aio context, the iocb pointed to by
+.I iocbs
+contains an improperly initialized iocb,
+.TP
+.B EBADF
+The iocb contains a file descriptor that does not exist.
+.TP
+.B EINVAL
+The file specified in the iocb does not support the given io operation.
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_getevents(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_getevents.1 b/man/io_getevents.1
new file mode 100644
index 0000000..27730b9
--- /dev/null
+++ b/man/io_getevents.1
@@ -0,0 +1,29 @@
+./"/* io_getevents:
+./" * Attempts to read at least min_nr events and up to nr events from
+./" * the completion queue for the aio_context specified by ctx_id. May
+./" * fail with -EINVAL if ctx_id is invalid, if min_nr is out of range,
+./" * if nr is out of range, if when is out of range. May fail with
+./" * -EFAULT if any of the memory specified to is invalid. May return
+./" * 0 or < min_nr if no events are available and the timeout specified
+./" * by when has elapsed, where when == NULL specifies an infinite
+./" * timeout. Note that the timeout pointed to by when is relative and
+./" * will be updated if not NULL and the operation blocks. Will fail
+./" * with -ENOSYS if not implemented.
+./" */
+./"asmlinkage long sys_io_getevents(io_context_t ctx_id,
+./" long min_nr,
+./" long nr,
+./" struct io_event *events,
+./" struct timespec *timeout)
+./"
+.TH io_getevents 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_getevents \- read resulting events from io requests
+.SH SYNOPSIS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.sp
+.BI "int io_getevents(io_context_t " ctx ", long " min_nr ", long " nr ", struct io_events *" events "[], struct timespec *" timeout ");"
+
+
diff --git a/man/io_getevents.3 b/man/io_getevents.3
new file mode 100644
index 0000000..8e9ddc8
--- /dev/null
+++ b/man/io_getevents.3
@@ -0,0 +1,79 @@
+./"/* io_getevents:
+./" * Attempts to read at least min_nr events and up to nr events from
+./" * the completion queue for the aio_context specified by ctx_id. May
+./" * fail with -EINVAL if ctx_id is invalid, if min_nr is out of range,
+./" * if nr is out of range, if when is out of range. May fail with
+./" * -EFAULT if any of the memory specified to is invalid. May return
+./" * 0 or < min_nr if no events are available and the timeout specified
+./" * by when has elapsed, where when == NULL specifies an infinite
+./" * timeout. Note that the timeout pointed to by when is relative and
+./" * will be updated if not NULL and the operation blocks. Will fail
+./" * with -ENOSYS if not implemented.
+./" */
+./"asmlinkage long sys_io_getevents(io_context_t ctx_id,
+./" long min_nr,
+./" long nr,
+./" struct io_event *events,
+./" struct timespec *timeout)
+./"
+.TH io_getevents 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_getevents \- Read resulting events from io requests
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <libaio.h>
+.br
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.sp
+struct io_event {
+ unsigned PADDED(data, __pad1);
+ unsigned PADDED(obj, __pad2);
+ unsigned PADDED(res, __pad3);
+ unsigned PADDED(res2, __pad4);
+};
+.sp
+.BI "int io_getevents(io_context_t " ctx ", long " nr ", struct io_event *" events "[], struct timespec *" timeout ");"
+
+.fi
+.SH DESCRIPTION
+Attempts to read up to nr events from
+the completion queue for the aio_context specified by ctx.
+.SH "RETURN VALUES"
+May return
+0 if no events are available and the timeout specified
+by when has elapsed, where when == NULL specifies an infinite
+timeout. Note that the timeout pointed to by when is relative and
+will be updated if not NULL and the operation blocks. Will fail
+with ENOSYS if not implemented.
+.SH ERRORS
+.TP
+.B EINVAL
+if ctx_id is invalid, if min_nr is out of range,
+if nr is out of range, if when is out of range.
+.TP
+.B EFAULT
+if any of the memory specified to is invalid.
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_prep_fsync.3 b/man/io_prep_fsync.3
new file mode 100644
index 0000000..4cf935a
--- /dev/null
+++ b/man/io_prep_fsync.3
@@ -0,0 +1,89 @@
+./" static inline void io_prep_fsync(struct iocb *iocb, int fd)
+./" {
+./" memset(iocb, 0, sizeof(*iocb));
+./" iocb->aio_fildes = fd;
+./" iocb->aio_lio_opcode = IO_CMD_FSYNC;
+./" iocb->aio_reqprio = 0;
+./" }
+.TH io_prep_fsync 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+io_prep_fsync \- Synchronize a file's complete in-core state with that on disk
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "static inline void io_prep_fsync(struct iocb *iocb, int fd)"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.sp
+.fi
+.SH DESCRIPTION
+This is an inline convenience function for setting up an iocbv for a FSYNC request.
+.br
+The file for which
+.TP
+.IR "iocb->aio_fildes = fd"
+is a descriptor is set up with
+the command
+.TP
+.IR "iocb->aio_lio_opcode = IO_CMD_FSYNC:
+.
+.PP
+The io_prep_fsync() function shall set up an IO_CMD_FSYNC operation
+to asynchronously force all I/O
+operations associated with the file indicated by the file
+descriptor aio_fildes member of the iocb structure referenced by
+the iocb argument and queued at the time of the call to
+io_submit() to the synchronized I/O completion state. The function
+call shall return when the synchronization request has been
+initiated or queued to the file or device (even when the data
+cannot be synchronized immediately).
+
+All currently queued I/O operations shall be completed as if by a call
+to fsync(); that is, as defined for synchronized I/O file
+integrity completion. If the
+operation queued by io_prep_fsync() fails, then, as for fsync(),
+outstanding I/O operations are not guaranteed to have
+been completed.
+
+If io_prep_fsync() succeeds, then it is only the I/O that was queued
+at the time of the call to io_submit() that is guaranteed to be
+forced to the relevant completion state. The completion of
+subsequent I/O on the file descriptor is not guaranteed to be
+completed in a synchronized fashion.
+.PP
+This function returns immediately . To schedule the operation, the
+function
+.IR io_submit
+must be called.
+.PP
+Simultaneous asynchronous operations using the same iocb produce
+undefined results.
+.SH "RETURN VALUES"
+None
+.SH ERRORS
+None
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_prep_pread.3 b/man/io_prep_pread.3
new file mode 100644
index 0000000..5938aec
--- /dev/null
+++ b/man/io_prep_pread.3
@@ -0,0 +1,79 @@
+./" static inline void io_prep_pread(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+./" {
+./" memset(iocb, 0, sizeof(*iocb));
+./" iocb->aio_fildes = fd;
+./" iocb->aio_lio_opcode = IO_CMD_PREAD;
+./" iocb->aio_reqprio = 0;
+./" iocb->u.c.buf = buf;
+./" iocb->u.c.nbytes = count;
+./" iocb->u.c.offset = offset;
+./" }
+.TH io_prep_pread 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+io_prep_pread \- Set up asynchronous read
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.sp
+.br
+.B #include <libaio.h>
+.br
+.sp
+.BI "inline void io_prep_pread(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.fi
+.SH DESCRIPTION
+.IR io_prep_pread
+is an inline convenience function designed to facilitate the initialization of
+the iocb for an asynchronous read operation.
+
+The first
+.TP
+.IR "iocb->u.c.nbytes = count"
+bytes of the file for which
+.TP
+.IR "iocb->aio_fildes = fd"
+is a descriptor are written to the buffer
+starting at
+.TP
+.IR "iocb->u.c.buf = buf"
+.
+.br
+Reading starts at the absolute position
+.TP
+.IR "ioc->u.c.offset = offset"
+in the file.
+.PP
+This function returns immediately . To schedule the operation, the
+function
+.IR io_submit
+must be called.
+.PP
+Simultaneous asynchronous operations using the same iocb produce
+undefined results.
+.SH "RETURN VALUES"
+None
+.SH ERRORS
+None
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_prep_pwrite.3 b/man/io_prep_pwrite.3
new file mode 100644
index 0000000..68b3500
--- /dev/null
+++ b/man/io_prep_pwrite.3
@@ -0,0 +1,77 @@
+./" static inline void io_prep_pwrite(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+./" {
+./" memset(iocb, 0, sizeof(*iocb));
+./" iocb->aio_fildes = fd;
+./" iocb->aio_lio_opcode = IO_CMD_PWRITE;
+./" iocb->aio_reqprio = 0;
+./" iocb->u.c.buf = buf;
+./" iocb->u.c.nbytes = count;
+./" iocb->u.c.offset = offset;
+./" }
+.TH io_prep_pwrite 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+io_prep_pwrite \- Set up iocb for asynchronous writes
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "inline void io_prep_pwrite(struct iocb *iocb, int fd, void *buf, size_t count, long long offset)
+"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.fi
+.SH DESCRIPTION
+io_prep_write is a convenicence function for setting up parallel writes.
+
+The first
+.TP
+.IR "iocb->u.c.nbytes = count"
+bytes of the file for which
+.TP
+.IR "iocb->aio_fildes = fd"
+is a descriptor are written from the buffer
+starting at
+.TP
+.IR "iocb->u.c.buf = buf"
+.
+.br
+Writing starts at the absolute position
+.TP
+.IR "ioc->u.c.offset = offset"
+in the file.
+.PP
+This function returns immediately . To schedule the operation, the
+function
+.IR io_submit
+must be called.
+.PP
+Simultaneous asynchronous operations using the same iocb produce
+undefined results.
+.SH "RETURN VALUES"
+None
+.SH ERRORS
+None
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_queue_init.3 b/man/io_queue_init.3
new file mode 100644
index 0000000..317f631
--- /dev/null
+++ b/man/io_queue_init.3
@@ -0,0 +1,63 @@
+.TH io_queue_init 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_queue_init \- Initialize asynchronous io state machine
+
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "int io_queue_init(int maxevents, io_context_t *ctx );"
+.sp
+.fi
+.SH DESCRIPTION
+.B io_queue_init
+Attempts to create an aio context capable of receiving at least
+.IR maxevents
+events.
+.IR ctx
+must point to an aio context that already exists and must be initialized
+to
+.IR 0
+before the call.
+If the operation is successful, *cxtp is filled with the resulting handle.
+.SH "RETURN VALUES"
+On success,
+.B io_queue_init
+returns 0. Otherwise, -error is return, where
+error is one of the Exxx values defined in the Errors section.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I maxevents
+is <= 0 or
+.IR ctx
+is an invalid memory locattion.
+.TP
+.B ENOSYS
+Not implemented
+.TP
+.B EAGAIN
+.IR "maxevents > max_aio_reqs"
+where max_aio_reqs is a tunable value.
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_queue_release.3 b/man/io_queue_release.3
new file mode 100644
index 0000000..06b9ec0
--- /dev/null
+++ b/man/io_queue_release.3
@@ -0,0 +1,48 @@
+.TH io_queue_release 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_queue_release \- Release the context associated with the userspace handle
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.br
+.sp
+.BI "int io_queue_release(io_context_t ctx)"
+.sp
+.SH DESCRIPTION
+.B io_queue_release
+destroys the context associated with the userspace handle. May cancel any outstanding
+AIOs and block on completion.
+
+.B cts.
+.SH "RETURN VALUES"
+On success,
+.B io_queue_release
+returns 0. Otherwise, -error is return, where
+error is one of the Exxx values defined in the Errors section.
+.SH ERRORS
+.TP
+.B EINVAL
+.I ctx
+refers to an unitialized aio context, the iocb pointed to by
+.I iocbs
+contains an improperly initialized iocb,
+.TP
+.B ENOSYS
+Not implemented
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
+
diff --git a/man/io_queue_run.3 b/man/io_queue_run.3
new file mode 100644
index 0000000..57dd417
--- /dev/null
+++ b/man/io_queue_run.3
@@ -0,0 +1,50 @@
+.TH io_queue_run 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_queue_run \- Handle completed io requests
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "int io_queue_run(io_context_t ctx );"
+.sp
+.fi
+.SH DESCRIPTION
+.B io_queue_run
+Attempts to read all the events events from
+the completion queue for the aio_context specified by ctx_id.
+.SH "RETURN VALUES"
+May return
+0 if no events are available.
+Will fail with -ENOSYS if not implemented.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I ctx
+refers to an unitialized aio context, the iocb pointed to by
+.I iocbs
+contains an improperly initialized iocb,
+.TP
+.B ENOSYS
+Not implemented
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_queue_wait.3 b/man/io_queue_wait.3
new file mode 100644
index 0000000..2306663
--- /dev/null
+++ b/man/io_queue_wait.3
@@ -0,0 +1,56 @@
+.TH io_queue_wait 2 2002-09-03 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_queue_wait \- Wait for io requests to complete
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "int io_queue_wait(io_context_t ctx, const struct timespec *timeout);"
+.fi
+.SH DESCRIPTION
+Attempts to read an event from
+the completion queue for the aio_context specified by ctx_id.
+.SH "RETURN VALUES"
+May return
+0 if no events are available and the timeout specified
+by when has elapsed, where when == NULL specifies an infinite
+timeout. Note that the timeout pointed to by when is relative and
+will be updated if not NULL and the operation blocks. Will fail
+with -ENOSYS if not implemented.
+.SH "RETURN VALUES"
+On success,
+.B io_queue_wait
+returns 0. Otherwise, -error is return, where
+error is one of the Exxx values defined in the Errors section.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I ctx
+refers to an unitialized aio context, the iocb pointed to by
+.I iocbs
+contains an improperly initialized iocb,
+.TP
+.B ENOSYS
+Not implemented
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_set_callback(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_set_callback.3 b/man/io_set_callback.3
new file mode 100644
index 0000000..a8ca789
--- /dev/null
+++ b/man/io_set_callback.3
@@ -0,0 +1,44 @@
+./"static inline void io_set_callback(struct iocb *iocb, io_callback_t cb)
+.TH io_set_callback 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+io_set_callback \- Set up io completion callback function
+.SH SYNOPSYS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "static inline void io_set_callback(struct iocb *iocb, io_callback_t cb)"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.sp
+typedef void (*io_callback_t)(io_context_t ctx, struct iocb *iocb, long res, long res2);
+.sp
+.fi
+.SH DESCRIPTION
+The callback is not done if the caller uses raw events from
+io_getevents, only with the library helpers
+.SH "RETURN VALUES"
+.SH ERRORS
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_submit(3),
+.BR errno(3)
diff --git a/man/io_setup.1 b/man/io_setup.1
new file mode 100644
index 0000000..68690e1
--- /dev/null
+++ b/man/io_setup.1
@@ -0,0 +1,15 @@
+./"/* sys_io_setup:
+./" * Create an aio_context capable of receiving at least nr_events.
+./" * ctxp must not point to an aio_context that already exists, and
+./" * must be initialized to 0 prior to the call. On successful
+./" * creation of the aio_context, *ctxp is filled in with the resulting
+./" * handle. May fail with -EINVAL if *ctxp is not initialized,
+./" * if the specified nr_events exceeds internal limits. May fail
+./" * with -EAGAIN if the specified nr_events exceeds the user's limit
+./" * of available events. May fail with -ENOMEM if insufficient kernel
+./" * resources are available. May fail with -EFAULT if an invalid
+./" * pointer is passed for ctxp. Will fail with -ENOSYS if not
+./" * implemented.
+./" */
+./" -- note: libaio is actually providing io_queue_init and io_queue_grow
+./" as separate functions. For now io_setup is the same as io_queue_grow.
diff --git a/man/io_submit.1 b/man/io_submit.1
new file mode 100644
index 0000000..f66e80f
--- /dev/null
+++ b/man/io_submit.1
@@ -0,0 +1,109 @@
+.TH io_submit 2 2002-09-02 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_submit \- submit io requests
+.SH SYNOPSIS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.LP
+.BI "int io_submit(io_context_t " ctx ", long " nr ", struct iocb *" iocbs "[]);"
+.SH DESCRIPTION
+.B io_submit
+submits to the io_context
+.I ctx
+up to
+.I nr
+I/O requests pointed to by the vector
+.IR iocbs .
+
+The
+.B iocb
+structure is defined as something like
+.sp
+.RS
+.nf
+struct iocb {
+ void *data;
+.\" unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.fi
+.RE
+.sp
+.I data
+is a an opaque pointer which will upon completion be returned in the
+.B io_event
+structure by
+.BR io_getevents (2).
+.\" and io_wait(2)
+Callers will typically use this to point directly or indirectly to a
+callback function.
+.sp
+.I aio_lio_opcode
+is the I/O operation requested. Callers will typically set this and the
+arguments to the I/O operation calling the
+.BR io_prep_ (3)
+function corresponding to the operation.
+.sp
+.I aio_reqprio
+is the priority of the request. Higher values have more priority; the
+normal priority is 0.
+.sp
+.I aio_fildes
+is the file descriptor for the I/O operation.
+Callers will typically set this and the
+arguments to the I/O operation calling the
+.BR io_prep_ *(3)
+function corresponding to the operation.
+.sp
+The caller may not modify the contents or resubmit a submitted
+.B iocb
+structure until after the operation completes or is canceled.
+The implementation of
+.BR io_submit (2)
+is permitted to modify reserved fields of the
+.B iocb
+structure.
+.SH "RETURN VALUES"
+If able to submit at least one iocb,
+.B io_submit
+returns the number of iocbs submitted successfully. Otherwise,
+.RI - error
+is returned, where
+.I error
+is one of the Exxx values defined in the Errors section.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I nr
+is negative,
+.I ctx
+refers to an uninitialized aio context, the iocb pointed to by
+.IR iocbs [0]
+is improperly initialized or specifies an unsupported operation.
+.TP
+.B EBADF
+The iocb pointed to by
+.IR iocbs [0]
+contains a file descriptor that does not exist.
+.TP
+.B EAGAIN
+Insufficient resources were available to queue any operations.
+.SH "SEE ALSO"
+.BR io_setup (2),
+.BR io_destroy (2),
+.BR io_getevents (2),
+.\".BR io_wait (2),
+.BR io_prep_pread (3),
+.BR io_prep_pwrite (3),
+.BR io_prep_fsync (3),
+.BR io_prep_fdsync (3),
+.BR io_prep_noop (3),
+.BR io_cancel (2),
+.BR errno (3)
diff --git a/man/io_submit.3 b/man/io_submit.3
new file mode 100644
index 0000000..b6966ef
--- /dev/null
+++ b/man/io_submit.3
@@ -0,0 +1,135 @@
+./"/* sys_io_submit:
+./" * Queue the nr iocbs pointed to by iocbpp for processing. Returns
+./" * the number of iocbs queued. May return -EINVAL if the aio_context
+./" * specified by ctx_id is invalid, if nr is < 0, if the iocb at
+./" * *iocbpp[0] is not properly initialized, if the operation specified
+./" * is invalid for the file descriptor in the iocb. May fail with
+./" * -EFAULT if any of the data structures point to invalid data. May
+./" * fail with -EBADF if the file descriptor specified in the first
+./" * iocb is invalid. May fail with -EAGAIN if insufficient resources
+./" * are available to queue any iocbs. Will return 0 if nr is 0. Will
+./" * fail with -ENOSYS if not implemented.
+./" */
+.TH io_submit 2 2002-09-02 "Linux 2.4" "Linux AIO"
+.SH NAME
+io_submit \- Submit io requests
+.SH SYNOPSIS
+.nf
+.B #include <errno.h>
+.br
+.sp
+.B #include <libaio.h>
+.br
+.sp
+.BI "int io_submit(io_context_t " ctx ", long " nr ", struct iocb *" iocbs "[]);"
+.sp
+struct iocb {
+ void *data;
+ unsigned key;
+ short aio_lio_opcode;
+ short aio_reqprio;
+ int aio_fildes;
+};
+.fi
+.SH DESCRIPTION
+.B io_submit
+submits
+.I nr
+iocbs for processing for a given io context ctx.
+
+The
+.IR "io_submit"
+function can be used to enqueue an arbitrary
+number of read and write requests at one time. The requests can all be
+meant for the same file, all for different files or every solution in
+between.
+
+.IR "io_submit"
+gets the
+.IR "nr"
+requests from the array pointed to
+by
+.IR "iocbs"
+. The operation to be performed is determined by the
+.IR "aio_lio_opcode"
+member in each element of
+.IR "iocbs"
+. If this
+field is
+.B "IO_CMD_PREAD"
+a read operation is enqueued, similar to a call
+of
+.IR "io_prep_pread"
+for this element of the array (except that the way
+the termination is signalled is different, as we will see below). If
+the
+.IR "aio_lio_opcode"
+member is
+.B "IO_CMD_PWRITE"
+a write operation
+is enqueued. Otherwise the
+.IR "aio_lio_opcode"
+must be
+.B "IO_CMD_NOP"
+in which case this element of
+.IR "iocbs"
+is simply ignored. This
+``operation'' is useful in situations where one has a fixed array of
+.IR "struct iocb"
+elements from which only a few need to be handled at
+a time. Another situation is where the
+.IR "io_submit"
+call was
+canceled before all requests are processed and the remaining requests have to be reissued.
+
+The other members of each element of the array pointed to by
+.IR "iocbs"
+must have values suitable for the operation as described in
+the documentation for
+.IR "io_prep_pread"
+and
+.IR "io_prep_pwrite"
+above.
+
+The function returns immediately after
+having enqueued all the requests.
+On success,
+.B io_submit
+returns the number of iocbs submitted successfully. Otherwise, -error is return, where
+error is one of the Exxx values defined in the Errors section.
+.PP
+If an error is detected, then the behavior is undefined.
+.PP
+Simultaneous asynchronous operations using the same iocb produce
+undefined results.
+.SH ERRORS
+.TP
+.B EFAULT
+.I iocbs
+referenced data outside of the program's accessible address space.
+.TP
+.B EINVAL
+.I ctx
+refers to an unitialized aio context, the iocb pointed to by
+.I iocbs
+contains an improperly initialized iocb,
+.TP
+.B EBADF
+The iocb contains a file descriptor that does not exist.
+.TP
+.B EINVAL
+The file specified in the iocb does not support the given io operation.
+.SH "SEE ALSO"
+.BR io(3),
+.BR io_cancel(3),
+.BR io_fsync(3),
+.BR io_getevents(3),
+.BR io_prep_fsync(3),
+.BR io_prep_pread(3),
+.BR io_prep_pwrite(3),
+.BR io_queue_init(3),
+.BR io_queue_release(3),
+.BR io_queue_run(3),
+.BR io_queue_wait(3),
+.BR io_set_callback(3),
+.BR errno(3)
diff --git a/man/lio_listio.3 b/man/lio_listio.3
new file mode 100644
index 0000000..9b5b5e4
--- /dev/null
+++ b/man/lio_listio.3
@@ -0,0 +1,229 @@
+.TH lio_listio 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+lio_listio - List directed I/O
+.SH SYNOPSYS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.LP
+.BI "int lio_listio (int mode, struct aiocb *const list[], int nent, struct sigevent *sig)"
+.nf
+.SH DESCRIPTION
+
+Besides these functions with the more or less traditional interface,
+POSIX.1b also defines a function which can initiate more than one
+operation at a time, and which can handle freely mixed read and write
+operations. It is therefore similar to a combination of
+.IR readv
+and
+.IR "writev"
+.
+
+The
+.IR "lio_listio"
+function can be used to enqueue an arbitrary
+number of read and write requests at one time. The requests can all be
+meant for the same file, all for different files or every solution in
+between.
+
+.IR "lio_listio"
+gets the
+.IR "nent"
+requests from the array pointed to
+by
+.IR "list"
+. The operation to be performed is determined by the
+.IR "aio_lio_opcode"
+member in each element of
+.IR "list"
+. If this
+field is
+.B "LIO_READ"
+a read operation is enqueued, similar to a call
+of
+.IR "aio_read"
+for this element of the array (except that the way
+the termination is signalled is different, as we will see below). If
+the
+.IR "aio_lio_opcode"
+member is
+.B "LIO_WRITE"
+a write operation
+is enqueued. Otherwise the
+.IR "aio_lio_opcode"
+must be
+.B "LIO_NOP"
+in which case this element of
+.IR "list"
+is simply ignored. This
+``operation'' is useful in situations where one has a fixed array of
+.IR "struct aiocb"
+elements from which only a few need to be handled at
+a time. Another situation is where the
+.IR "lio_listio"
+call was
+canceled before all requests are processed and the remaining requests have to be reissued.
+
+The other members of each element of the array pointed to by
+.IR "list"
+must have values suitable for the operation as described in
+the documentation for
+.IR "aio_read"
+and
+.IR "aio_write"
+above.
+
+The
+.IR "mode"
+argument determines how
+.IR "lio_listio"
+behaves after
+having enqueued all the requests. If
+.IR "mode"
+is
+.B "LIO_WAIT"
+it
+waits until all requests terminated. Otherwise
+.IR "mode"
+must be
+.B "LIO_NOWAIT"
+and in this case the function returns immediately after
+having enqueued all the requests. In this case the caller gets a
+notification of the termination of all requests according to the
+.IR "sig"
+parameter. If
+.IR "sig"
+is
+.B "NULL"
+no notification is
+send. Otherwise a signal is sent or a thread is started, just as
+described in the description for
+.IR "aio_read"
+or
+.IR "aio_write"
+.
+
+When the sources are compiled with
+.B "_FILE_OFFSET_BITS == 64"
+, this
+function is in fact
+.IR "lio_listio64"
+since the LFS interface
+transparently replaces the normal implementation.
+.SH "RETURN VALUES"
+If
+.IR "mode"
+is
+.B "LIO_WAIT"
+, the return value of
+.IR "lio_listio"
+is
+.IR 0
+when all requests completed successfully. Otherwise the
+function return
+.IR 1
+and
+.IR "errno"
+is set accordingly. To find
+out which request or requests failed one has to use the
+.IR "aio_error"
+function on all the elements of the array
+.IR "list"
+.
+
+In case
+.IR "mode"
+is
+.B "LIO_NOWAIT"
+, the function returns
+.IR 0
+if
+all requests were enqueued correctly. The current state of the requests
+can be found using
+.IR "aio_error"
+and
+.IR "aio_return"
+as described
+above. If
+.IR "lio_listio"
+returns
+.IR -1
+in this mode, the
+global variable
+.IR "errno"
+is set accordingly. If a request did not
+yet terminate, a call to
+.IR "aio_error"
+returns
+.B "EINPROGRESS"
+. If
+the value is different, the request is finished and the error value (or
+
+.IR 0
+) is returned and the result of the operation can be retrieved
+using
+.IR "aio_return"
+.
+.SH ERRORS
+Possible values for
+.IR "errno"
+are:
+
+.TP
+.B EAGAIN
+The resources necessary to queue all the requests are not available at
+the moment. The error status for each element of
+.IR "list"
+must be
+checked to determine which request failed.
+
+Another reason could be that the system wide limit of AIO requests is
+exceeded. This cannot be the case for the implementation on GNU systems
+since no arbitrary limits exist.
+.TP
+.B EINVAL
+The
+.IR "mode"
+parameter is invalid or
+.IR "nent"
+is larger than
+.B "AIO_LISTIO_MAX"
+.
+.TP
+.B EIO
+One or more of the request's I/O operations failed. The error status of
+each request should be checked to determine which one failed.
+.TP
+.B ENOSYS
+The
+.IR "lio_listio"
+function is not supported.
+.PP
+
+If the
+.IR "mode"
+parameter is
+.B "LIO_NOWAIT"
+and the caller cancels
+a request, the error status for this request returned by
+.IR "aio_error"
+is
+.B "ECANCELED"
+.
+.SH "SEE ALSO"
+.BR aio(3),
+.BR aio_cancel(3),
+.BR aio_cancel64(3),
+.BR aio_error(3),
+.BR aio_error64(3),
+.BR aio_fsync(3),
+.BR aio_fsync64(3),
+.BR aio_init(3),
+.BR aio_read(3),
+.BR aio_read64(3),
+.BR aio_return(3),
+.BR aio_return64(3),
+.BR aio_suspend(3),
+.BR aio_suspend64(3),
+.BR aio_write(3),
+.BR aio_write64(3)
diff --git a/man/lio_listio64.3 b/man/lio_listio64.3
new file mode 100644
index 0000000..97f6955
--- /dev/null
+++ b/man/lio_listio64.3
@@ -0,0 +1,39 @@
+.TH lio_listio64 3 2002-09-12 "Linux 2.4" Linux AIO"
+.SH NAME
+lio_listio64 \- List directed I/O
+.SH SYNOPSYS
+.B #include <errno.h>
+.br
+.B #include <libaio.h>
+.LP
+.BI "int lio_listio64 (int mode, struct aiocb *const list[], int nent, struct sigevent *sig)"
+.nf
+.SH DESCRIPTION
+This function is similar to the
+.IR "code{lio_listio"
+function. The only
+difference is that on
+.IR "32 bit"
+machines, the file descriptor should
+be opened in the large file mode. Internally,
+.IR "lio_listio64"
+uses
+functionality equivalent to
+.IR lseek64"
+to position the file descriptor correctly for the reading or
+writing, as opposed to
+.IR "lseek"
+functionality used in
+.IR "lio_listio".
+
+When the sources are compiled with
+.IR "_FILE_OFFSET_BITS == 64"
+, this
+function is available under the name
+.IR "lio_listio"
+and so
+transparently replaces the interface for small files on 32 bit
+machines.
+.SH "RETURN VALUES"
+.SH ERRORS
+.SH "SEE ALSO"
diff --git a/raw_syscall.c b/raw_syscall.c
new file mode 100644
index 0000000..c3fe4b8
--- /dev/null
+++ b/raw_syscall.c
@@ -0,0 +1,19 @@
+#include "syscall.h"
+
+#if defined(__ia64__)
+/* based on code from glibc by Jes Sorensen */
+__asm__(".text\n"
+ ".globl __ia64_aio_raw_syscall\n"
+ ".proc __ia64_aio_raw_syscall\n"
+ "__ia64_aio_raw_syscall:\n"
+ "alloc r2=ar.pfs,1,0,8,0\n"
+ "mov r15=r32\n"
+ "break 0x100000\n"
+ ";;"
+ "br.ret.sptk.few b0\n"
+ ".size __ia64_aio_raw_syscall, . - __ia64_aio_raw_syscall\n"
+ ".endp __ia64_aio_raw_syscall"
+);
+#endif
+
+;
diff --git a/syscall-alpha.h b/syscall-alpha.h
new file mode 100644
index 0000000..467b74f
--- /dev/null
+++ b/syscall-alpha.h
@@ -0,0 +1,209 @@
+#define __NR_io_setup 398
+#define __NR_io_destroy 399
+#define __NR_io_getevents 400
+#define __NR_io_submit 401
+#define __NR_io_cancel 402
+
+#define inline_syscall_r0_asm
+#define inline_syscall_r0_out_constraint "=v"
+
+#define inline_syscall_clobbers \
+ "$1", "$2", "$3", "$4", "$5", "$6", "$7", "$8", \
+ "$22", "$23", "$24", "$25", "$27", "$28", "memory"
+
+#define inline_syscall0(name, args...) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_19 __asm__("$19"); \
+ \
+ _sc_0 = name; \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19) \
+ : "0"(_sc_0) \
+ : inline_syscall_clobbers, \
+ "$16", "$17", "$18", "$20", "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall1(name,arg1) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_19 __asm__("$19"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16) \
+ : "0"(_sc_0), "2"(_sc_16) \
+ : inline_syscall_clobbers, \
+ "$17", "$18", "$20", "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall2(name,arg1,arg2) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_17 __asm__("$17"); \
+ register long _sc_19 __asm__("$19"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ _sc_17 = (long) (arg2); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3 %4" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16), "=r"(_sc_17) \
+ : "0"(_sc_0), "2"(_sc_16), "3"(_sc_17) \
+ : inline_syscall_clobbers, \
+ "$18", "$20", "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall3(name,arg1,arg2,arg3) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_17 __asm__("$17"); \
+ register long _sc_18 __asm__("$18"); \
+ register long _sc_19 __asm__("$19"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ _sc_17 = (long) (arg2); \
+ _sc_18 = (long) (arg3); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3 %4 %5" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16), "=r"(_sc_17), \
+ "=r"(_sc_18) \
+ : "0"(_sc_0), "2"(_sc_16), "3"(_sc_17), \
+ "4"(_sc_18) \
+ : inline_syscall_clobbers, "$20", "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall4(name,arg1,arg2,arg3,arg4) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_17 __asm__("$17"); \
+ register long _sc_18 __asm__("$18"); \
+ register long _sc_19 __asm__("$19"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ _sc_17 = (long) (arg2); \
+ _sc_18 = (long) (arg3); \
+ _sc_19 = (long) (arg4); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3 %4 %5 %6" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16), "=r"(_sc_17), \
+ "=r"(_sc_18) \
+ : "0"(_sc_0), "2"(_sc_16), "3"(_sc_17), \
+ "4"(_sc_18), "1"(_sc_19) \
+ : inline_syscall_clobbers, "$20", "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall5(name,arg1,arg2,arg3,arg4,arg5) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_17 __asm__("$17"); \
+ register long _sc_18 __asm__("$18"); \
+ register long _sc_19 __asm__("$19"); \
+ register long _sc_20 __asm__("$20"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ _sc_17 = (long) (arg2); \
+ _sc_18 = (long) (arg3); \
+ _sc_19 = (long) (arg4); \
+ _sc_20 = (long) (arg5); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3 %4 %5 %6 %7" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16), "=r"(_sc_17), \
+ "=r"(_sc_18), "=r"(_sc_20) \
+ : "0"(_sc_0), "2"(_sc_16), "3"(_sc_17), \
+ "4"(_sc_18), "1"(_sc_19), "5"(_sc_20) \
+ : inline_syscall_clobbers, "$21"); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define inline_syscall6(name,arg1,arg2,arg3,arg4,arg5,arg6) \
+{ \
+ register long _sc_0 inline_syscall_r0_asm; \
+ register long _sc_16 __asm__("$16"); \
+ register long _sc_17 __asm__("$17"); \
+ register long _sc_18 __asm__("$18"); \
+ register long _sc_19 __asm__("$19"); \
+ register long _sc_20 __asm__("$20"); \
+ register long _sc_21 __asm__("$21"); \
+ \
+ _sc_0 = name; \
+ _sc_16 = (long) (arg1); \
+ _sc_17 = (long) (arg2); \
+ _sc_18 = (long) (arg3); \
+ _sc_19 = (long) (arg4); \
+ _sc_20 = (long) (arg5); \
+ _sc_21 = (long) (arg6); \
+ __asm__ __volatile__ \
+ ("callsys # %0 %1 <= %2 %3 %4 %5 %6 %7 %8" \
+ : inline_syscall_r0_out_constraint (_sc_0), \
+ "=r"(_sc_19), "=r"(_sc_16), "=r"(_sc_17), \
+ "=r"(_sc_18), "=r"(_sc_20), "=r"(_sc_21) \
+ : "0"(_sc_0), "2"(_sc_16), "3"(_sc_17), "4"(_sc_18), \
+ "1"(_sc_19), "5"(_sc_20), "6"(_sc_21) \
+ : inline_syscall_clobbers); \
+ _sc_ret = _sc_0, _sc_err = _sc_19; \
+}
+
+#define INLINE_SYSCALL1(name, nr, args...) \
+({ \
+ long _sc_ret, _sc_err; \
+ inline_syscall##nr(__NR_##name, args); \
+ if (_sc_err != 0) \
+ { \
+ _sc_ret = -(_sc_ret); \
+ } \
+ _sc_ret; \
+})
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) \
+{ \
+ return (type)INLINE_SYSCALL1(sname, 1, arg1); \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1,type2 arg2) \
+{ \
+ return (type)INLINE_SYSCALL1(sname, 2, arg1, arg2); \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1,type2 arg2,type3 arg3) \
+{ \
+ return (type)INLINE_SYSCALL1(sname, 3, arg1, arg2, arg3); \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type fname (type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
+{ \
+ return (type)INLINE_SYSCALL1(sname, 4, arg1, arg2, arg3, arg4); \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4, \
+ type5,arg5) \
+type fname (type1 arg1,type2 arg2,type3 arg3,type4 arg4,type5 arg5) \
+{ \
+ return (type)INLINE_SYSCALL1(sname, 5, arg1, arg2, arg3, arg4, arg5);\
+}
diff --git a/syscall-arm.h b/syscall-arm.h
new file mode 100644
index 0000000..4a9b665
--- /dev/null
+++ b/syscall-arm.h
@@ -0,0 +1,116 @@
+/*
+ * linux/include/asm-arm/unistd.h
+ *
+ * Copyright (C) 2001-2005 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * Please forward _all_ changes to this file to rmk@arm.linux.org.uk,
+ * no matter what the change is. Thanks!
+ */
+
+#define __NR_OABI_SYSCALL_BASE 0x900000
+
+#if defined(__thumb__) || defined(__ARM_EABI__)
+#define __NR_SYSCALL_BASE 0
+#else
+#define __NR_SYSCALL_BASE __NR_OABI_SYSCALL_BASE
+#endif
+
+#define __NR_io_setup (__NR_SYSCALL_BASE+243)
+#define __NR_io_destroy (__NR_SYSCALL_BASE+244)
+#define __NR_io_getevents (__NR_SYSCALL_BASE+245)
+#define __NR_io_submit (__NR_SYSCALL_BASE+246)
+#define __NR_io_cancel (__NR_SYSCALL_BASE+247)
+
+#define __sys2(x) #x
+#define __sys1(x) __sys2(x)
+
+#if defined(__thumb__) || defined(__ARM_EABI__)
+#define __SYS_REG(name) register long __sysreg __asm__("r7") = __NR_##name;
+#define __SYS_REG_LIST(regs...) "r" (__sysreg) , ##regs
+#define __syscall(name) "swi\t0"
+#else
+#define __SYS_REG(name)
+#define __SYS_REG_LIST(regs...) regs
+#define __syscall(name) "swi\t" __sys1(__NR_##name) ""
+#endif
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) { \
+ __SYS_REG(sname) \
+ register long __r0 __asm__("r0") = (long)arg1; \
+ register long __res_r0 __asm__("r0"); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_r0) \
+ : __SYS_REG_LIST( "0" (__r0) ) \
+ : "memory" ); \
+ return (type) __res_r0; \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1,type2 arg2) { \
+ __SYS_REG(sname) \
+ register long __r0 __asm__("r0") = (long)arg1; \
+ register long __r1 __asm__("r1") = (long)arg2; \
+ register long __res_r0 __asm__("r0"); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_r0) \
+ : __SYS_REG_LIST( "0" (__r0), "r" (__r1) ) \
+ : "memory" ); \
+ return (type) __res_r0; \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1,type2 arg2,type3 arg3) { \
+ __SYS_REG(sname) \
+ register long __r0 __asm__("r0") = (long)arg1; \
+ register long __r1 __asm__("r1") = (long)arg2; \
+ register long __r2 __asm__("r2") = (long)arg3; \
+ register long __res_r0 __asm__("r0"); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_r0) \
+ : __SYS_REG_LIST( "0" (__r0), "r" (__r1), "r" (__r2) ) \
+ : "memory" ); \
+ return (type) __res_r0; \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4)\
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4) { \
+ __SYS_REG(sname) \
+ register long __r0 __asm__("r0") = (long)arg1; \
+ register long __r1 __asm__("r1") = (long)arg2; \
+ register long __r2 __asm__("r2") = (long)arg3; \
+ register long __r3 __asm__("r3") = (long)arg4; \
+ register long __res_r0 __asm__("r0"); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_r0) \
+ : __SYS_REG_LIST( "0" (__r0), "r" (__r1), "r" (__r2), "r" (__r3) ) \
+ : "memory" ); \
+ return (type) __res_r0; \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5) {\
+ __SYS_REG(sname) \
+ register long __r0 __asm__("r0") = (long)arg1; \
+ register long __r1 __asm__("r1") = (long)arg2; \
+ register long __r2 __asm__("r2") = (long)arg3; \
+ register long __r3 __asm__("r3") = (long)arg4; \
+ register long __r4 __asm__("r4") = (long)arg5; \
+ register long __res_r0 __asm__("r0"); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_r0) \
+ : __SYS_REG_LIST( "0" (__r0), "r" (__r1), "r" (__r2), \
+ "r" (__r3), "r" (__r4) ) \
+ : "memory" ); \
+ return (type) __res_r0; \
+}
+
diff --git a/syscall-arm64.h b/syscall-arm64.h
new file mode 100644
index 0000000..9c93c9b
--- /dev/null
+++ b/syscall-arm64.h
@@ -0,0 +1,123 @@
+/*
+ * linux/include/asm-arm/unistd.h
+ *
+ * Copyright (C) 2001-2005 Russell King
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * Please forward _all_ changes to this file to rmk@arm.linux.org.uk,
+ * no matter what the change is. Thanks!
+ */
+
+// Syscall hooks for Sanitizer tools.
+// They help sanitizers update their knowledge of the process memory state
+// before and after a system call.
+#if defined(ADDRESS_SANITIZER) || defined(HWADDRESS_SANITIZER) || defined(MEMORY_SANITIZER) || defined(THREAD_SANITIZER)
+#include <sanitizer/linux_syscall_hooks.h>
+#define PRE(name, ...) __sanitizer_syscall_pre_##name(__VA_ARGS__)
+#define POST(name, res, ...) __sanitizer_syscall_post_##name(res, __VA_ARGS__)
+#else
+#define PRE(...)
+#define POST(...)
+#endif
+
+#define __NR_io_setup 0
+#define __NR_io_destroy 1
+#define __NR_io_submit 2
+#define __NR_io_cancel 3
+#define __NR_io_getevents 4
+
+#define __sys2(x) #x
+#define __sys1(x) __sys2(x)
+
+#define __SYS_REG(name) register long __sysreg __asm__("w8") = __NR_##name;
+#define __SYS_REG_LIST(regs...) "r" (__sysreg) , ##regs
+#define __syscall(name) "svc\t#0"
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) { \
+ __SYS_REG(sname) \
+ register long __x0 __asm__("x0") = (long)arg1; \
+ register long __res_x0 __asm__("x0"); \
+ PRE(sname, arg1); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_x0) \
+ : __SYS_REG_LIST( "0" (__x0) ) \
+ : "memory" ); \
+ POST(sname, __res_x0, arg1); \
+ return (type) __res_x0; \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1,type2 arg2) { \
+ __SYS_REG(sname) \
+ register long __x0 __asm__("x0") = (long)arg1; \
+ register long __x1 __asm__("x1") = (long)arg2; \
+ register long __res_x0 __asm__("x0"); \
+ PRE(sname, arg1, arg2); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_x0) \
+ : __SYS_REG_LIST( "0" (__x0), "r" (__x1) ) \
+ : "memory" ); \
+ POST(sname, __res_x0, arg1, arg2); \
+ return (type) __res_x0; \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1,type2 arg2,type3 arg3) { \
+ __SYS_REG(sname) \
+ register long __x0 __asm__("x0") = (long)arg1; \
+ register long __x1 __asm__("x1") = (long)arg2; \
+ register long __x2 __asm__("x2") = (long)arg3; \
+ register long __res_x0 __asm__("x0"); \
+ PRE(sname, arg1, arg2, arg3); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_x0) \
+ : __SYS_REG_LIST( "0" (__x0), "r" (__x1), "r" (__x2) ) \
+ : "memory" ); \
+ POST(sname, __res_x0, arg1, arg2, arg3); \
+ return (type) __res_x0; \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4)\
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4) { \
+ __SYS_REG(sname) \
+ register long __x0 __asm__("x0") = (long)arg1; \
+ register long __x1 __asm__("x1") = (long)arg2; \
+ register long __x2 __asm__("x2") = (long)arg3; \
+ register long __x3 __asm__("x3") = (long)arg4; \
+ register long __res_x0 __asm__("x0"); \
+ PRE(sname, arg1, arg2, arg3, arg4); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_x0) \
+ : __SYS_REG_LIST( "0" (__x0), "r" (__x1), "r" (__x2), "r" (__x3) ) \
+ : "memory" ); \
+ POST(sname, __res_x0, arg1, arg2, arg3, arg4); \
+ return (type) __res_x0; \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5) {\
+ __SYS_REG(sname) \
+ register long __x0 __asm__("x0") = (long)arg1; \
+ register long __x1 __asm__("x1") = (long)arg2; \
+ register long __x2 __asm__("x2") = (long)arg3; \
+ register long __x3 __asm__("x3") = (long)arg4; \
+ register long __x4 __asm__("x4") = (long)arg5; \
+ register long __res_x0 __asm__("x0"); \
+ PRE(sname, arg1, arg2, arg3, arg4, arg5); \
+ __asm__ __volatile__ ( \
+ __syscall(sname) \
+ : "=r" (__res_x0) \
+ : __SYS_REG_LIST( "0" (__x0), "r" (__x1), "r" (__x2), \
+ "r" (__x3), "r" (__x4) ) \
+ : "memory" ); \
+ POST(sname, __res_x0, arg1, arg2, arg3, arg4, arg5); \
+ return (type) __res_x0; \
+}
diff --git a/syscall-i386.h b/syscall-i386.h
new file mode 100644
index 0000000..a2b8d02
--- /dev/null
+++ b/syscall-i386.h
@@ -0,0 +1,72 @@
+#define __NR_io_setup 245
+#define __NR_io_destroy 246
+#define __NR_io_getevents 247
+#define __NR_io_submit 248
+#define __NR_io_cancel 249
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) \
+{ \
+long __res; \
+__asm__ volatile ("xchgl %%edi,%%ebx\n" \
+ "int $0x80\n" \
+ "xchgl %%edi,%%ebx" \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1))); \
+return __res; \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1,type2 arg2) \
+{ \
+long __res; \
+__asm__ volatile ("xchgl %%edi,%%ebx\n" \
+ "int $0x80\n" \
+ "xchgl %%edi,%%ebx" \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"c" ((long)(arg2))); \
+return __res; \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1,type2 arg2,type3 arg3) \
+{ \
+long __res; \
+__asm__ volatile ("xchgl %%edi,%%ebx\n" \
+ "int $0x80\n" \
+ "xchgl %%edi,%%ebx" \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"c" ((long)(arg2)), \
+ "d" ((long)(arg3))); \
+return __res; \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type fname (type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
+{ \
+long __res; \
+__asm__ volatile ("xchgl %%edi,%%ebx\n" \
+ "int $0x80\n" \
+ "xchgl %%edi,%%ebx" \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"c" ((long)(arg2)), \
+ "d" ((long)(arg3)),"S" ((long)(arg4))); \
+return __res; \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4, \
+ type5,arg5) \
+type fname (type1 arg1,type2 arg2,type3 arg3,type4 arg4,type5 arg5) \
+{ \
+long __res; \
+long tmp = 0; \
+__asm__ volatile ("movl %%ebx,%7\n" \
+ "movl %2,%%ebx\n" \
+ "int $0x80\n" \
+ "movl %7,%%ebx" \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"rm" ((long)(arg1)),"c" ((long)(arg2)), \
+ "d" ((long)(arg3)),"S" ((long)(arg4)),"D" ((long)(arg5)), \
+ "m" (tmp)); \
+return __res; \
+}
diff --git a/syscall-ia64.h b/syscall-ia64.h
new file mode 100644
index 0000000..52ce9dd
--- /dev/null
+++ b/syscall-ia64.h
@@ -0,0 +1,45 @@
+#define __NR_io_setup 1238
+#define __NR_io_destroy 1239
+#define __NR_io_getevents 1240
+#define __NR_io_submit 1241
+#define __NR_io_cancel 1242
+
+#define __ia64_raw_syscall(fname, sname) \
+ __asm__ (".text\n" \
+ ".globl " SYMSTR(fname) "\n" \
+ ".proc " SYMSTR(fname) "\n" \
+ SYMSTR(fname) ":\n" \
+ " mov r15=" SYMSTR( __NR_ ## sname ) "\n" \
+ " break 0x100000\n" \
+ " ;;\n" \
+ " cmp.eq p6,p0=-1,r10\n" \
+ " ;;\n" \
+ " (p6) sub r8=0,r8\n" \
+ " br.ret.sptk.few b0\n" \
+ ".size " SYMSTR(fname) ", . - " SYMSTR(fname) "\n" \
+ ".endp " SYMSTR(fname) "\n" \
+ );
+
+#define io_syscall0(type, name) \
+ extern type name(void); \
+ __ia64_raw_syscall(name);
+
+#define io_syscall1(type, fname, sname, type1, arg1) \
+ extern type fname(type1 arg1); \
+ __ia64_raw_syscall(fname, sname);
+
+#define io_syscall2(type, fname, sname, type1, arg1, type2, arg2) \
+ extern type fname(type1 arg1, type2 arg2); \
+ __ia64_raw_syscall(fname, sname);
+
+#define io_syscall3(type, fname, sname, type1, arg1, type2, arg2, type3, arg3) \
+ extern type fname(type1 arg1, type2 arg2, type3 arg3); \
+ __ia64_raw_syscall(fname, sname);
+
+#define io_syscall4(type, fname, sname, type1, arg1, type2, arg2, type3, arg3, type4, arg4) \
+ extern type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4); \
+ __ia64_raw_syscall(fname, sname);
+
+#define io_syscall5(type, fname, sname, type1, arg1, type2, arg2, type3, arg3, type4, arg4, type5, arg5) \
+ extern type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5); \
+ __ia64_raw_syscall(fname, sname);
diff --git a/syscall-ppc.h b/syscall-ppc.h
new file mode 100644
index 0000000..ca70dd2
--- /dev/null
+++ b/syscall-ppc.h
@@ -0,0 +1,94 @@
+#define __NR_io_setup 227
+#define __NR_io_destroy 228
+#define __NR_io_getevents 229
+#define __NR_io_submit 230
+#define __NR_io_cancel 231
+
+/* On powerpc a system call basically clobbers the same registers like a
+ * function call, with the exception of LR (which is needed for the
+ * "sc; bnslr" sequence) and CR (where only CR0.SO is clobbered to signal
+ * an error return status).
+ */
+
+#define __syscall_nr(nr, type, name, args...) \
+ unsigned long __sc_ret, __sc_err; \
+ { \
+ register unsigned long __sc_0 __asm__ ("r0"); \
+ register unsigned long __sc_3 __asm__ ("r3"); \
+ register unsigned long __sc_4 __asm__ ("r4"); \
+ register unsigned long __sc_5 __asm__ ("r5"); \
+ register unsigned long __sc_6 __asm__ ("r6"); \
+ register unsigned long __sc_7 __asm__ ("r7"); \
+ register unsigned long __sc_8 __asm__ ("r8"); \
+ \
+ __sc_loadargs_##nr(name, args); \
+ __asm__ __volatile__ \
+ ("sc \n\t" \
+ "mfcr %0 " \
+ : "=&r" (__sc_0), \
+ "=&r" (__sc_3), "=&r" (__sc_4), \
+ "=&r" (__sc_5), "=&r" (__sc_6), \
+ "=&r" (__sc_7), "=&r" (__sc_8) \
+ : __sc_asm_input_##nr \
+ : "cr0", "ctr", "memory", \
+ "r9", "r10","r11", "r12"); \
+ __sc_ret = __sc_3; \
+ __sc_err = __sc_0; \
+ } \
+ if (__sc_err & 0x10000000) return -((int)__sc_ret); \
+ return (type) __sc_ret
+
+#define __sc_loadargs_0(name, dummy...) \
+ __sc_0 = __NR_##name
+#define __sc_loadargs_1(name, arg1) \
+ __sc_loadargs_0(name); \
+ __sc_3 = (unsigned long) (arg1)
+#define __sc_loadargs_2(name, arg1, arg2) \
+ __sc_loadargs_1(name, arg1); \
+ __sc_4 = (unsigned long) (arg2)
+#define __sc_loadargs_3(name, arg1, arg2, arg3) \
+ __sc_loadargs_2(name, arg1, arg2); \
+ __sc_5 = (unsigned long) (arg3)
+#define __sc_loadargs_4(name, arg1, arg2, arg3, arg4) \
+ __sc_loadargs_3(name, arg1, arg2, arg3); \
+ __sc_6 = (unsigned long) (arg4)
+#define __sc_loadargs_5(name, arg1, arg2, arg3, arg4, arg5) \
+ __sc_loadargs_4(name, arg1, arg2, arg3, arg4); \
+ __sc_7 = (unsigned long) (arg5)
+
+#define __sc_asm_input_0 "0" (__sc_0)
+#define __sc_asm_input_1 __sc_asm_input_0, "1" (__sc_3)
+#define __sc_asm_input_2 __sc_asm_input_1, "2" (__sc_4)
+#define __sc_asm_input_3 __sc_asm_input_2, "3" (__sc_5)
+#define __sc_asm_input_4 __sc_asm_input_3, "4" (__sc_6)
+#define __sc_asm_input_5 __sc_asm_input_4, "5" (__sc_7)
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) \
+{ \
+ __syscall_nr(1, type, sname, arg1); \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1, type2 arg2) \
+{ \
+ __syscall_nr(2, type, sname, arg1, arg2); \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1, type2 arg2, type3 arg3) \
+{ \
+ __syscall_nr(3, type, sname, arg1, arg2, arg3); \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
+{ \
+ __syscall_nr(4, type, sname, arg1, arg2, arg3, arg4); \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4,type5,arg5) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4, type5 arg5) \
+{ \
+ __syscall_nr(5, type, sname, arg1, arg2, arg3, arg4, arg5); \
+}
diff --git a/syscall-s390.h b/syscall-s390.h
new file mode 100644
index 0000000..3ec5ee3
--- /dev/null
+++ b/syscall-s390.h
@@ -0,0 +1,131 @@
+#define __NR_io_setup 243
+#define __NR_io_destroy 244
+#define __NR_io_getevents 245
+#define __NR_io_submit 246
+#define __NR_io_cancel 247
+
+#define io_svc_clobber "1", "cc", "memory"
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) { \
+ register type1 __arg1 asm("2") = arg1; \
+ register long __svcres asm("2"); \
+ long __res; \
+ __asm__ __volatile__ ( \
+ " .if %1 < 256\n" \
+ " svc %b1\n" \
+ " .else\n" \
+ " la %%r1,%1\n" \
+ " .svc 0\n" \
+ " .endif" \
+ : "=d" (__svcres) \
+ : "i" (__NR_##sname), \
+ "0" (__arg1) \
+ : io_svc_clobber ); \
+ __res = __svcres; \
+ return (type) __res; \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1, type2 arg2) { \
+ register type1 __arg1 asm("2") = arg1; \
+ register type2 __arg2 asm("3") = arg2; \
+ register long __svcres asm("2"); \
+ long __res; \
+ __asm__ __volatile__ ( \
+ " .if %1 < 256\n" \
+ " svc %b1\n" \
+ " .else\n" \
+ " la %%r1,%1\n" \
+ " svc 0\n" \
+ " .endif" \
+ : "=d" (__svcres) \
+ : "i" (__NR_##sname), \
+ "0" (__arg1), \
+ "d" (__arg2) \
+ : io_svc_clobber ); \
+ __res = __svcres; \
+ return (type) __res; \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2, \
+ type3,arg3) \
+type fname(type1 arg1, type2 arg2, type3 arg3) { \
+ register type1 __arg1 asm("2") = arg1; \
+ register type2 __arg2 asm("3") = arg2; \
+ register type3 __arg3 asm("4") = arg3; \
+ register long __svcres asm("2"); \
+ long __res; \
+ __asm__ __volatile__ ( \
+ " .if %1 < 256\n" \
+ " svc %b1\n" \
+ " .else\n" \
+ " la %%r1,%1\n" \
+ " svc 0\n" \
+ " .endif" \
+ : "=d" (__svcres) \
+ : "i" (__NR_##sname), \
+ "0" (__arg1), \
+ "d" (__arg2), \
+ "d" (__arg3) \
+ : io_svc_clobber ); \
+ __res = __svcres; \
+ return (type) __res; \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2, \
+ type3,arg3,type4,arg4) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4) { \
+ register type1 __arg1 asm("2") = arg1; \
+ register type2 __arg2 asm("3") = arg2; \
+ register type3 __arg3 asm("4") = arg3; \
+ register type4 __arg4 asm("5") = arg4; \
+ register long __svcres asm("2"); \
+ long __res; \
+ __asm__ __volatile__ ( \
+ " .if %1 < 256\n" \
+ " svc %b1\n" \
+ " .else\n" \
+ " la %%r1,%1\n" \
+ " svc 0\n" \
+ " .endif" \
+ : "=d" (__svcres) \
+ : "i" (__NR_##sname), \
+ "0" (__arg1), \
+ "d" (__arg2), \
+ "d" (__arg3), \
+ "d" (__arg4) \
+ : io_svc_clobber ); \
+ __res = __svcres; \
+ return (type) __res; \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2, \
+ type3,arg3,type4,arg4,type5,arg5) \
+type fname(type1 arg1, type2 arg2, type3 arg3, type4 arg4, \
+ type5 arg5) { \
+ register type1 __arg1 asm("2") = arg1; \
+ register type2 __arg2 asm("3") = arg2; \
+ register type3 __arg3 asm("4") = arg3; \
+ register type4 __arg4 asm("5") = arg4; \
+ register type5 __arg5 asm("6") = arg5; \
+ register long __svcres asm("2"); \
+ long __res; \
+ __asm__ __volatile__ ( \
+ " .if %1 < 256\n" \
+ " svc %b1\n" \
+ " .else\n" \
+ " la %%r1,%1\n" \
+ " svc 0\n" \
+ " .endif" \
+ : "=d" (__svcres) \
+ : "i" (__NR_##sname), \
+ "0" (__arg1), \
+ "d" (__arg2), \
+ "d" (__arg3), \
+ "d" (__arg4), \
+ "d" (__arg5) \
+ : io_svc_clobber ); \
+ __res = __svcres; \
+ return (type) __res; \
+}
diff --git a/syscall-x86_64.h b/syscall-x86_64.h
new file mode 100644
index 0000000..8972cab
--- /dev/null
+++ b/syscall-x86_64.h
@@ -0,0 +1,85 @@
+// Syscall hooks for Sanitizer tools.
+// They help sanitizers update their knowledge of the process memory state
+// before and after a system call.
+#if defined(ADDRESS_SANITIZER) || defined(HWADDRESS_SANITIZER) || defined(MEMORY_SANITIZER) || defined(THREAD_SANITIZER)
+#include <sanitizer/linux_syscall_hooks.h>
+#define PRE(name, ...) __sanitizer_syscall_pre_##name(__VA_ARGS__)
+#define POST(name, res, ...) __sanitizer_syscall_post_##name(res, __VA_ARGS__)
+#else
+#define PRE(...)
+#define POST(...)
+#endif
+
+#define __NR_io_setup 206
+#define __NR_io_destroy 207
+#define __NR_io_getevents 208
+#define __NR_io_submit 209
+#define __NR_io_cancel 210
+
+#define __syscall_clobber "r11","rcx","memory"
+#define __syscall "syscall"
+
+#define io_syscall1(type,fname,sname,type1,arg1) \
+type fname(type1 arg1) \
+{ \
+long __res; \
+ PRE(sname, arg1); \
+__asm__ volatile (__syscall \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)) : __syscall_clobber ); \
+ POST(sname, __res, arg1); \
+return __res; \
+}
+
+#define io_syscall2(type,fname,sname,type1,arg1,type2,arg2) \
+type fname(type1 arg1,type2 arg2) \
+{ \
+long __res; \
+ PRE(sname, arg1, arg2); \
+__asm__ volatile (__syscall \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"S" ((long)(arg2)) : __syscall_clobber ); \
+ POST(sname, __res, arg1, arg2); \
+return __res; \
+}
+
+#define io_syscall3(type,fname,sname,type1,arg1,type2,arg2,type3,arg3) \
+type fname(type1 arg1,type2 arg2,type3 arg3) \
+{ \
+long __res; \
+ PRE(sname, arg1, arg2, arg3); \
+__asm__ volatile (__syscall \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"S" ((long)(arg2)), \
+ "d" ((long)(arg3)) : __syscall_clobber); \
+ POST(sname, __res, arg1, arg2, arg3); \
+return __res; \
+}
+
+#define io_syscall4(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4) \
+type fname (type1 arg1, type2 arg2, type3 arg3, type4 arg4) \
+{ \
+long __res; \
+ PRE(sname, arg1, arg2, arg3, arg4); \
+__asm__ volatile ("movq %5,%%r10 ;" __syscall \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"S" ((long)(arg2)), \
+ "d" ((long)(arg3)),"g" ((long)(arg4)) : __syscall_clobber,"r10" ); \
+ POST(sname, __res, arg1, arg2, arg3, arg4); \
+return __res; \
+}
+
+#define io_syscall5(type,fname,sname,type1,arg1,type2,arg2,type3,arg3,type4,arg4, \
+ type5,arg5) \
+type fname (type1 arg1,type2 arg2,type3 arg3,type4 arg4,type5 arg5) \
+{ \
+long __res; \
+ PRE(sname, arg1, arg2, arg3, arg4, arg5); \
+__asm__ volatile ("movq %5,%%r10 ; movq %6,%%r8 ; " __syscall \
+ : "=a" (__res) \
+ : "0" (__NR_##sname),"D" ((long)(arg1)),"S" ((long)(arg2)), \
+ "d" ((long)(arg3)),"g" ((long)(arg4)),"g" ((long)(arg5)) : \
+ __syscall_clobber,"r8","r10" ); \
+ POST(sname, __res, arg1, arg2, arg3, arg4, arg5); \
+return __res; \
+}
diff --git a/syscall.h b/syscall.h
new file mode 100644
index 0000000..dc3326d
--- /dev/null
+++ b/syscall.h
@@ -0,0 +1,31 @@
+#include <sys/syscall.h>
+#include <unistd.h>
+
+#define _SYMSTR(str) #str
+#define SYMSTR(str) _SYMSTR(str)
+
+#define SYMVER(compat_sym, orig_sym, ver_sym) \
+ __asm__(".symver " SYMSTR(compat_sym) "," SYMSTR(orig_sym) "@LIBAIO_" SYMSTR(ver_sym));
+
+#define DEFSYMVER(compat_sym, orig_sym, ver_sym) \
+ __asm__(".symver " SYMSTR(compat_sym) "," SYMSTR(orig_sym) "@@LIBAIO_" SYMSTR(ver_sym));
+
+#if defined(__i386__)
+#include "syscall-i386.h"
+#elif defined(__x86_64__)
+#include "syscall-x86_64.h"
+#elif defined(__ia64__)
+#include "syscall-ia64.h"
+#elif defined(__PPC__)
+#include "syscall-ppc.h"
+#elif defined(__s390__)
+#include "syscall-s390.h"
+#elif defined(__alpha__)
+#include "syscall-alpha.h"
+#elif defined(__arm__)
+#include "syscall-arm.h"
+#elif defined(__aarch64__)
+#include "syscall-arm64.h"
+#else
+#error "add syscall-arch.h"
+#endif
diff --git a/vsys_def.h b/vsys_def.h
new file mode 100644
index 0000000..13d032e
--- /dev/null
+++ b/vsys_def.h
@@ -0,0 +1,24 @@
+/* libaio Linux async I/O interface
+ Copyright 2002 Red Hat, Inc.
+
+ This library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2 of the License, or (at your option) any later version.
+
+ This library is distributed in the hope that it will be useful,
+ but WITHOUT ANY WARRANTY; without even the implied warranty of
+ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with this library; if not, write to the Free Software
+ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ */
+extern int vsys_io_setup(unsigned nr_reqs, io_context_t *ctxp);
+extern int vsys_io_destroy(io_context_t ctx);
+extern int vsys_io_submit(io_context_t ctx, long nr, struct iocb *iocbs[]);
+extern int vsys_io_cancel(io_context_t ctx, struct iocb *iocb);
+extern int vsys_io_wait(io_context_t ctx, struct iocb *iocb, const struct timespec *when);
+extern int vsys_io_getevents(io_context_t ctx_id, long nr, struct io_event *events, const struct timespec *timeout);
+
