Changeset - r28519:990ea5120e84
[Not reviewed]
master
0 2 9
Patric Stout - 11 months ago 2024-01-18 22:13:20
truebrain@openttd.org
Add: monocypher 4.0.2

Monocypher will take care of all our encryption needs; as most
OSes and vcpkg doesn't have it available, we vendor it.
11 files changed with 4616 insertions and 0 deletions:
0 comments (0 inline, 0 general)
README.md
Show inline comments
 
@@ -194,9 +194,12 @@ See `src/3rdparty/opengl/khrplatform.h` 
 
The catch2 implementation in `src/3rdparty/catch2` is licensed under the Boost Software License, Version 1.0.
 
See `src/3rdparty/catch2/LICENSE.txt` for the complete license text.
 

	
 
The icu scriptrun implementation in `src/3rdparty/icu` is licensed under the Unicode license.
 
See `src/3rdparty/icu/LICENSE` for the complete license text.
 

	
 
The monocypher implementation in `src/3rdparty/monocypher` is licensed under the 2-clause BSD and CC-0 license.
 
See src/3rdparty/monocypher/LICENSE.md` for the complete license text.
 

	
 
## 4.0 Credits
 

	
 
See [CREDITS.md](./CREDITS.md)
src/3rdparty/CMakeLists.txt
Show inline comments
 
add_subdirectory(catch2)
 
add_subdirectory(fmt)
 
add_subdirectory(icu)
 
add_subdirectory(md5)
 
add_subdirectory(monocypher)
 
add_subdirectory(squirrel)
 
add_subdirectory(nlohmann)
 
add_subdirectory(opengl)
src/3rdparty/monocypher/AUTHORS.md
Show inline comments
 
new file 100644
 
Designers
 
---------
 

	
 
- **ChaCha20:** Daniel J. Bernstein.
 
- **Poly1305:** Daniel J. Bernstein.
 
- **BLAKE2:**   Jean-Philippe Aumasson, Christian Winnerlein, Samuel Neves,
 
                and Zooko Wilcox-O'Hearn.
 
- **Argon2:**   Alex Biryukov, Daniel Dinu, and Dmitry Khovratovich.
 
- **X25519:**   Daniel J. Bernstein.
 
- **EdDSA:**    Daniel J. Bernstein, Bo-Yin Yang, Niels Duif, Peter
 
                Schwabe, and Tanja Lange.
 

	
 
Implementors
 
------------
 

	
 
- **ChaCha20:** Loup Vaillant, implemented from spec.
 
- **Poly1305:** Loup Vaillant, implemented from spec.
 
- **BLAKE2b:**  Loup Vaillant, implemented from spec.
 
- **Argon2i:**  Loup Vaillant, implemented from spec.
 
- **X25519:**   Daniel J. Bernstein, taken and packaged from SUPERCOP
 
                ref10.
 
- **EdDSA:**    Loup Vaillant, with bits and pieces from SUPERCOP ref10.
 

	
 
Test suite
 
----------
 

	
 
Designed and implemented by Loup Vaillant, using _libsodium_ (by many
 
authors), and _ed25519-donna_ (by Andrew Moon —floodyberry).
 

	
 
Manual
 
------
 

	
 
Loup Vaillant, Fabio Scotoni, and Michael Savage.
 

	
 
- Loup Vaillant did a first draft.
 
- Fabio Scotoni rewrote the manual into proper man pages (and
 
  substantially changed it in the process).
 
- Michael Savage did extensive editing and proofreading.
 

	
 
Thanks
 
------
 

	
 
Fabio Scotoni provided much needed advice about testing, interface,
 
packaging, and the general direction of the whole project.  He also
 
redesigned the monocypher.org style sheets.
 

	
 
Mike Pechkin and André Maroneze found bugs in earlier versions of
 
Monocypher.
 

	
 
Andrew Moon clarified carry propagation in modular arithmetic, and
 
provided advice and code that significantly simplified and improved
 
Elligator2 mappings.
 

	
 
Mike Hamburg explained comb algorithms, including the signed
 
all-bits-set comb described in his 2012 paper, Fast and compact
 
elliptic-curve cryptography.  This made EdDSA signatures over twice as
 
fast.
 

	
 
Samuel Lucas found many typos in both the manual and the website.
 

	
 
Jens Alfke added some #ifdefs that enabled Monocypher to compile into
 
a C++ namespace, preventing symbol collisions with similarly-named
 
functions in other crypto libraries.
src/3rdparty/monocypher/CHANGELOG.md
Show inline comments
 
new file 100644
 
4.0.2
 
-----
 
2023/08/24
 

	
 
- Fixed multiple-lanes Argon2.
 
- Improved Poly1305 performance.
 
- Improved Argon2 performance.
 
- Makefiles no longer override standard environment variables.
 

	
 

	
 
4.0.1
 
-----
 
2023/03/06
 

	
 
- Fixed Ed25519 secret key size in function prototype.
 
- Fixed soname (should have been changed in 4.0.0)
 
- Added convenience sub-targets to makefile.
 
- Briefly specified wire format of Elligator and incremental AEAD.
 

	
 

	
 
4.0.0
 
-----
 
2023/02/20
 

	
 
- Fixed unsafe signature API.
 
- Simpler, more flexible low-level signature API.
 
- Fully specified, consensus-friendly signatures.
 
- Added Argon2d and Argon2id, support multiple lanes.
 
- Added safe and fast streaming AEAD.
 
- Added HKDF-SHA-512 and documented BLAKE2b KDF.
 
- More consistent and memorable function names.
 
- POSIX makefile.
 

	
 

	
 
3.1.3
 
-----
 
2022/04/25
 

	
 
- Fixed many typos in the documentation.
 
- Fixed buffer overflow in speed benchmarks.
 
- Fixed some MSVC warnings.
 
- Fixed a minor violation of the Elligator2 reverse map specs.
 
- Added `change-prefix.sh` to help changing the `crypto_` prefix.
 
- Added the `MONOCYPHER_CPP_NAMESPACE` preprocessor definition to
 
  support namespaces for C++.
 
- Deprecated `crypto_key_exchange()`
 
- Use GitHub actions to automate the regular test suite.
 

	
 

	
 
3.1.2
 
-----
 
2020/12/27
 

	
 
- Addressed issues from Cure53's audit:
 
  - MON-01-001: Clarified which CSPRNG to use on Darwin.
 
  - MON-01-002: Won't fix (nonce handling is a core design decision).
 
  - MON-01-004: Compared with Kleshni's implementation.
 
  - MON-01-005: Split a dedicated "advanced" folder in the manual.
 
- Quality assurance for 2^255-19 arithmetic (elliptic curves):
 
  - Documented carry propagation.
 
  - Enforced slightly safer invariants.
 
- Improved the speed of EdDSA signature generation.
 
- Made the vectors.h header more compact and easier to modify.
 
- TIS-CI integration.
 
- Added speed benchmark for ed25519-donna.
 
- Documented lengths limits of `crypto_ietf_chacha20()`
 

	
 

	
 
3.1.1
 
-----
 
2020/06/15
 

	
 
- Various documentation fixes.
 
- Fixed various compiler warnings.
 
- Fixed some integer overflows (16-bit platforms only).
 

	
 

	
 
3.1.0
 
-----
 
2020/04/03
 

	
 
- Added Elligator 2 mappings (hash to curve, curve to hash).
 
- Added OPRF support (with scalar inversion).
 
- Added Edwards25519 -> Curve25519 conversions.
 

	
 

	
 
3.0.0
 
-----
 
2020/01/19
 

	
 
- Deprecated the incremental AEAD interface.
 
- Deprecated the incremental Chacha20, added a direct interface.
 
- Added IETF Chacha20 (96-bit nonce), as described in RFC 8439.
 
- Moved deprecated interfaces to a separate `src/deprecated` folder.
 
- Removed the `ED25519_SHA512` preprocessor flag.
 
- `crypto_x25519()` and `crypto_key_exchange()` now return `void`.
 
- Added a custom hash interface to EdDSA.  Several instances of EdDSA
 
  can share the same binary.
 
- Added optional support for HMAC SHA-512.
 
- Moved SHA-512 operations to `src/optional/monocypher-ed25519.(h|c)`.
 
- Optional support for Ed25519 no longer requires a preprocessor flag.
 
  Add `src/optional/monocypher-ed25519.(h|c)` to your project instead.
 

	
 

	
 
2.0.6
 
-----
 
2019/10/21
 

	
 
- Added the `BLAKE2_NO_UNROLLING` preprocessor definition. Activating
 
  it makes the binary about 5KB smaller and speeds up processing times
 
  on many embedded processors.
 
- Reduced the stack usage of signature verification by about
 
  40%. Signature verification now fits in smaller machines.
 
- Fixed many implicit casts warnings.
 
- Fixed the manual here and there.
 
- Lots of small nitpicks.
 

	
 

	
 
2.0.5
 
-----
 
2018/08/23
 

	
 
- Faster EdDSA signatures and verification.  Like, 4 times as fast.
 

	
 

	
 
2.0.4
 
-----
 
2018/06/24
 

	
 
- Corrected a critical vulnerability found by Mike Pechkin in EdDSA,
 
  where crypto_check() was accepting invalid signatures.  The current
 
  fix removes a buggy optimisation, effectively halving the performance
 
  of EdDSA.
 
- The test suite no longer tries to allocate zero bytes (some platforms
 
  fail such an allocation).
 

	
 
2.0.3
 
-----
 
2018/06/16
 

	
 
- Corrected undefined behaviour in BLAKE2b.
 
- Improved the test suite (faster, better coverage).
 

	
 
2.0.2
 
-----
 
2018/04/23
 

	
 
- Corrected a couple failures to wipe secret buffers.
 
- Corrected a bug that prevented compilation in Ed25519 mode.
 
- Adjusted the number of test vectors in the test suite.
 
- Improved tests for incremental interfaces.
 
- Replaced the GNU all permissive licence by a public domain dedication
 
  (Creative Commons CC-0).  The BSD licence remains as a fallback.
 

	
 
2.0.1
 
-----
 
2018/03/07
 

	
 
- Followed a systematic pattern for the loading code of symmetric
 
  crypto.  It is now easier to review.
 
- Tweaked Poly1305 code to make it easier to prove correct.
 

	
 
2.0.0
 
-----
 
2018/02/14
 

	
 
- Changed the authenticated encryption format.  It now conforms to
 
  RFC 7539, with one exception: it uses XChacha20 initialisation instead
 
  of the IETF version of Chacha20.  This new format conforms to
 
  libsodium's `crypto_aead_xchacha20poly1305_ietf_encrypt`.
 
- Removed `crypto_lock_encrypt()` and `crypto_lock_auth()`.
 
- Renamed `crypto_lock_aead_auth()` to `crypto_lock_auth_ad()`.
 
- Renamed `crypto_unlock_aead_auth()` to `crypto_unlock_auth_ad()`.
 
- Added `crypto_lock_auth_message()` and `crypto_unlock_auth_message()`.
 
- Renamed `crypto_aead_lock` to `crypto_lock_aead`.
 
- Renamed `crypto_aead_unlock` to `crypto_unlock_aead`.
 

	
 
The format change facilitates optimisation by aligning data to block
 
boundaries.  The API changes increase consistency.
 

	
 
1.1.0
 
-----
 
2018/02/06
 

	
 
- Rewrote the manual into proper man pages.
 
- Added incremental interfaces for authenticated encryption and
 
  signatures.
 
- Replaced `crypto_memcmp()` by 3 fixed size buffer comparisons (16, 32,
 
  and 64 bytes), to make sure the generated code remains constant time.
 
- A couple breaking API changes, easily fixed by renaming the affected
 
  functions.
 

	
 
1.0.1
 
-----
 
2017/07/23
 

	
 
- Optimised the loading and unloading code of the symmetric crypto
 
  (BLAKE2b, SHA-512, Chacha20, and Poly1305).
 
- Fused self-contained tests together for easier analysis with Frama-C
 
  and the TIS interpreter.
 

	
 
1.0
 
---
 
2017/07/18
 

	
 
- Renamed `crypto_chacha20_Xinit` to `crypto_chacha20_x_init`, for
 
  consistency reasons (snake case everywhere).
 
- Fixed signed integer overflow detected by UBSan.
 
- Doubled the speed of EdDSA by performing the scalar product in
 
  Montgomery space.
 

	
 
0.8
 
---
 
2017/07/06
 

	
 
- Added about a hundred lines of code to improve performance of public
 
  key cryptography.  Diffie-Hellman is now 20% faster than before.
 
  The effects are less pronounced for EdDSA.
 
- Added random self-consistency tests.
 
- Added a speed benchmark against libsodium.
 

	
 
0.7
 
---
 
2017/06/07
 

	
 
- Slightly changed the authenticated encryption API.  Functions are
 
  now all in "detached" mode.  The reason is better support for
 
  authenticated encryption _without_ additional data.
 
- Rewrote BLAKE2b from spec so it can use the same licence as
 
  everything else.
 
- Added random tests that compare Monocypher with libsodium and
 
  ed25519-donna.
 
- Added explicit support for Frama-C analysis (this doesn't affect the
 
  source code).
 

	
 
0.6
 
---
 
2017/03/17
 

	
 
- Fixed incorrect Poly1305 output on empty messages.  (Found by Mike
 
  Pechkin.)
 

	
 
0.5
 
---
 
2017/03/10
 

	
 
- Fixed many undefined behaviours in Curve25519 that occur whenever
 
  we perform a left shift on a signed negative integer.  It doesn't
 
  affect the generated code, but you never know.  (Found with Frama-C
 
  by André Maroneze.)
 

	
 
Fun fact: TweetNaCl and ref10 have the same bug.  Libsodium has
 
corrected the issue, though.
 

	
 
For those who don't comprehend the magnitude of this madness, the
 
expression `-1 << 3` is undefined in C.  This is explained in
 
section 6.5.7(§4) of the C11 standard.
 

	
 
0.4
 
---
 
2017/03/09
 

	
 
- Fixed critical bug causing Argon2i to fail whenever it uses more
 
  than 512 blocks.  It was reading uninitialised memory and the
 
  results were incorrect.  (Found by Mike Pechkin.)
 
- Fixed an undefined behaviour in Curve25519 (`fe_tobytes()`).  It was
 
  accessing uninitialised memory before throwing it away.  It didn't
 
  affect the compiled code nor the results, but you never know.
 
  (Found with [Frama-C](http://frama-c.com) by André Maroneze.)
 

	
 
0.3
 
---
 
2017/02/27
 

	
 
- Got the invariants of Poly1305 right and put them in the comments.
 
  There was no bug, but that was lucky (turned out the IETF test
 
  vectors were designed to trigger the bugs I was afraid of).
 
- Simplified Poly1305 finalisation (replaced conditional subtraction
 
  by a carry propagation).
 
- Made a few cosmetic changes here and there.
 

	
 
0.2
 
---
 
????/??/??
 

	
 
- Public interface significantly reworked. Removed redundant, hard to
 
  mess up constructions.
 
- Added AEAD.
 
- Sped up Curve25519 by a factor of more than 6 (switched to ref10
 
  arithmetic).
 
- Added various test vectors and completed the consistency tests.
 

	
 
0.1
 
---
 
2016/??/??
src/3rdparty/monocypher/CMakeLists.txt
Show inline comments
 
new file 100644
 
add_files(
 
    monocypher-ed25519.cpp
 
    monocypher-ed25519.h
 
    monocypher.cpp
 
    monocypher.h
 
)
src/3rdparty/monocypher/LICENCE.md
Show inline comments
 
new file 100644
 
Monocypher as a whole is dual-licensed.  Choose whichever licence you
 
want from the two licences listed below.
 

	
 
The first licence is a regular 2-clause BSD licence.  The second licence
 
is the CC-0 from Creative Commons. It is intended to release Monocypher
 
to the public domain.  The BSD licence serves as a fallback option.
 

	
 
See the individual files for specific information about who contributed
 
to what file during which years.  See below for special notes.
 

	
 
Licence 1 (2-clause BSD)
 
------------------------
 

	
 
Copyright (c) 2017-2023, Loup Vaillant
 
Copyright (c) 2017-2019, Michael Savage
 
Copyright (c) 2017-2023, Fabio Scotoni
 
All rights reserved.
 

	
 
Redistribution and use in source and binary forms, with or without
 
modification, are permitted provided that the following conditions are
 
met:
 

	
 
1. Redistributions of source code must retain the above copyright
 
   notice, this list of conditions and the following disclaimer.
 

	
 
2. Redistributions in binary form must reproduce the above copyright
 
   notice, this list of conditions and the following disclaimer in the
 
   documentation and/or other materials provided with the
 
   distribution.
 

	
 
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 
"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 
LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 
A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 
HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 
LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 
THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 
(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 
OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 

	
 

	
 
Licence 2 (CC-0)
 
----------------
 

	
 
> CREATIVE COMMONS CORPORATION IS NOT A LAW FIRM AND DOES NOT PROVIDE
 
> LEGAL SERVICES. DISTRIBUTION OF THIS DOCUMENT DOES NOT CREATE AN
 
> ATTORNEY-CLIENT RELATIONSHIP. CREATIVE COMMONS PROVIDES THIS
 
> INFORMATION ON AN "AS-IS" BASIS. CREATIVE COMMONS MAKES NO WARRANTIES
 
> REGARDING THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS
 
> PROVIDED HEREUNDER, AND DISCLAIMS LIABILITY FOR DAMAGES RESULTING FROM
 
> THE USE OF THIS DOCUMENT OR THE INFORMATION OR WORKS PROVIDED
 
> HEREUNDER.
 

	
 
### Statement of Purpose
 

	
 
The laws of most jurisdictions throughout the world automatically confer
 
exclusive Copyright and Related Rights (defined below) upon the creator
 
and subsequent owner(s) (each and all, an "owner") of an original work
 
of authorship and/or a database (each, a "Work").
 

	
 
Certain owners wish to permanently relinquish those rights to a Work for
 
the purpose of contributing to a commons of creative, cultural and
 
scientific works ("Commons") that the public can reliably and without
 
fear of later claims of infringement build upon, modify, incorporate in
 
other works, reuse and redistribute as freely as possible in any form
 
whatsoever and for any purposes, including without limitation commercial
 
purposes. These owners may contribute to the Commons to promote the
 
ideal of a free culture and the further production of creative, cultural
 
and scientific works, or to gain reputation or greater distribution for
 
their Work in part through the use and efforts of others.
 

	
 
For these and/or other purposes and motivations, and without any
 
expectation of additional consideration or compensation, the person
 
associating CC0 with a Work (the "Affirmer"), to the extent that he or
 
she is an owner of Copyright and Related Rights in the Work, voluntarily
 
elects to apply CC0 to the Work and publicly distribute the Work under
 
its terms, with knowledge of his or her Copyright and Related Rights in
 
the Work and the meaning and intended legal effect of CC0 on those
 
rights.
 

	
 
1. **Copyright and Related Rights.** A Work made available under CC0 may
 
   be protected by copyright and related or neighboring rights
 
   ("Copyright and Related Rights"). Copyright and Related Rights
 
   include, but are not limited to, the following:
 

	
 
   - the right to reproduce, adapt, distribute, perform, display,
 
     communicate, and translate a Work;
 
   - moral rights retained by the original author(s) and/or
 
     performer(s); publicity and privacy rights pertaining to a person's
 
     image or likeness depicted in a Work;
 
   - rights protecting against unfair competition in regards to a Work,
 
     subject to the limitations in paragraph 4(a), below;
 
   - rights protecting the extraction, dissemination, use and reuse of
 
     data in a Work;
 
   - database rights (such as those arising under Directive 96/9/EC of
 
     the European Parliament and of the Council of 11 March 1996 on the
 
     legal protection of databases, and under any national
 
     implementation thereof, including any amended or successor version
 
     of such directive); and
 
   - other similar, equivalent or corresponding rights throughout the
 
     world based on applicable law or treaty, and any national
 
     implementations thereof.
 

	
 
2. **Waiver.** To the greatest extent permitted by, but not in
 
   contravention of, applicable law, Affirmer hereby overtly, fully,
 
   permanently, irrevocably and unconditionally waives, abandons, and
 
   surrenders all of Affirmer's Copyright and Related Rights and
 
   associated claims and causes of action, whether now known or unknown
 
   (including existing as well as future claims and causes of action),
 
   in the Work (i) in all territories worldwide, (ii) for the maximum
 
   duration provided by applicable law or treaty (including future time
 
   extensions), (iii) in any current or future medium and for any number
 
   of copies, and (iv) for any purpose whatsoever, including without
 
   limitation commercial, advertising or promotional purposes (the
 
   "Waiver"). Affirmer makes the Waiver for the benefit of each member
 
   of the public at large and to the detriment of Affirmer's heirs and
 
   successors, fully intending that such Waiver shall not be subject to
 
   revocation, rescission, cancellation, termination, or any other legal
 
   or equitable action to disrupt the quiet enjoyment of the Work by the
 
   public as contemplated by Affirmer's express Statement of Purpose.
 

	
 
3. **Public License Fallback.** Should any part of the Waiver for any
 
   reason be judged legally invalid or ineffective under applicable law,
 
   then the Waiver shall be preserved to the maximum extent permitted
 
   taking into account Affirmer's express Statement of Purpose. In
 
   addition, to the extent the Waiver is so judged Affirmer hereby
 
   grants to each affected person a royalty-free, non transferable, non
 
   sublicensable, non exclusive, irrevocable and unconditional license
 
   to exercise Affirmer's Copyright and Related Rights in the Work (i)
 
   in all territories worldwide, (ii) for the maximum duration provided
 
   by applicable law or treaty (including future time extensions), (iii)
 
   in any current or future medium and for any number of copies, and
 
   (iv) for any purpose whatsoever, including without limitation
 
   commercial, advertising or promotional purposes (the "License"). The
 
   License shall be deemed effective as of the date CC0 was applied by
 
   Affirmer to the Work. Should any part of the License for any reason
 
   be judged legally invalid or ineffective under applicable law, such
 
   partial invalidity or ineffectiveness shall not invalidate the
 
   remainder of the License, and in such case Affirmer hereby affirms
 
   that he or she will not (i) exercise any of his or her remaining
 
   Copyright and Related Rights in the Work or (ii) assert any
 
   associated claims and causes of action with respect to the Work, in
 
   either case contrary to Affirmer's express Statement of Purpose.
 

	
 
4. **Limitations and Disclaimers.**
 

	
 
   - No trademark or patent rights held by Affirmer are waived,
 
     abandoned, surrendered, licensed or otherwise affected by this
 
     document.
 
   - Affirmer offers the Work as-is and makes no representations or
 
     warranties of any kind concerning the Work, express, implied,
 
     statutory or otherwise, including without limitation warranties of
 
     title, merchantability, fitness for a particular purpose, non
 
     infringement, or the absence of latent or other defects, accuracy,
 
     or the present or absence of errors, whether or not discoverable,
 
     all to the greatest extent permissible under applicable law.
 
   - Affirmer disclaims responsibility for clearing rights of other
 
     persons that may apply to the Work or any use thereof, including
 
     without limitation any person's Copyright and Related Rights in the
 
     Work. Further, Affirmer disclaims responsibility for obtaining any
 
     necessary consents, permissions or other rights required for any
 
     use of the Work.
 
   - Affirmer understands and acknowledges that Creative Commons is not
 
     a party to this document and has no duty or obligation with respect
 
     to this CC0 or use of the Work.
src/3rdparty/monocypher/README.md
Show inline comments
 
new file 100644
 
Monocypher
 
----------
 

	
 
Monocypher is an easy to use, easy to deploy, auditable crypto library
 
written in portable C.  It approaches the size of [TweetNaCl][] and the
 
speed of [libsodium][].
 

	
 
[Official site.](https://monocypher.org/)
 
[Official releases.](https://monocypher.org/download/)
 

	
 
[libsodium]: https://libsodium.org
 
[TweetNaCl]: https://tweetnacl.cr.yp.to/
 

	
 

	
 
Features
 
--------
 

	
 
- [Authenticated Encryption][AEAD] with XChaCha20 and Poly1305
 
  (RFC 8439).
 
- [Hashing and key derivation][HASH] with BLAKE2b (and [SHA-512][]).
 
- [Password Hashing][PWH] with Argon2.
 
- [Public Key Cryptography][PKC] with X25519 key exchanges.
 
- [Public Key Signatures][EDDSA] with EdDSA and [Ed25519][].
 
- [Steganography and PAKE][STEG] with [Elligator 2][ELLI].
 

	
 
[AEAD]:    https://monocypher.org/manual/aead
 
[HASH]:    https://monocypher.org/manual/blake2
 
[SHA-512]: https://monocypher.org/manual/sha-512
 
[PWH]:     https://monocypher.org/manual/argon2
 
[PKC]:     https://monocypher.org/manual/x25519
 
[EDDSA]:   https://monocypher.org/manual/eddsa
 
[Ed25519]: https://monocypher.org/manual/ed25519
 
[STEG]:    https://monocypher.org/manual/elligator
 
[ELLI]:    https://elligator.org
 

	
 

	
 
Manual
 
------
 

	
 
The manual can be found at https://monocypher.org/manual/, and in the
 
`doc/` folder.
 

	
 

	
 
Installation
 
------------
 

	
 
### Option 1: grab the sources
 

	
 
The easiest way to use Monocypher is to include `src/monocypher.h` and
 
`src/monocypher.c` directly into your project.  They compile as C (since
 
C99) and C++ (since C++98).
 

	
 
If you need the optional SHA-512 or Ed25519, grab
 
`src/optional/monocypher-ed25519.h` and
 
`src/optional/monocypher-ed25519.c` as well.
 

	
 
### Option 2: grab the library
 

	
 
Run `make`, then grab the `src/monocypher.h` header and either the
 
`lib/libmonocypher.a` or `lib/libmonocypher.so` library.  The default
 
compiler is `gcc -std=c99`, and the default flags are `-pedantic -Wall
 
-Wextra -O3 -march=native`.  If they don't work on your platform, you
 
can change them like this:
 

	
 
    $ make CC="clang -std=c11" CFLAGS="-O2"
 

	
 
### Option 3: install it on your system
 

	
 
Run `make`, then `make install` as root. This will install Monocypher in
 
`/usr/local` by default. This can be changed with `PREFIX` and
 
`DESTDIR`:
 

	
 
    $ make install PREFIX="/opt"
 

	
 
Once installed, you may use `pkg-config` to compile and link your
 
program.  For instance:
 

	
 
    $ gcc program.c $(pkg-config monocypher --cflags) -c
 
    $ gcc program.o $(pkg-config monocypher --libs)   -o program
 

	
 
If for any reason you wish to avoid installing the man pages or the
 
`pkg-config` file, you can use the following installation sub targets
 
instead: `install-lib`, `install-doc`, and `install-pc`.
 

	
 

	
 
Test suite
 
----------
 

	
 
    $ make test
 

	
 
It should display a nice printout of all the tests, ending with "All
 
tests OK!". If you see "failure" or "Error" anywhere, something has gone
 
wrong.
 

	
 
*Do not* use Monocypher without running those tests at least once.
 

	
 
The same test suite can be run under Clang sanitisers and Valgrind, and
 
be checked for code coverage:
 

	
 
    $ tests/test.sh
 
    $ tests/coverage.sh
 

	
 

	
 
### Serious auditing
 

	
 
The code may be analysed more formally with [Frama-c][] and the
 
[TIS interpreter][TIS].  To analyse the code with Frama-c, run:
 

	
 
    $ tests/formal-analysis.sh
 
    $ tests/frama-c.sh
 

	
 
This will have Frama-c parse, and analyse the code, then launch a GUI.
 
You must have Frama-c installed.  See `frama-c.sh` for the recommended
 
settings.  To run the code under the TIS interpreter, run
 

	
 
    $ tests/formal-analysis.sh
 
    $ tis-interpreter.sh --cc -Dvolatile= tests/formal-analysis/*.c
 

	
 
Notes:
 

	
 
- `tis-interpreter.sh` is part of TIS.  If it is not in your path,
 
  adjust the command accordingly.
 

	
 
- The TIS interpreter sometimes fails to evaluate correct programs when
 
  they use the `volatile` keyword (which is only used as an attempt to
 
  prevent dead store elimination for memory wipes).  The `-cc
 
  -Dvolatile=` option works around that bug by ignoring `volatile`
 
  altogether.
 

	
 
[Frama-c]:https://frama-c.com/
 
[TIS]: https://trust-in-soft.com/tis-interpreter/
 

	
 

	
 
Customisation
 
-------------
 

	
 
Monocypher has optional compatibility with Ed25519. To have that, add
 
`monocypher-ed25519.h` and `monocypher-ed25519.c` provided in
 
`src/optional` to your project.  If you compile or install Monocypher
 
with the makefile, they will be automatically included.
 

	
 
Monocypher also has the `BLAKE2_NO_UNROLLING` preprocessor flag, which
 
is activated by compiling monocypher.c with the `-DBLAKE2_NO_UNROLLING`
 
option.
 

	
 
The `-DBLAKE2_NO_UNROLLING` option is a performance tweak.  By default,
 
Monocypher unrolls the BLAKE2b inner loop, because doing so is over 25%
 
faster on modern processors.  Some embedded processors however, run the
 
unrolled loop _slower_ (possibly because of the cost of fetching 5KB of
 
additional code).  If you're using an embedded platform, try this
 
option.  The binary will be about 5KB smaller, and in some cases faster.
 

	
 
The `MONOCYPHER_CPP_NAMESPACE` preprocessor definition allows C++ users
 
who compile Monocypher as C++ to wrap it in a namespace. When it is not
 
defined (the default), we assume Monocypher is compiled as C, and an
 
`extern "C"` declaration is added when we detect that the header is
 
included in C++ code.
 

	
 
The `change-prefix.sh` script can rename all functions by replacing
 
`crypto_` by a chosen prefix, so you can avoid name clashes. For
 
instance, the following command changes all instances of `crypto_` by
 
`foobar_` (note the absence of the underscore):
 

	
 
    ./change-prefix.sh foobar
src/3rdparty/monocypher/monocypher-ed25519.cpp
Show inline comments
 
new file 100644
 
// Monocypher version 4.0.2
 
//
 
// This file is dual-licensed.  Choose whichever licence you want from
 
// the two licences listed below.
 
//
 
// The first licence is a regular 2-clause BSD licence.  The second licence
 
// is the CC-0 from Creative Commons. It is intended to release Monocypher
 
// to the public domain.  The BSD licence serves as a fallback option.
 
//
 
// SPDX-License-Identifier: BSD-2-Clause OR CC0-1.0
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Copyright (c) 2017-2019, Loup Vaillant
 
// All rights reserved.
 
//
 
//
 
// Redistribution and use in source and binary forms, with or without
 
// modification, are permitted provided that the following conditions are
 
// met:
 
//
 
// 1. Redistributions of source code must retain the above copyright
 
//    notice, this list of conditions and the following disclaimer.
 
//
 
// 2. Redistributions in binary form must reproduce the above copyright
 
//    notice, this list of conditions and the following disclaimer in the
 
//    documentation and/or other materials provided with the
 
//    distribution.
 
//
 
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 
// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Written in 2017-2019 by Loup Vaillant
 
//
 
// To the extent possible under law, the author(s) have dedicated all copyright
 
// and related neighboring rights to this software to the public domain
 
// worldwide.  This software is distributed without any warranty.
 
//
 
// You should have received a copy of the CC0 Public Domain Dedication along
 
// with this software.  If not, see
 
// <https://creativecommons.org/publicdomain/zero/1.0/>
 

	
 
#include "monocypher-ed25519.h"
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
namespace MONOCYPHER_CPP_NAMESPACE {
 
#endif
 

	
 
/////////////////
 
/// Utilities ///
 
/////////////////
 
#define FOR(i, min, max)     for (size_t i = min; i < max; i++)
 
#define COPY(dst, src, size) FOR(_i_, 0, size) (dst)[_i_] = (src)[_i_]
 
#define ZERO(buf, size)      FOR(_i_, 0, size) (buf)[_i_] = 0
 
#define WIPE_CTX(ctx)        crypto_wipe(ctx   , sizeof(*(ctx)))
 
#define WIPE_BUFFER(buffer)  crypto_wipe(buffer, sizeof(buffer))
 
#define MIN(a, b)            ((a) <= (b) ? (a) : (b))
 
typedef uint8_t u8;
 
typedef uint64_t u64;
 

	
 
// Returns the smallest positive integer y such that
 
// (x + y) % pow_2  == 0
 
// Basically, it's how many bytes we need to add to "align" x.
 
// Only works when pow_2 is a power of 2.
 
// Note: we use ~x+1 instead of -x to avoid compiler warnings
 
static size_t align(size_t x, size_t pow_2)
 
{
 
	return (~x + 1) & (pow_2 - 1);
 
}
 

	
 
static u64 load64_be(const u8 s[8])
 
{
 
	return((u64)s[0] << 56)
 
		| ((u64)s[1] << 48)
 
		| ((u64)s[2] << 40)
 
		| ((u64)s[3] << 32)
 
		| ((u64)s[4] << 24)
 
		| ((u64)s[5] << 16)
 
		| ((u64)s[6] <<  8)
 
		|  (u64)s[7];
 
}
 

	
 
static void store64_be(u8 out[8], u64 in)
 
{
 
	out[0] = (in >> 56) & 0xff;
 
	out[1] = (in >> 48) & 0xff;
 
	out[2] = (in >> 40) & 0xff;
 
	out[3] = (in >> 32) & 0xff;
 
	out[4] = (in >> 24) & 0xff;
 
	out[5] = (in >> 16) & 0xff;
 
	out[6] = (in >>  8) & 0xff;
 
	out[7] =  in        & 0xff;
 
}
 

	
 
static void load64_be_buf (u64 *dst, const u8 *src, size_t size) {
 
	FOR(i, 0, size) { dst[i] = load64_be(src + i*8); }
 
}
 

	
 
///////////////
 
/// SHA 512 ///
 
///////////////
 
static u64 rot(u64 x, int c       ) { return (x >> c) | (x << (64 - c));   }
 
static u64 ch (u64 x, u64 y, u64 z) { return (x & y) ^ (~x & z);           }
 
static u64 maj(u64 x, u64 y, u64 z) { return (x & y) ^ ( x & z) ^ (y & z); }
 
static u64 big_sigma0(u64 x) { return rot(x, 28) ^ rot(x, 34) ^ rot(x, 39); }
 
static u64 big_sigma1(u64 x) { return rot(x, 14) ^ rot(x, 18) ^ rot(x, 41); }
 
static u64 lit_sigma0(u64 x) { return rot(x,  1) ^ rot(x,  8) ^ (x >> 7);   }
 
static u64 lit_sigma1(u64 x) { return rot(x, 19) ^ rot(x, 61) ^ (x >> 6);   }
 

	
 
static const u64 K[80] = {
 
	0x428a2f98d728ae22,0x7137449123ef65cd,0xb5c0fbcfec4d3b2f,0xe9b5dba58189dbbc,
 
	0x3956c25bf348b538,0x59f111f1b605d019,0x923f82a4af194f9b,0xab1c5ed5da6d8118,
 
	0xd807aa98a3030242,0x12835b0145706fbe,0x243185be4ee4b28c,0x550c7dc3d5ffb4e2,
 
	0x72be5d74f27b896f,0x80deb1fe3b1696b1,0x9bdc06a725c71235,0xc19bf174cf692694,
 
	0xe49b69c19ef14ad2,0xefbe4786384f25e3,0x0fc19dc68b8cd5b5,0x240ca1cc77ac9c65,
 
	0x2de92c6f592b0275,0x4a7484aa6ea6e483,0x5cb0a9dcbd41fbd4,0x76f988da831153b5,
 
	0x983e5152ee66dfab,0xa831c66d2db43210,0xb00327c898fb213f,0xbf597fc7beef0ee4,
 
	0xc6e00bf33da88fc2,0xd5a79147930aa725,0x06ca6351e003826f,0x142929670a0e6e70,
 
	0x27b70a8546d22ffc,0x2e1b21385c26c926,0x4d2c6dfc5ac42aed,0x53380d139d95b3df,
 
	0x650a73548baf63de,0x766a0abb3c77b2a8,0x81c2c92e47edaee6,0x92722c851482353b,
 
	0xa2bfe8a14cf10364,0xa81a664bbc423001,0xc24b8b70d0f89791,0xc76c51a30654be30,
 
	0xd192e819d6ef5218,0xd69906245565a910,0xf40e35855771202a,0x106aa07032bbd1b8,
 
	0x19a4c116b8d2d0c8,0x1e376c085141ab53,0x2748774cdf8eeb99,0x34b0bcb5e19b48a8,
 
	0x391c0cb3c5c95a63,0x4ed8aa4ae3418acb,0x5b9cca4f7763e373,0x682e6ff3d6b2b8a3,
 
	0x748f82ee5defb2fc,0x78a5636f43172f60,0x84c87814a1f0ab72,0x8cc702081a6439ec,
 
	0x90befffa23631e28,0xa4506cebde82bde9,0xbef9a3f7b2c67915,0xc67178f2e372532b,
 
	0xca273eceea26619c,0xd186b8c721c0c207,0xeada7dd6cde0eb1e,0xf57d4f7fee6ed178,
 
	0x06f067aa72176fba,0x0a637dc5a2c898a6,0x113f9804bef90dae,0x1b710b35131c471b,
 
	0x28db77f523047d84,0x32caab7b40c72493,0x3c9ebe0a15c9bebc,0x431d67c49c100d4c,
 
	0x4cc5d4becb3e42b6,0x597f299cfc657e2a,0x5fcb6fab3ad6faec,0x6c44198c4a475817
 
};
 

	
 
static void sha512_compress(crypto_sha512_ctx *ctx)
 
{
 
	u64 a = ctx->hash[0];    u64 b = ctx->hash[1];
 
	u64 c = ctx->hash[2];    u64 d = ctx->hash[3];
 
	u64 e = ctx->hash[4];    u64 f = ctx->hash[5];
 
	u64 g = ctx->hash[6];    u64 h = ctx->hash[7];
 

	
 
	FOR (j, 0, 16) {
 
		u64 in = K[j] + ctx->input[j];
 
		u64 t1 = big_sigma1(e) + ch (e, f, g) + h + in;
 
		u64 t2 = big_sigma0(a) + maj(a, b, c);
 
		h = g;  g = f;  f = e;  e = d  + t1;
 
		d = c;  c = b;  b = a;  a = t1 + t2;
 
	}
 
	size_t i16 = 0;
 
	FOR(i, 1, 5) {
 
		i16 += 16;
 
		FOR (j, 0, 16) {
 
			ctx->input[j] += lit_sigma1(ctx->input[(j- 2) & 15]);
 
			ctx->input[j] += lit_sigma0(ctx->input[(j-15) & 15]);
 
			ctx->input[j] +=            ctx->input[(j- 7) & 15];
 
			u64 in = K[i16 + j] + ctx->input[j];
 
			u64 t1 = big_sigma1(e) + ch (e, f, g) + h + in;
 
			u64 t2 = big_sigma0(a) + maj(a, b, c);
 
			h = g;  g = f;  f = e;  e = d  + t1;
 
			d = c;  c = b;  b = a;  a = t1 + t2;
 
		}
 
	}
 

	
 
	ctx->hash[0] += a;    ctx->hash[1] += b;
 
	ctx->hash[2] += c;    ctx->hash[3] += d;
 
	ctx->hash[4] += e;    ctx->hash[5] += f;
 
	ctx->hash[6] += g;    ctx->hash[7] += h;
 
}
 

	
 
// Write 1 input byte
 
static void sha512_set_input(crypto_sha512_ctx *ctx, u8 input)
 
{
 
	size_t word = ctx->input_idx >> 3;
 
	size_t byte = ctx->input_idx &  7;
 
	ctx->input[word] |= (u64)input << (8 * (7 - byte));
 
}
 

	
 
// Increment a 128-bit "word".
 
static void sha512_incr(u64 x[2], u64 y)
 
{
 
	x[1] += y;
 
	if (x[1] < y) {
 
		x[0]++;
 
	}
 
}
 

	
 
void crypto_sha512_init(crypto_sha512_ctx *ctx)
 
{
 
	ctx->hash[0] = 0x6a09e667f3bcc908;
 
	ctx->hash[1] = 0xbb67ae8584caa73b;
 
	ctx->hash[2] = 0x3c6ef372fe94f82b;
 
	ctx->hash[3] = 0xa54ff53a5f1d36f1;
 
	ctx->hash[4] = 0x510e527fade682d1;
 
	ctx->hash[5] = 0x9b05688c2b3e6c1f;
 
	ctx->hash[6] = 0x1f83d9abfb41bd6b;
 
	ctx->hash[7] = 0x5be0cd19137e2179;
 
	ctx->input_size[0] = 0;
 
	ctx->input_size[1] = 0;
 
	ctx->input_idx = 0;
 
	ZERO(ctx->input, 16);
 
}
 

	
 
void crypto_sha512_update(crypto_sha512_ctx *ctx,
 
                          const u8 *message, size_t message_size)
 
{
 
	// Avoid undefined NULL pointer increments with empty messages
 
	if (message_size == 0) {
 
		return;
 
	}
 

	
 
	// Align ourselves with word boundaries
 
	if ((ctx->input_idx & 7) != 0) {
 
		size_t nb_bytes = MIN(align(ctx->input_idx, 8), message_size);
 
		FOR (i, 0, nb_bytes) {
 
			sha512_set_input(ctx, message[i]);
 
			ctx->input_idx++;
 
		}
 
		message      += nb_bytes;
 
		message_size -= nb_bytes;
 
	}
 

	
 
	// Align ourselves with block boundaries
 
	if ((ctx->input_idx & 127) != 0) {
 
		size_t nb_words = MIN(align(ctx->input_idx, 128), message_size) >> 3;
 
		load64_be_buf(ctx->input + (ctx->input_idx >> 3), message, nb_words);
 
		ctx->input_idx += nb_words << 3;
 
		message        += nb_words << 3;
 
		message_size   -= nb_words << 3;
 
	}
 

	
 
	// Compress block if needed
 
	if (ctx->input_idx == 128) {
 
		sha512_incr(ctx->input_size, 1024); // size is in bits
 
		sha512_compress(ctx);
 
		ctx->input_idx = 0;
 
		ZERO(ctx->input, 16);
 
	}
 

	
 
	// Process the message block by block
 
	FOR (i, 0, message_size >> 7) { // number of blocks
 
		load64_be_buf(ctx->input, message, 16);
 
		sha512_incr(ctx->input_size, 1024); // size is in bits
 
		sha512_compress(ctx);
 
		ctx->input_idx = 0;
 
		ZERO(ctx->input, 16);
 
		message += 128;
 
	}
 
	message_size &= 127;
 

	
 
	if (message_size != 0) {
 
		// Remaining words
 
		size_t nb_words = message_size >> 3;
 
		load64_be_buf(ctx->input, message, nb_words);
 
		ctx->input_idx += nb_words << 3;
 
		message        += nb_words << 3;
 
		message_size   -= nb_words << 3;
 

	
 
		// Remaining bytes
 
		FOR (i, 0, message_size) {
 
			sha512_set_input(ctx, message[i]);
 
			ctx->input_idx++;
 
		}
 
	}
 
}
 

	
 
void crypto_sha512_final(crypto_sha512_ctx *ctx, u8 hash[64])
 
{
 
	// Add padding bit
 
	if (ctx->input_idx == 0) {
 
		ZERO(ctx->input, 16);
 
	}
 
	sha512_set_input(ctx, 128);
 

	
 
	// Update size
 
	sha512_incr(ctx->input_size, ctx->input_idx * 8);
 

	
 
	// Compress penultimate block (if any)
 
	if (ctx->input_idx > 111) {
 
		sha512_compress(ctx);
 
		ZERO(ctx->input, 14);
 
	}
 
	// Compress last block
 
	ctx->input[14] = ctx->input_size[0];
 
	ctx->input[15] = ctx->input_size[1];
 
	sha512_compress(ctx);
 

	
 
	// Copy hash to output (big endian)
 
	FOR (i, 0, 8) {
 
		store64_be(hash + i*8, ctx->hash[i]);
 
	}
 

	
 
	WIPE_CTX(ctx);
 
}
 

	
 
void crypto_sha512(u8 hash[64], const u8 *message, size_t message_size)
 
{
 
	crypto_sha512_ctx ctx;
 
	crypto_sha512_init  (&ctx);
 
	crypto_sha512_update(&ctx, message, message_size);
 
	crypto_sha512_final (&ctx, hash);
 
}
 

	
 
////////////////////
 
/// HMAC SHA 512 ///
 
////////////////////
 
void crypto_sha512_hmac_init(crypto_sha512_hmac_ctx *ctx,
 
                             const u8 *key, size_t key_size)
 
{
 
	// hash key if it is too long
 
	if (key_size > 128) {
 
		crypto_sha512(ctx->key, key, key_size);
 
		key      = ctx->key;
 
		key_size = 64;
 
	}
 
	// Compute inner key: padded key XOR 0x36
 
	FOR (i, 0, key_size)   { ctx->key[i] = key[i] ^ 0x36; }
 
	FOR (i, key_size, 128) { ctx->key[i] =          0x36; }
 
	// Start computing inner hash
 
	crypto_sha512_init  (&ctx->ctx);
 
	crypto_sha512_update(&ctx->ctx, ctx->key, 128);
 
}
 

	
 
void crypto_sha512_hmac_update(crypto_sha512_hmac_ctx *ctx,
 
                               const u8 *message, size_t message_size)
 
{
 
	crypto_sha512_update(&ctx->ctx, message, message_size);
 
}
 

	
 
void crypto_sha512_hmac_final(crypto_sha512_hmac_ctx *ctx, u8 hmac[64])
 
{
 
	// Finish computing inner hash
 
	crypto_sha512_final(&ctx->ctx, hmac);
 
	// Compute outer key: padded key XOR 0x5c
 
	FOR (i, 0, 128) {
 
		ctx->key[i] ^= 0x36 ^ 0x5c;
 
	}
 
	// Compute outer hash
 
	crypto_sha512_init  (&ctx->ctx);
 
	crypto_sha512_update(&ctx->ctx, ctx->key , 128);
 
	crypto_sha512_update(&ctx->ctx, hmac, 64);
 
	crypto_sha512_final (&ctx->ctx, hmac); // outer hash
 
	WIPE_CTX(ctx);
 
}
 

	
 
void crypto_sha512_hmac(u8 hmac[64], const u8 *key, size_t key_size,
 
                        const u8 *message, size_t message_size)
 
{
 
	crypto_sha512_hmac_ctx ctx;
 
	crypto_sha512_hmac_init  (&ctx, key, key_size);
 
	crypto_sha512_hmac_update(&ctx, message, message_size);
 
	crypto_sha512_hmac_final (&ctx, hmac);
 
}
 

	
 
////////////////////
 
/// HKDF SHA 512 ///
 
////////////////////
 
void crypto_sha512_hkdf_expand(u8       *okm,  size_t okm_size,
 
                               const u8 *prk,  size_t prk_size,
 
                               const u8 *info, size_t info_size)
 
{
 
	int not_first = 0;
 
	u8 ctr = 1;
 
	u8 blk[64];
 

	
 
	while (okm_size > 0) {
 
		size_t out_size = MIN(okm_size, sizeof(blk));
 

	
 
		crypto_sha512_hmac_ctx ctx;
 
		crypto_sha512_hmac_init(&ctx, prk , prk_size);
 
		if (not_first) {
 
			// For some reason HKDF uses some kind of CBC mode.
 
			// For some reason CTR mode alone wasn't enough.
 
			// Like what, they didn't trust HMAC in 2010?  Really??
 
			crypto_sha512_hmac_update(&ctx, blk , sizeof(blk));
 
		}
 
		crypto_sha512_hmac_update(&ctx, info, info_size);
 
		crypto_sha512_hmac_update(&ctx, &ctr, 1);
 
		crypto_sha512_hmac_final(&ctx, blk);
 

	
 
		COPY(okm, blk, out_size);
 

	
 
		not_first = 1;
 
		okm      += out_size;
 
		okm_size -= out_size;
 
		ctr++;
 
	}
 
}
 

	
 
void crypto_sha512_hkdf(u8       *okm , size_t okm_size,
 
                        const u8 *ikm , size_t ikm_size,
 
                        const u8 *salt, size_t salt_size,
 
                        const u8 *info, size_t info_size)
 
{
 
	// Extract
 
	u8 prk[64];
 
	crypto_sha512_hmac(prk, salt, salt_size, ikm, ikm_size);
 

	
 
	// Expand
 
	crypto_sha512_hkdf_expand(okm, okm_size, prk, sizeof(prk), info, info_size);
 
}
 

	
 
///////////////
 
/// Ed25519 ///
 
///////////////
 
void crypto_ed25519_key_pair(u8 secret_key[64], u8 public_key[32], u8 seed[32])
 
{
 
	u8 a[64];
 
	COPY(a, seed, 32);                      // a[ 0..31]  = seed
 
	crypto_wipe(seed, 32);
 
	COPY(secret_key, a, 32);                // secret key = seed
 
	crypto_sha512(a, a, 32);                // a[ 0..31]  = scalar
 
	crypto_eddsa_trim_scalar(a, a);         // a[ 0..31]  = trimmed scalar
 
	crypto_eddsa_scalarbase(public_key, a); // public key = [trimmed scalar]B
 
	COPY(secret_key + 32, public_key, 32);  // secret key includes public half
 
	WIPE_BUFFER(a);
 
}
 

	
 
static void hash_reduce(u8 h[32],
 
                        const u8 *a, size_t a_size,
 
                        const u8 *b, size_t b_size,
 
                        const u8 *c, size_t c_size,
 
                        const u8 *d, size_t d_size)
 
{
 
	u8 hash[64];
 
	crypto_sha512_ctx ctx;
 
	crypto_sha512_init  (&ctx);
 
	crypto_sha512_update(&ctx, a, a_size);
 
	crypto_sha512_update(&ctx, b, b_size);
 
	crypto_sha512_update(&ctx, c, c_size);
 
	crypto_sha512_update(&ctx, d, d_size);
 
	crypto_sha512_final (&ctx, hash);
 
	crypto_eddsa_reduce(h, hash);
 
}
 

	
 
static void ed25519_dom_sign(u8 signature [64], const u8 secret_key[32],
 
                             const u8 *dom,     size_t dom_size,
 
                             const u8 *message, size_t message_size)
 
{
 
	u8 a[64];  // secret scalar and prefix
 
	u8 r[32];  // secret deterministic "random" nonce
 
	u8 h[32];  // publically verifiable hash of the message (not wiped)
 
	u8 R[32];  // first half of the signature (allows overlapping inputs)
 
	const u8 *pk = secret_key + 32;
 

	
 
	crypto_sha512(a, secret_key, 32);
 
	crypto_eddsa_trim_scalar(a, a);
 
	hash_reduce(r, dom, dom_size, a + 32, 32, message, message_size, 0, 0);
 
	crypto_eddsa_scalarbase(R, r);
 
	hash_reduce(h, dom, dom_size, R, 32, pk, 32, message, message_size);
 
	COPY(signature, R, 32);
 
	crypto_eddsa_mul_add(signature + 32, h, a, r);
 

	
 
	WIPE_BUFFER(a);
 
	WIPE_BUFFER(r);
 
}
 

	
 
void crypto_ed25519_sign(u8 signature [64], const u8 secret_key[64],
 
                         const u8 *message, size_t message_size)
 
{
 
	ed25519_dom_sign(signature, secret_key, 0, 0, message, message_size);
 
}
 

	
 
int crypto_ed25519_check(const u8 signature[64], const u8 public_key[32],
 
                         const u8 *msg, size_t msg_size)
 
{
 
	u8 h_ram[32];
 
	hash_reduce(h_ram, signature, 32, public_key, 32, msg, msg_size, 0, 0);
 
	return crypto_eddsa_check_equation(signature, public_key, h_ram);
 
}
 

	
 
static const u8 domain[34] = "SigEd25519 no Ed25519 collisions\1";
 

	
 
void crypto_ed25519_ph_sign(uint8_t signature[64], const uint8_t secret_key[64],
 
                            const uint8_t message_hash[64])
 
{
 
	ed25519_dom_sign(signature, secret_key, domain, sizeof(domain),
 
	                 message_hash, 64);
 
}
 

	
 
int crypto_ed25519_ph_check(const uint8_t sig[64], const uint8_t pk[32],
 
                            const uint8_t msg_hash[64])
 
{
 
	u8 h_ram[32];
 
	hash_reduce(h_ram, domain, sizeof(domain), sig, 32, pk, 32, msg_hash, 64);
 
	return crypto_eddsa_check_equation(sig, pk, h_ram);
 
}
 

	
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
}
 
#endif
src/3rdparty/monocypher/monocypher-ed25519.h
Show inline comments
 
new file 100644
 
// Monocypher version 4.0.2
 
//
 
// This file is dual-licensed.  Choose whichever licence you want from
 
// the two licences listed below.
 
//
 
// The first licence is a regular 2-clause BSD licence.  The second licence
 
// is the CC-0 from Creative Commons. It is intended to release Monocypher
 
// to the public domain.  The BSD licence serves as a fallback option.
 
//
 
// SPDX-License-Identifier: BSD-2-Clause OR CC0-1.0
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Copyright (c) 2017-2019, Loup Vaillant
 
// All rights reserved.
 
//
 
//
 
// Redistribution and use in source and binary forms, with or without
 
// modification, are permitted provided that the following conditions are
 
// met:
 
//
 
// 1. Redistributions of source code must retain the above copyright
 
//    notice, this list of conditions and the following disclaimer.
 
//
 
// 2. Redistributions in binary form must reproduce the above copyright
 
//    notice, this list of conditions and the following disclaimer in the
 
//    documentation and/or other materials provided with the
 
//    distribution.
 
//
 
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 
// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Written in 2017-2019 by Loup Vaillant
 
//
 
// To the extent possible under law, the author(s) have dedicated all copyright
 
// and related neighboring rights to this software to the public domain
 
// worldwide.  This software is distributed without any warranty.
 
//
 
// You should have received a copy of the CC0 Public Domain Dedication along
 
// with this software.  If not, see
 
// <https://creativecommons.org/publicdomain/zero/1.0/>
 

	
 
#ifndef ED25519_H
 
#define ED25519_H
 

	
 
#include "monocypher.h"
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
namespace MONOCYPHER_CPP_NAMESPACE {
 
#elif defined(__cplusplus)
 
extern "C" {
 
#endif
 

	
 
////////////////////////
 
/// Type definitions ///
 
////////////////////////
 

	
 
// Do not rely on the size or content on any of those types,
 
// they may change without notice.
 
typedef struct {
 
	uint64_t hash[8];
 
	uint64_t input[16];
 
	uint64_t input_size[2];
 
	size_t   input_idx;
 
} crypto_sha512_ctx;
 

	
 
typedef struct {
 
	uint8_t key[128];
 
	crypto_sha512_ctx ctx;
 
} crypto_sha512_hmac_ctx;
 

	
 

	
 
// SHA 512
 
// -------
 
void crypto_sha512_init  (crypto_sha512_ctx *ctx);
 
void crypto_sha512_update(crypto_sha512_ctx *ctx,
 
                          const uint8_t *message, size_t  message_size);
 
void crypto_sha512_final (crypto_sha512_ctx *ctx, uint8_t hash[64]);
 
void crypto_sha512(uint8_t hash[64],
 
                   const uint8_t *message, size_t message_size);
 

	
 
// SHA 512 HMAC
 
// ------------
 
void crypto_sha512_hmac_init(crypto_sha512_hmac_ctx *ctx,
 
                             const uint8_t *key, size_t key_size);
 
void crypto_sha512_hmac_update(crypto_sha512_hmac_ctx *ctx,
 
                               const uint8_t *message, size_t  message_size);
 
void crypto_sha512_hmac_final(crypto_sha512_hmac_ctx *ctx, uint8_t hmac[64]);
 
void crypto_sha512_hmac(uint8_t hmac[64],
 
                        const uint8_t *key    , size_t key_size,
 
                        const uint8_t *message, size_t message_size);
 

	
 
// SHA 512 HKDF
 
// ------------
 
void crypto_sha512_hkdf_expand(uint8_t       *okm,  size_t okm_size,
 
                               const uint8_t *prk,  size_t prk_size,
 
                               const uint8_t *info, size_t info_size);
 
void crypto_sha512_hkdf(uint8_t       *okm , size_t okm_size,
 
                        const uint8_t *ikm , size_t ikm_size,
 
                        const uint8_t *salt, size_t salt_size,
 
                        const uint8_t *info, size_t info_size);
 

	
 
// Ed25519
 
// -------
 
// Signatures (EdDSA with curve25519 + SHA-512)
 
// --------------------------------------------
 
void crypto_ed25519_key_pair(uint8_t secret_key[64],
 
                             uint8_t public_key[32],
 
                             uint8_t seed[32]);
 
void crypto_ed25519_sign(uint8_t        signature [64],
 
                         const uint8_t  secret_key[64],
 
                         const uint8_t *message, size_t message_size);
 
int crypto_ed25519_check(const uint8_t  signature [64],
 
                         const uint8_t  public_key[32],
 
                         const uint8_t *message, size_t message_size);
 

	
 
// Pre-hash variants
 
void crypto_ed25519_ph_sign(uint8_t       signature   [64],
 
                            const uint8_t secret_key  [64],
 
                            const uint8_t message_hash[64]);
 
int crypto_ed25519_ph_check(const uint8_t signature   [64],
 
                            const uint8_t public_key  [32],
 
                            const uint8_t message_hash[64]);
 

	
 
#ifdef __cplusplus
 
}
 
#endif
 

	
 
#endif // ED25519_H
src/3rdparty/monocypher/monocypher.cpp
Show inline comments
 
new file 100644
 
// Monocypher version 4.0.2
 
//
 
// This file is dual-licensed.  Choose whichever licence you want from
 
// the two licences listed below.
 
//
 
// The first licence is a regular 2-clause BSD licence.  The second licence
 
// is the CC-0 from Creative Commons. It is intended to release Monocypher
 
// to the public domain.  The BSD licence serves as a fallback option.
 
//
 
// SPDX-License-Identifier: BSD-2-Clause OR CC0-1.0
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Copyright (c) 2017-2020, Loup Vaillant
 
// All rights reserved.
 
//
 
//
 
// Redistribution and use in source and binary forms, with or without
 
// modification, are permitted provided that the following conditions are
 
// met:
 
//
 
// 1. Redistributions of source code must retain the above copyright
 
//    notice, this list of conditions and the following disclaimer.
 
//
 
// 2. Redistributions in binary form must reproduce the above copyright
 
//    notice, this list of conditions and the following disclaimer in the
 
//    documentation and/or other materials provided with the
 
//    distribution.
 
//
 
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 
// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Written in 2017-2020 by Loup Vaillant
 
//
 
// To the extent possible under law, the author(s) have dedicated all copyright
 
// and related neighboring rights to this software to the public domain
 
// worldwide.  This software is distributed without any warranty.
 
//
 
// You should have received a copy of the CC0 Public Domain Dedication along
 
// with this software.  If not, see
 
// <https://creativecommons.org/publicdomain/zero/1.0/>
 

	
 
#include "monocypher.h"
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
namespace MONOCYPHER_CPP_NAMESPACE {
 
#endif
 

	
 
/////////////////
 
/// Utilities ///
 
/////////////////
 
#define FOR_T(type, i, start, end) for (type i = (start); i < (end); i++)
 
#define FOR(i, start, end)         FOR_T(size_t, i, start, end)
 
#define COPY(dst, src, size)       FOR(_i_, 0, size) (dst)[_i_] = (src)[_i_]
 
#define ZERO(buf, size)            FOR(_i_, 0, size) (buf)[_i_] = 0
 
#define WIPE_CTX(ctx)              crypto_wipe(ctx   , sizeof(*(ctx)))
 
#define WIPE_BUFFER(buffer)        crypto_wipe(buffer, sizeof(buffer))
 
#define MIN(a, b)                  ((a) <= (b) ? (a) : (b))
 
#define MAX(a, b)                  ((a) >= (b) ? (a) : (b))
 

	
 
typedef int8_t   i8;
 
typedef uint8_t  u8;
 
typedef int16_t  i16;
 
typedef uint32_t u32;
 
typedef int32_t  i32;
 
typedef int64_t  i64;
 
typedef uint64_t u64;
 

	
 
static const u8 zero[128] = {0};
 

	
 
// returns the smallest positive integer y such that
 
// (x + y) % pow_2  == 0
 
// Basically, y is the "gap" missing to align x.
 
// Only works when pow_2 is a power of 2.
 
// Note: we use ~x+1 instead of -x to avoid compiler warnings
 
static size_t gap(size_t x, size_t pow_2)
 
{
 
	return (~x + 1) & (pow_2 - 1);
 
}
 

	
 
static u32 load24_le(const u8 s[3])
 
{
 
	return
 
		((u32)s[0] <<  0) |
 
		((u32)s[1] <<  8) |
 
		((u32)s[2] << 16);
 
}
 

	
 
static u32 load32_le(const u8 s[4])
 
{
 
	return
 
		((u32)s[0] <<  0) |
 
		((u32)s[1] <<  8) |
 
		((u32)s[2] << 16) |
 
		((u32)s[3] << 24);
 
}
 

	
 
static u64 load64_le(const u8 s[8])
 
{
 
	return load32_le(s) | ((u64)load32_le(s+4) << 32);
 
}
 

	
 
static void store32_le(u8 out[4], u32 in)
 
{
 
	out[0] =  in        & 0xff;
 
	out[1] = (in >>  8) & 0xff;
 
	out[2] = (in >> 16) & 0xff;
 
	out[3] = (in >> 24) & 0xff;
 
}
 

	
 
static void store64_le(u8 out[8], u64 in)
 
{
 
	store32_le(out    , (u32)in );
 
	store32_le(out + 4, in >> 32);
 
}
 

	
 
static void load32_le_buf (u32 *dst, const u8 *src, size_t size) {
 
	FOR(i, 0, size) { dst[i] = load32_le(src + i*4); }
 
}
 
static void load64_le_buf (u64 *dst, const u8 *src, size_t size) {
 
	FOR(i, 0, size) { dst[i] = load64_le(src + i*8); }
 
}
 
static void store32_le_buf(u8 *dst, const u32 *src, size_t size) {
 
	FOR(i, 0, size) { store32_le(dst + i*4, src[i]); }
 
}
 
static void store64_le_buf(u8 *dst, const u64 *src, size_t size) {
 
	FOR(i, 0, size) { store64_le(dst + i*8, src[i]); }
 
}
 

	
 
static u64 rotr64(u64 x, u64 n) { return (x >> n) ^ (x << (64 - n)); }
 
static u32 rotl32(u32 x, u32 n) { return (x << n) ^ (x >> (32 - n)); }
 

	
 
static int neq0(u64 diff)
 
{
 
	// constant time comparison to zero
 
	// return diff != 0 ? -1 : 0
 
	u64 half = (diff >> 32) | ((u32)diff);
 
	return (1 & ((half - 1) >> 32)) - 1;
 
}
 

	
 
static u64 x16(const u8 a[16], const u8 b[16])
 
{
 
	return (load64_le(a + 0) ^ load64_le(b + 0))
 
		|  (load64_le(a + 8) ^ load64_le(b + 8));
 
}
 
static u64 x32(const u8 a[32],const u8 b[32]){return x16(a,b)| x16(a+16, b+16);}
 
static u64 x64(const u8 a[64],const u8 b[64]){return x32(a,b)| x32(a+32, b+32);}
 
int crypto_verify16(const u8 a[16], const u8 b[16]){ return neq0(x16(a, b)); }
 
int crypto_verify32(const u8 a[32], const u8 b[32]){ return neq0(x32(a, b)); }
 
int crypto_verify64(const u8 a[64], const u8 b[64]){ return neq0(x64(a, b)); }
 

	
 
void crypto_wipe(void *secret, size_t size)
 
{
 
	volatile u8 *v_secret = (u8*)secret;
 
	ZERO(v_secret, size);
 
}
 

	
 
/////////////////
 
/// Chacha 20 ///
 
/////////////////
 
#define QUARTERROUND(a, b, c, d)	\
 
	a += b;  d = rotl32(d ^ a, 16); \
 
	c += d;  b = rotl32(b ^ c, 12); \
 
	a += b;  d = rotl32(d ^ a,  8); \
 
	c += d;  b = rotl32(b ^ c,  7)
 

	
 
static void chacha20_rounds(u32 out[16], const u32 in[16])
 
{
 
	// The temporary variables make Chacha20 10% faster.
 
	u32 t0  = in[ 0];  u32 t1  = in[ 1];  u32 t2  = in[ 2];  u32 t3  = in[ 3];
 
	u32 t4  = in[ 4];  u32 t5  = in[ 5];  u32 t6  = in[ 6];  u32 t7  = in[ 7];
 
	u32 t8  = in[ 8];  u32 t9  = in[ 9];  u32 t10 = in[10];  u32 t11 = in[11];
 
	u32 t12 = in[12];  u32 t13 = in[13];  u32 t14 = in[14];  u32 t15 = in[15];
 

	
 
	FOR (i, 0, 10) { // 20 rounds, 2 rounds per loop.
 
		QUARTERROUND(t0, t4, t8 , t12); // column 0
 
		QUARTERROUND(t1, t5, t9 , t13); // column 1
 
		QUARTERROUND(t2, t6, t10, t14); // column 2
 
		QUARTERROUND(t3, t7, t11, t15); // column 3
 
		QUARTERROUND(t0, t5, t10, t15); // diagonal 0
 
		QUARTERROUND(t1, t6, t11, t12); // diagonal 1
 
		QUARTERROUND(t2, t7, t8 , t13); // diagonal 2
 
		QUARTERROUND(t3, t4, t9 , t14); // diagonal 3
 
	}
 
	out[ 0] = t0;   out[ 1] = t1;   out[ 2] = t2;   out[ 3] = t3;
 
	out[ 4] = t4;   out[ 5] = t5;   out[ 6] = t6;   out[ 7] = t7;
 
	out[ 8] = t8;   out[ 9] = t9;   out[10] = t10;  out[11] = t11;
 
	out[12] = t12;  out[13] = t13;  out[14] = t14;  out[15] = t15;
 
}
 

	
 
static const u8 *chacha20_constant = (const u8*)"expand 32-byte k"; // 16 bytes
 

	
 
void crypto_chacha20_h(u8 out[32], const u8 key[32], const u8 in [16])
 
{
 
	u32 block[16];
 
	load32_le_buf(block     , chacha20_constant, 4);
 
	load32_le_buf(block +  4, key              , 8);
 
	load32_le_buf(block + 12, in               , 4);
 

	
 
	chacha20_rounds(block, block);
 

	
 
	// prevent reversal of the rounds by revealing only half of the buffer.
 
	store32_le_buf(out   , block   , 4); // constant
 
	store32_le_buf(out+16, block+12, 4); // counter and nonce
 
	WIPE_BUFFER(block);
 
}
 

	
 
u64 crypto_chacha20_djb(u8 *cipher_text, const u8 *plain_text,
 
                        size_t text_size, const u8 key[32], const u8 nonce[8],
 
                        u64 ctr)
 
{
 
	u32 input[16];
 
	load32_le_buf(input     , chacha20_constant, 4);
 
	load32_le_buf(input +  4, key              , 8);
 
	load32_le_buf(input + 14, nonce            , 2);
 
	input[12] = (u32) ctr;
 
	input[13] = (u32)(ctr >> 32);
 

	
 
	// Whole blocks
 
	u32    pool[16];
 
	size_t nb_blocks = text_size >> 6;
 
	FOR (i, 0, nb_blocks) {
 
		chacha20_rounds(pool, input);
 
		if (plain_text != 0) {
 
			FOR (j, 0, 16) {
 
				u32 p = pool[j] + input[j];
 
				store32_le(cipher_text, p ^ load32_le(plain_text));
 
				cipher_text += 4;
 
				plain_text  += 4;
 
			}
 
		} else {
 
			FOR (j, 0, 16) {
 
				u32 p = pool[j] + input[j];
 
				store32_le(cipher_text, p);
 
				cipher_text += 4;
 
			}
 
		}
 
		input[12]++;
 
		if (input[12] == 0) {
 
			input[13]++;
 
		}
 
	}
 
	text_size &= 63;
 

	
 
	// Last (incomplete) block
 
	if (text_size > 0) {
 
		if (plain_text == 0) {
 
			plain_text = zero;
 
		}
 
		chacha20_rounds(pool, input);
 
		u8 tmp[64];
 
		FOR (i, 0, 16) {
 
			store32_le(tmp + i*4, pool[i] + input[i]);
 
		}
 
		FOR (i, 0, text_size) {
 
			cipher_text[i] = tmp[i] ^ plain_text[i];
 
		}
 
		WIPE_BUFFER(tmp);
 
	}
 
	ctr = input[12] + ((u64)input[13] << 32) + (text_size > 0);
 

	
 
	WIPE_BUFFER(pool);
 
	WIPE_BUFFER(input);
 
	return ctr;
 
}
 

	
 
u32 crypto_chacha20_ietf(u8 *cipher_text, const u8 *plain_text,
 
                         size_t text_size,
 
                         const u8 key[32], const u8 nonce[12], u32 ctr)
 
{
 
	u64 big_ctr = ctr + ((u64)load32_le(nonce) << 32);
 
	return (u32)crypto_chacha20_djb(cipher_text, plain_text, text_size,
 
	                                key, nonce + 4, big_ctr);
 
}
 

	
 
u64 crypto_chacha20_x(u8 *cipher_text, const u8 *plain_text,
 
                      size_t text_size,
 
                      const u8 key[32], const u8 nonce[24], u64 ctr)
 
{
 
	u8 sub_key[32];
 
	crypto_chacha20_h(sub_key, key, nonce);
 
	ctr = crypto_chacha20_djb(cipher_text, plain_text, text_size,
 
	                          sub_key, nonce + 16, ctr);
 
	WIPE_BUFFER(sub_key);
 
	return ctr;
 
}
 

	
 
/////////////////
 
/// Poly 1305 ///
 
/////////////////
 

	
 
// h = (h + c) * r
 
// preconditions:
 
//   ctx->h <= 4_ffffffff_ffffffff_ffffffff_ffffffff
 
//   ctx->r <=   0ffffffc_0ffffffc_0ffffffc_0fffffff
 
//   end    <= 1
 
// Postcondition:
 
//   ctx->h <= 4_ffffffff_ffffffff_ffffffff_ffffffff
 
static void poly_blocks(crypto_poly1305_ctx *ctx, const u8 *in,
 
                        size_t nb_blocks, unsigned end)
 
{
 
	// Local all the things!
 
	const u32 r0 = ctx->r[0];
 
	const u32 r1 = ctx->r[1];
 
	const u32 r2 = ctx->r[2];
 
	const u32 r3 = ctx->r[3];
 
	const u32 rr0 = (r0 >> 2) * 5;  // lose 2 bits...
 
	const u32 rr1 = (r1 >> 2) + r1; // rr1 == (r1 >> 2) * 5
 
	const u32 rr2 = (r2 >> 2) + r2; // rr1 == (r2 >> 2) * 5
 
	const u32 rr3 = (r3 >> 2) + r3; // rr1 == (r3 >> 2) * 5
 
	const u32 rr4 = r0 & 3;         // ...recover 2 bits
 
	u32 h0 = ctx->h[0];
 
	u32 h1 = ctx->h[1];
 
	u32 h2 = ctx->h[2];
 
	u32 h3 = ctx->h[3];
 
	u32 h4 = ctx->h[4];
 

	
 
	FOR (i, 0, nb_blocks) {
 
		// h + c, without carry propagation
 
		const u64 s0 = (u64)h0 + load32_le(in);  in += 4;
 
		const u64 s1 = (u64)h1 + load32_le(in);  in += 4;
 
		const u64 s2 = (u64)h2 + load32_le(in);  in += 4;
 
		const u64 s3 = (u64)h3 + load32_le(in);  in += 4;
 
		const u32 s4 =      h4 + end;
 

	
 
		// (h + c) * r, without carry propagation
 
		const u64 x0 = s0*r0+ s1*rr3+ s2*rr2+ s3*rr1+ s4*rr0;
 
		const u64 x1 = s0*r1+ s1*r0 + s2*rr3+ s3*rr2+ s4*rr1;
 
		const u64 x2 = s0*r2+ s1*r1 + s2*r0 + s3*rr3+ s4*rr2;
 
		const u64 x3 = s0*r3+ s1*r2 + s2*r1 + s3*r0 + s4*rr3;
 
		const u32 x4 =                                s4*rr4;
 

	
 
		// partial reduction modulo 2^130 - 5
 
		const u32 u5 = x4 + (x3 >> 32); // u5 <= 7ffffff5
 
		const u64 u0 = (u5 >>  2) * 5 + (x0 & 0xffffffff);
 
		const u64 u1 = (u0 >> 32)     + (x1 & 0xffffffff) + (x0 >> 32);
 
		const u64 u2 = (u1 >> 32)     + (x2 & 0xffffffff) + (x1 >> 32);
 
		const u64 u3 = (u2 >> 32)     + (x3 & 0xffffffff) + (x2 >> 32);
 
		const u32 u4 = (u3 >> 32)     + (u5 & 3); // u4 <= 4
 

	
 
		// Update the hash
 
		h0 = u0 & 0xffffffff;
 
		h1 = u1 & 0xffffffff;
 
		h2 = u2 & 0xffffffff;
 
		h3 = u3 & 0xffffffff;
 
		h4 = u4;
 
	}
 
	ctx->h[0] = h0;
 
	ctx->h[1] = h1;
 
	ctx->h[2] = h2;
 
	ctx->h[3] = h3;
 
	ctx->h[4] = h4;
 
}
 

	
 
void crypto_poly1305_init(crypto_poly1305_ctx *ctx, const u8 key[32])
 
{
 
	ZERO(ctx->h, 5); // Initial hash is zero
 
	ctx->c_idx = 0;
 
	// load r and pad (r has some of its bits cleared)
 
	load32_le_buf(ctx->r  , key   , 4);
 
	load32_le_buf(ctx->pad, key+16, 4);
 
	FOR (i, 0, 1) { ctx->r[i] &= 0x0fffffff; }
 
	FOR (i, 1, 4) { ctx->r[i] &= 0x0ffffffc; }
 
}
 

	
 
void crypto_poly1305_update(crypto_poly1305_ctx *ctx,
 
                            const u8 *message, size_t message_size)
 
{
 
	// Avoid undefined NULL pointer increments with empty messages
 
	if (message_size == 0) {
 
		return;
 
	}
 

	
 
	// Align ourselves with block boundaries
 
	size_t aligned = MIN(gap(ctx->c_idx, 16), message_size);
 
	FOR (i, 0, aligned) {
 
		ctx->c[ctx->c_idx] = *message;
 
		ctx->c_idx++;
 
		message++;
 
		message_size--;
 
	}
 

	
 
	// If block is complete, process it
 
	if (ctx->c_idx == 16) {
 
		poly_blocks(ctx, ctx->c, 1, 1);
 
		ctx->c_idx = 0;
 
	}
 

	
 
	// Process the message block by block
 
	size_t nb_blocks = message_size >> 4;
 
	poly_blocks(ctx, message, nb_blocks, 1);
 
	message      += nb_blocks << 4;
 
	message_size &= 15;
 

	
 
	// remaining bytes (we never complete a block here)
 
	FOR (i, 0, message_size) {
 
		ctx->c[ctx->c_idx] = message[i];
 
		ctx->c_idx++;
 
	}
 
}
 

	
 
void crypto_poly1305_final(crypto_poly1305_ctx *ctx, u8 mac[16])
 
{
 
	// Process the last block (if any)
 
	// We move the final 1 according to remaining input length
 
	// (this will add less than 2^130 to the last input block)
 
	if (ctx->c_idx != 0) {
 
		ZERO(ctx->c + ctx->c_idx, 16 - ctx->c_idx);
 
		ctx->c[ctx->c_idx] = 1;
 
		poly_blocks(ctx, ctx->c, 1, 0);
 
	}
 

	
 
	// check if we should subtract 2^130-5 by performing the
 
	// corresponding carry propagation.
 
	u64 c = 5;
 
	FOR (i, 0, 4) {
 
		c  += ctx->h[i];
 
		c >>= 32;
 
	}
 
	c += ctx->h[4];
 
	c  = (c >> 2) * 5; // shift the carry back to the beginning
 
	// c now indicates how many times we should subtract 2^130-5 (0 or 1)
 
	FOR (i, 0, 4) {
 
		c += (u64)ctx->h[i] + ctx->pad[i];
 
		store32_le(mac + i*4, (u32)c);
 
		c = c >> 32;
 
	}
 
	WIPE_CTX(ctx);
 
}
 

	
 
void crypto_poly1305(u8     mac[16],  const u8 *message,
 
                     size_t message_size, const u8  key[32])
 
{
 
	crypto_poly1305_ctx ctx;
 
	crypto_poly1305_init  (&ctx, key);
 
	crypto_poly1305_update(&ctx, message, message_size);
 
	crypto_poly1305_final (&ctx, mac);
 
}
 

	
 
////////////////
 
/// BLAKE2 b ///
 
////////////////
 
static const u64 iv[8] = {
 
	0x6a09e667f3bcc908, 0xbb67ae8584caa73b,
 
	0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
 
	0x510e527fade682d1, 0x9b05688c2b3e6c1f,
 
	0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
 
};
 

	
 
static void blake2b_compress(crypto_blake2b_ctx *ctx, int is_last_block)
 
{
 
	static const u8 sigma[12][16] = {
 
		{  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 },
 
		{ 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 },
 
		{ 11,  8, 12,  0,  5,  2, 15, 13, 10, 14,  3,  6,  7,  1,  9,  4 },
 
		{  7,  9,  3,  1, 13, 12, 11, 14,  2,  6,  5, 10,  4,  0, 15,  8 },
 
		{  9,  0,  5,  7,  2,  4, 10, 15, 14,  1, 11, 12,  6,  8,  3, 13 },
 
		{  2, 12,  6, 10,  0, 11,  8,  3,  4, 13,  7,  5, 15, 14,  1,  9 },
 
		{ 12,  5,  1, 15, 14, 13,  4, 10,  0,  7,  6,  3,  9,  2,  8, 11 },
 
		{ 13, 11,  7, 14, 12,  1,  3,  9,  5,  0, 15,  4,  8,  6,  2, 10 },
 
		{  6, 15, 14,  9, 11,  3,  0,  8, 12,  2, 13,  7,  1,  4, 10,  5 },
 
		{ 10,  2,  8,  4,  7,  6,  1,  5, 15, 11,  9, 14,  3, 12, 13,  0 },
 
		{  0,  1,  2,  3,  4,  5,  6,  7,  8,  9, 10, 11, 12, 13, 14, 15 },
 
		{ 14, 10,  4,  8,  9, 15, 13,  6,  1, 12,  0,  2, 11,  7,  5,  3 },
 
	};
 

	
 
	// increment input offset
 
	u64   *x = ctx->input_offset;
 
	size_t y = ctx->input_idx;
 
	x[0] += y;
 
	if (x[0] < y) {
 
		x[1]++;
 
	}
 

	
 
	// init work vector
 
	u64 v0 = ctx->hash[0];  u64 v8  = iv[0];
 
	u64 v1 = ctx->hash[1];  u64 v9  = iv[1];
 
	u64 v2 = ctx->hash[2];  u64 v10 = iv[2];
 
	u64 v3 = ctx->hash[3];  u64 v11 = iv[3];
 
	u64 v4 = ctx->hash[4];  u64 v12 = iv[4] ^ ctx->input_offset[0];
 
	u64 v5 = ctx->hash[5];  u64 v13 = iv[5] ^ ctx->input_offset[1];
 
	u64 v6 = ctx->hash[6];  u64 v14 = iv[6] ^ (u64)~(is_last_block - 1);
 
	u64 v7 = ctx->hash[7];  u64 v15 = iv[7];
 

	
 
	// mangle work vector
 
	u64 *input = ctx->input;
 
#define BLAKE2_G(a, b, c, d, x, y)	\
 
	a += b + x;  d = rotr64(d ^ a, 32); \
 
	c += d;      b = rotr64(b ^ c, 24); \
 
	a += b + y;  d = rotr64(d ^ a, 16); \
 
	c += d;      b = rotr64(b ^ c, 63)
 
#define BLAKE2_ROUND(i)	\
 
	BLAKE2_G(v0, v4, v8 , v12, input[sigma[i][ 0]], input[sigma[i][ 1]]); \
 
	BLAKE2_G(v1, v5, v9 , v13, input[sigma[i][ 2]], input[sigma[i][ 3]]); \
 
	BLAKE2_G(v2, v6, v10, v14, input[sigma[i][ 4]], input[sigma[i][ 5]]); \
 
	BLAKE2_G(v3, v7, v11, v15, input[sigma[i][ 6]], input[sigma[i][ 7]]); \
 
	BLAKE2_G(v0, v5, v10, v15, input[sigma[i][ 8]], input[sigma[i][ 9]]); \
 
	BLAKE2_G(v1, v6, v11, v12, input[sigma[i][10]], input[sigma[i][11]]); \
 
	BLAKE2_G(v2, v7, v8 , v13, input[sigma[i][12]], input[sigma[i][13]]); \
 
	BLAKE2_G(v3, v4, v9 , v14, input[sigma[i][14]], input[sigma[i][15]])
 

	
 
#ifdef BLAKE2_NO_UNROLLING
 
	FOR (i, 0, 12) {
 
		BLAKE2_ROUND(i);
 
	}
 
#else
 
	BLAKE2_ROUND(0);  BLAKE2_ROUND(1);  BLAKE2_ROUND(2);  BLAKE2_ROUND(3);
 
	BLAKE2_ROUND(4);  BLAKE2_ROUND(5);  BLAKE2_ROUND(6);  BLAKE2_ROUND(7);
 
	BLAKE2_ROUND(8);  BLAKE2_ROUND(9);  BLAKE2_ROUND(10); BLAKE2_ROUND(11);
 
#endif
 

	
 
	// update hash
 
	ctx->hash[0] ^= v0 ^ v8;   ctx->hash[1] ^= v1 ^ v9;
 
	ctx->hash[2] ^= v2 ^ v10;  ctx->hash[3] ^= v3 ^ v11;
 
	ctx->hash[4] ^= v4 ^ v12;  ctx->hash[5] ^= v5 ^ v13;
 
	ctx->hash[6] ^= v6 ^ v14;  ctx->hash[7] ^= v7 ^ v15;
 
}
 

	
 
void crypto_blake2b_keyed_init(crypto_blake2b_ctx *ctx, size_t hash_size,
 
                               const u8 *key, size_t key_size)
 
{
 
	// initial hash
 
	COPY(ctx->hash, iv, 8);
 
	ctx->hash[0] ^= 0x01010000 ^ (key_size << 8) ^ hash_size;
 

	
 
	ctx->input_offset[0] = 0;  // beginning of the input, no offset
 
	ctx->input_offset[1] = 0;  // beginning of the input, no offset
 
	ctx->hash_size       = hash_size;
 
	ctx->input_idx       = 0;
 
	ZERO(ctx->input, 16);
 

	
 
	// if there is a key, the first block is that key (padded with zeroes)
 
	if (key_size > 0) {
 
		u8 key_block[128] = {0};
 
		COPY(key_block, key, key_size);
 
		// same as calling crypto_blake2b_update(ctx, key_block , 128)
 
		load64_le_buf(ctx->input, key_block, 16);
 
		ctx->input_idx = 128;
 
	}
 
}
 

	
 
void crypto_blake2b_init(crypto_blake2b_ctx *ctx, size_t hash_size)
 
{
 
	crypto_blake2b_keyed_init(ctx, hash_size, 0, 0);
 
}
 

	
 
void crypto_blake2b_update(crypto_blake2b_ctx *ctx,
 
                           const u8 *message, size_t message_size)
 
{
 
	// Avoid undefined NULL pointer increments with empty messages
 
	if (message_size == 0) {
 
		return;
 
	}
 

	
 
	// Align with word boundaries
 
	if ((ctx->input_idx & 7) != 0) {
 
		size_t nb_bytes = MIN(gap(ctx->input_idx, 8), message_size);
 
		size_t word     = ctx->input_idx >> 3;
 
		size_t byte     = ctx->input_idx & 7;
 
		FOR (i, 0, nb_bytes) {
 
			ctx->input[word] |= (u64)message[i] << ((byte + i) << 3);
 
		}
 
		ctx->input_idx += nb_bytes;
 
		message        += nb_bytes;
 
		message_size   -= nb_bytes;
 
	}
 

	
 
	// Align with block boundaries (faster than byte by byte)
 
	if ((ctx->input_idx & 127) != 0) {
 
		size_t nb_words = MIN(gap(ctx->input_idx, 128), message_size) >> 3;
 
		load64_le_buf(ctx->input + (ctx->input_idx >> 3), message, nb_words);
 
		ctx->input_idx += nb_words << 3;
 
		message        += nb_words << 3;
 
		message_size   -= nb_words << 3;
 
	}
 

	
 
	// Process block by block
 
	size_t nb_blocks = message_size >> 7;
 
	FOR (i, 0, nb_blocks) {
 
		if (ctx->input_idx == 128) {
 
			blake2b_compress(ctx, 0);
 
		}
 
		load64_le_buf(ctx->input, message, 16);
 
		message += 128;
 
		ctx->input_idx = 128;
 
	}
 
	message_size &= 127;
 

	
 
	if (message_size != 0) {
 
		// Compress block & flush input buffer as needed
 
		if (ctx->input_idx == 128) {
 
			blake2b_compress(ctx, 0);
 
			ctx->input_idx = 0;
 
		}
 
		if (ctx->input_idx == 0) {
 
			ZERO(ctx->input, 16);
 
		}
 
		// Fill remaining words (faster than byte by byte)
 
		size_t nb_words = message_size >> 3;
 
		load64_le_buf(ctx->input, message, nb_words);
 
		ctx->input_idx += nb_words << 3;
 
		message        += nb_words << 3;
 
		message_size   -= nb_words << 3;
 

	
 
		// Fill remaining bytes
 
		FOR (i, 0, message_size) {
 
			size_t word = ctx->input_idx >> 3;
 
			size_t byte = ctx->input_idx & 7;
 
			ctx->input[word] |= (u64)message[i] << (byte << 3);
 
			ctx->input_idx++;
 
		}
 
	}
 
}
 

	
 
void crypto_blake2b_final(crypto_blake2b_ctx *ctx, u8 *hash)
 
{
 
	blake2b_compress(ctx, 1); // compress the last block
 
	size_t hash_size = MIN(ctx->hash_size, 64);
 
	size_t nb_words  = hash_size >> 3;
 
	store64_le_buf(hash, ctx->hash, nb_words);
 
	FOR (i, nb_words << 3, hash_size) {
 
		hash[i] = (ctx->hash[i >> 3] >> (8 * (i & 7))) & 0xff;
 
	}
 
	WIPE_CTX(ctx);
 
}
 

	
 
void crypto_blake2b_keyed(u8 *hash,          size_t hash_size,
 
                          const u8 *key,     size_t key_size,
 
                          const u8 *message, size_t message_size)
 
{
 
	crypto_blake2b_ctx ctx;
 
	crypto_blake2b_keyed_init(&ctx, hash_size, key, key_size);
 
	crypto_blake2b_update    (&ctx, message, message_size);
 
	crypto_blake2b_final     (&ctx, hash);
 
}
 

	
 
void crypto_blake2b(u8 *hash, size_t hash_size, const u8 *msg, size_t msg_size)
 
{
 
	crypto_blake2b_keyed(hash, hash_size, 0, 0, msg, msg_size);
 
}
 

	
 
//////////////
 
/// Argon2 ///
 
//////////////
 
// references to R, Z, Q etc. come from the spec
 

	
 
// Argon2 operates on 1024 byte blocks.
 
typedef struct { u64 a[128]; } blk;
 

	
 
// updates a BLAKE2 hash with a 32 bit word, little endian.
 
static void blake_update_32(crypto_blake2b_ctx *ctx, u32 input)
 
{
 
	u8 buf[4];
 
	store32_le(buf, input);
 
	crypto_blake2b_update(ctx, buf, 4);
 
	WIPE_BUFFER(buf);
 
}
 

	
 
static void blake_update_32_buf(crypto_blake2b_ctx *ctx,
 
                                const u8 *buf, u32 size)
 
{
 
	blake_update_32(ctx, size);
 
	crypto_blake2b_update(ctx, buf, size);
 
}
 

	
 

	
 
static void copy_block(blk *o,const blk*in){FOR(i, 0, 128) o->a[i]  = in->a[i];}
 
static void  xor_block(blk *o,const blk*in){FOR(i, 0, 128) o->a[i] ^= in->a[i];}
 

	
 
// Hash with a virtually unlimited digest size.
 
// Doesn't extract more entropy than the base hash function.
 
// Mainly used for filling a whole kilobyte block with pseudo-random bytes.
 
// (One could use a stream cipher with a seed hash as the key, but
 
//  this would introduce another dependency —and point of failure.)
 
static void extended_hash(u8       *digest, u32 digest_size,
 
                          const u8 *input , u32 input_size)
 
{
 
	crypto_blake2b_ctx ctx;
 
	crypto_blake2b_init  (&ctx, MIN(digest_size, 64));
 
	blake_update_32      (&ctx, digest_size);
 
	crypto_blake2b_update(&ctx, input, input_size);
 
	crypto_blake2b_final (&ctx, digest);
 

	
 
	if (digest_size > 64) {
 
		// the conversion to u64 avoids integer overflow on
 
		// ludicrously big hash sizes.
 
		u32 r   = (u32)(((u64)digest_size + 31) >> 5) - 2;
 
		u32 i   =  1;
 
		u32 in  =  0;
 
		u32 out = 32;
 
		while (i < r) {
 
			// Input and output overlap. This is intentional
 
			crypto_blake2b(digest + out, 64, digest + in, 64);
 
			i   +=  1;
 
			in  += 32;
 
			out += 32;
 
		}
 
		crypto_blake2b(digest + out, digest_size - (32 * r), digest + in , 64);
 
	}
 
}
 

	
 
#define LSB(x) ((u64)(u32)x)
 
#define G(a, b, c, d)	\
 
	a += b + ((LSB(a) * LSB(b)) << 1);  d ^= a;  d = rotr64(d, 32); \
 
	c += d + ((LSB(c) * LSB(d)) << 1);  b ^= c;  b = rotr64(b, 24); \
 
	a += b + ((LSB(a) * LSB(b)) << 1);  d ^= a;  d = rotr64(d, 16); \
 
	c += d + ((LSB(c) * LSB(d)) << 1);  b ^= c;  b = rotr64(b, 63)
 
#define ROUND(v0,  v1,  v2,  v3,  v4,  v5,  v6,  v7,	\
 
              v8,  v9, v10, v11, v12, v13, v14, v15)	\
 
	G(v0, v4,  v8, v12);  G(v1, v5,  v9, v13); \
 
	G(v2, v6, v10, v14);  G(v3, v7, v11, v15); \
 
	G(v0, v5, v10, v15);  G(v1, v6, v11, v12); \
 
	G(v2, v7,  v8, v13);  G(v3, v4,  v9, v14)
 

	
 
// Core of the compression function G.  Computes Z from R in place.
 
static void g_rounds(blk *b)
 
{
 
	// column rounds (work_block = Q)
 
	for (int i = 0; i < 128; i += 16) {
 
		ROUND(b->a[i   ], b->a[i+ 1], b->a[i+ 2], b->a[i+ 3],
 
		      b->a[i+ 4], b->a[i+ 5], b->a[i+ 6], b->a[i+ 7],
 
		      b->a[i+ 8], b->a[i+ 9], b->a[i+10], b->a[i+11],
 
		      b->a[i+12], b->a[i+13], b->a[i+14], b->a[i+15]);
 
	}
 
	// row rounds (b = Z)
 
	for (int i = 0; i < 16; i += 2) {
 
		ROUND(b->a[i   ], b->a[i+ 1], b->a[i+ 16], b->a[i+ 17],
 
		      b->a[i+32], b->a[i+33], b->a[i+ 48], b->a[i+ 49],
 
		      b->a[i+64], b->a[i+65], b->a[i+ 80], b->a[i+ 81],
 
		      b->a[i+96], b->a[i+97], b->a[i+112], b->a[i+113]);
 
	}
 
}
 

	
 
const crypto_argon2_extras crypto_argon2_no_extras = { 0, 0, 0, 0 };
 

	
 
void crypto_argon2(u8 *hash, u32 hash_size, void *work_area,
 
                   crypto_argon2_config config,
 
                   crypto_argon2_inputs inputs,
 
                   crypto_argon2_extras extras)
 
{
 
	const u32 segment_size = config.nb_blocks / config.nb_lanes / 4;
 
	const u32 lane_size    = segment_size * 4;
 
	const u32 nb_blocks    = lane_size * config.nb_lanes; // rounding down
 

	
 
	// work area seen as blocks (must be suitably aligned)
 
	blk *blocks = (blk*)work_area;
 
	{
 
		u8 initial_hash[72]; // 64 bytes plus 2 words for future hashes
 
		crypto_blake2b_ctx ctx;
 
		crypto_blake2b_init (&ctx, 64);
 
		blake_update_32     (&ctx, config.nb_lanes ); // p: number of "threads"
 
		blake_update_32     (&ctx, hash_size);
 
		blake_update_32     (&ctx, config.nb_blocks);
 
		blake_update_32     (&ctx, config.nb_passes);
 
		blake_update_32     (&ctx, 0x13);             // v: version number
 
		blake_update_32     (&ctx, config.algorithm); // y: Argon2i, Argon2d...
 
		blake_update_32_buf (&ctx, inputs.pass, inputs.pass_size);
 
		blake_update_32_buf (&ctx, inputs.salt, inputs.salt_size);
 
		blake_update_32_buf (&ctx, extras.key,  extras.key_size);
 
		blake_update_32_buf (&ctx, extras.ad,   extras.ad_size);
 
		crypto_blake2b_final(&ctx, initial_hash); // fill 64 first bytes only
 

	
 
		// fill first 2 blocks of each lane
 
		u8 hash_area[1024];
 
		FOR_T(u32, l, 0, config.nb_lanes) {
 
			FOR_T(u32, i, 0, 2) {
 
				store32_le(initial_hash + 64, i); // first  additional word
 
				store32_le(initial_hash + 68, l); // second additional word
 
				extended_hash(hash_area, 1024, initial_hash, 72);
 
				load64_le_buf(blocks[l * lane_size + i].a, hash_area, 128);
 
			}
 
		}
 

	
 
		WIPE_BUFFER(initial_hash);
 
		WIPE_BUFFER(hash_area);
 
	}
 

	
 
	// Argon2i and Argon2id start with constant time indexing
 
	int constant_time = config.algorithm != CRYPTO_ARGON2_D;
 

	
 
	// Fill (and re-fill) the rest of the blocks
 
	//
 
	// Note: even though each segment within the same slice can be
 
	// computed in parallel, (one thread per lane), we are computing
 
	// them sequentially, because Monocypher doesn't support threads.
 
	//
 
	// Yet optimal performance (and therefore security) requires one
 
	// thread per lane. The only reason Monocypher supports multiple
 
	// lanes is compatibility.
 
	blk tmp;
 
	FOR_T(u32, pass, 0, config.nb_passes) {
 
		FOR_T(u32, slice, 0, 4) {
 
			// On the first slice of the first pass,
 
			// blocks 0 and 1 are already filled, hence pass_offset.
 
			u32 pass_offset  = pass == 0 && slice == 0 ? 2 : 0;
 
			u32 slice_offset = slice * segment_size;
 

	
 
			// Argon2id switches back to non-constant time indexing
 
			// after the first two slices of the first pass
 
			if (slice == 2 && config.algorithm == CRYPTO_ARGON2_ID) {
 
				constant_time = 0;
 
			}
 

	
 
			// Each iteration of the following loop may be performed in
 
			// a separate thread.  All segments must be fully completed
 
			// before we start filling the next slice.
 
			FOR_T(u32, segment, 0, config.nb_lanes) {
 
				blk index_block;
 
				u32 index_ctr = 1;
 
				FOR_T (u32, block, pass_offset, segment_size) {
 
					// Current and previous blocks
 
					u32  lane_offset   = segment * lane_size;
 
					blk *segment_start = blocks + lane_offset + slice_offset;
 
					blk *current       = segment_start + block;
 
					blk *previous      =
 
						block == 0 && slice_offset == 0
 
						? segment_start + lane_size - 1
 
						: segment_start + block - 1;
 

	
 
					u64 index_seed;
 
					if (constant_time) {
 
						if (block == pass_offset || (block % 128) == 0) {
 
							// Fill or refresh deterministic indices block
 

	
 
							// seed the beginning of the block...
 
							ZERO(index_block.a, 128);
 
							index_block.a[0] = pass;
 
							index_block.a[1] = segment;
 
							index_block.a[2] = slice;
 
							index_block.a[3] = nb_blocks;
 
							index_block.a[4] = config.nb_passes;
 
							index_block.a[5] = config.algorithm;
 
							index_block.a[6] = index_ctr;
 
							index_ctr++;
 

	
 
							// ... then shuffle it
 
							copy_block(&tmp, &index_block);
 
							g_rounds  (&index_block);
 
							xor_block (&index_block, &tmp);
 
							copy_block(&tmp, &index_block);
 
							g_rounds  (&index_block);
 
							xor_block (&index_block, &tmp);
 
						}
 
						index_seed = index_block.a[block % 128];
 
					} else {
 
						index_seed = previous->a[0];
 
					}
 

	
 
					// Establish the reference set.  *Approximately* comprises:
 
					// - The last 3 slices (if they exist yet)
 
					// - The already constructed blocks in the current segment
 
					u32 next_slice   = ((slice + 1) % 4) * segment_size;
 
					u32 window_start = pass == 0 ? 0     : next_slice;
 
					u32 nb_segments  = pass == 0 ? slice : 3;
 
					u64 lane         =
 
						pass == 0 && slice == 0
 
						? segment
 
						: (index_seed >> 32) % config.nb_lanes;
 
					u32 window_size  =
 
						nb_segments * segment_size +
 
						(lane  == segment ? block-1 :
 
						 block == 0       ? (u32)-1 : 0);
 

	
 
					// Find reference block
 
					u64  j1        = index_seed & 0xffffffff; // block selector
 
					u64  x         = (j1 * j1)         >> 32;
 
					u64  y         = (window_size * x) >> 32;
 
					u64  z         = (window_size - 1) - y;
 
					u64  ref       = (window_start + z) % lane_size;
 
					u32  index     = lane * lane_size + (u32)ref;
 
					blk *reference = blocks + index;
 

	
 
					// Shuffle the previous & reference block
 
					// into the current block
 
					copy_block(&tmp, previous);
 
					xor_block (&tmp, reference);
 
					if (pass == 0) { copy_block(current, &tmp); }
 
					else           { xor_block (current, &tmp); }
 
					g_rounds  (&tmp);
 
					xor_block (current, &tmp);
 
				}
 
			}
 
		}
 
	}
 

	
 
	// Wipe temporary block
 
	volatile u64* p = tmp.a;
 
	ZERO(p, 128);
 

	
 
	// XOR last blocks of each lane
 
	blk *last_block = blocks + lane_size - 1;
 
	FOR_T (u32, lane, 1, config.nb_lanes) {
 
		blk *next_block = last_block + lane_size;
 
		xor_block(next_block, last_block);
 
		last_block = next_block;
 
	}
 

	
 
	// Serialize last block
 
	u8 final_block[1024];
 
	store64_le_buf(final_block, last_block->a, 128);
 

	
 
	// Wipe work area
 
	p = (u64*)work_area;
 
	ZERO(p, 128 * nb_blocks);
 

	
 
	// Hash the very last block with H' into the output hash
 
	extended_hash(hash, hash_size, final_block, 1024);
 
	WIPE_BUFFER(final_block);
 
}
 

	
 
////////////////////////////////////
 
/// Arithmetic modulo 2^255 - 19 ///
 
////////////////////////////////////
 
//  Originally taken from SUPERCOP's ref10 implementation.
 
//  A bit bigger than TweetNaCl, over 4 times faster.
 

	
 
// field element
 
typedef i32 fe[10];
 

	
 
// field constants
 
//
 
// fe_one      : 1
 
// sqrtm1      : sqrt(-1)
 
// d           :     -121665 / 121666
 
// D2          : 2 * -121665 / 121666
 
// lop_x, lop_y: low order point in Edwards coordinates
 
// ufactor     : -sqrt(-1) * 2
 
// A2          : 486662^2  (A squared)
 
static const fe fe_one  = {1};
 
static const fe sqrtm1  = {
 
	-32595792, -7943725, 9377950, 3500415, 12389472,
 
	-272473, -25146209, -2005654, 326686, 11406482,
 
};
 
static const fe d       = {
 
	-10913610, 13857413, -15372611, 6949391, 114729,
 
	-8787816, -6275908, -3247719, -18696448, -12055116,
 
};
 
static const fe D2      = {
 
	-21827239, -5839606, -30745221, 13898782, 229458,
 
	15978800, -12551817, -6495438, 29715968, 9444199,
 
};
 
static const fe lop_x   = {
 
	21352778, 5345713, 4660180, -8347857, 24143090,
 
	14568123, 30185756, -12247770, -33528939, 8345319,
 
};
 
static const fe lop_y   = {
 
	-6952922, -1265500, 6862341, -7057498, -4037696,
 
	-5447722, 31680899, -15325402, -19365852, 1569102,
 
};
 
static const fe ufactor = {
 
	-1917299, 15887451, -18755900, -7000830, -24778944,
 
	544946, -16816446, 4011309, -653372, 10741468,
 
};
 
static const fe A2      = {
 
	12721188, 3529, 0, 0, 0, 0, 0, 0, 0, 0,
 
};
 

	
 
static void fe_0(fe h) {           ZERO(h  , 10); }
 
static void fe_1(fe h) { h[0] = 1; ZERO(h+1,  9); }
 

	
 
static void fe_copy(fe h,const fe f           ){FOR(i,0,10) h[i] =  f[i];      }
 
static void fe_neg (fe h,const fe f           ){FOR(i,0,10) h[i] = -f[i];      }
 
static void fe_add (fe h,const fe f,const fe g){FOR(i,0,10) h[i] = f[i] + g[i];}
 
static void fe_sub (fe h,const fe f,const fe g){FOR(i,0,10) h[i] = f[i] - g[i];}
 

	
 
static void fe_cswap(fe f, fe g, int b)
 
{
 
	i32 mask = -b; // -1 = 0xffffffff
 
	FOR (i, 0, 10) {
 
		i32 x = (f[i] ^ g[i]) & mask;
 
		f[i] = f[i] ^ x;
 
		g[i] = g[i] ^ x;
 
	}
 
}
 

	
 
static void fe_ccopy(fe f, const fe g, int b)
 
{
 
	i32 mask = -b; // -1 = 0xffffffff
 
	FOR (i, 0, 10) {
 
		i32 x = (f[i] ^ g[i]) & mask;
 
		f[i] = f[i] ^ x;
 
	}
 
}
 

	
 

	
 
// Signed carry propagation
 
// ------------------------
 
//
 
// Let t be a number.  It can be uniquely decomposed thus:
 
//
 
//    t = h*2^26 + l
 
//    such that -2^25 <= l < 2^25
 
//
 
// Let c = (t + 2^25) / 2^26            (rounded down)
 
//     c = (h*2^26 + l + 2^25) / 2^26   (rounded down)
 
//     c =  h   +   (l + 2^25) / 2^26   (rounded down)
 
//     c =  h                           (exactly)
 
// Because 0 <= l + 2^25 < 2^26
 
//
 
// Let u = t          - c*2^26
 
//     u = h*2^26 + l - h*2^26
 
//     u = l
 
// Therefore, -2^25 <= u < 2^25
 
//
 
// Additionally, if |t| < x, then |h| < x/2^26 (rounded down)
 
//
 
// Notations:
 
// - In C, 1<<25 means 2^25.
 
// - In C, x>>25 means floor(x / (2^25)).
 
// - All of the above applies with 25 & 24 as well as 26 & 25.
 
//
 
//
 
// Note on negative right shifts
 
// -----------------------------
 
//
 
// In C, x >> n, where x is a negative integer, is implementation
 
// defined.  In practice, all platforms do arithmetic shift, which is
 
// equivalent to division by 2^26, rounded down.  Some compilers, like
 
// GCC, even guarantee it.
 
//
 
// If we ever stumble upon a platform that does not propagate the sign
 
// bit (we won't), visible failures will show at the slightest test, and
 
// the signed shifts can be replaced by the following:
 
//
 
//     typedef struct { i64 x:39; } s25;
 
//     typedef struct { i64 x:38; } s26;
 
//     i64 shift25(i64 x) { s25 s; s.x = ((u64)x)>>25; return s.x; }
 
//     i64 shift26(i64 x) { s26 s; s.x = ((u64)x)>>26; return s.x; }
 
//
 
// Current compilers cannot optimise this, causing a 30% drop in
 
// performance.  Fairly expensive for something that never happens.
 
//
 
//
 
// Precondition
 
// ------------
 
//
 
// |t0|       < 2^63
 
// |t1|..|t9| < 2^62
 
//
 
// Algorithm
 
// ---------
 
// c   = t0 + 2^25 / 2^26   -- |c|  <= 2^36
 
// t0 -= c * 2^26           -- |t0| <= 2^25
 
// t1 += c                  -- |t1| <= 2^63
 
//
 
// c   = t4 + 2^25 / 2^26   -- |c|  <= 2^36
 
// t4 -= c * 2^26           -- |t4| <= 2^25
 
// t5 += c                  -- |t5| <= 2^63
 
//
 
// c   = t1 + 2^24 / 2^25   -- |c|  <= 2^38
 
// t1 -= c * 2^25           -- |t1| <= 2^24
 
// t2 += c                  -- |t2| <= 2^63
 
//
 
// c   = t5 + 2^24 / 2^25   -- |c|  <= 2^38
 
// t5 -= c * 2^25           -- |t5| <= 2^24
 
// t6 += c                  -- |t6| <= 2^63
 
//
 
// c   = t2 + 2^25 / 2^26   -- |c|  <= 2^37
 
// t2 -= c * 2^26           -- |t2| <= 2^25        < 1.1 * 2^25  (final t2)
 
// t3 += c                  -- |t3| <= 2^63
 
//
 
// c   = t6 + 2^25 / 2^26   -- |c|  <= 2^37
 
// t6 -= c * 2^26           -- |t6| <= 2^25        < 1.1 * 2^25  (final t6)
 
// t7 += c                  -- |t7| <= 2^63
 
//
 
// c   = t3 + 2^24 / 2^25   -- |c|  <= 2^38
 
// t3 -= c * 2^25           -- |t3| <= 2^24        < 1.1 * 2^24  (final t3)
 
// t4 += c                  -- |t4| <= 2^25 + 2^38 < 2^39
 
//
 
// c   = t7 + 2^24 / 2^25   -- |c|  <= 2^38
 
// t7 -= c * 2^25           -- |t7| <= 2^24        < 1.1 * 2^24  (final t7)
 
// t8 += c                  -- |t8| <= 2^63
 
//
 
// c   = t4 + 2^25 / 2^26   -- |c|  <= 2^13
 
// t4 -= c * 2^26           -- |t4| <= 2^25        < 1.1 * 2^25  (final t4)
 
// t5 += c                  -- |t5| <= 2^24 + 2^13 < 1.1 * 2^24  (final t5)
 
//
 
// c   = t8 + 2^25 / 2^26   -- |c|  <= 2^37
 
// t8 -= c * 2^26           -- |t8| <= 2^25        < 1.1 * 2^25  (final t8)
 
// t9 += c                  -- |t9| <= 2^63
 
//
 
// c   = t9 + 2^24 / 2^25   -- |c|  <= 2^38
 
// t9 -= c * 2^25           -- |t9| <= 2^24        < 1.1 * 2^24  (final t9)
 
// t0 += c * 19             -- |t0| <= 2^25 + 2^38*19 < 2^44
 
//
 
// c   = t0 + 2^25 / 2^26   -- |c|  <= 2^18
 
// t0 -= c * 2^26           -- |t0| <= 2^25        < 1.1 * 2^25  (final t0)
 
// t1 += c                  -- |t1| <= 2^24 + 2^18 < 1.1 * 2^24  (final t1)
 
//
 
// Postcondition
 
// -------------
 
//   |t0|, |t2|, |t4|, |t6|, |t8|  <  1.1 * 2^25
 
//   |t1|, |t3|, |t5|, |t7|, |t9|  <  1.1 * 2^24
 
#define FE_CARRY	\
 
	i64 c; \
 
	c = (t0 + ((i64)1<<25)) >> 26;  t0 -= c * ((i64)1 << 26);  t1 += c; \
 
	c = (t4 + ((i64)1<<25)) >> 26;  t4 -= c * ((i64)1 << 26);  t5 += c; \
 
	c = (t1 + ((i64)1<<24)) >> 25;  t1 -= c * ((i64)1 << 25);  t2 += c; \
 
	c = (t5 + ((i64)1<<24)) >> 25;  t5 -= c * ((i64)1 << 25);  t6 += c; \
 
	c = (t2 + ((i64)1<<25)) >> 26;  t2 -= c * ((i64)1 << 26);  t3 += c; \
 
	c = (t6 + ((i64)1<<25)) >> 26;  t6 -= c * ((i64)1 << 26);  t7 += c; \
 
	c = (t3 + ((i64)1<<24)) >> 25;  t3 -= c * ((i64)1 << 25);  t4 += c; \
 
	c = (t7 + ((i64)1<<24)) >> 25;  t7 -= c * ((i64)1 << 25);  t8 += c; \
 
	c = (t4 + ((i64)1<<25)) >> 26;  t4 -= c * ((i64)1 << 26);  t5 += c; \
 
	c = (t8 + ((i64)1<<25)) >> 26;  t8 -= c * ((i64)1 << 26);  t9 += c; \
 
	c = (t9 + ((i64)1<<24)) >> 25;  t9 -= c * ((i64)1 << 25);  t0 += c * 19; \
 
	c = (t0 + ((i64)1<<25)) >> 26;  t0 -= c * ((i64)1 << 26);  t1 += c; \
 
	h[0]=(i32)t0;  h[1]=(i32)t1;  h[2]=(i32)t2;  h[3]=(i32)t3;  h[4]=(i32)t4; \
 
	h[5]=(i32)t5;  h[6]=(i32)t6;  h[7]=(i32)t7;  h[8]=(i32)t8;  h[9]=(i32)t9
 

	
 
// Decodes a field element from a byte buffer.
 
// mask specifies how many bits we ignore.
 
// Traditionally we ignore 1. It's useful for EdDSA,
 
// which uses that bit to denote the sign of x.
 
// Elligator however uses positive representatives,
 
// which means ignoring 2 bits instead.
 
static void fe_frombytes_mask(fe h, const u8 s[32], unsigned nb_mask)
 
{
 
	u32 mask = 0xffffff >> nb_mask;
 
	i64 t0 =  load32_le(s);                    // t0 < 2^32
 
	i64 t1 =  load24_le(s +  4) << 6;          // t1 < 2^30
 
	i64 t2 =  load24_le(s +  7) << 5;          // t2 < 2^29
 
	i64 t3 =  load24_le(s + 10) << 3;          // t3 < 2^27
 
	i64 t4 =  load24_le(s + 13) << 2;          // t4 < 2^26
 
	i64 t5 =  load32_le(s + 16);               // t5 < 2^32
 
	i64 t6 =  load24_le(s + 20) << 7;          // t6 < 2^31
 
	i64 t7 =  load24_le(s + 23) << 5;          // t7 < 2^29
 
	i64 t8 =  load24_le(s + 26) << 4;          // t8 < 2^28
 
	i64 t9 = (load24_le(s + 29) & mask) << 2;  // t9 < 2^25
 
	FE_CARRY;                                  // Carry precondition OK
 
}
 

	
 
static void fe_frombytes(fe h, const u8 s[32])
 
{
 
	fe_frombytes_mask(h, s, 1);
 
}
 

	
 

	
 
// Precondition
 
//   |h[0]|, |h[2]|, |h[4]|, |h[6]|, |h[8]|  <  1.1 * 2^25
 
//   |h[1]|, |h[3]|, |h[5]|, |h[7]|, |h[9]|  <  1.1 * 2^24
 
//
 
// Therefore, |h| < 2^255-19
 
// There are two possibilities:
 
//
 
// - If h is positive, all we need to do is reduce its individual
 
//   limbs down to their tight positive range.
 
// - If h is negative, we also need to add 2^255-19 to it.
 
//   Or just remove 19 and chop off any excess bit.
 
static void fe_tobytes(u8 s[32], const fe h)
 
{
 
	i32 t[10];
 
	COPY(t, h, 10);
 
	i32 q = (19 * t[9] + (((i32) 1) << 24)) >> 25;
 
	//                 |t9|                    < 1.1 * 2^24
 
	//  -1.1 * 2^24  <  t9                     < 1.1 * 2^24
 
	//  -21  * 2^24  <  19 * t9                < 21  * 2^24
 
	//  -2^29        <  19 * t9 + 2^24         < 2^29
 
	//  -2^29 / 2^25 < (19 * t9 + 2^24) / 2^25 < 2^29 / 2^25
 
	//  -16          < (19 * t9 + 2^24) / 2^25 < 16
 
	FOR (i, 0, 5) {
 
		q += t[2*i  ]; q >>= 26; // q = 0 or -1
 
		q += t[2*i+1]; q >>= 25; // q = 0 or -1
 
	}
 
	// q =  0 iff h >= 0
 
	// q = -1 iff h <  0
 
	// Adding q * 19 to h reduces h to its proper range.
 
	q *= 19;  // Shift carry back to the beginning
 
	FOR (i, 0, 5) {
 
		t[i*2  ] += q;  q = t[i*2  ] >> 26;  t[i*2  ] -= q * ((i32)1 << 26);
 
		t[i*2+1] += q;  q = t[i*2+1] >> 25;  t[i*2+1] -= q * ((i32)1 << 25);
 
	}
 
	// h is now fully reduced, and q represents the excess bit.
 

	
 
	store32_le(s +  0, ((u32)t[0] >>  0) | ((u32)t[1] << 26));
 
	store32_le(s +  4, ((u32)t[1] >>  6) | ((u32)t[2] << 19));
 
	store32_le(s +  8, ((u32)t[2] >> 13) | ((u32)t[3] << 13));
 
	store32_le(s + 12, ((u32)t[3] >> 19) | ((u32)t[4] <<  6));
 
	store32_le(s + 16, ((u32)t[5] >>  0) | ((u32)t[6] << 25));
 
	store32_le(s + 20, ((u32)t[6] >>  7) | ((u32)t[7] << 19));
 
	store32_le(s + 24, ((u32)t[7] >> 13) | ((u32)t[8] << 12));
 
	store32_le(s + 28, ((u32)t[8] >> 20) | ((u32)t[9] <<  6));
 

	
 
	WIPE_BUFFER(t);
 
}
 

	
 
// Precondition
 
// -------------
 
//   |f0|, |f2|, |f4|, |f6|, |f8|  <  1.65 * 2^26
 
//   |f1|, |f3|, |f5|, |f7|, |f9|  <  1.65 * 2^25
 
//
 
//   |g0|, |g2|, |g4|, |g6|, |g8|  <  1.65 * 2^26
 
//   |g1|, |g3|, |g5|, |g7|, |g9|  <  1.65 * 2^25
 
static void fe_mul_small(fe h, const fe f, i32 g)
 
{
 
	i64 t0 = f[0] * (i64) g;  i64 t1 = f[1] * (i64) g;
 
	i64 t2 = f[2] * (i64) g;  i64 t3 = f[3] * (i64) g;
 
	i64 t4 = f[4] * (i64) g;  i64 t5 = f[5] * (i64) g;
 
	i64 t6 = f[6] * (i64) g;  i64 t7 = f[7] * (i64) g;
 
	i64 t8 = f[8] * (i64) g;  i64 t9 = f[9] * (i64) g;
 
	// |t0|, |t2|, |t4|, |t6|, |t8|  <  1.65 * 2^26 * 2^31  < 2^58
 
	// |t1|, |t3|, |t5|, |t7|, |t9|  <  1.65 * 2^25 * 2^31  < 2^57
 

	
 
	FE_CARRY; // Carry precondition OK
 
}
 

	
 
// Precondition
 
// -------------
 
//   |f0|, |f2|, |f4|, |f6|, |f8|  <  1.65 * 2^26
 
//   |f1|, |f3|, |f5|, |f7|, |f9|  <  1.65 * 2^25
 
//
 
//   |g0|, |g2|, |g4|, |g6|, |g8|  <  1.65 * 2^26
 
//   |g1|, |g3|, |g5|, |g7|, |g9|  <  1.65 * 2^25
 
static void fe_mul(fe h, const fe f, const fe g)
 
{
 
	// Everything is unrolled and put in temporary variables.
 
	// We could roll the loop, but that would make curve25519 twice as slow.
 
	i32 f0 = f[0]; i32 f1 = f[1]; i32 f2 = f[2]; i32 f3 = f[3]; i32 f4 = f[4];
 
	i32 f5 = f[5]; i32 f6 = f[6]; i32 f7 = f[7]; i32 f8 = f[8]; i32 f9 = f[9];
 
	i32 g0 = g[0]; i32 g1 = g[1]; i32 g2 = g[2]; i32 g3 = g[3]; i32 g4 = g[4];
 
	i32 g5 = g[5]; i32 g6 = g[6]; i32 g7 = g[7]; i32 g8 = g[8]; i32 g9 = g[9];
 
	i32 F1 = f1*2; i32 F3 = f3*2; i32 F5 = f5*2; i32 F7 = f7*2; i32 F9 = f9*2;
 
	i32 G1 = g1*19;  i32 G2 = g2*19;  i32 G3 = g3*19;
 
	i32 G4 = g4*19;  i32 G5 = g5*19;  i32 G6 = g6*19;
 
	i32 G7 = g7*19;  i32 G8 = g8*19;  i32 G9 = g9*19;
 
	// |F1|, |F3|, |F5|, |F7|, |F9|  <  1.65 * 2^26
 
	// |G0|, |G2|, |G4|, |G6|, |G8|  <  2^31
 
	// |G1|, |G3|, |G5|, |G7|, |G9|  <  2^30
 

	
 
	i64 t0 = f0*(i64)g0 + F1*(i64)G9 + f2*(i64)G8 + F3*(i64)G7 + f4*(i64)G6
 
	       + F5*(i64)G5 + f6*(i64)G4 + F7*(i64)G3 + f8*(i64)G2 + F9*(i64)G1;
 
	i64 t1 = f0*(i64)g1 + f1*(i64)g0 + f2*(i64)G9 + f3*(i64)G8 + f4*(i64)G7
 
	       + f5*(i64)G6 + f6*(i64)G5 + f7*(i64)G4 + f8*(i64)G3 + f9*(i64)G2;
 
	i64 t2 = f0*(i64)g2 + F1*(i64)g1 + f2*(i64)g0 + F3*(i64)G9 + f4*(i64)G8
 
	       + F5*(i64)G7 + f6*(i64)G6 + F7*(i64)G5 + f8*(i64)G4 + F9*(i64)G3;
 
	i64 t3 = f0*(i64)g3 + f1*(i64)g2 + f2*(i64)g1 + f3*(i64)g0 + f4*(i64)G9
 
	       + f5*(i64)G8 + f6*(i64)G7 + f7*(i64)G6 + f8*(i64)G5 + f9*(i64)G4;
 
	i64 t4 = f0*(i64)g4 + F1*(i64)g3 + f2*(i64)g2 + F3*(i64)g1 + f4*(i64)g0
 
	       + F5*(i64)G9 + f6*(i64)G8 + F7*(i64)G7 + f8*(i64)G6 + F9*(i64)G5;
 
	i64 t5 = f0*(i64)g5 + f1*(i64)g4 + f2*(i64)g3 + f3*(i64)g2 + f4*(i64)g1
 
	       + f5*(i64)g0 + f6*(i64)G9 + f7*(i64)G8 + f8*(i64)G7 + f9*(i64)G6;
 
	i64 t6 = f0*(i64)g6 + F1*(i64)g5 + f2*(i64)g4 + F3*(i64)g3 + f4*(i64)g2
 
	       + F5*(i64)g1 + f6*(i64)g0 + F7*(i64)G9 + f8*(i64)G8 + F9*(i64)G7;
 
	i64 t7 = f0*(i64)g7 + f1*(i64)g6 + f2*(i64)g5 + f3*(i64)g4 + f4*(i64)g3
 
	       + f5*(i64)g2 + f6*(i64)g1 + f7*(i64)g0 + f8*(i64)G9 + f9*(i64)G8;
 
	i64 t8 = f0*(i64)g8 + F1*(i64)g7 + f2*(i64)g6 + F3*(i64)g5 + f4*(i64)g4
 
	       + F5*(i64)g3 + f6*(i64)g2 + F7*(i64)g1 + f8*(i64)g0 + F9*(i64)G9;
 
	i64 t9 = f0*(i64)g9 + f1*(i64)g8 + f2*(i64)g7 + f3*(i64)g6 + f4*(i64)g5
 
	       + f5*(i64)g4 + f6*(i64)g3 + f7*(i64)g2 + f8*(i64)g1 + f9*(i64)g0;
 
	// t0 < 0.67 * 2^61
 
	// t1 < 0.41 * 2^61
 
	// t2 < 0.52 * 2^61
 
	// t3 < 0.32 * 2^61
 
	// t4 < 0.38 * 2^61
 
	// t5 < 0.22 * 2^61
 
	// t6 < 0.23 * 2^61
 
	// t7 < 0.13 * 2^61
 
	// t8 < 0.09 * 2^61
 
	// t9 < 0.03 * 2^61
 

	
 
	FE_CARRY; // Everything below 2^62, Carry precondition OK
 
}
 

	
 
// Precondition
 
// -------------
 
//   |f0|, |f2|, |f4|, |f6|, |f8|  <  1.65 * 2^26
 
//   |f1|, |f3|, |f5|, |f7|, |f9|  <  1.65 * 2^25
 
//
 
// Note: we could use fe_mul() for this, but this is significantly faster
 
static void fe_sq(fe h, const fe f)
 
{
 
	i32 f0 = f[0]; i32 f1 = f[1]; i32 f2 = f[2]; i32 f3 = f[3]; i32 f4 = f[4];
 
	i32 f5 = f[5]; i32 f6 = f[6]; i32 f7 = f[7]; i32 f8 = f[8]; i32 f9 = f[9];
 
	i32 f0_2  = f0*2;   i32 f1_2  = f1*2;   i32 f2_2  = f2*2;   i32 f3_2 = f3*2;
 
	i32 f4_2  = f4*2;   i32 f5_2  = f5*2;   i32 f6_2  = f6*2;   i32 f7_2 = f7*2;
 
	i32 f5_38 = f5*38;  i32 f6_19 = f6*19;  i32 f7_38 = f7*38;
 
	i32 f8_19 = f8*19;  i32 f9_38 = f9*38;
 
	// |f0_2| , |f2_2| , |f4_2| , |f6_2| , |f8_2|  <  1.65 * 2^27
 
	// |f1_2| , |f3_2| , |f5_2| , |f7_2| , |f9_2|  <  1.65 * 2^26
 
	// |f5_38|, |f6_19|, |f7_38|, |f8_19|, |f9_38| <  2^31
 

	
 
	i64 t0 = f0  *(i64)f0    + f1_2*(i64)f9_38 + f2_2*(i64)f8_19
 
	       + f3_2*(i64)f7_38 + f4_2*(i64)f6_19 + f5  *(i64)f5_38;
 
	i64 t1 = f0_2*(i64)f1    + f2  *(i64)f9_38 + f3_2*(i64)f8_19
 
	       + f4  *(i64)f7_38 + f5_2*(i64)f6_19;
 
	i64 t2 = f0_2*(i64)f2    + f1_2*(i64)f1    + f3_2*(i64)f9_38
 
	       + f4_2*(i64)f8_19 + f5_2*(i64)f7_38 + f6  *(i64)f6_19;
 
	i64 t3 = f0_2*(i64)f3    + f1_2*(i64)f2    + f4  *(i64)f9_38
 
	       + f5_2*(i64)f8_19 + f6  *(i64)f7_38;
 
	i64 t4 = f0_2*(i64)f4    + f1_2*(i64)f3_2  + f2  *(i64)f2
 
	       + f5_2*(i64)f9_38 + f6_2*(i64)f8_19 + f7  *(i64)f7_38;
 
	i64 t5 = f0_2*(i64)f5    + f1_2*(i64)f4    + f2_2*(i64)f3
 
	       + f6  *(i64)f9_38 + f7_2*(i64)f8_19;
 
	i64 t6 = f0_2*(i64)f6    + f1_2*(i64)f5_2  + f2_2*(i64)f4
 
	       + f3_2*(i64)f3    + f7_2*(i64)f9_38 + f8  *(i64)f8_19;
 
	i64 t7 = f0_2*(i64)f7    + f1_2*(i64)f6    + f2_2*(i64)f5
 
	       + f3_2*(i64)f4    + f8  *(i64)f9_38;
 
	i64 t8 = f0_2*(i64)f8    + f1_2*(i64)f7_2  + f2_2*(i64)f6
 
	       + f3_2*(i64)f5_2  + f4  *(i64)f4    + f9  *(i64)f9_38;
 
	i64 t9 = f0_2*(i64)f9    + f1_2*(i64)f8    + f2_2*(i64)f7
 
	       + f3_2*(i64)f6    + f4  *(i64)f5_2;
 
	// t0 < 0.67 * 2^61
 
	// t1 < 0.41 * 2^61
 
	// t2 < 0.52 * 2^61
 
	// t3 < 0.32 * 2^61
 
	// t4 < 0.38 * 2^61
 
	// t5 < 0.22 * 2^61
 
	// t6 < 0.23 * 2^61
 
	// t7 < 0.13 * 2^61
 
	// t8 < 0.09 * 2^61
 
	// t9 < 0.03 * 2^61
 

	
 
	FE_CARRY;
 
}
 

	
 
//  Parity check.  Returns 0 if even, 1 if odd
 
static int fe_isodd(const fe f)
 
{
 
	u8 s[32];
 
	fe_tobytes(s, f);
 
	u8 isodd = s[0] & 1;
 
	WIPE_BUFFER(s);
 
	return isodd;
 
}
 

	
 
// Returns 1 if equal, 0 if not equal
 
static int fe_isequal(const fe f, const fe g)
 
{
 
	u8 fs[32];
 
	u8 gs[32];
 
	fe_tobytes(fs, f);
 
	fe_tobytes(gs, g);
 
	int isdifferent = crypto_verify32(fs, gs);
 
	WIPE_BUFFER(fs);
 
	WIPE_BUFFER(gs);
 
	return 1 + isdifferent;
 
}
 

	
 
// Inverse square root.
 
// Returns true if x is a square, false otherwise.
 
// After the call:
 
//   isr = sqrt(1/x)        if x is a non-zero square.
 
//   isr = sqrt(sqrt(-1)/x) if x is not a square.
 
//   isr = 0                if x is zero.
 
// We do not guarantee the sign of the square root.
 
//
 
// Notes:
 
// Let quartic = x^((p-1)/4)
 
//
 
// x^((p-1)/2) = chi(x)
 
// quartic^2   = chi(x)
 
// quartic     = sqrt(chi(x))
 
// quartic     = 1 or -1 or sqrt(-1) or -sqrt(-1)
 
//
 
// Note that x is a square if quartic is 1 or -1
 
// There are 4 cases to consider:
 
//
 
// if   quartic         = 1  (x is a square)
 
// then x^((p-1)/4)     = 1
 
//      x^((p-5)/4) * x = 1
 
//      x^((p-5)/4)     = 1/x
 
//      x^((p-5)/8)     = sqrt(1/x) or -sqrt(1/x)
 
//
 
// if   quartic                = -1  (x is a square)
 
// then x^((p-1)/4)            = -1
 
//      x^((p-5)/4) * x        = -1
 
//      x^((p-5)/4)            = -1/x
 
//      x^((p-5)/8)            = sqrt(-1)   / sqrt(x)
 
//      x^((p-5)/8) * sqrt(-1) = sqrt(-1)^2 / sqrt(x)
 
//      x^((p-5)/8) * sqrt(-1) = -1/sqrt(x)
 
//      x^((p-5)/8) * sqrt(-1) = -sqrt(1/x) or sqrt(1/x)
 
//
 
// if   quartic         = sqrt(-1)  (x is not a square)
 
// then x^((p-1)/4)     = sqrt(-1)
 
//      x^((p-5)/4) * x = sqrt(-1)
 
//      x^((p-5)/4)     = sqrt(-1)/x
 
//      x^((p-5)/8)     = sqrt(sqrt(-1)/x) or -sqrt(sqrt(-1)/x)
 
//
 
// Note that the product of two non-squares is always a square:
 
//   For any non-squares a and b, chi(a) = -1 and chi(b) = -1.
 
//   Since chi(x) = x^((p-1)/2), chi(a)*chi(b) = chi(a*b) = 1.
 
//   Therefore a*b is a square.
 
//
 
//   Since sqrt(-1) and x are both non-squares, their product is a
 
//   square, and we can compute their square root.
 
//
 
// if   quartic                = -sqrt(-1)  (x is not a square)
 
// then x^((p-1)/4)            = -sqrt(-1)
 
//      x^((p-5)/4) * x        = -sqrt(-1)
 
//      x^((p-5)/4)            = -sqrt(-1)/x
 
//      x^((p-5)/8)            = sqrt(-sqrt(-1)/x)
 
//      x^((p-5)/8)            = sqrt( sqrt(-1)/x) * sqrt(-1)
 
//      x^((p-5)/8) * sqrt(-1) = sqrt( sqrt(-1)/x) * sqrt(-1)^2
 
//      x^((p-5)/8) * sqrt(-1) = sqrt( sqrt(-1)/x) * -1
 
//      x^((p-5)/8) * sqrt(-1) = -sqrt(sqrt(-1)/x) or sqrt(sqrt(-1)/x)
 
static int invsqrt(fe isr, const fe x)
 
{
 
	fe t0, t1, t2;
 

	
 
	// t0 = x^((p-5)/8)
 
	// Can be achieved with a simple double & add ladder,
 
	// but it would be slower.
 
	fe_sq(t0, x);
 
	fe_sq(t1,t0);                     fe_sq(t1, t1);    fe_mul(t1, x, t1);
 
	fe_mul(t0, t0, t1);
 
	fe_sq(t0, t0);                                      fe_mul(t0, t1, t0);
 
	fe_sq(t1, t0);  FOR (i, 1,   5) { fe_sq(t1, t1); }  fe_mul(t0, t1, t0);
 
	fe_sq(t1, t0);  FOR (i, 1,  10) { fe_sq(t1, t1); }  fe_mul(t1, t1, t0);
 
	fe_sq(t2, t1);  FOR (i, 1,  20) { fe_sq(t2, t2); }  fe_mul(t1, t2, t1);
 
	fe_sq(t1, t1);  FOR (i, 1,  10) { fe_sq(t1, t1); }  fe_mul(t0, t1, t0);
 
	fe_sq(t1, t0);  FOR (i, 1,  50) { fe_sq(t1, t1); }  fe_mul(t1, t1, t0);
 
	fe_sq(t2, t1);  FOR (i, 1, 100) { fe_sq(t2, t2); }  fe_mul(t1, t2, t1);
 
	fe_sq(t1, t1);  FOR (i, 1,  50) { fe_sq(t1, t1); }  fe_mul(t0, t1, t0);
 
	fe_sq(t0, t0);  FOR (i, 1,   2) { fe_sq(t0, t0); }  fe_mul(t0, t0, x);
 

	
 
	// quartic = x^((p-1)/4)
 
	i32 *quartic = t1;
 
	fe_sq (quartic, t0);
 
	fe_mul(quartic, quartic, x);
 

	
 
	i32 *check = t2;
 
	fe_0  (check);          int z0 = fe_isequal(x      , check);
 
	fe_1  (check);          int p1 = fe_isequal(quartic, check);
 
	fe_neg(check, check );  int m1 = fe_isequal(quartic, check);
 
	fe_neg(check, sqrtm1);  int ms = fe_isequal(quartic, check);
 

	
 
	// if quartic == -1 or sqrt(-1)
 
	// then  isr = x^((p-1)/4) * sqrt(-1)
 
	// else  isr = x^((p-1)/4)
 
	fe_mul(isr, t0, sqrtm1);
 
	fe_ccopy(isr, t0, 1 - (m1 | ms));
 

	
 
	WIPE_BUFFER(t0);
 
	WIPE_BUFFER(t1);
 
	WIPE_BUFFER(t2);
 
	return p1 | m1 | z0;
 
}
 

	
 
// Inverse in terms of inverse square root.
 
// Requires two additional squarings to get rid of the sign.
 
//
 
//   1/x = x * (+invsqrt(x^2))^2
 
//       = x * (-invsqrt(x^2))^2
 
//
 
// A fully optimised exponentiation by p-1 would save 6 field
 
// multiplications, but it would require more code.
 
static void fe_invert(fe out, const fe x)
 
{
 
	fe tmp;
 
	fe_sq(tmp, x);
 
	invsqrt(tmp, tmp);
 
	fe_sq(tmp, tmp);
 
	fe_mul(out, tmp, x);
 
	WIPE_BUFFER(tmp);
 
}
 

	
 
// trim a scalar for scalar multiplication
 
void crypto_eddsa_trim_scalar(u8 out[32], const u8 in[32])
 
{
 
	COPY(out, in, 32);
 
	out[ 0] &= 248;
 
	out[31] &= 127;
 
	out[31] |= 64;
 
}
 

	
 
// get bit from scalar at position i
 
static int scalar_bit(const u8 s[32], int i)
 
{
 
	if (i < 0) { return 0; } // handle -1 for sliding windows
 
	return (s[i>>3] >> (i&7)) & 1;
 
}
 

	
 
///////////////
 
/// X-25519 /// Taken from SUPERCOP's ref10 implementation.
 
///////////////
 
static void scalarmult(u8 q[32], const u8 scalar[32], const u8 p[32],
 
                       int nb_bits)
 
{
 
	// computes the scalar product
 
	fe x1;
 
	fe_frombytes(x1, p);
 

	
 
	// computes the actual scalar product (the result is in x2 and z2)
 
	fe x2, z2, x3, z3, t0, t1;
 
	// Montgomery ladder
 
	// In projective coordinates, to avoid divisions: x = X / Z
 
	// We don't care about the y coordinate, it's only 1 bit of information
 
	fe_1(x2);        fe_0(z2); // "zero" point
 
	fe_copy(x3, x1); fe_1(z3); // "one"  point
 
	int swap = 0;
 
	for (int pos = nb_bits-1; pos >= 0; --pos) {
 
		// constant time conditional swap before ladder step
 
		int b = scalar_bit(scalar, pos);
 
		swap ^= b; // xor trick avoids swapping at the end of the loop
 
		fe_cswap(x2, x3, swap);
 
		fe_cswap(z2, z3, swap);
 
		swap = b;  // anticipates one last swap after the loop
 

	
 
		// Montgomery ladder step: replaces (P2, P3) by (P2*2, P2+P3)
 
		// with differential addition
 
		fe_sub(t0, x3, z3);
 
		fe_sub(t1, x2, z2);
 
		fe_add(x2, x2, z2);
 
		fe_add(z2, x3, z3);
 
		fe_mul(z3, t0, x2);
 
		fe_mul(z2, z2, t1);
 
		fe_sq (t0, t1    );
 
		fe_sq (t1, x2    );
 
		fe_add(x3, z3, z2);
 
		fe_sub(z2, z3, z2);
 
		fe_mul(x2, t1, t0);
 
		fe_sub(t1, t1, t0);
 
		fe_sq (z2, z2    );
 
		fe_mul_small(z3, t1, 121666);
 
		fe_sq (x3, x3    );
 
		fe_add(t0, t0, z3);
 
		fe_mul(z3, x1, z2);
 
		fe_mul(z2, t1, t0);
 
	}
 
	// last swap is necessary to compensate for the xor trick
 
	// Note: after this swap, P3 == P2 + P1.
 
	fe_cswap(x2, x3, swap);
 
	fe_cswap(z2, z3, swap);
 

	
 
	// normalises the coordinates: x == X / Z
 
	fe_invert(z2, z2);
 
	fe_mul(x2, x2, z2);
 
	fe_tobytes(q, x2);
 

	
 
	WIPE_BUFFER(x1);
 
	WIPE_BUFFER(x2);  WIPE_BUFFER(z2);  WIPE_BUFFER(t0);
 
	WIPE_BUFFER(x3);  WIPE_BUFFER(z3);  WIPE_BUFFER(t1);
 
}
 

	
 
void crypto_x25519(u8       raw_shared_secret[32],
 
                   const u8 your_secret_key  [32],
 
                   const u8 their_public_key [32])
 
{
 
	// restrict the possible scalar values
 
	u8 e[32];
 
	crypto_eddsa_trim_scalar(e, your_secret_key);
 
	scalarmult(raw_shared_secret, e, their_public_key, 255);
 
	WIPE_BUFFER(e);
 
}
 

	
 
void crypto_x25519_public_key(u8       public_key[32],
 
                              const u8 secret_key[32])
 
{
 
	static const u8 base_point[32] = {9};
 
	crypto_x25519(public_key, secret_key, base_point);
 
}
 

	
 
///////////////////////////
 
/// Arithmetic modulo L ///
 
///////////////////////////
 
static const u32 L[8] = {
 
	0x5cf5d3ed, 0x5812631a, 0xa2f79cd6, 0x14def9de,
 
	0x00000000, 0x00000000, 0x00000000, 0x10000000,
 
};
 

	
 
//  p = a*b + p
 
static void multiply(u32 p[16], const u32 a[8], const u32 b[8])
 
{
 
	FOR (i, 0, 8) {
 
		u64 carry = 0;
 
		FOR (j, 0, 8) {
 
			carry  += p[i+j] + (u64)a[i] * b[j];
 
			p[i+j]  = (u32)carry;
 
			carry >>= 32;
 
		}
 
		p[i+8] = (u32)carry;
 
	}
 
}
 

	
 
static int is_above_l(const u32 x[8])
 
{
 
	// We work with L directly, in a 2's complement encoding
 
	// (-L == ~L + 1)
 
	u64 carry = 1;
 
	FOR (i, 0, 8) {
 
		carry  += (u64)x[i] + (~L[i] & 0xffffffff);
 
		carry >>= 32;
 
	}
 
	return (int)carry; // carry is either 0 or 1
 
}
 

	
 
// Final reduction modulo L, by conditionally removing L.
 
// if x < l     , then r = x
 
// if l <= x 2*l, then r = x-l
 
// otherwise the result will be wrong
 
static void remove_l(u32 r[8], const u32 x[8])
 
{
 
	u64 carry = (u64)is_above_l(x);
 
	u32 mask  = ~(u32)carry + 1; // carry == 0 or 1
 
	FOR (i, 0, 8) {
 
		carry += (u64)x[i] + (~L[i] & mask);
 
		r[i]   = (u32)carry;
 
		carry >>= 32;
 
	}
 
}
 

	
 
// Full reduction modulo L (Barrett reduction)
 
static void mod_l(u8 reduced[32], const u32 x[16])
 
{
 
	static const u32 r[9] = {
 
		0x0a2c131b,0xed9ce5a3,0x086329a7,0x2106215d,
 
		0xffffffeb,0xffffffff,0xffffffff,0xffffffff,0xf,
 
	};
 
	// xr = x * r
 
	u32 xr[25] = {0};
 
	FOR (i, 0, 9) {
 
		u64 carry = 0;
 
		FOR (j, 0, 16) {
 
			carry  += xr[i+j] + (u64)r[i] * x[j];
 
			xr[i+j] = (u32)carry;
 
			carry >>= 32;
 
		}
 
		xr[i+16] = (u32)carry;
 
	}
 
	// xr = floor(xr / 2^512) * L
 
	// Since the result is guaranteed to be below 2*L,
 
	// it is enough to only compute the first 256 bits.
 
	// The division is performed by saying xr[i+16]. (16 * 32 = 512)
 
	ZERO(xr, 8);
 
	FOR (i, 0, 8) {
 
		u64 carry = 0;
 
		FOR (j, 0, 8-i) {
 
			carry   += xr[i+j] + (u64)xr[i+16] * L[j];
 
			xr[i+j] = (u32)carry;
 
			carry >>= 32;
 
		}
 
	}
 
	// xr = x - xr
 
	u64 carry = 1;
 
	FOR (i, 0, 8) {
 
		carry  += (u64)x[i] + (~xr[i] & 0xffffffff);
 
		xr[i]   = (u32)carry;
 
		carry >>= 32;
 
	}
 
	// Final reduction modulo L (conditional subtraction)
 
	remove_l(xr, xr);
 
	store32_le_buf(reduced, xr, 8);
 

	
 
	WIPE_BUFFER(xr);
 
}
 

	
 
void crypto_eddsa_reduce(u8 reduced[32], const u8 expanded[64])
 
{
 
	u32 x[16];
 
	load32_le_buf(x, expanded, 16);
 
	mod_l(reduced, x);
 
	WIPE_BUFFER(x);
 
}
 

	
 
// r = (a * b) + c
 
void crypto_eddsa_mul_add(u8 r[32],
 
                          const u8 a[32], const u8 b[32], const u8 c[32])
 
{
 
	u32 A[8];  load32_le_buf(A, a, 8);
 
	u32 B[8];  load32_le_buf(B, b, 8);
 
	u32 p[16]; load32_le_buf(p, c, 8);  ZERO(p + 8, 8);
 
	multiply(p, A, B);
 
	mod_l(r, p);
 
	WIPE_BUFFER(p);
 
	WIPE_BUFFER(A);
 
	WIPE_BUFFER(B);
 
}
 

	
 
///////////////
 
/// Ed25519 ///
 
///////////////
 

	
 
// Point (group element, ge) in a twisted Edwards curve,
 
// in extended projective coordinates.
 
// ge        : x  = X/Z, y  = Y/Z, T  = XY/Z
 
// ge_cached : Yp = X+Y, Ym = X-Y, T2 = T*D2
 
// ge_precomp: Z  = 1
 
typedef struct { fe X;  fe Y;  fe Z; fe T;  } ge;
 
typedef struct { fe Yp; fe Ym; fe Z; fe T2; } ge_cached;
 
typedef struct { fe Yp; fe Ym;       fe T2; } ge_precomp;
 

	
 
static void ge_zero(ge *p)
 
{
 
	fe_0(p->X);
 
	fe_1(p->Y);
 
	fe_1(p->Z);
 
	fe_0(p->T);
 
}
 

	
 
static void ge_tobytes(u8 s[32], const ge *h)
 
{
 
	fe recip, x, y;
 
	fe_invert(recip, h->Z);
 
	fe_mul(x, h->X, recip);
 
	fe_mul(y, h->Y, recip);
 
	fe_tobytes(s, y);
 
	s[31] ^= fe_isodd(x) << 7;
 

	
 
	WIPE_BUFFER(recip);
 
	WIPE_BUFFER(x);
 
	WIPE_BUFFER(y);
 
}
 

	
 
// h = -s, where s is a point encoded in 32 bytes
 
//
 
// Variable time!  Inputs must not be secret!
 
// => Use only to *check* signatures.
 
//
 
// From the specifications:
 
//   The encoding of s contains y and the sign of x
 
//   x = sqrt((y^2 - 1) / (d*y^2 + 1))
 
// In extended coordinates:
 
//   X = x, Y = y, Z = 1, T = x*y
 
//
 
//    Note that num * den is a square iff num / den is a square
 
//    If num * den is not a square, the point was not on the curve.
 
// From the above:
 
//   Let num =   y^2 - 1
 
//   Let den = d*y^2 + 1
 
//   x = sqrt((y^2 - 1) / (d*y^2 + 1))
 
//   x = sqrt(num / den)
 
//   x = sqrt(num^2 / (num * den))
 
//   x = num * sqrt(1 / (num * den))
 
//
 
// Therefore, we can just compute:
 
//   num =   y^2 - 1
 
//   den = d*y^2 + 1
 
//   isr = invsqrt(num * den)  // abort if not square
 
//   x   = num * isr
 
// Finally, negate x if its sign is not as specified.
 
static int ge_frombytes_neg_vartime(ge *h, const u8 s[32])
 
{
 
	fe_frombytes(h->Y, s);
 
	fe_1(h->Z);
 
	fe_sq (h->T, h->Y);        // t =   y^2
 
	fe_mul(h->X, h->T, d   );  // x = d*y^2
 
	fe_sub(h->T, h->T, h->Z);  // t =   y^2 - 1
 
	fe_add(h->X, h->X, h->Z);  // x = d*y^2 + 1
 
	fe_mul(h->X, h->T, h->X);  // x = (y^2 - 1) * (d*y^2 + 1)
 
	int is_square = invsqrt(h->X, h->X);
 
	if (!is_square) {
 
		return -1;             // Not on the curve, abort
 
	}
 
	fe_mul(h->X, h->T, h->X);  // x = sqrt((y^2 - 1) / (d*y^2 + 1))
 
	if (fe_isodd(h->X) == (s[31] >> 7)) {
 
		fe_neg(h->X, h->X);
 
	}
 
	fe_mul(h->T, h->X, h->Y);
 
	return 0;
 
}
 

	
 
static void ge_cache(ge_cached *c, const ge *p)
 
{
 
	fe_add (c->Yp, p->Y, p->X);
 
	fe_sub (c->Ym, p->Y, p->X);
 
	fe_copy(c->Z , p->Z      );
 
	fe_mul (c->T2, p->T, D2  );
 
}
 

	
 
// Internal buffers are not wiped! Inputs must not be secret!
 
// => Use only to *check* signatures.
 
static void ge_add(ge *s, const ge *p, const ge_cached *q)
 
{
 
	fe a, b;
 
	fe_add(a   , p->Y, p->X );
 
	fe_sub(b   , p->Y, p->X );
 
	fe_mul(a   , a   , q->Yp);
 
	fe_mul(b   , b   , q->Ym);
 
	fe_add(s->Y, a   , b    );
 
	fe_sub(s->X, a   , b    );
 

	
 
	fe_add(s->Z, p->Z, p->Z );
 
	fe_mul(s->Z, s->Z, q->Z );
 
	fe_mul(s->T, p->T, q->T2);
 
	fe_add(a   , s->Z, s->T );
 
	fe_sub(b   , s->Z, s->T );
 

	
 
	fe_mul(s->T, s->X, s->Y);
 
	fe_mul(s->X, s->X, b   );
 
	fe_mul(s->Y, s->Y, a   );
 
	fe_mul(s->Z, a   , b   );
 
}
 

	
 
// Internal buffers are not wiped! Inputs must not be secret!
 
// => Use only to *check* signatures.
 
static void ge_sub(ge *s, const ge *p, const ge_cached *q)
 
{
 
	ge_cached neg;
 
	fe_copy(neg.Ym, q->Yp);
 
	fe_copy(neg.Yp, q->Ym);
 
	fe_copy(neg.Z , q->Z );
 
	fe_neg (neg.T2, q->T2);
 
	ge_add(s, p, &neg);
 
}
 

	
 
static void ge_madd(ge *s, const ge *p, const ge_precomp *q, fe a, fe b)
 
{
 
	fe_add(a   , p->Y, p->X );
 
	fe_sub(b   , p->Y, p->X );
 
	fe_mul(a   , a   , q->Yp);
 
	fe_mul(b   , b   , q->Ym);
 
	fe_add(s->Y, a   , b    );
 
	fe_sub(s->X, a   , b    );
 

	
 
	fe_add(s->Z, p->Z, p->Z );
 
	fe_mul(s->T, p->T, q->T2);
 
	fe_add(a   , s->Z, s->T );
 
	fe_sub(b   , s->Z, s->T );
 

	
 
	fe_mul(s->T, s->X, s->Y);
 
	fe_mul(s->X, s->X, b   );
 
	fe_mul(s->Y, s->Y, a   );
 
	fe_mul(s->Z, a   , b   );
 
}
 

	
 
// Internal buffers are not wiped! Inputs must not be secret!
 
// => Use only to *check* signatures.
 
static void ge_msub(ge *s, const ge *p, const ge_precomp *q, fe a, fe b)
 
{
 
	ge_precomp neg;
 
	fe_copy(neg.Ym, q->Yp);
 
	fe_copy(neg.Yp, q->Ym);
 
	fe_neg (neg.T2, q->T2);
 
	ge_madd(s, p, &neg, a, b);
 
}
 

	
 
static void ge_double(ge *s, const ge *p, ge *q)
 
{
 
	fe_sq (q->X, p->X);
 
	fe_sq (q->Y, p->Y);
 
	fe_sq (q->Z, p->Z);          // qZ = pZ^2
 
	fe_mul_small(q->Z, q->Z, 2); // qZ = pZ^2 * 2
 
	fe_add(q->T, p->X, p->Y);
 
	fe_sq (s->T, q->T);
 
	fe_add(q->T, q->Y, q->X);
 
	fe_sub(q->Y, q->Y, q->X);
 
	fe_sub(q->X, s->T, q->T);
 
	fe_sub(q->Z, q->Z, q->Y);
 

	
 
	fe_mul(s->X, q->X , q->Z);
 
	fe_mul(s->Y, q->T , q->Y);
 
	fe_mul(s->Z, q->Y , q->Z);
 
	fe_mul(s->T, q->X , q->T);
 
}
 

	
 
// 5-bit signed window in cached format (Niels coordinates, Z=1)
 
static const ge_precomp b_window[8] = {
 
	{{25967493,-14356035,29566456,3660896,-12694345,
 
	  4014787,27544626,-11754271,-6079156,2047605,},
 
	 {-12545711,934262,-2722910,3049990,-727428,
 
	  9406986,12720692,5043384,19500929,-15469378,},
 
	 {-8738181,4489570,9688441,-14785194,10184609,
 
	  -12363380,29287919,11864899,-24514362,-4438546,},},
 
	{{15636291,-9688557,24204773,-7912398,616977,
 
	  -16685262,27787600,-14772189,28944400,-1550024,},
 
	 {16568933,4717097,-11556148,-1102322,15682896,
 
	  -11807043,16354577,-11775962,7689662,11199574,},
 
	 {30464156,-5976125,-11779434,-15670865,23220365,
 
	  15915852,7512774,10017326,-17749093,-9920357,},},
 
	{{10861363,11473154,27284546,1981175,-30064349,
 
	  12577861,32867885,14515107,-15438304,10819380,},
 
	 {4708026,6336745,20377586,9066809,-11272109,
 
	  6594696,-25653668,12483688,-12668491,5581306,},
 
	 {19563160,16186464,-29386857,4097519,10237984,
 
	  -4348115,28542350,13850243,-23678021,-15815942,},},
 
	{{5153746,9909285,1723747,-2777874,30523605,
 
	  5516873,19480852,5230134,-23952439,-15175766,},
 
	 {-30269007,-3463509,7665486,10083793,28475525,
 
	  1649722,20654025,16520125,30598449,7715701,},
 
	 {28881845,14381568,9657904,3680757,-20181635,
 
	  7843316,-31400660,1370708,29794553,-1409300,},},
 
	{{-22518993,-6692182,14201702,-8745502,-23510406,
 
	  8844726,18474211,-1361450,-13062696,13821877,},
 
	 {-6455177,-7839871,3374702,-4740862,-27098617,
 
	  -10571707,31655028,-7212327,18853322,-14220951,},
 
	 {4566830,-12963868,-28974889,-12240689,-7602672,
 
	  -2830569,-8514358,-10431137,2207753,-3209784,},},
 
	{{-25154831,-4185821,29681144,7868801,-6854661,
 
	  -9423865,-12437364,-663000,-31111463,-16132436,},
 
	 {25576264,-2703214,7349804,-11814844,16472782,
 
	  9300885,3844789,15725684,171356,6466918,},
 
	 {23103977,13316479,9739013,-16149481,817875,
 
	  -15038942,8965339,-14088058,-30714912,16193877,},},
 
	{{-33521811,3180713,-2394130,14003687,-16903474,
 
	  -16270840,17238398,4729455,-18074513,9256800,},
 
	 {-25182317,-4174131,32336398,5036987,-21236817,
 
	  11360617,22616405,9761698,-19827198,630305,},
 
	 {-13720693,2639453,-24237460,-7406481,9494427,
 
	  -5774029,-6554551,-15960994,-2449256,-14291300,},},
 
	{{-3151181,-5046075,9282714,6866145,-31907062,
 
	  -863023,-18940575,15033784,25105118,-7894876,},
 
	 {-24326370,15950226,-31801215,-14592823,-11662737,
 
	  -5090925,1573892,-2625887,2198790,-15804619,},
 
	 {-3099351,10324967,-2241613,7453183,-5446979,
 
	  -2735503,-13812022,-16236442,-32461234,-12290683,},},
 
};
 

	
 
// Incremental sliding windows (left to right)
 
// Based on Roberto Maria Avanzi[2005]
 
typedef struct {
 
	i16 next_index; // position of the next signed digit
 
	i8  next_digit; // next signed digit (odd number below 2^window_width)
 
	u8  next_check; // point at which we must check for a new window
 
} slide_ctx;
 

	
 
static void slide_init(slide_ctx *ctx, const u8 scalar[32])
 
{
 
	// scalar is guaranteed to be below L, either because we checked (s),
 
	// or because we reduced it modulo L (h_ram). L is under 2^253, so
 
	// so bits 253 to 255 are guaranteed to be zero. No need to test them.
 
	//
 
	// Note however that L is very close to 2^252, so bit 252 is almost
 
	// always zero.  If we were to start at bit 251, the tests wouldn't
 
	// catch the off-by-one error (constructing one that does would be
 
	// prohibitively expensive).
 
	//
 
	// We should still check bit 252, though.
 
	int i = 252;
 
	while (i > 0 && scalar_bit(scalar, i) == 0) {
 
		i--;
 
	}
 
	ctx->next_check = (u8)(i + 1);
 
	ctx->next_index = -1;
 
	ctx->next_digit = -1;
 
}
 

	
 
static int slide_step(slide_ctx *ctx, int width, int i, const u8 scalar[32])
 
{
 
	if (i == ctx->next_check) {
 
		if (scalar_bit(scalar, i) == scalar_bit(scalar, i - 1)) {
 
			ctx->next_check--;
 
		} else {
 
			// compute digit of next window
 
			int w = MIN(width, i + 1);
 
			int v = -(scalar_bit(scalar, i) << (w-1));
 
			FOR_T (int, j, 0, w-1) {
 
				v += scalar_bit(scalar, i-(w-1)+j) << j;
 
			}
 
			v += scalar_bit(scalar, i-w);
 
			int lsb = v & (~v + 1); // smallest bit of v
 
			int s   =               // log2(lsb)
 
				(((lsb & 0xAA) != 0) << 0) |
 
				(((lsb & 0xCC) != 0) << 1) |
 
				(((lsb & 0xF0) != 0) << 2);
 
			ctx->next_index  = (i16)(i-(w-1)+s);
 
			ctx->next_digit  = (i8) (v >> s   );
 
			ctx->next_check -= (u8) w;
 
		}
 
	}
 
	return i == ctx->next_index ? ctx->next_digit: 0;
 
}
 

	
 
#define P_W_WIDTH 3 // Affects the size of the stack
 
#define B_W_WIDTH 5 // Affects the size of the binary
 
#define P_W_SIZE  (1<<(P_W_WIDTH-2))
 

	
 
int crypto_eddsa_check_equation(const u8 signature[64], const u8 public_key[32],
 
                                const u8 h[32])
 
{
 
	ge minus_A; // -public_key
 
	ge minus_R; // -first_half_of_signature
 
	const u8 *s = signature + 32;
 

	
 
	// Check that A and R are on the curve
 
	// Check that 0 <= S < L (prevents malleability)
 
	// *Allow* non-cannonical encoding for A and R
 
	{
 
		u32 s32[8];
 
		load32_le_buf(s32, s, 8);
 
		if (ge_frombytes_neg_vartime(&minus_A, public_key) ||
 
		    ge_frombytes_neg_vartime(&minus_R, signature)  ||
 
		    is_above_l(s32)) {
 
			return -1;
 
		}
 
	}
 

	
 
	// look-up table for minus_A
 
	ge_cached lutA[P_W_SIZE];
 
	{
 
		ge minus_A2, tmp;
 
		ge_double(&minus_A2, &minus_A, &tmp);
 
		ge_cache(&lutA[0], &minus_A);
 
		FOR (i, 1, P_W_SIZE) {
 
			ge_add(&tmp, &minus_A2, &lutA[i-1]);
 
			ge_cache(&lutA[i], &tmp);
 
		}
 
	}
 

	
 
	// sum = [s]B - [h]A
 
	// Merged double and add ladder, fused with sliding
 
	slide_ctx h_slide;  slide_init(&h_slide, h);
 
	slide_ctx s_slide;  slide_init(&s_slide, s);
 
	int i = MAX(h_slide.next_check, s_slide.next_check);
 
	ge *sum = &minus_A; // reuse minus_A for the sum
 
	ge_zero(sum);
 
	while (i >= 0) {
 
		ge tmp;
 
		ge_double(sum, sum, &tmp);
 
		int h_digit = slide_step(&h_slide, P_W_WIDTH, i, h);
 
		int s_digit = slide_step(&s_slide, B_W_WIDTH, i, s);
 
		if (h_digit > 0) { ge_add(sum, sum, &lutA[ h_digit / 2]); }
 
		if (h_digit < 0) { ge_sub(sum, sum, &lutA[-h_digit / 2]); }
 
		fe t1, t2;
 
		if (s_digit > 0) { ge_madd(sum, sum, b_window +  s_digit/2, t1, t2); }
 
		if (s_digit < 0) { ge_msub(sum, sum, b_window + -s_digit/2, t1, t2); }
 
		i--;
 
	}
 

	
 
	// Compare [8](sum-R) and the zero point
 
	// The multiplication by 8 eliminates any low-order component
 
	// and ensures consistency with batched verification.
 
	ge_cached cached;
 
	u8 check[32];
 
	static const u8 zero_point[32] = {1}; // Point of order 1
 
	ge_cache(&cached, &minus_R);
 
	ge_add(sum, sum, &cached);
 
	ge_double(sum, sum, &minus_R); // reuse minus_R as temporary
 
	ge_double(sum, sum, &minus_R); // reuse minus_R as temporary
 
	ge_double(sum, sum, &minus_R); // reuse minus_R as temporary
 
	ge_tobytes(check, sum);
 
	return crypto_verify32(check, zero_point);
 
}
 

	
 
// 5-bit signed comb in cached format (Niels coordinates, Z=1)
 
static const ge_precomp b_comb_low[8] = {
 
	{{-6816601,-2324159,-22559413,124364,18015490,
 
	  8373481,19993724,1979872,-18549925,9085059,},
 
	 {10306321,403248,14839893,9633706,8463310,
 
	  -8354981,-14305673,14668847,26301366,2818560,},
 
	 {-22701500,-3210264,-13831292,-2927732,-16326337,
 
	  -14016360,12940910,177905,12165515,-2397893,},},
 
	{{-12282262,-7022066,9920413,-3064358,-32147467,
 
	  2927790,22392436,-14852487,2719975,16402117,},
 
	 {-7236961,-4729776,2685954,-6525055,-24242706,
 
	  -15940211,-6238521,14082855,10047669,12228189,},
 
	 {-30495588,-12893761,-11161261,3539405,-11502464,
 
	  16491580,-27286798,-15030530,-7272871,-15934455,},},
 
	{{17650926,582297,-860412,-187745,-12072900,
 
	  -10683391,-20352381,15557840,-31072141,-5019061,},
 
	 {-6283632,-2259834,-4674247,-4598977,-4089240,
 
	  12435688,-31278303,1060251,6256175,10480726,},
 
	 {-13871026,2026300,-21928428,-2741605,-2406664,
 
	  -8034988,7355518,15733500,-23379862,7489131,},},
 
	{{6883359,695140,23196907,9644202,-33430614,
 
	  11354760,-20134606,6388313,-8263585,-8491918,},
 
	 {-7716174,-13605463,-13646110,14757414,-19430591,
 
	  -14967316,10359532,-11059670,-21935259,12082603,},
 
	 {-11253345,-15943946,10046784,5414629,24840771,
 
	  8086951,-6694742,9868723,15842692,-16224787,},},
 
	{{9639399,11810955,-24007778,-9320054,3912937,
 
	  -9856959,996125,-8727907,-8919186,-14097242,},
 
	 {7248867,14468564,25228636,-8795035,14346339,
 
	  8224790,6388427,-7181107,6468218,-8720783,},
 
	 {15513115,15439095,7342322,-10157390,18005294,
 
	  -7265713,2186239,4884640,10826567,7135781,},},
 
	{{-14204238,5297536,-5862318,-6004934,28095835,
 
	  4236101,-14203318,1958636,-16816875,3837147,},
 
	 {-5511166,-13176782,-29588215,12339465,15325758,
 
	  -15945770,-8813185,11075932,-19608050,-3776283,},
 
	 {11728032,9603156,-4637821,-5304487,-7827751,
 
	  2724948,31236191,-16760175,-7268616,14799772,},},
 
	{{-28842672,4840636,-12047946,-9101456,-1445464,
 
	  381905,-30977094,-16523389,1290540,12798615,},
 
	 {27246947,-10320914,14792098,-14518944,5302070,
 
	  -8746152,-3403974,-4149637,-27061213,10749585,},
 
	 {25572375,-6270368,-15353037,16037944,1146292,
 
	  32198,23487090,9585613,24714571,-1418265,},},
 
	{{19844825,282124,-17583147,11004019,-32004269,
 
	  -2716035,6105106,-1711007,-21010044,14338445,},
 
	 {8027505,8191102,-18504907,-12335737,25173494,
 
	  -5923905,15446145,7483684,-30440441,10009108,},
 
	 {-14134701,-4174411,10246585,-14677495,33553567,
 
	  -14012935,23366126,15080531,-7969992,7663473,},},
 
};
 

	
 
static const ge_precomp b_comb_high[8] = {
 
	{{33055887,-4431773,-521787,6654165,951411,
 
	  -6266464,-5158124,6995613,-5397442,-6985227,},
 
	 {4014062,6967095,-11977872,3960002,8001989,
 
	  5130302,-2154812,-1899602,-31954493,-16173976,},
 
	 {16271757,-9212948,23792794,731486,-25808309,
 
	  -3546396,6964344,-4767590,10976593,10050757,},},
 
	{{2533007,-4288439,-24467768,-12387405,-13450051,
 
	  14542280,12876301,13893535,15067764,8594792,},
 
	 {20073501,-11623621,3165391,-13119866,13188608,
 
	  -11540496,-10751437,-13482671,29588810,2197295,},
 
	 {-1084082,11831693,6031797,14062724,14748428,
 
	  -8159962,-20721760,11742548,31368706,13161200,},},
 
	{{2050412,-6457589,15321215,5273360,25484180,
 
	  124590,-18187548,-7097255,-6691621,-14604792,},
 
	 {9938196,2162889,-6158074,-1711248,4278932,
 
	  -2598531,-22865792,-7168500,-24323168,11746309,},
 
	 {-22691768,-14268164,5965485,9383325,20443693,
 
	  5854192,28250679,-1381811,-10837134,13717818,},},
 
	{{-8495530,16382250,9548884,-4971523,-4491811,
 
	  -3902147,6182256,-12832479,26628081,10395408,},
 
	 {27329048,-15853735,7715764,8717446,-9215518,
 
	  -14633480,28982250,-5668414,4227628,242148,},
 
	 {-13279943,-7986904,-7100016,8764468,-27276630,
 
	  3096719,29678419,-9141299,3906709,11265498,},},
 
	{{11918285,15686328,-17757323,-11217300,-27548967,
 
	  4853165,-27168827,6807359,6871949,-1075745,},
 
	 {-29002610,13984323,-27111812,-2713442,28107359,
 
	  -13266203,6155126,15104658,3538727,-7513788,},
 
	 {14103158,11233913,-33165269,9279850,31014152,
 
	  4335090,-1827936,4590951,13960841,12787712,},},
 
	{{1469134,-16738009,33411928,13942824,8092558,
 
	  -8778224,-11165065,1437842,22521552,-2792954,},
 
	 {31352705,-4807352,-25327300,3962447,12541566,
 
	  -9399651,-27425693,7964818,-23829869,5541287,},
 
	 {-25732021,-6864887,23848984,3039395,-9147354,
 
	  6022816,-27421653,10590137,25309915,-1584678,},},
 
	{{-22951376,5048948,31139401,-190316,-19542447,
 
	  -626310,-17486305,-16511925,-18851313,-12985140,},
 
	 {-9684890,14681754,30487568,7717771,-10829709,
 
	  9630497,30290549,-10531496,-27798994,-13812825,},
 
	 {5827835,16097107,-24501327,12094619,7413972,
 
	  11447087,28057551,-1793987,-14056981,4359312,},},
 
	{{26323183,2342588,-21887793,-1623758,-6062284,
 
	  2107090,-28724907,9036464,-19618351,-13055189,},
 
	 {-29697200,14829398,-4596333,14220089,-30022969,
 
	  2955645,12094100,-13693652,-5941445,7047569,},
 
	 {-3201977,14413268,-12058324,-16417589,-9035655,
 
	  -7224648,9258160,1399236,30397584,-5684634,},},
 
};
 

	
 
static void lookup_add(ge *p, ge_precomp *tmp_c, fe tmp_a, fe tmp_b,
 
                       const ge_precomp comb[8], const u8 scalar[32], int i)
 
{
 
	u8 teeth = (u8)((scalar_bit(scalar, i)          ) +
 
	                (scalar_bit(scalar, i + 32) << 1) +
 
	                (scalar_bit(scalar, i + 64) << 2) +
 
	                (scalar_bit(scalar, i + 96) << 3));
 
	u8 high  = teeth >> 3;
 
	u8 index = (teeth ^ (high - 1)) & 7;
 
	FOR (j, 0, 8) {
 
		i32 select = 1 & (((j ^ index) - 1) >> 8);
 
		fe_ccopy(tmp_c->Yp, comb[j].Yp, select);
 
		fe_ccopy(tmp_c->Ym, comb[j].Ym, select);
 
		fe_ccopy(tmp_c->T2, comb[j].T2, select);
 
	}
 
	fe_neg(tmp_a, tmp_c->T2);
 
	fe_cswap(tmp_c->T2, tmp_a    , high ^ 1);
 
	fe_cswap(tmp_c->Yp, tmp_c->Ym, high ^ 1);
 
	ge_madd(p, p, tmp_c, tmp_a, tmp_b);
 
}
 

	
 
// p = [scalar]B, where B is the base point
 
static void ge_scalarmult_base(ge *p, const u8 scalar[32])
 
{
 
	// twin 4-bits signed combs, from Mike Hamburg's
 
	// Fast and compact elliptic-curve cryptography (2012)
 
	// 1 / 2 modulo L
 
	static const u8 half_mod_L[32] = {
 
		247,233,122,46,141,49,9,44,107,206,123,81,239,124,111,10,
 
		0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,8,
 
	};
 
	// (2^256 - 1) / 2 modulo L
 
	static const u8 half_ones[32] = {
 
		142,74,204,70,186,24,118,107,184,231,190,57,250,173,119,99,
 
		255,255,255,255,255,255,255,255,255,255,255,255,255,255,255,7,
 
	};
 

	
 
	// All bits set form: 1 means 1, 0 means -1
 
	u8 s_scalar[32];
 
	crypto_eddsa_mul_add(s_scalar, scalar, half_mod_L, half_ones);
 

	
 
	// Double and add ladder
 
	fe tmp_a, tmp_b;  // temporaries for addition
 
	ge_precomp tmp_c; // temporary for comb lookup
 
	ge tmp_d;         // temporary for doubling
 
	fe_1(tmp_c.Yp);
 
	fe_1(tmp_c.Ym);
 
	fe_0(tmp_c.T2);
 

	
 
	// Save a double on the first iteration
 
	ge_zero(p);
 
	lookup_add(p, &tmp_c, tmp_a, tmp_b, b_comb_low , s_scalar, 31);
 
	lookup_add(p, &tmp_c, tmp_a, tmp_b, b_comb_high, s_scalar, 31+128);
 
	// Regular double & add for the rest
 
	for (int i = 30; i >= 0; i--) {
 
		ge_double(p, p, &tmp_d);
 
		lookup_add(p, &tmp_c, tmp_a, tmp_b, b_comb_low , s_scalar, i);
 
		lookup_add(p, &tmp_c, tmp_a, tmp_b, b_comb_high, s_scalar, i+128);
 
	}
 
	// Note: we could save one addition at the end if we assumed the
 
	// scalar fit in 252 bits.  Which it does in practice if it is
 
	// selected at random.  However, non-random, non-hashed scalars
 
	// *can* overflow 252 bits in practice.  Better account for that
 
	// than leaving that kind of subtle corner case.
 

	
 
	WIPE_BUFFER(tmp_a);  WIPE_CTX(&tmp_d);
 
	WIPE_BUFFER(tmp_b);  WIPE_CTX(&tmp_c);
 
	WIPE_BUFFER(s_scalar);
 
}
 

	
 
void crypto_eddsa_scalarbase(u8 point[32], const u8 scalar[32])
 
{
 
	ge P;
 
	ge_scalarmult_base(&P, scalar);
 
	ge_tobytes(point, &P);
 
	WIPE_CTX(&P);
 
}
 

	
 
void crypto_eddsa_key_pair(u8 secret_key[64], u8 public_key[32], u8 seed[32])
 
{
 
	// To allow overlaps, observable writes happen in this order:
 
	// 1. seed
 
	// 2. secret_key
 
	// 3. public_key
 
	u8 a[64];
 
	COPY(a, seed, 32);
 
	crypto_wipe(seed, 32);
 
	COPY(secret_key, a, 32);
 
	crypto_blake2b(a, 64, a, 32);
 
	crypto_eddsa_trim_scalar(a, a);
 
	crypto_eddsa_scalarbase(secret_key + 32, a);
 
	COPY(public_key, secret_key + 32, 32);
 
	WIPE_BUFFER(a);
 
}
 

	
 
static void hash_reduce(u8 h[32],
 
                        const u8 *a, size_t a_size,
 
                        const u8 *b, size_t b_size,
 
                        const u8 *c, size_t c_size)
 
{
 
	u8 hash[64];
 
	crypto_blake2b_ctx ctx;
 
	crypto_blake2b_init  (&ctx, 64);
 
	crypto_blake2b_update(&ctx, a, a_size);
 
	crypto_blake2b_update(&ctx, b, b_size);
 
	crypto_blake2b_update(&ctx, c, c_size);
 
	crypto_blake2b_final (&ctx, hash);
 
	crypto_eddsa_reduce(h, hash);
 
}
 

	
 
// Digital signature of a message with from a secret key.
 
//
 
// The secret key comprises two parts:
 
// - The seed that generates the key (secret_key[ 0..31])
 
// - The public key                  (secret_key[32..63])
 
//
 
// The seed and the public key are bundled together to make sure users
 
// don't use mismatched seeds and public keys, which would instantly
 
// leak the secret scalar and allow forgeries (allowing this to happen
 
// has resulted in critical vulnerabilities in the wild).
 
//
 
// The seed is hashed to derive the secret scalar and a secret prefix.
 
// The sole purpose of the prefix is to generate a secret random nonce.
 
// The properties of that nonce must be as follows:
 
// - Unique: we need a different one for each message.
 
// - Secret: third parties must not be able to predict it.
 
// - Random: any detectable bias would break all security.
 
//
 
// There are two ways to achieve these properties.  The obvious one is
 
// to simply generate a random number.  Here that would be a parameter
 
// (Monocypher doesn't have an RNG).  It works, but then users may reuse
 
// the nonce by accident, which _also_ leaks the secret scalar and
 
// allows forgeries.  This has happened in the wild too.
 
//
 
// This is no good, so instead we generate that nonce deterministically
 
// by reducing modulo L a hash of the secret prefix and the message.
 
// The secret prefix makes the nonce unpredictable, the message makes it
 
// unique, and the hash/reduce removes all bias.
 
//
 
// The cost of that safety is hashing the message twice.  If that cost
 
// is unacceptable, there are two alternatives:
 
//
 
// - Signing a hash of the message instead of the message itself.  This
 
//   is fine as long as the hash is collision resistant. It is not
 
//   compatible with existing "pure" signatures, but at least it's safe.
 
//
 
// - Using a random nonce.  Please exercise **EXTREME CAUTION** if you
 
//   ever do that.  It is absolutely **critical** that the nonce is
 
//   really an unbiased random number between 0 and L-1, never reused,
 
//   and wiped immediately.
 
//
 
//   To lower the likelihood of complete catastrophe if the RNG is
 
//   either flawed or misused, you can hash the RNG output together with
 
//   the secret prefix and the beginning of the message, and use the
 
//   reduction of that hash instead of the RNG output itself.  It's not
 
//   foolproof (you'd need to hash the whole message) but it helps.
 
//
 
// Signing a message involves the following operations:
 
//
 
//   scalar, prefix = HASH(secret_key)
 
//   r              = HASH(prefix || message) % L
 
//   R              = [r]B
 
//   h              = HASH(R || public_key || message) % L
 
//   S              = ((h * a) + r) % L
 
//   signature      = R || S
 
void crypto_eddsa_sign(u8 signature [64], const u8 secret_key[64],
 
                       const u8 *message, size_t message_size)
 
{
 
	u8 a[64];  // secret scalar and prefix
 
	u8 r[32];  // secret deterministic "random" nonce
 
	u8 h[32];  // publically verifiable hash of the message (not wiped)
 
	u8 R[32];  // first half of the signature (allows overlapping inputs)
 

	
 
	crypto_blake2b(a, 64, secret_key, 32);
 
	crypto_eddsa_trim_scalar(a, a);
 
	hash_reduce(r, a + 32, 32, message, message_size, 0, 0);
 
	crypto_eddsa_scalarbase(R, r);
 
	hash_reduce(h, R, 32, secret_key + 32, 32, message, message_size);
 
	COPY(signature, R, 32);
 
	crypto_eddsa_mul_add(signature + 32, h, a, r);
 

	
 
	WIPE_BUFFER(a);
 
	WIPE_BUFFER(r);
 
}
 

	
 
// To check the signature R, S of the message M with the public key A,
 
// there are 3 steps:
 
//
 
//   compute h = HASH(R || A || message) % L
 
//   check that A is on the curve.
 
//   check that R == [s]B - [h]A
 
//
 
// The last two steps are done in crypto_eddsa_check_equation()
 
int crypto_eddsa_check(const u8  signature[64], const u8 public_key[32],
 
                       const u8 *message, size_t message_size)
 
{
 
	u8 h[32];
 
	hash_reduce(h, signature, 32, public_key, 32, message, message_size);
 
	return crypto_eddsa_check_equation(signature, public_key, h);
 
}
 

	
 
/////////////////////////
 
/// EdDSA <--> X25519 ///
 
/////////////////////////
 
void crypto_eddsa_to_x25519(u8 x25519[32], const u8 eddsa[32])
 
{
 
	// (u, v) = ((1+y)/(1-y), sqrt(-486664)*u/x)
 
	// Only converting y to u, the sign of x is ignored.
 
	fe t1, t2;
 
	fe_frombytes(t2, eddsa);
 
	fe_add(t1, fe_one, t2);
 
	fe_sub(t2, fe_one, t2);
 
	fe_invert(t2, t2);
 
	fe_mul(t1, t1, t2);
 
	fe_tobytes(x25519, t1);
 
	WIPE_BUFFER(t1);
 
	WIPE_BUFFER(t2);
 
}
 

	
 
void crypto_x25519_to_eddsa(u8 eddsa[32], const u8 x25519[32])
 
{
 
	// (x, y) = (sqrt(-486664)*u/v, (u-1)/(u+1))
 
	// Only converting u to y, x is assumed positive.
 
	fe t1, t2;
 
	fe_frombytes(t2, x25519);
 
	fe_sub(t1, t2, fe_one);
 
	fe_add(t2, t2, fe_one);
 
	fe_invert(t2, t2);
 
	fe_mul(t1, t1, t2);
 
	fe_tobytes(eddsa, t1);
 
	WIPE_BUFFER(t1);
 
	WIPE_BUFFER(t2);
 
}
 

	
 
/////////////////////////////////////////////
 
/// Dirty ephemeral public key generation ///
 
/////////////////////////////////////////////
 

	
 
// Those functions generates a public key, *without* clearing the
 
// cofactor.  Sending that key over the network leaks 3 bits of the
 
// private key.  Use only to generate ephemeral keys that will be hidden
 
// with crypto_curve_to_hidden().
 
//
 
// The public key is otherwise compatible with crypto_x25519(), which
 
// properly clears the cofactor.
 
//
 
// Note that the distribution of the resulting public keys is almost
 
// uniform.  Flipping the sign of the v coordinate (not provided by this
 
// function), covers the entire key space almost perfectly, where
 
// "almost" means a 2^-128 bias (undetectable).  This uniformity is
 
// needed to ensure the proper randomness of the resulting
 
// representatives (once we apply crypto_curve_to_hidden()).
 
//
 
// Recall that Curve25519 has order C = 2^255 + e, with e < 2^128 (not
 
// to be confused with the prime order of the main subgroup, L, which is
 
// 8 times less than that).
 
//
 
// Generating all points would require us to multiply a point of order C
 
// (the base point plus any point of order 8) by all scalars from 0 to
 
// C-1.  Clamping limits us to scalars between 2^254 and 2^255 - 1. But
 
// by negating the resulting point at random, we also cover scalars from
 
// -2^255 + 1 to -2^254 (which modulo C is congruent to e+1 to 2^254 + e).
 
//
 
// In practice:
 
// - Scalars from 0         to e + 1     are never generated
 
// - Scalars from 2^255     to 2^255 + e are never generated
 
// - Scalars from 2^254 + 1 to 2^254 + e are generated twice
 
//
 
// Since e < 2^128, detecting this bias requires observing over 2^100
 
// representatives from a given source (this will never happen), *and*
 
// recovering enough of the private key to determine that they do, or do
 
// not, belong to the biased set (this practically requires solving
 
// discrete logarithm, which is conjecturally intractable).
 
//
 
// In practice, this means the bias is impossible to detect.
 

	
 
// s + (x*L) % 8*L
 
// Guaranteed to fit in 256 bits iff s fits in 255 bits.
 
//   L             < 2^253
 
//   x%8           < 2^3
 
//   L * (x%8)     < 2^255
 
//   s             < 2^255
 
//   s + L * (x%8) < 2^256
 
static void add_xl(u8 s[32], u8 x)
 
{
 
	u64 mod8  = x & 7;
 
	u64 carry = 0;
 
	FOR (i , 0, 8) {
 
		carry = carry + load32_le(s + 4*i) + L[i] * mod8;
 
		store32_le(s + 4*i, (u32)carry);
 
		carry >>= 32;
 
	}
 
}
 

	
 
// "Small" dirty ephemeral key.
 
// Use if you need to shrink the size of the binary, and can afford to
 
// slow down by a factor of two (compared to the fast version)
 
//
 
// This version works by decoupling the cofactor from the main factor.
 
//
 
// - The trimmed scalar determines the main factor
 
// - The clamped bits of the scalar determine the cofactor.
 
//
 
// Cofactor and main factor are combined into a single scalar, which is
 
// then multiplied by a point of order 8*L (unlike the base point, which
 
// has prime order).  That "dirty" base point is the addition of the
 
// regular base point (9), and a point of order 8.
 
void crypto_x25519_dirty_small(u8 public_key[32], const u8 secret_key[32])
 
{
 
	// Base point of order 8*L
 
	// Raw scalar multiplication with it does not clear the cofactor,
 
	// and the resulting public key will reveal 3 bits of the scalar.
 
	//
 
	// The low order component of this base point  has been chosen
 
	// to yield the same results as crypto_x25519_dirty_fast().
 
	static const u8 dirty_base_point[32] = {
 
		0xd8, 0x86, 0x1a, 0xa2, 0x78, 0x7a, 0xd9, 0x26,
 
		0x8b, 0x74, 0x74, 0xb6, 0x82, 0xe3, 0xbe, 0xc3,
 
		0xce, 0x36, 0x9a, 0x1e, 0x5e, 0x31, 0x47, 0xa2,
 
		0x6d, 0x37, 0x7c, 0xfd, 0x20, 0xb5, 0xdf, 0x75,
 
	};
 
	// separate the main factor & the cofactor of the scalar
 
	u8 scalar[32];
 
	crypto_eddsa_trim_scalar(scalar, secret_key);
 

	
 
	// Separate the main factor and the cofactor
 
	//
 
	// The scalar is trimmed, so its cofactor is cleared.  The three
 
	// least significant bits however still have a main factor.  We must
 
	// remove it for X25519 compatibility.
 
	//
 
	//   cofactor = lsb * L            (modulo 8*L)
 
	//   combined = scalar + cofactor  (modulo 8*L)
 
	add_xl(scalar, secret_key[0]);
 
	scalarmult(public_key, scalar, dirty_base_point, 256);
 
	WIPE_BUFFER(scalar);
 
}
 

	
 
// Select low order point
 
// We're computing the [cofactor]lop scalar multiplication, where:
 
//
 
//   cofactor = tweak & 7.
 
//   lop      = (lop_x, lop_y)
 
//   lop_x    = sqrt((sqrt(d + 1) + 1) / d)
 
//   lop_y    = -lop_x * sqrtm1
 
//
 
// The low order point has order 8. There are 4 such points.  We've
 
// chosen the one whose both coordinates are positive (below p/2).
 
// The 8 low order points are as follows:
 
//
 
// [0]lop = ( 0       ,  1    )
 
// [1]lop = ( lop_x   ,  lop_y)
 
// [2]lop = ( sqrt(-1), -0    )
 
// [3]lop = ( lop_x   , -lop_y)
 
// [4]lop = (-0       , -1    )
 
// [5]lop = (-lop_x   , -lop_y)
 
// [6]lop = (-sqrt(-1),  0    )
 
// [7]lop = (-lop_x   ,  lop_y)
 
//
 
// The x coordinate is either 0, sqrt(-1), lop_x, or their opposite.
 
// The y coordinate is either 0,      -1 , lop_y, or their opposite.
 
// The pattern for both is the same, except for a rotation of 2 (modulo 8)
 
//
 
// This helper function captures the pattern, and we can use it thus:
 
//
 
//    select_lop(x, lop_x, sqrtm1, cofactor);
 
//    select_lop(y, lop_y, fe_one, cofactor + 2);
 
//
 
// This is faster than an actual scalar multiplication,
 
// and requires less code than naive constant time look up.
 
static void select_lop(fe out, const fe x, const fe k, u8 cofactor)
 
{
 
	fe tmp;
 
	fe_0(out);
 
	fe_ccopy(out, k  , (cofactor >> 1) & 1); // bit 1
 
	fe_ccopy(out, x  , (cofactor >> 0) & 1); // bit 0
 
	fe_neg  (tmp, out);
 
	fe_ccopy(out, tmp, (cofactor >> 2) & 1); // bit 2
 
	WIPE_BUFFER(tmp);
 
}
 

	
 
// "Fast" dirty ephemeral key
 
// We use this one by default.
 
//
 
// This version works by performing a regular scalar multiplication,
 
// then add a low order point.  The scalar multiplication is done in
 
// Edwards space for more speed (*2 compared to the "small" version).
 
// The cost is a bigger binary for programs that don't also sign messages.
 
void crypto_x25519_dirty_fast(u8 public_key[32], const u8 secret_key[32])
 
{
 
	// Compute clean scalar multiplication
 
	u8 scalar[32];
 
	ge pk;
 
	crypto_eddsa_trim_scalar(scalar, secret_key);
 
	ge_scalarmult_base(&pk, scalar);
 

	
 
	// Compute low order point
 
	fe t1, t2;
 
	select_lop(t1, lop_x, sqrtm1, secret_key[0]);
 
	select_lop(t2, lop_y, fe_one, secret_key[0] + 2);
 
	ge_precomp low_order_point;
 
	fe_add(low_order_point.Yp, t2, t1);
 
	fe_sub(low_order_point.Ym, t2, t1);
 
	fe_mul(low_order_point.T2, t2, t1);
 
	fe_mul(low_order_point.T2, low_order_point.T2, D2);
 

	
 
	// Add low order point to the public key
 
	ge_madd(&pk, &pk, &low_order_point, t1, t2);
 

	
 
	// Convert to Montgomery u coordinate (we ignore the sign)
 
	fe_add(t1, pk.Z, pk.Y);
 
	fe_sub(t2, pk.Z, pk.Y);
 
	fe_invert(t2, t2);
 
	fe_mul(t1, t1, t2);
 

	
 
	fe_tobytes(public_key, t1);
 

	
 
	WIPE_BUFFER(t1);    WIPE_CTX(&pk);
 
	WIPE_BUFFER(t2);    WIPE_CTX(&low_order_point);
 
	WIPE_BUFFER(scalar);
 
}
 

	
 
///////////////////
 
/// Elligator 2 ///
 
///////////////////
 
static const fe A = {486662};
 

	
 
// Elligator direct map
 
//
 
// Computes the point corresponding to a representative, encoded in 32
 
// bytes (little Endian).  Since positive representatives fits in 254
 
// bits, The two most significant bits are ignored.
 
//
 
// From the paper:
 
// w = -A / (fe(1) + non_square * r^2)
 
// e = chi(w^3 + A*w^2 + w)
 
// u = e*w - (fe(1)-e)*(A//2)
 
// v = -e * sqrt(u^3 + A*u^2 + u)
 
//
 
// We ignore v because we don't need it for X25519 (the Montgomery
 
// ladder only uses u).
 
//
 
// Note that e is either 0, 1 or -1
 
// if e = 0    u = 0  and v = 0
 
// if e = 1    u = w
 
// if e = -1   u = -w - A = w * non_square * r^2
 
//
 
// Let r1 = non_square * r^2
 
// Let r2 = 1 + r1
 
// Note that r2 cannot be zero, -1/non_square is not a square.
 
// We can (tediously) verify that:
 
//   w^3 + A*w^2 + w = (A^2*r1 - r2^2) * A / r2^3
 
// Therefore:
 
//   chi(w^3 + A*w^2 + w) = chi((A^2*r1 - r2^2) * (A / r2^3))
 
//   chi(w^3 + A*w^2 + w) = chi((A^2*r1 - r2^2) * (A / r2^3)) * 1
 
//   chi(w^3 + A*w^2 + w) = chi((A^2*r1 - r2^2) * (A / r2^3)) * chi(r2^6)
 
//   chi(w^3 + A*w^2 + w) = chi((A^2*r1 - r2^2) * (A / r2^3)  *     r2^6)
 
//   chi(w^3 + A*w^2 + w) = chi((A^2*r1 - r2^2) *  A * r2^3)
 
// Corollary:
 
//   e =  1 if (A^2*r1 - r2^2) *  A * r2^3) is a non-zero square
 
//   e = -1 if (A^2*r1 - r2^2) *  A * r2^3) is not a square
 
//   Note that w^3 + A*w^2 + w (and therefore e) can never be zero:
 
//     w^3 + A*w^2 + w = w * (w^2 + A*w + 1)
 
//     w^3 + A*w^2 + w = w * (w^2 + A*w + A^2/4 - A^2/4 + 1)
 
//     w^3 + A*w^2 + w = w * (w + A/2)^2        - A^2/4 + 1)
 
//     which is zero only if:
 
//       w = 0                   (impossible)
 
//       (w + A/2)^2 = A^2/4 - 1 (impossible, because A^2/4-1 is not a square)
 
//
 
// Let isr   = invsqrt((A^2*r1 - r2^2) *  A * r2^3)
 
//     isr   = sqrt(1        / ((A^2*r1 - r2^2) *  A * r2^3)) if e =  1
 
//     isr   = sqrt(sqrt(-1) / ((A^2*r1 - r2^2) *  A * r2^3)) if e = -1
 
//
 
// if e = 1
 
//   let u1 = -A * (A^2*r1 - r2^2) * A * r2^2 * isr^2
 
//       u1 = w
 
//       u1 = u
 
//
 
// if e = -1
 
//   let ufactor = -non_square * sqrt(-1) * r^2
 
//   let vfactor = sqrt(ufactor)
 
//   let u2 = -A * (A^2*r1 - r2^2) * A * r2^2 * isr^2 * ufactor
 
//       u2 = w * -1 * -non_square * r^2
 
//       u2 = w * non_square * r^2
 
//       u2 = u
 
void crypto_elligator_map(u8 curve[32], const u8 hidden[32])
 
{
 
	fe r, u, t1, t2, t3;
 
	fe_frombytes_mask(r, hidden, 2); // r is encoded in 254 bits.
 
	fe_sq(r, r);
 
	fe_add(t1, r, r);
 
	fe_add(u, t1, fe_one);
 
	fe_sq (t2, u);
 
	fe_mul(t3, A2, t1);
 
	fe_sub(t3, t3, t2);
 
	fe_mul(t3, t3, A);
 
	fe_mul(t1, t2, u);
 
	fe_mul(t1, t3, t1);
 
	int is_square = invsqrt(t1, t1);
 
	fe_mul(u, r, ufactor);
 
	fe_ccopy(u, fe_one, is_square);
 
	fe_sq (t1, t1);
 
	fe_mul(u, u, A);
 
	fe_mul(u, u, t3);
 
	fe_mul(u, u, t2);
 
	fe_mul(u, u, t1);
 
	fe_neg(u, u);
 
	fe_tobytes(curve, u);
 

	
 
	WIPE_BUFFER(t1);  WIPE_BUFFER(r);
 
	WIPE_BUFFER(t2);  WIPE_BUFFER(u);
 
	WIPE_BUFFER(t3);
 
}
 

	
 
// Elligator inverse map
 
//
 
// Computes the representative of a point, if possible.  If not, it does
 
// nothing and returns -1.  Note that the success of the operation
 
// depends only on the point (more precisely its u coordinate).  The
 
// tweak parameter is used only upon success
 
//
 
// The tweak should be a random byte.  Beyond that, its contents are an
 
// implementation detail. Currently, the tweak comprises:
 
// - Bit  1  : sign of the v coordinate (0 if positive, 1 if negative)
 
// - Bit  2-5: not used
 
// - Bits 6-7: random padding
 
//
 
// From the paper:
 
// Let sq = -non_square * u * (u+A)
 
// if sq is not a square, or u = -A, there is no mapping
 
// Assuming there is a mapping:
 
//    if v is positive: r = sqrt(-u     / (non_square * (u+A)))
 
//    if v is negative: r = sqrt(-(u+A) / (non_square * u    ))
 
//
 
// We compute isr = invsqrt(-non_square * u * (u+A))
 
// if it wasn't a square, abort.
 
// else, isr = sqrt(-1 / (non_square * u * (u+A))
 
//
 
// If v is positive, we return isr * u:
 
//   isr * u = sqrt(-1 / (non_square * u * (u+A)) * u
 
//   isr * u = sqrt(-u / (non_square * (u+A))
 
//
 
// If v is negative, we return isr * (u+A):
 
//   isr * (u+A) = sqrt(-1     / (non_square * u * (u+A)) * (u+A)
 
//   isr * (u+A) = sqrt(-(u+A) / (non_square * u)
 
int crypto_elligator_rev(u8 hidden[32], const u8 public_key[32], u8 tweak)
 
{
 
	fe t1, t2, t3;
 
	fe_frombytes(t1, public_key);    // t1 = u
 

	
 
	fe_add(t2, t1, A);               // t2 = u + A
 
	fe_mul(t3, t1, t2);
 
	fe_mul_small(t3, t3, -2);
 
	int is_square = invsqrt(t3, t3); // t3 = sqrt(-1 / non_square * u * (u+A))
 
	if (is_square) {
 
		// The only variable time bit.  This ultimately reveals how many
 
		// tries it took us to find a representable key.
 
		// This does not affect security as long as we try keys at random.
 

	
 
		fe_ccopy    (t1, t2, tweak & 1); // multiply by u if v is positive,
 
		fe_mul      (t3, t1, t3);        // multiply by u+A otherwise
 
		fe_mul_small(t1, t3, 2);
 
		fe_neg      (t2, t3);
 
		fe_ccopy    (t3, t2, fe_isodd(t1));
 
		fe_tobytes(hidden, t3);
 

	
 
		// Pad with two random bits
 
		hidden[31] |= tweak & 0xc0;
 
	}
 

	
 
	WIPE_BUFFER(t1);
 
	WIPE_BUFFER(t2);
 
	WIPE_BUFFER(t3);
 
	return is_square - 1;
 
}
 

	
 
void crypto_elligator_key_pair(u8 hidden[32], u8 secret_key[32], u8 seed[32])
 
{
 
	u8 pk [32]; // public key
 
	u8 buf[64]; // seed + representative
 
	COPY(buf + 32, seed, 32);
 
	do {
 
		crypto_chacha20_djb(buf, 0, 64, buf+32, zero, 0);
 
		crypto_x25519_dirty_fast(pk, buf); // or the "small" version
 
	} while(crypto_elligator_rev(buf+32, pk, buf[32]));
 
	// Note that the return value of crypto_elligator_rev() is
 
	// independent from its tweak parameter.
 
	// Therefore, buf[32] is not actually reused.  Either we loop one
 
	// more time and buf[32] is used for the new seed, or we succeeded,
 
	// and buf[32] becomes the tweak parameter.
 

	
 
	crypto_wipe(seed, 32);
 
	COPY(hidden    , buf + 32, 32);
 
	COPY(secret_key, buf     , 32);
 
	WIPE_BUFFER(buf);
 
	WIPE_BUFFER(pk);
 
}
 

	
 
///////////////////////
 
/// Scalar division ///
 
///////////////////////
 

	
 
// Montgomery reduction.
 
// Divides x by (2^256), and reduces the result modulo L
 
//
 
// Precondition:
 
//   x < L * 2^256
 
// Constants:
 
//   r = 2^256                 (makes division by r trivial)
 
//   k = (r * (1/r) - 1) // L  (1/r is computed modulo L   )
 
// Algorithm:
 
//   s = (x * k) % r
 
//   t = x + s*L      (t is always a multiple of r)
 
//   u = (t/r) % L    (u is always below 2*L, conditional subtraction is enough)
 
static void redc(u32 u[8], u32 x[16])
 
{
 
	static const u32 k[8] = {
 
		0x12547e1b, 0xd2b51da3, 0xfdba84ff, 0xb1a206f2,
 
		0xffa36bea, 0x14e75438, 0x6fe91836, 0x9db6c6f2,
 
	};
 

	
 
	// s = x * k (modulo 2^256)
 
	// This is cheaper than the full multiplication.
 
	u32 s[8] = {0};
 
	FOR (i, 0, 8) {
 
		u64 carry = 0;
 
		FOR (j, 0, 8-i) {
 
			carry  += s[i+j] + (u64)x[i] * k[j];
 
			s[i+j]  = (u32)carry;
 
			carry >>= 32;
 
		}
 
	}
 
	u32 t[16] = {0};
 
	multiply(t, s, L);
 

	
 
	// t = t + x
 
	u64 carry = 0;
 
	FOR (i, 0, 16) {
 
		carry  += (u64)t[i] + x[i];
 
		t[i]    = (u32)carry;
 
		carry >>= 32;
 
	}
 

	
 
	// u = (t / 2^256) % L
 
	// Note that t / 2^256 is always below 2*L,
 
	// So a constant time conditional subtraction is enough
 
	remove_l(u, t+8);
 

	
 
	WIPE_BUFFER(s);
 
	WIPE_BUFFER(t);
 
}
 

	
 
void crypto_x25519_inverse(u8 blind_salt [32], const u8 private_key[32],
 
                           const u8 curve_point[32])
 
{
 
	static const  u8 Lm2[32] = { // L - 2
 
		0xeb, 0xd3, 0xf5, 0x5c, 0x1a, 0x63, 0x12, 0x58,
 
		0xd6, 0x9c, 0xf7, 0xa2, 0xde, 0xf9, 0xde, 0x14,
 
		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
 
		0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x10,
 
	};
 
	// 1 in Montgomery form
 
	u32 m_inv [8] = {
 
		0x8d98951d, 0xd6ec3174, 0x737dcf70, 0xc6ef5bf4,
 
		0xfffffffe, 0xffffffff, 0xffffffff, 0x0fffffff,
 
	};
 

	
 
	u8 scalar[32];
 
	crypto_eddsa_trim_scalar(scalar, private_key);
 

	
 
	// Convert the scalar in Montgomery form
 
	// m_scl = scalar * 2^256 (modulo L)
 
	u32 m_scl[8];
 
	{
 
		u32 tmp[16];
 
		ZERO(tmp, 8);
 
		load32_le_buf(tmp+8, scalar, 8);
 
		mod_l(scalar, tmp);
 
		load32_le_buf(m_scl, scalar, 8);
 
		WIPE_BUFFER(tmp); // Wipe ASAP to save stack space
 
	}
 

	
 
	// Compute the inverse
 
	u32 product[16];
 
	for (int i = 252; i >= 0; i--) {
 
		ZERO(product, 16);
 
		multiply(product, m_inv, m_inv);
 
		redc(m_inv, product);
 
		if (scalar_bit(Lm2, i)) {
 
			ZERO(product, 16);
 
			multiply(product, m_inv, m_scl);
 
			redc(m_inv, product);
 
		}
 
	}
 
	// Convert the inverse *out* of Montgomery form
 
	// scalar = m_inv / 2^256 (modulo L)
 
	COPY(product, m_inv, 8);
 
	ZERO(product + 8, 8);
 
	redc(m_inv, product);
 
	store32_le_buf(scalar, m_inv, 8); // the *inverse* of the scalar
 

	
 
	// Clear the cofactor of scalar:
 
	//   cleared = scalar * (3*L + 1)      (modulo 8*L)
 
	//   cleared = scalar + scalar * 3 * L (modulo 8*L)
 
	// Note that (scalar * 3) is reduced modulo 8, so we only need the
 
	// first byte.
 
	add_xl(scalar, scalar[0] * 3);
 

	
 
	// Recall that 8*L < 2^256. However it is also very close to
 
	// 2^255. If we spanned the ladder over 255 bits, random tests
 
	// wouldn't catch the off-by-one error.
 
	scalarmult(blind_salt, scalar, curve_point, 256);
 

	
 
	WIPE_BUFFER(scalar);   WIPE_BUFFER(m_scl);
 
	WIPE_BUFFER(product);  WIPE_BUFFER(m_inv);
 
}
 

	
 
////////////////////////////////
 
/// Authenticated encryption ///
 
////////////////////////////////
 
static void lock_auth(u8 mac[16], const u8  auth_key[32],
 
                      const u8 *ad         , size_t ad_size,
 
                      const u8 *cipher_text, size_t text_size)
 
{
 
	u8 sizes[16]; // Not secret, not wiped
 
	store64_le(sizes + 0, ad_size);
 
	store64_le(sizes + 8, text_size);
 
	crypto_poly1305_ctx poly_ctx;           // auto wiped...
 
	crypto_poly1305_init  (&poly_ctx, auth_key);
 
	crypto_poly1305_update(&poly_ctx, ad         , ad_size);
 
	crypto_poly1305_update(&poly_ctx, zero       , gap(ad_size, 16));
 
	crypto_poly1305_update(&poly_ctx, cipher_text, text_size);
 
	crypto_poly1305_update(&poly_ctx, zero       , gap(text_size, 16));
 
	crypto_poly1305_update(&poly_ctx, sizes      , 16);
 
	crypto_poly1305_final (&poly_ctx, mac); // ...here
 
}
 

	
 
void crypto_aead_init_x(crypto_aead_ctx *ctx,
 
                        u8 const key[32], const u8 nonce[24])
 
{
 
	crypto_chacha20_h(ctx->key, key, nonce);
 
	COPY(ctx->nonce, nonce + 16, 8);
 
	ctx->counter = 0;
 
}
 

	
 
void crypto_aead_init_djb(crypto_aead_ctx *ctx,
 
                          const u8 key[32], const u8 nonce[8])
 
{
 
	COPY(ctx->key  , key  , 32);
 
	COPY(ctx->nonce, nonce,  8);
 
	ctx->counter = 0;
 
}
 

	
 
void crypto_aead_init_ietf(crypto_aead_ctx *ctx,
 
                           const u8 key[32], const u8 nonce[12])
 
{
 
	COPY(ctx->key  , key      , 32);
 
	COPY(ctx->nonce, nonce + 4,  8);
 
	ctx->counter = (u64)load32_le(nonce) << 32;
 
}
 

	
 
void crypto_aead_write(crypto_aead_ctx *ctx, u8 *cipher_text, u8 mac[16],
 
                       const u8 *ad,         size_t ad_size,
 
                       const u8 *plain_text, size_t text_size)
 
{
 
	u8 auth_key[64]; // the last 32 bytes are used for rekeying.
 
	crypto_chacha20_djb(auth_key, 0, 64, ctx->key, ctx->nonce, ctx->counter);
 
	crypto_chacha20_djb(cipher_text, plain_text, text_size,
 
	                    ctx->key, ctx->nonce, ctx->counter + 1);
 
	lock_auth(mac, auth_key, ad, ad_size, cipher_text, text_size);
 
	COPY(ctx->key, auth_key + 32, 32);
 
	WIPE_BUFFER(auth_key);
 
}
 

	
 
int crypto_aead_read(crypto_aead_ctx *ctx, u8 *plain_text, const u8 mac[16],
 
                     const u8 *ad,          size_t ad_size,
 
                     const u8 *cipher_text, size_t text_size)
 
{
 
	u8 auth_key[64]; // the last 32 bytes are used for rekeying.
 
	u8 real_mac[16];
 
	crypto_chacha20_djb(auth_key, 0, 64, ctx->key, ctx->nonce, ctx->counter);
 
	lock_auth(real_mac, auth_key, ad, ad_size, cipher_text, text_size);
 
	int mismatch = crypto_verify16(mac, real_mac);
 
	if (!mismatch) {
 
		crypto_chacha20_djb(plain_text, cipher_text, text_size,
 
		                    ctx->key, ctx->nonce, ctx->counter + 1);
 
		COPY(ctx->key, auth_key + 32, 32);
 
	}
 
	WIPE_BUFFER(auth_key);
 
	WIPE_BUFFER(real_mac);
 
	return mismatch;
 
}
 

	
 
void crypto_aead_lock(u8 *cipher_text, u8 mac[16], const u8 key[32],
 
                      const u8  nonce[24], const u8 *ad, size_t ad_size,
 
                      const u8 *plain_text, size_t text_size)
 
{
 
	crypto_aead_ctx ctx;
 
	crypto_aead_init_x(&ctx, key, nonce);
 
	crypto_aead_write(&ctx, cipher_text, mac, ad, ad_size,
 
	                  plain_text, text_size);
 
	crypto_wipe(&ctx, sizeof(ctx));
 
}
 

	
 
int crypto_aead_unlock(u8 *plain_text, const u8  mac[16], const u8 key[32],
 
                       const u8 nonce[24], const u8 *ad, size_t ad_size,
 
                       const u8 *cipher_text, size_t text_size)
 
{
 
	crypto_aead_ctx ctx;
 
	crypto_aead_init_x(&ctx, key, nonce);
 
	int mismatch = crypto_aead_read(&ctx, plain_text, mac, ad, ad_size,
 
	                                cipher_text, text_size);
 
	crypto_wipe(&ctx, sizeof(ctx));
 
	return mismatch;
 
}
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
}
 
#endif
src/3rdparty/monocypher/monocypher.h
Show inline comments
 
new file 100644
 
// Monocypher version 4.0.2
 
//
 
// This file is dual-licensed.  Choose whichever licence you want from
 
// the two licences listed below.
 
//
 
// The first licence is a regular 2-clause BSD licence.  The second licence
 
// is the CC-0 from Creative Commons. It is intended to release Monocypher
 
// to the public domain.  The BSD licence serves as a fallback option.
 
//
 
// SPDX-License-Identifier: BSD-2-Clause OR CC0-1.0
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Copyright (c) 2017-2019, Loup Vaillant
 
// All rights reserved.
 
//
 
//
 
// Redistribution and use in source and binary forms, with or without
 
// modification, are permitted provided that the following conditions are
 
// met:
 
//
 
// 1. Redistributions of source code must retain the above copyright
 
//    notice, this list of conditions and the following disclaimer.
 
//
 
// 2. Redistributions in binary form must reproduce the above copyright
 
//    notice, this list of conditions and the following disclaimer in the
 
//    documentation and/or other materials provided with the
 
//    distribution.
 
//
 
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
 
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
 
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
 
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
 
// HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
 
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
 
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
 
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
 
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
 
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
//
 
// ------------------------------------------------------------------------
 
//
 
// Written in 2017-2019 by Loup Vaillant
 
//
 
// To the extent possible under law, the author(s) have dedicated all copyright
 
// and related neighboring rights to this software to the public domain
 
// worldwide.  This software is distributed without any warranty.
 
//
 
// You should have received a copy of the CC0 Public Domain Dedication along
 
// with this software.  If not, see
 
// <https://creativecommons.org/publicdomain/zero/1.0/>
 

	
 
#ifndef MONOCYPHER_H
 
#define MONOCYPHER_H
 

	
 
#include <stddef.h>
 
#include <stdint.h>
 

	
 
#ifdef MONOCYPHER_CPP_NAMESPACE
 
namespace MONOCYPHER_CPP_NAMESPACE {
 
#elif defined(__cplusplus)
 
extern "C" {
 
#endif
 

	
 
// Constant time comparisons
 
// -------------------------
 

	
 
// Return 0 if a and b are equal, -1 otherwise
 
int crypto_verify16(const uint8_t a[16], const uint8_t b[16]);
 
int crypto_verify32(const uint8_t a[32], const uint8_t b[32]);
 
int crypto_verify64(const uint8_t a[64], const uint8_t b[64]);
 

	
 

	
 
// Erase sensitive data
 
// --------------------
 
void crypto_wipe(void *secret, size_t size);
 

	
 

	
 
// Authenticated encryption
 
// ------------------------
 
void crypto_aead_lock(uint8_t       *cipher_text,
 
                      uint8_t        mac  [16],
 
                      const uint8_t  key  [32],
 
                      const uint8_t  nonce[24],
 
                      const uint8_t *ad,         size_t ad_size,
 
                      const uint8_t *plain_text, size_t text_size);
 
int crypto_aead_unlock(uint8_t       *plain_text,
 
                       const uint8_t  mac  [16],
 
                       const uint8_t  key  [32],
 
                       const uint8_t  nonce[24],
 
                       const uint8_t *ad,          size_t ad_size,
 
                       const uint8_t *cipher_text, size_t text_size);
 

	
 
// Authenticated stream
 
// --------------------
 
typedef struct {
 
	uint64_t counter;
 
	uint8_t  key[32];
 
	uint8_t  nonce[8];
 
} crypto_aead_ctx;
 

	
 
void crypto_aead_init_x(crypto_aead_ctx *ctx,
 
                        const uint8_t key[32], const uint8_t nonce[24]);
 
void crypto_aead_init_djb(crypto_aead_ctx *ctx,
 
                          const uint8_t key[32], const uint8_t nonce[8]);
 
void crypto_aead_init_ietf(crypto_aead_ctx *ctx,
 
                           const uint8_t key[32], const uint8_t nonce[12]);
 

	
 
void crypto_aead_write(crypto_aead_ctx *ctx,
 
                       uint8_t         *cipher_text,
 
                       uint8_t          mac[16],
 
                       const uint8_t   *ad        , size_t ad_size,
 
                       const uint8_t   *plain_text, size_t text_size);
 
int crypto_aead_read(crypto_aead_ctx *ctx,
 
                     uint8_t         *plain_text,
 
                     const uint8_t    mac[16],
 
                     const uint8_t   *ad        , size_t ad_size,
 
                     const uint8_t   *cipher_text, size_t text_size);
 

	
 

	
 
// General purpose hash (BLAKE2b)
 
// ------------------------------
 

	
 
// Direct interface
 
void crypto_blake2b(uint8_t *hash,          size_t hash_size,
 
                    const uint8_t *message, size_t message_size);
 

	
 
void crypto_blake2b_keyed(uint8_t *hash,          size_t hash_size,
 
                          const uint8_t *key,     size_t key_size,
 
                          const uint8_t *message, size_t message_size);
 

	
 
// Incremental interface
 
typedef struct {
 
	// Do not rely on the size or contents of this type,
 
	// for they may change without notice.
 
	uint64_t hash[8];
 
	uint64_t input_offset[2];
 
	uint64_t input[16];
 
	size_t   input_idx;
 
	size_t   hash_size;
 
} crypto_blake2b_ctx;
 

	
 
void crypto_blake2b_init(crypto_blake2b_ctx *ctx, size_t hash_size);
 
void crypto_blake2b_keyed_init(crypto_blake2b_ctx *ctx, size_t hash_size,
 
                               const uint8_t *key, size_t key_size);
 
void crypto_blake2b_update(crypto_blake2b_ctx *ctx,
 
                           const uint8_t *message, size_t message_size);
 
void crypto_blake2b_final(crypto_blake2b_ctx *ctx, uint8_t *hash);
 

	
 

	
 
// Password key derivation (Argon2)
 
// --------------------------------
 
#define CRYPTO_ARGON2_D  0
 
#define CRYPTO_ARGON2_I  1
 
#define CRYPTO_ARGON2_ID 2
 

	
 
typedef struct {
 
	uint32_t algorithm;  // Argon2d, Argon2i, Argon2id
 
	uint32_t nb_blocks;  // memory hardness, >= 8 * nb_lanes
 
	uint32_t nb_passes;  // CPU hardness, >= 1 (>= 3 recommended for Argon2i)
 
	uint32_t nb_lanes;   // parallelism level (single threaded anyway)
 
} crypto_argon2_config;
 

	
 
typedef struct {
 
	const uint8_t *pass;
 
	const uint8_t *salt;
 
	uint32_t pass_size;
 
	uint32_t salt_size;  // 16 bytes recommended
 
} crypto_argon2_inputs;
 

	
 
typedef struct {
 
	const uint8_t *key; // may be NULL if no key
 
	const uint8_t *ad;  // may be NULL if no additional data
 
	uint32_t key_size;  // 0 if no key (32 bytes recommended otherwise)
 
	uint32_t ad_size;   // 0 if no additional data
 
} crypto_argon2_extras;
 

	
 
extern const crypto_argon2_extras crypto_argon2_no_extras;
 

	
 
void crypto_argon2(uint8_t *hash, uint32_t hash_size, void *work_area,
 
                   crypto_argon2_config config,
 
                   crypto_argon2_inputs inputs,
 
                   crypto_argon2_extras extras);
 

	
 

	
 
// Key exchange (X-25519)
 
// ----------------------
 

	
 
// Shared secrets are not quite random.
 
// Hash them to derive an actual shared key.
 
void crypto_x25519_public_key(uint8_t       public_key[32],
 
                              const uint8_t secret_key[32]);
 
void crypto_x25519(uint8_t       raw_shared_secret[32],
 
                   const uint8_t your_secret_key  [32],
 
                   const uint8_t their_public_key [32]);
 

	
 
// Conversion to EdDSA
 
void crypto_x25519_to_eddsa(uint8_t eddsa[32], const uint8_t x25519[32]);
 

	
 
// scalar "division"
 
// Used for OPRF.  Be aware that exponential blinding is less secure
 
// than Diffie-Hellman key exchange.
 
void crypto_x25519_inverse(uint8_t       blind_salt [32],
 
                           const uint8_t private_key[32],
 
                           const uint8_t curve_point[32]);
 

	
 
// "Dirty" versions of x25519_public_key().
 
// Use with crypto_elligator_rev().
 
// Leaks 3 bits of the private key.
 
void crypto_x25519_dirty_small(uint8_t pk[32], const uint8_t sk[32]);
 
void crypto_x25519_dirty_fast (uint8_t pk[32], const uint8_t sk[32]);
 

	
 

	
 
// Signatures
 
// ----------
 

	
 
// EdDSA with curve25519 + BLAKE2b
 
void crypto_eddsa_key_pair(uint8_t secret_key[64],
 
                           uint8_t public_key[32],
 
                           uint8_t seed[32]);
 
void crypto_eddsa_sign(uint8_t        signature [64],
 
                       const uint8_t  secret_key[64],
 
                       const uint8_t *message, size_t message_size);
 
int crypto_eddsa_check(const uint8_t  signature [64],
 
                       const uint8_t  public_key[32],
 
                       const uint8_t *message, size_t message_size);
 

	
 
// Conversion to X25519
 
void crypto_eddsa_to_x25519(uint8_t x25519[32], const uint8_t eddsa[32]);
 

	
 
// EdDSA building blocks
 
void crypto_eddsa_trim_scalar(uint8_t out[32], const uint8_t in[32]);
 
void crypto_eddsa_reduce(uint8_t reduced[32], const uint8_t expanded[64]);
 
void crypto_eddsa_mul_add(uint8_t r[32],
 
                          const uint8_t a[32],
 
                          const uint8_t b[32],
 
                          const uint8_t c[32]);
 
void crypto_eddsa_scalarbase(uint8_t point[32], const uint8_t scalar[32]);
 
int crypto_eddsa_check_equation(const uint8_t signature[64],
 
                                const uint8_t public_key[32],
 
                                const uint8_t h_ram[32]);
 

	
 

	
 
// Chacha20
 
// --------
 

	
 
// Specialised hash.
 
// Used to hash X25519 shared secrets.
 
void crypto_chacha20_h(uint8_t       out[32],
 
                       const uint8_t key[32],
 
                       const uint8_t in [16]);
 

	
 
// Unauthenticated stream cipher.
 
// Don't forget to add authentication.
 
uint64_t crypto_chacha20_djb(uint8_t       *cipher_text,
 
                             const uint8_t *plain_text,
 
                             size_t         text_size,
 
                             const uint8_t  key[32],
 
                             const uint8_t  nonce[8],
 
                             uint64_t       ctr);
 
uint32_t crypto_chacha20_ietf(uint8_t       *cipher_text,
 
                              const uint8_t *plain_text,
 
                              size_t         text_size,
 
                              const uint8_t  key[32],
 
                              const uint8_t  nonce[12],
 
                              uint32_t       ctr);
 
uint64_t crypto_chacha20_x(uint8_t       *cipher_text,
 
                           const uint8_t *plain_text,
 
                           size_t         text_size,
 
                           const uint8_t  key[32],
 
                           const uint8_t  nonce[24],
 
                           uint64_t       ctr);
 

	
 

	
 
// Poly 1305
 
// ---------
 

	
 
// This is a *one time* authenticator.
 
// Disclosing the mac reveals the key.
 
// See crypto_lock() on how to use it properly.
 

	
 
// Direct interface
 
void crypto_poly1305(uint8_t        mac[16],
 
                     const uint8_t *message, size_t message_size,
 
                     const uint8_t  key[32]);
 

	
 
// Incremental interface
 
typedef struct {
 
	// Do not rely on the size or contents of this type,
 
	// for they may change without notice.
 
	uint8_t  c[16];  // chunk of the message
 
	size_t   c_idx;  // How many bytes are there in the chunk.
 
	uint32_t r  [4]; // constant multiplier (from the secret key)
 
	uint32_t pad[4]; // random number added at the end (from the secret key)
 
	uint32_t h  [5]; // accumulated hash
 
} crypto_poly1305_ctx;
 

	
 
void crypto_poly1305_init  (crypto_poly1305_ctx *ctx, const uint8_t key[32]);
 
void crypto_poly1305_update(crypto_poly1305_ctx *ctx,
 
                            const uint8_t *message, size_t message_size);
 
void crypto_poly1305_final (crypto_poly1305_ctx *ctx, uint8_t mac[16]);
 

	
 

	
 
// Elligator 2
 
// -----------
 

	
 
// Elligator mappings proper
 
void crypto_elligator_map(uint8_t curve [32], const uint8_t hidden[32]);
 
int  crypto_elligator_rev(uint8_t hidden[32], const uint8_t curve [32],
 
                          uint8_t tweak);
 

	
 
// Easy to use key pair generation
 
void crypto_elligator_key_pair(uint8_t hidden[32], uint8_t secret_key[32],
 
                               uint8_t seed[32]);
 

	
 
#ifdef __cplusplus
 
}
 
#endif
 

	
 
#endif // MONOCYPHER_H
0 comments (0 inline, 0 general)