957c61cbbf
This change upgrades to GCC 12.3 and GNU binutils 2.42. The GNU linker appears to have changed things so that only a single de-duplicated str table is present in the binary, and it gets placed wherever the linker wants, regardless of what the linker script says. To cope with that we need to stop using .ident to embed licenses. As such, this change does significant work to revamp how third party licenses are defined in the codebase, using `.section .notice,"aR",@progbits`. This new GCC 12.3 toolchain has support for GNU indirect functions. It lets us support __target_clones__ for the first time. This is used for optimizing the performance of libc string functions such as strlen and friends so far on x86, by ensuring AVX systems favor a second codepath that uses VEX encoding. It shaves some latency off certain operations. It's a useful feature to have for scientific computing for the reasons explained by the test/libcxx/openmp_test.cc example which compiles for fifteen different microarchitectures. Thanks to the upgrades, it's now also possible to use newer instruction sets, such as AVX512FP16, VNNI. Cosmo now uses the %gs register on x86 by default for TLS. Doing it is helpful for any program that links `cosmo_dlopen()`. Such programs had to recompile their binaries at startup to change the TLS instructions. That's not great, since it means every page in the executable needs to be faulted. The work of rewriting TLS-related x86 opcodes, is moved to fixupobj.com instead. This is great news for MacOS x86 users, since we previously needed to morph the binary every time for that platform but now that's no longer necessary. The only platforms where we need fixup of TLS x86 opcodes at runtime are now Windows, OpenBSD, and NetBSD. On Windows we morph TLS to point deeper into the TIB, based on a TlsAlloc assignment, and on OpenBSD/NetBSD we morph %gs back into %fs since the kernels do not allow us to specify a value for the %gs register. OpenBSD users are now required to use APE Loader to run Cosmo binaries and assimilation is no longer possible. OpenBSD kernel needs to change to allow programs to specify a value for the %gs register, or it needs to stop marking executable pages loaded by the kernel as mimmutable(). This release fixes __constructor__, .ctor, .init_array, and lastly the .preinit_array so they behave the exact same way as glibc. We no longer use hex constants to define math.h symbols like M_PI. |
||
---|---|---|
.. | ||
argon2.c | ||
argon2.h | ||
blake2-impl.h | ||
blake2.h | ||
blake2b.c | ||
blamka-round-ref.h | ||
BUILD.mk | ||
core.c | ||
core.h | ||
encoding.c | ||
encoding.h | ||
LICENSE | ||
README.md | ||
ref.c |
Argon2
This is the reference C implementation of Argon2, the password-hashing function that won the Password Hashing Competition (PHC).
Argon2 is a password-hashing function that summarizes the state of the art in the design of memory-hard functions and can be used to hash passwords for credential storage, key derivation, or other applications.
It has a simple design aimed at the highest memory filling rate and effective use of multiple computing units, while still providing defense against tradeoff attacks (by exploiting the cache and memory organization of the recent processors).
Argon2 has three variants: Argon2i, Argon2d, and Argon2id. Argon2d is faster and uses data-depending memory access, which makes it highly resistant against GPU cracking attacks and suitable for applications with no threats from side-channel timing attacks (eg. cryptocurrencies). Argon2i instead uses data-independent memory access, which is preferred for password hashing and password-based key derivation, but it is slower as it makes more passes over the memory to protect from tradeoff attacks. Argon2id is a hybrid of Argon2i and Argon2d, using a combination of data-depending and data-independent memory accesses, which gives some of Argon2i's resistance to side-channel cache timing attacks and much of Argon2d's resistance to GPU cracking attacks.
Argon2i, Argon2d, and Argon2id are parametrized by:
- A time cost, which defines the amount of computation realized and therefore the execution time, given in number of iterations
- A memory cost, which defines the memory usage, given in kibibytes
- A parallelism degree, which defines the number of parallel threads
The Argon2 document gives detailed specs and design rationale.
Please report bugs as issues on this repository.
Usage
make
builds the executable argon2
, the static library libargon2.a
,
and the shared library libargon2.so
(or on macOS, the dynamic library
libargon2.dylib
-- make sure to specify the installation prefix when
you compile: make PREFIX=/usr
). Make sure to run make test
to verify
that your build produces valid results. sudo make install PREFIX=/usr
installs it to your system.
Command-line utility
argon2
is a command-line utility to test specific Argon2 instances
on your system. To show usage instructions, run
./argon2 -h
as
Usage: ./argon2 [-h] salt [-i|-d|-id] [-t iterations] [-m memory] [-p parallelism] [-l hash length] [-e|-r] [-v (10|13)]
Password is read from stdin
Parameters:
salt The salt to use, at least 8 characters
-i Use Argon2i (this is the default)
-d Use Argon2d instead of Argon2i
-id Use Argon2id instead of Argon2i
-t N Sets the number of iterations to N (default = 3)
-m N Sets the memory usage of 2^N KiB (default 12)
-p N Sets parallelism to N threads (default 1)
-l N Sets hash output length to N bytes (default 32)
-e Output only encoded hash
-r Output only the raw bytes of the hash
-v (10|13) Argon2 version (defaults to the most recent version, currently 13)
-h Print argon2 usage
For example, to hash "password" using "somesalt" as a salt and doing 2 iterations, consuming 64 MiB, using four parallel threads and an output hash of 24 bytes
$ echo -n "password" | ./argon2 somesalt -t 2 -m 16 -p 4 -l 24
Type: Argon2i
Iterations: 2
Memory: 65536 KiB
Parallelism: 4
Hash: 45d7ac72e76f242b20b77b9bf9bf9d5915894e669a24e6c6
Encoded: $argon2i$v=19$m=65536,t=2,p=4$c29tZXNhbHQ$RdescudvJCsgt3ub+b+dWRWJTmaaJObG
0.188 seconds
Verification ok
Library
libargon2
provides an API to both low-level and high-level functions
for using Argon2.
The example program below hashes the string "password" with Argon2i
using the high-level API and then using the low-level API. While the
high-level API takes the three cost parameters (time, memory, and
parallelism), the password input buffer, the salt input buffer, and the
output buffers, the low-level API takes in these and additional parameters
, as defined in include/argon2.h
.
There are many additional parameters, but we will highlight three of them here.
-
The
secret
parameter, which is used for keyed hashing. This allows a secret key to be input at hashing time (from some external location) and be folded into the value of the hash. This means that even if your salts and hashes are compromised, an attacker cannot brute-force to find the password without the key. -
The
ad
parameter, which is used to fold any additional data into the hash value. Functionally, this behaves almost exactly like thesecret
orsalt
parameters; thead
parameter is folding into the value of the hash. However, this parameter is used for different data. Thesalt
should be a random string stored alongside your password. Thesecret
should be a random key only usable at hashing time. Thead
is for any other data. -
The
flags
parameter, which determines which memory should be securely erased. This is useful if you want to securely delete thepwd
orsecret
fields right after they are used. To do this setflags
to eitherARGON2_FLAG_CLEAR_PASSWORD
orARGON2_FLAG_CLEAR_SECRET
. To change how internal memory is cleared, change the global flagFLAG_clear_internal_memory
(defaults to clearing internal memory).
Here the time cost t_cost
is set to 2 iterations, the
memory cost m_cost
is set to 216 kibibytes (64 mebibytes),
and parallelism is set to 1 (single-thread).
Compile for example as gcc test.c libargon2.a -Isrc -o test
, if the program
below is named test.c
and placed in the project's root directory.
#include "third_party/argon2/argon2.h"
#define HASHLEN 32
#define SALTLEN 16
#define PWD "password"
int main(void)
{
uint8_t hash1[HASHLEN];
uint8_t hash2[HASHLEN];
uint8_t salt[SALTLEN];
memset( salt, 0x00, SALTLEN );
uint8_t *pwd = (uint8_t *)strdup(PWD);
uint32_t pwdlen = strlen((char *)pwd);
uint32_t t_cost = 2; // 2-pass computation
uint32_t m_cost = (1<<16); // 64 mebibytes memory usage
uint32_t parallelism = 1; // number of threads and lanes
// high-level API
argon2i_hash_raw(t_cost, m_cost, parallelism, pwd, pwdlen, salt, SALTLEN, hash1, HASHLEN);
// low-level API
argon2_context context = {
hash2, /* output array, at least HASHLEN in size */
HASHLEN, /* digest length */
pwd, /* password array */
pwdlen, /* password length */
salt, /* salt array */
SALTLEN, /* salt length */
NULL, 0, /* optional secret data */
NULL, 0, /* optional associated data */
t_cost, m_cost, parallelism, parallelism,
ARGON2_VERSION_13, /* algorithm version */
NULL, NULL, /* custom memory allocation / deallocation functions */
/* by default only internal memory is cleared (pwd is not wiped) */
ARGON2_DEFAULT_FLAGS
};
int rc = argon2i_ctx( &context );
if(ARGON2_OK != rc) {
printf("Error: %s\n", argon2_error_message(rc));
exit(1);
}
free(pwd);
for( int i=0; i<HASHLEN; ++i ) printf( "%02x", hash1[i] ); printf( "\n" );
if (memcmp(hash1, hash2, HASHLEN)) {
for( int i=0; i<HASHLEN; ++i ) {
printf( "%02x", hash2[i] );
}
printf("\nfail\n");
}
else printf("ok\n");
return 0;
}
To use Argon2d instead of Argon2i call argon2d_hash_raw
instead of
argon2i_hash_raw
using the high-level API, and argon2d
instead of
argon2i
using the low-level API. Similarly for Argon2id, call argon2id_hash_raw
and argon2id
.
To produce the crypt-like encoding rather than the raw hash, call
argon2i_hash_encoded
for Argon2i, argon2d_hash_encoded
for Argon2d, and
argon2id_hash_encoded
for Argon2id
See include/argon2.h
for API details.
Note: in this example the salt is set to the all-0x00
string for the
sake of simplicity, but in your application you should use a random salt.
Benchmarks
make bench
creates the executable bench
, which measures the execution
time of various Argon2 instances:
$ ./bench
Argon2d 1 iterations 1 MiB 1 threads: 5.91 cpb 5.91 Mcycles
Argon2i 1 iterations 1 MiB 1 threads: 4.64 cpb 4.64 Mcycles
0.0041 seconds
Argon2d 1 iterations 1 MiB 2 threads: 2.76 cpb 2.76 Mcycles
Argon2i 1 iterations 1 MiB 2 threads: 2.87 cpb 2.87 Mcycles
0.0038 seconds
Argon2d 1 iterations 1 MiB 4 threads: 3.25 cpb 3.25 Mcycles
Argon2i 1 iterations 1 MiB 4 threads: 3.57 cpb 3.57 Mcycles
0.0048 seconds
(...)
Argon2d 1 iterations 4096 MiB 2 threads: 2.15 cpb 8788.08 Mcycles
Argon2i 1 iterations 4096 MiB 2 threads: 2.15 cpb 8821.59 Mcycles
13.0112 seconds
Argon2d 1 iterations 4096 MiB 4 threads: 1.79 cpb 7343.72 Mcycles
Argon2i 1 iterations 4096 MiB 4 threads: 2.72 cpb 11124.86 Mcycles
19.3974 seconds
(...)
Bindings
Bindings are available for the following languages (make sure to read their documentation):
- Android (Java/Kotlin) by @lambdapioneer
- Dart by @tmthecoder
- Elixir by @riverrun
- Erlang by @ergenius
- Go by @tvdburgt
- Haskell by @hvr
- JavaScript (native), by @ranisalt
- JavaScript (native), by @jdconley
- JavaScript (ffi), by @cjlarose
- JavaScript (browser), by @antelle
- JVM by @phXql
- JVM (with keyed hashing) by @kosprov
- Lua (native) by @thibaultCha
- Lua (ffi) by @thibaultCha
- OCaml by @Khady
- Python (native), by @flamewow
- Python (ffi), by @hynek
- Python (ffi, with keyed hashing), by @thusoy
- Python (ffi, with keyed hashing), by @ultrahorizon
- R by @wrathematics
- Ruby by @technion
- Rust by @quininer
- Rust by @bcmyers
- C#/.NET CoreCLR by @kmaragon
- Perl by @leont
- mruby by @Asmod4n
- Swift by @ImKcat
- Swift by @tmthecoder
Test suite
There are two sets of test suites. One is a low level test for the hash function, the other tests the higher level API. Both of these are built and executed by running:
make test
Intellectual property
Except for the components listed below, the Argon2 code in this repository is copyright (c) 2015 Daniel Dinu, Dmitry Khovratovich (main authors), Jean-Philippe Aumasson and Samuel Neves, and dual licensed under the CC0 License and the Apache 2.0 License. For more info see the LICENSE file.
The string encoding routines in src/encoding.c
are
copyright (c) 2015 Thomas Pornin, and under
CC0 License.
The BLAKE2 code in src/blake2/
is copyright (c) Samuel
Neves, 2013-2015, and under
CC0 License.
All licenses are therefore GPL-compatible.