Commit graph

32 commits

Author SHA1 Message Date
Manuel Pégourié-Gonnard 7124fb63be Use zeroize function from new platform_util 2018-05-22 16:05:33 +02:00
Manuel Pégourié-Gonnard 2df4bfe803 Fix typo in comments 2018-05-22 13:39:01 +02:00
Manuel Pégourié-Gonnard 565e4e0fb2 Use more appropriate type for local variable 2018-05-22 13:30:28 +02:00
Manuel Pégourié-Gonnard 08c337d058 Remove useless parameter from function 2018-05-22 13:18:01 +02:00
Manuel Pégourié-Gonnard 89924ddc7e Wipe sensitive info from the stack 2018-05-22 13:07:07 +02:00
Manuel Pégourié-Gonnard 12e2fbdf29 Style adjustments 2018-05-22 13:01:09 +02:00
Manuel Pégourié-Gonnard d418b0dcba Fix typo in comment 2018-05-22 12:56:11 +02:00
Manuel Pégourié-Gonnard 366e1b0464 aria: fix comment on aria_a function
The new version of the comment has been generated by the following python3
script, when the first constant is copy-pasted from RFC 5794 2.4.3.

 #!/usr/bin/python3

RFC_A = """
      y0  = x3 ^ x4 ^ x6 ^ x8  ^ x9  ^ x13 ^ x14,
      y1  = x2 ^ x5 ^ x7 ^ x8  ^ x9  ^ x12 ^ x15,
      y2  = x1 ^ x4 ^ x6 ^ x10 ^ x11 ^ x12 ^ x15,
      y3  = x0 ^ x5 ^ x7 ^ x10 ^ x11 ^ x13 ^ x14,
      y4  = x0 ^ x2 ^ x5 ^ x8  ^ x11 ^ x14 ^ x15,
      y5  = x1 ^ x3 ^ x4 ^ x9  ^ x10 ^ x14 ^ x15,
      y6  = x0 ^ x2 ^ x7 ^ x9  ^ x10 ^ x12 ^ x13,
      y7  = x1 ^ x3 ^ x6 ^ x8  ^ x11 ^ x12 ^ x13,
      y8  = x0 ^ x1 ^ x4 ^ x7  ^ x10 ^ x13 ^ x15,
      y9  = x0 ^ x1 ^ x5 ^ x6  ^ x11 ^ x12 ^ x14,
      y10 = x2 ^ x3 ^ x5 ^ x6  ^ x8  ^ x13 ^ x15,
      y11 = x2 ^ x3 ^ x4 ^ x7  ^ x9  ^ x12 ^ x14,
      y12 = x1 ^ x2 ^ x6 ^ x7  ^ x9  ^ x11 ^ x12,
      y13 = x0 ^ x3 ^ x6 ^ x7  ^ x8  ^ x10 ^ x13,
      y14 = x0 ^ x3 ^ x4 ^ x5  ^ x9  ^ x11 ^ x14,
      y15 = x1 ^ x2 ^ x4 ^ x5  ^ x8  ^ x10 ^ x15.
"""

matrix = []
for l in RFC_A.split('\n')[1:-1]:
    rhs = l.split('=')[1][:-1]
    row = tuple(hex(int(t[2:]))[2:] for t in rhs.split('^'))
    matrix.append(row)

out = {}
out['a'] = tuple(''.join(w) for w in zip(*(matrix[0:4])))
out['b'] = tuple(''.join(w) for w in zip(*(matrix[4:8])))
out['c'] = tuple(''.join(w) for w in zip(*(matrix[8:12])))
out['d'] = tuple(''.join(w) for w in zip(*(matrix[12:])))

out2 = {}
for o, r in out.items():
    row = list(r)
    for i in range(len(r) - 1):
        w1 = row[i]
        if len(set(w1)) == 2:
            w2 = row[i+1]
            nw1 = nw2 = ''
            for j in range(len(w1)):
                if w1[j] in nw1:
                    nw1 += w2[j]
                    nw2 += w1[j]
                else:
                    nw1 += w1[j]
                    nw2 += w2[j]
            row[i] = nw1
            row[i+1] = nw2

    out2[o] = row

for o in 'abcd':
    print(o,   '=', ' + '.join(out[o]))
    print(' ', '=', ' + '.join(out2[o]))
2018-03-01 14:48:10 +01:00
Manuel Pégourié-Gonnard 21662148f7 aria: improve compiler compat by using __asm
gcc --std=c99 doesn't like the shorter "asm" (this broke all.sh)
2018-03-01 11:28:51 +01:00
Manuel Pégourié-Gonnard 2078725feb aria: check arm arch version for asm
rev and rev16 are only supported from v6 (all profiles) and up.

arm-none-eabi-gcc picks a lower architecture version by default, which means
before this commit it would fail to build (assembler error) unless you
manually specified -march=armv6-m -mthumb or similar, which broke all.sh.

Source for version-checking macros:
- GCC/Clang: use the -E -dM - </dev/null trick
- armcc5: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0472k/chr1359125007083.html
- armclang 6: http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0774g/chr1383660321827.html

Tested with the following script:

 #!/bin/sh

set -eu

ARMCLANG="env ARM_TOOL_VARIANT=ult $ARMC6_BIN_DIR/armclang"

build() {
    echo "$@"
    "$@" -Iinclude -c library/aria.c

}

build arm-none-eabi-gcc
build arm-none-eabi-gcc -march=armv5

build clang --target=arm-none-eabi
build clang --target=arm-none-eabi -march=armv5

build armcc
build armcc --gnu
build armcc --cpu=5T
build armcc --cpu=5T --gnu

build $ARMCLANG --target=arm-arm-none-eabi

check_asm() {
    rm -f aria.o
    build "$@"
    arm-none-eabi-objdump -d aria.o | grep rev16
}

check_asm arm-none-eabi-gcc -march=armv6-m -mthumb
check_asm arm-none-eabi-gcc -march=armv7-m -mthumb
check_asm arm-none-eabi-gcc -march=armv8-m.base -mthumb

check_asm arm-none-eabi-gcc -march=armv7-a -mthumb
check_asm arm-none-eabi-gcc -march=armv8-a -mthumb
check_asm arm-none-eabi-gcc -march=armv7-a -marm
check_asm arm-none-eabi-gcc -march=armv8-a -marm

check_asm clang --target=arm-none-eabi -march=armv6-m
check_asm clang --target=arm-none-eabi -march=armv7-a
check_asm clang --target=arm-none-eabi -march=armv7-m
check_asm clang --target=arm-none-eabi -march=armv7-r
check_asm clang --target=arm-none-eabi -march=armv8-a

check_asm armcc -O0 --cpu=6-M
check_asm armcc -O0 --cpu=7-M
check_asm armcc -O0 --cpu=6
check_asm armcc -O0 --cpu=7-A

check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv6-m
check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv7-a
check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv7-m
check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv7-r
check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv8-a
check_asm $ARMCLANG --target=arm-arm-none-eabi -march=armv8-m.base
2018-03-01 11:28:51 +01:00
Manuel Pégourié-Gonnard 7fc08795c1 aria: more whitespace fixes 2018-03-01 09:33:20 +01:00
Manuel Pégourié-Gonnard 5ad88b6d0d aria: define constants for block size and max rounds 2018-03-01 09:25:31 +01:00
Manuel Pégourié-Gonnard f3a46a9b4f aria: fix some typos in comments 2018-03-01 09:25:05 +01:00
Manuel Pégourié-Gonnard c0bb66f47e aria: improve compiler inline compatibility 2018-03-01 09:25:05 +01:00
Manuel Pégourié-Gonnard 4231e7f46f Fix some whitespace and other style issues
In addition to whitespace:
- wrapped a few long lines
- added parenthesis to return statements
2018-02-28 11:34:01 +01:00
Manuel Pégourié-Gonnard 377b2b624d aria: optimize byte perms on Arm
Use specific instructions for moving bytes around in a word. This speeds
things up, and as a side-effect, slightly lowers code size.

ARIA_P3 and ARIA_P1 are now 1 single-cycle instruction each (those
instructions are available in all architecture versions starting from v6-M).
Note: ARIA_P3 was already translated to a single instruction by Clang 3.8 and
armclang 6.5, but not arm-gcc 5.4 nor armcc 5.06.

ARIA_P2 is already efficiently translated to the minimal number of
instruction (1 in ARM mode, 2 in thumb mode) by all tested compilers

Manually compiled and inspected generated code with the following compilers:
arm-gcc 5.4, clang 3.8, armcc 5.06 (with and without --gnu), armclang 6.5.

Size reduction (arm-none-eabi-gcc -march=armv6-m -mthumb -Os): 5288 -> 5044 B

Effect on executing time of self-tests on a few boards:
FRDM-K64F   (Cortex-M4):    444 ->  385 us (-13%)
LPC1768     (Cortex-M3):    488 ->  432 us (-11%)
FRDM-KL64Z  (Cortex-M0):   1429 -> 1134 us (-20%)

Measured using a config.h with no cipher mode and the following program with
aria.c and aria.h copy-pasted to the online compiler:

 #include "mbed.h"
 #include "aria.h"

int main() {
    Timer t;
    t.start();
    int ret = mbedtls_aria_self_test(0);
    t.stop();
    printf("ret = %d; time = %d us\n", ret, t.read_us());
}
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard fb0e4f0d1a aria: optimise byte perms on Intel
(A similar commit for Arm follows.)

Use specific instructions for moving bytes around in a word. This speeds
things up, and as a side-effect, slightly lowers code size.

ARIA_P3 (aka reverse byte order) is now 1 instruction on x86, which speeds up
key schedule. (Clang 3.8 finds this but GCC 5.4 doesn't.)

I couldn't find an Intel equivalent of ARM's ret16 (aka ARIA_P1), so I made it
two instructions, which is still much better than the code generated with
the previous mask-shift-or definition, and speeds up en/decryption. (Neither
Clang 3.8 nor GCC 5.4 find this.)

Before:
O	aria.o	ins
s	7976	43,865
2	10520	37,631
3	13040	28,146

After:
O	aria.o	ins
s	7768	33,497
2	9816	28,268
3	11432	20,829

For measurement method, see previous commit:
"aria: turn macro into static inline function"
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard cac5008b17 aria: define P3 macro
This will allow to replace it with an optimised implementation later
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard f205a012b8 aria: comment implementation of A transform
The line-by-line comments were generated using the following Python 3 script:

 #!/usr/bin/python3

class Atom:
    def __init__(self, val):
        self.v = val

    def __str__(self):
        return self.v

    def p1(self):
        v = self.v
        return Atom(v[1] + v[0] + v[3] + v[2])

    def p2(self):
        v = self.v
        return Atom(v[2] + v[3] + v[0] + v[1])

    def __xor__(self, other):
        return Sum(self.tuple() + other.tuple())

    def tuple(self):
        return (self,)

class Sum:
    def __init__(self, terms):
        self.t = terms
        assert(type(terms) == tuple)
        for t in terms:
            assert(type(t) == Atom)

    def __str__(self):
        return '+'.join(sorted((str(t) for t in self.t),
                        key=lambda v: int(v, 16)))

    def p1(self):
        return Sum(tuple(t.p1() for t in self.t))

    def p2(self):
        return Sum(tuple(t.p2() for t in self.t))

    def tuple(self):
        return self.t

    def __xor__(self, other):
        return Sum(self.t + other.tuple())

class LoggingDict(dict):
    def __setitem__(self, key, val):
        print(key, '=', val)
        dict.__setitem__(self, key, val)

    def set(self, key, val):
        dict.__setitem__(self, key, val)

env = LoggingDict()

env.set('ra', Atom('0123'))
env.set('rb', Atom('4567'))
env.set('rc', Atom('89ab'))
env.set('rd', Atom('cdef'))
env.set('ARIA_P1', lambda x: x.p1())
env.set('ARIA_P2', lambda x: x.p2())

code = """
ta  =   rb;
rb  =   ra;
ra  =   ARIA_P2( ta );
tb  =   ARIA_P2( rd );
rd  =   ARIA_P1( rc );
rc  =   ARIA_P1( tb );
ta  ^=  rd;
tc  =   ARIA_P2( rb );
ta  =   ARIA_P1( ta ) ^ tc ^ rc;
tb  ^=  ARIA_P2( rd );
tc  ^=  ARIA_P1( ra );
rb  ^=  ta ^ tb;
tb  =   ARIA_P2( tb ) ^ ta;
ra  ^=  ARIA_P1( tb );
ta  =   ARIA_P2( ta );
rd  ^=  ARIA_P1( ta ) ^ tc;
tc  =   ARIA_P2( tc );
rc  ^=  ARIA_P1( tc ) ^ ta;
"""

exec(code, env)
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard 35ad891aee aria: internal names closer to standard document 2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard 64744f88b6 aria: define SLA() as sl(a())
This decreases the size with -Os by nearly 1k while
not hurting performance too much with -O2 and -O3

Before:
O	aria.o	ins
s	8784	41,408
2	11112	37,001
3	13096	27,438

After:
O	aria.o	ins
s	7976	43,865
2	10520	37,631
3	13040	28,146

(See previous commit for measurement details.)
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard 8c76a9489e aria: turn macro into static inline function
Besides documenting types better and so on, this give the compiler more room
to optimise either for size or performance.

Here are some before/after measurements of:
- size of aria.o in bytes (less is better)
- instruction count for the selftest function (less is better)
with various -O flags.

Before:
O	aria.o	ins
s	10896	37,256
2	11176	37,199
3	12248	27,752

After:
O	aria.o	ins
s	8784	41,408
2	11112	37,001
3	13096	27,438

The new version allows the compiler to reach smaller size with -Os while
maintaining (actually slightly improving) performance with -O2 and -O3.

Measurements were done on x86_64 (but since this is mainly about inlining
code, this should transpose well to other platforms) using the following
helper program and script, after disabling CBC, CFB and CTR in config.h, in
order to focus on the core functions.

==> st.c <==
 #include "mbedtls/aria.h"

int main( void ) {
    return mbedtls_aria_self_test( 0 );
}

==> p.sh <==
 #!/bin/sh

set -eu

ccount () {
    (
    valgrind --tool=callgrind --dump-line=no --callgrind-out-file=/dev/null --collect-atstart=no --toggle-collect=main $1
    ) 2>&1 | sed -n -e 's/.*refs: *\([0-9,]*\)/\1/p'
}

printf "O\taria.o\tins\n"
for O in s 2 3; do
    GCC="gcc -Wall -Wextra -Werror -Iinclude"

    $GCC -O$O -c library/aria.c
    $GCC -O1 st.c aria.o -o st
   ./st

    SIZE=$( du -b aria.o | cut -f1 )
    INS=$( ccount ./st )

    printf "$O\t$SIZE\t$INS\n"
done
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard a41ecdabed aria: closer to usual comment style
We're not absolutely consistent in the rest of the library, but we tend to use
C99-style comments less often.

Change to use C89-style comments everywhere except for end-of-line comments
2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard 56453937a1 aria: use mbedtls_zeroize() 2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard c76ceb677b aria: move conditional outside of loop 2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard 9cc89248fe aria: use unsigned type for bit count 2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard e1ad7491c5 aria: clean up interface of internal macros 2018-02-27 12:39:12 +01:00
Manuel Pégourié-Gonnard a6d639e553 aria: improve some comments & internal names 2018-02-27 12:39:12 +01:00
Markku-Juhani O. Saarinen 6ba68d4a3b ARIA init and free 2018-02-27 12:39:12 +01:00
Markku-Juhani O. Saarinen 3c0b53b2b0 ARIA build integration 2018-02-27 12:39:12 +01:00
Markku-Juhani O. Saarinen 259fa60f6c ARIA test vectors for CBC CFB CTR modes 2018-02-27 12:39:12 +01:00
Markku-Juhani O. Saarinen 41efbaabc9 ARIA cipher implementation 2018-02-27 12:39:12 +01:00