ARMv8 crypto support for AES with openssl backend #585
Description
We currently evaluate sqlcipher to use on aarch64 (on a Toradex Colibri IMX8X to be precise: https://www.toradex.com/computer-on-modules/colibri-arm-family/nxp-imx-8x)
The SoC has hardware crypt acceleration available, which generally also works when using openssl (v3.2.6), for example:
With hardware acceleration:
openssl speed -elapsed -evp aes-256-cbc
You have chosen to measure elapsed time instead of user CPU time.
Doing AES-256-CBC ops for 3s on 16 size blocks: 10988117 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 64 size blocks: 6580238 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 256 size blocks: 2512161 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 1024 size blocks: 728052 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 8192 size blocks: 95419 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 16384 size blocks: 47871 AES-256-CBC ops in 3.00s
version: 3.2.6
built on: Tue Aug 1 23:00:00 2023 UTC
options: bn(64,64)
compiler: aarch64-poky-linux-gcc -march=armv8-a+crypto -mbranch-protection=standard -fstack-protector-strong -O2 -D_FORTIFY_SOURCE=2 -Wformat -Wformat-security -Werror=format-security --sysroot=recipe-sysroot -O2 -pipe -g -feliminate-unused-debug-types -fcanon-prefix-map -fmacro-prefix-map= -fdebug-prefix-map= -fmacro-prefix-map= -fdebug-prefix-map= -fdebug-prefix-map= -fmacro-prefix-map= -fdebug-prefix-map= -DOPENSSL_USE_NODELETE -DOPENSSL_PIC -DOPENSSL_BUILDING_OPENSSL -DNDEBUG
CPUINFO: OPENSSL_armcap=0xbd
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
AES-256-CBC 58603.29k 140378.41k 214371.07k 248508.42k 260557.48k 261439.49k
Without hardware, as a reference:
OPENSSL_armcap=0 openssl speed -elapsed -evp aes-256-cbc
You have chosen to measure elapsed time instead of user CPU time.
Doing AES-256-CBC ops for 3s on 16 size blocks: 3179890 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 64 size blocks: 948452 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 256 size blocks: 251006 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 1024 size blocks: 63676 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 8192 size blocks: 7990 AES-256-CBC ops in 3.00s
Doing AES-256-CBC ops for 3s on 16384 size blocks: 3993 AES-256-CBC ops in 3.00s
version: 3.2.6
built on: Tue Aug 1 23:00:00 2023 UTC
options: bn(64,64)
compiler: aarch64-poky-linux-gcc -march=armv8-a+crypto -mbranch-protection=standard -fstack-protector-strong -O2 -D_FORTIFY_SOURCE=2 -Wformat -Wformat-security -Werror=format-security --sysroot=recipe-sysroot -O2 -pipe -g -feliminate-unused-debug-types -fcanon-prefix-map -fmacro-prefix-map= -fdebug-prefix-map= -fmacro-prefix-map= -fdebug-prefix-map= -fdebug-prefix-map= -fmacro-prefix-map= -fdebug-prefix-map= -DOPENSSL_USE_NODELETE -DOPENSSL_PIC -DOPENSSL_BUILDING_OPENSSL -DNDEBUG
CPUINFO: OPENSSL_armcap=0x0 env:0
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes 16384 bytes
AES-256-CBC 16959.41k 20233.64k 21419.18k 21734.74k 21818.03k 21807.10k
When we compile sqlcipher and run some tests, we get the same speed, regardless if we set OPENSSL_armcap=0 or not.
Is there a way to actually verify that sqlcipher was compiled correctly/uses hw crypto in openssl?