LibreSSL flavor doesn't use aesni

[h-2]

Coming here from #222

LibreSSL does not seem to be able to make use of AES-NI:

	speed
openssl -evp	390 MB/s
openssl	86 MB/s
libressl -evp	55 MB/s
libressl	55 MB/s

(all of the above is 8KB, aes-128-cbc)

Long log:

root@fire:~ # openssl speed -evp aes-128-cbc
OpenSSL 1.0.2k-freebsd  26 Jan 2017
built on: date not available
options:bn(64,64) rc4(16x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc     259552.04k   346362.11k   368229.12k   385205.63k   390116.57k

root@fire:~ # openssl speed  aes-128-cbc
OpenSSL 1.0.2k-freebsd  26 Jan 2017
built on: date not available
options:bn(64,64) rc4(16x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128 cbc      32988.44k    36508.98k    37855.04k    85214.55k    86029.65k

root@fire:~ # /usr/local/bin/openssl speed -evp aes-128-cbc
LibreSSL 2.6.4
built on: date not available
options:bn(64,64) rc4(ptr,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: information not available
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc      45461.70k    52489.94k    54501.63k    55128.06k    55235.93k

root@fire:~ # /usr/local/bin/openssl speed aes-128-cbc
LibreSSL 2.6.4
built on: date not available
options:bn(64,64) rc4(ptr,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: information not available
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128 cbc      50298.05k    53927.98k    54856.77k    55211.87k    55160.47k

Also note how the performance differs between the (block?) sizes in OpenSSL, but not in LibreSSL ❓

VERSIONS:
OPNsense 18.1.5-amd64
FreeBSD 11.1-RELEASE-p8
LibreSSL 2.6.4

[fichtner]

I looked into this. LibreSSL 2.6 seems to have broken their assembler inclusion switch in the portable version. First attempts to undo this ended in compiler errors...

[h-2]

Thanks for looking into this!

LibreSSL 2.6 seems to have broken their assembler inclusion switch in the portable version. First attempts to undo this ended in compiler errors...

Yeah, I heard that the "for other OS" - versions of OpenBSD tools receive a lot less love, but I would have hoped for something like LibreSSL to have upstream regression testing...

[fichtner]

Even OpenSSL was shipped without assembler bits in FreeBSD ports for a while... because it is a difficult regression to observe as it passes all checks... :)

Will try again tomorrow.

[fichtner]

Here's a report of the same issue, but with inconclusive outcome https://forums.freebsd.org/threads/openssl-almost-10x-faster-than-libressl.64497/

[fichtner]

Will be fixed in 18.1.7, or you can build from ports: opnsense/ports@167e07f

# /usr/local/bin/openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 5804203 aes-128 cbc's in 3.01s
Doing aes-128 cbc for 3s on 64 size blocks: 1538558 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 256 size blocks: 398692 aes-128 cbc's in 3.03s
Doing aes-128 cbc for 3s on 1024 size blocks: 259775 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 8192 size blocks: 33573 aes-128 cbc's in 3.05s
LibreSSL 2.6.4
built on: date not available
options:bn(64,64) rc4(8x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: information not available
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128 cbc      30875.34k    32822.57k    33670.98k    88669.87k    90035.40k


# /usr/local/bin/openssl speed -evp aes-128-cbc
Doing aes-128-cbc for 3s on 16 size blocks: 42103543 aes-128-cbc's in 2.99s
Doing aes-128-cbc for 3s on 64 size blocks: 15075790 aes-128-cbc's in 3.03s
Doing aes-128-cbc for 3s on 256 size blocks: 4999768 aes-128-cbc's in 3.07s
Doing aes-128-cbc for 3s on 1024 size blocks: 1347537 aes-128-cbc's in 3.09s
Doing aes-128-cbc for 3s on 8192 size blocks: 169767 aes-128-cbc's in 3.05s
LibreSSL 2.6.4
built on: date not available
options:bn(64,64) rc4(8x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: information not available
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc     225138.53k   318301.22k   416876.33k   447150.30k   456445.13k


# openssl speed -evp aes-128-cbc
Doing aes-128-cbc for 3s on 16 size blocks: 46193717 aes-128-cbc's in 3.01s
Doing aes-128-cbc for 3s on 64 size blocks: 15384777 aes-128-cbc's in 2.99s
Doing aes-128-cbc for 3s on 256 size blocks: 4979313 aes-128-cbc's in 3.04s
Doing aes-128-cbc for 3s on 1024 size blocks: 1350271 aes-128-cbc's in 3.09s
Doing aes-128-cbc for 3s on 8192 size blocks: 167022 aes-128-cbc's in 3.00s
OpenSSL 1.0.2k-freebsd  26 Jan 2017
built on: date not available
options:bn(64,64) rc4(8x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128-cbc     245726.58k   329065.52k   419439.92k   448057.52k   456081.41k


# openssl speed aes-128-cbc
Doing aes-128 cbc for 3s on 16 size blocks: 5905486 aes-128 cbc's in 3.02s
Doing aes-128 cbc for 3s on 64 size blocks: 1558719 aes-128 cbc's in 3.01s
Doing aes-128 cbc for 3s on 256 size blocks: 399193 aes-128 cbc's in 3.01s
Doing aes-128 cbc for 3s on 1024 size blocks: 262289 aes-128 cbc's in 3.00s
Doing aes-128 cbc for 3s on 8192 size blocks: 33245 aes-128 cbc's in 3.02s
OpenSSL 1.0.2k-freebsd  26 Jan 2017
built on: date not available
options:bn(64,64) rc4(8x,int) des(idx,cisc,16,int) aes(partial) idea(int) blowfish(idx) 
compiler: clang
The 'numbers' are in 1000s of bytes per second processed.
type             16 bytes     64 bytes    256 bytes   1024 bytes   8192 bytes
aes-128 cbc      31251.77k    33166.30k    33975.99k    89527.98k    90077.28k

Thanks for the report!