Page MenuHome GnuPG

AES block modes for PPC
Needs ReviewPublic

Authored by slandden on Jun 3 2019, 5:38 AM.

Details

Summary

libgcrypt's decision to use a 128-bit counter for CTR mode is highly unusual and I can't find that in any RFCs (they either use 64-bit or 32-bit counters), but I implemented it to pass the tests.

I ran all the tests on ppc64el and amd64

see T4529

I need to disclaim that these benchmarks are all wrong because these chips have a "turbo" mode, so the CPUs run at a faster clock speed for short bursts of time.

AES | nanosecs/byte mebibytes/sec cycles/byte auto Mhz

 ECB enc |      2.84 ns/B     336.1 MiB/s      5.38 c/B      1895
 ECB dec |      2.89 ns/B     330.6 MiB/s      5.47 c/B      1895
 CBC enc |      1.05 ns/B     908.3 MiB/s      1.99 c/B      1895
 CBC dec |     0.221 ns/B      4315 MiB/s     0.419 c/B      1895
 CFB enc |      4.41 ns/B     216.4 MiB/s      8.35 c/B      1895
 CFB dec |      4.88 ns/B     195.3 MiB/s      9.26 c/B      1895
 OFB enc |      5.06 ns/B     188.4 MiB/s      9.59 c/B      1895
 OFB dec |      5.07 ns/B     188.2 MiB/s      9.60 c/B      1895
 CTR enc |     0.218 ns/B      4374 MiB/s     0.413 c/B      1895
 CTR dec |     0.219 ns/B      4349 MiB/s     0.416 c/B      1895
 XTS enc |     0.681 ns/B      1400 MiB/s      1.29 c/B      1895
 XTS dec |     0.687 ns/B      1387 MiB/s      1.30 c/B      1895
 CCM enc |      4.21 ns/B     226.4 MiB/s      5.32 c/B      1264
 CCM dec |      4.21 ns/B     226.7 MiB/s      5.32 c/B      1264
CCM auth |      3.99 ns/B     239.2 MiB/s      5.04 c/B      1264
 EAX enc |      4.20 ns/B     227.2 MiB/s      5.30 c/B      1264
 EAX dec |      4.21 ns/B     226.5 MiB/s      5.32 c/B      1264
EAX auth |      3.97 ns/B     239.9 MiB/s      5.02 c/B      1264
 GCM enc |     19.81 ns/B     48.14 MiB/s     25.03 c/B      1264
 GCM dec |     19.79 ns/B     48.18 MiB/s     25.01 c/B      1264
GCM auth |     19.55 ns/B     48.78 MiB/s     24.71 c/B      1264
 OCB enc |     17.53 ns/B     54.41 MiB/s     14.77 c/B     842.4
 OCB dec |     13.89 ns/B     68.67 MiB/s     17.55 c/B      1263
OCB auth |      9.14 ns/B     104.4 MiB/s     11.54 c/B      1264

big-endian:
AES | nanosecs/byte mebibytes/sec cycles/byte auto Mhz

 ECB enc |      2.17 ns/B     440.1 MiB/s      4.11 c/B      1895
 ECB dec |      2.49 ns/B     382.3 MiB/s      4.73 c/B      1895
 CBC enc |      1.05 ns/B     908.8 MiB/s      1.99 c/B      1895
 CBC dec |     0.201 ns/B      4748 MiB/s     0.381 c/B      1895
 CFB enc |      2.07 ns/B     460.2 MiB/s      3.93 c/B      1895
 CFB dec |      2.07 ns/B     460.0 MiB/s      3.93 c/B      1895
 OFB enc |      2.54 ns/B     375.3 MiB/s      4.82 c/B      1895
 OFB dec |      2.11 ns/B     451.9 MiB/s      4.00 c/B      1895
 CTR enc |     0.207 ns/B      4609 MiB/s     0.261 c/B      1264
 CTR dec |     0.207 ns/B      4611 MiB/s     0.261 c/B      1264
 XTS enc |     0.564 ns/B      1689 MiB/s      1.07 c/B      1895
 XTS dec |     0.562 ns/B      1697 MiB/s      1.07 c/B      1895
 CCM enc |      2.28 ns/B     419.1 MiB/s      4.31 c/B      1895
 CCM dec |      2.28 ns/B     417.9 MiB/s      4.33 c/B      1895
CCM auth |      2.07 ns/B     459.9 MiB/s      3.93 c/B      1895
 EAX enc |      2.28 ns/B     418.8 MiB/s      4.32 c/B      1895
 EAX dec |      2.28 ns/B     418.4 MiB/s      4.32 c/B      1895
EAX auth |      2.07 ns/B     460.2 MiB/s      3.93 c/B      1895
 GCM enc |      3.17 ns/B     301.1 MiB/s      6.00 c/B      1895
 GCM dec |      3.17 ns/B     300.9 MiB/s      4.01 c/B      1264
GCM auth |      2.96 ns/B     322.0 MiB/s      5.61 c/B      1895
 OCB enc |      2.08 ns/B     458.1 MiB/s      3.95 c/B      1895
 OCB dec |      2.07 ns/B     461.8 MiB/s      3.91 c/B      1895
OCB auth |      2.08 ns/B     458.4 MiB/s      3.94 c/B      1895
Test Plan

see above benchmarks, is tested by tests/basic.

Diff Detail

Lint
Lint Skipped
Unit
Unit Tests Skipped

Event Timeline

slandden retitled this revision from SOME of the AES block modes for PPC to AES block modes for PPC.
slandden edited the summary of this revision. (Show Details)
slandden edited the test plan for this revision. (Show Details)
slandden added reviewers: six-heiko, jukivili.

added CTR mode support

ECB is not bulk optimized in libgcrypt. I've send patches to add this in past but this was rejected on grounds that ECB is insecure and should not be used.

128-bit CTR mode is implemented in for example OpenSSL. Difference here is that libgcrypt bulk functions do the full 128bit counter increments while OpenSSL bulk function does 32bit counter and upper 96bits are handled in CRYPTO_ctr128_encrypt_ctr32 function. I'll probably implement ctr bulk functions as ctr32 in future and use same ctr32->ctr128 wrapper approach as you have used here.

The openssl version is a 64-bit counter (at least for ppc8), not 32-bit.

I've added few new CTR test vectors to tests/basic.c for checking 32-bit and 64-bit carry overflow cases, rC971d372f512ff6805d5b8b54e9ac1446f3f66643

fix ctr mode when counter overflows.

Have you considered working on bulk CFB-decryption and OCB-enc/dec? Those are the block cipher modes used by GnuPG (OCB is new AEAD mode to be used starting with 2.3).

Regarding OCB: I do not want to touch a patent-encumbered algorithm (3 more years) which claims to force only GPL usage of libgcrypt[1].

[1] section 1.6 https://web.cs.ucdavis.edu/~rogaway/ocb/license1.pdf

fix building with hard ware acceleration off.

I can't see any specific claim to the GPL. License 1 grants a royality free license for all open source implementations defined by the OSI. This includes the LGPL.
If you use Libgcrypt in non-open-source software you may get a free license using License 2.

I will work on OCB mode, eventually. Perhaps you could review what I have, but leave T4529 open until OCB mode is completed.