[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aB7o2SiwNLfBTgQk@gondor.apana.org.au>
Date: Sat, 10 May 2025 13:49:13 +0800
From: Herbert Xu <herbert@...dor.apana.org.au>
To: Eric Biggers <ebiggers@...nel.org>
Cc: Venkat Rao Bagalkote <venkat88@...ux.ibm.com>,
Thorsten Leemhuis <linux@...mhuis.info>,
Linux Crypto Mailing List <linux-crypto@...r.kernel.org>,
LKML <linux-kernel@...r.kernel.org>,
Linux Next Mailing List <linux-next@...r.kernel.org>,
Madhavan Srinivasan <maddy@...ux.ibm.com>,
Stephen Rothwell <sfr@...b.auug.org.au>,
Danny Tsen <dtsen@...ux.ibm.com>, linuxppc-dev@...ts.ozlabs.org
Subject: Re: [v2 PATCH] crypto: powerpc/poly1305 - Add poly1305_emit_arch
wrapper
On Fri, May 09, 2025 at 10:33:08PM -0700, Eric Biggers wrote:
>
> Yes, the PowerPC Poly1305 code incorrectly uses VSX without first checking
> crypto_simd_usable(). And PowerPC also doesn't support VSX in softirqs, or at
> least it doesn't claim to (it doesn't override may_use_simd(), so it gets the
> default from include/asm-generic/simd.h which returns false in softirq context).
> Maybe add 'depends on BROKEN' to CRYPTO_POLY1305_P10 for now, and give the
> PowerPC folks (Cc'ed) a chance to fix this before removing the code.
I just noticed something weird with this code, running a speed
test using "modprobe tcrypt mode=217" shows that the p10 version
of poly1305 is way slower than the generic:
qemu P9 CPU:
May 10 13:36:46 test-p10 kernel: [ 59.585264][ T374] tcrypt: testing speed of multibuffer rfc7539esp(chacha20,poly1305) (rfc7539esp(chacha20-generic,poly1305-generic)) encryption
May 10 13:36:46 test-p10 kernel: [ 59.586011][ T374] tcrypt: test 0 (288 bit key, 16 byte blocks): 1 operation in 1374 cycles (16 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.587446][ T374] tcrypt: test 1 (288 bit key, 64 byte blocks): 1 operation in 1359 cycles (64 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.588025][ T374] tcrypt: test 2 (288 bit key, 256 byte blocks): 1 operation in 1778 cycles (256 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.588639][ T374] tcrypt: test 3 (288 bit key, 512 byte blocks): 1 operation in 2323 cycles (512 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.589342][ T374] tcrypt: test 4 (288 bit key, 1024 byte blocks): 1 operation in 31624 cycles (1024 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.594178][ T374] tcrypt: test 5 (288 bit key, 1420 byte blocks): 1 operation in 4408 cycles (1420 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.595317][ T374] tcrypt: test 6 (288 bit key, 4096 byte blocks): 1 operation in 9719 cycles (4096 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.597512][ T374] tcrypt: test 7 (288 bit key, 8192 byte blocks): 1 operation in 20168 cycles (8192 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.604616][ T374] tcrypt: testing speed of multibuffer rfc7539esp(chacha20,poly1305) (rfc7539esp(chacha20-generic,poly1305-generic)) decryption
May 10 13:36:46 test-p10 kernel: [ 59.604916][ T374] tcrypt: test 0 (288 bit key, 16 byte blocks): 1 operation in 1356 cycles (16 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.605564][ T374] tcrypt: test 1 (288 bit key, 64 byte blocks): 1 operation in 1393 cycles (64 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.608308][ T374] tcrypt: test 2 (288 bit key, 256 byte blocks): 1 operation in 1845 cycles (256 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.609002][ T374] tcrypt: test 3 (288 bit key, 512 byte blocks): 1 operation in 2392 cycles (512 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.612109][ T374] tcrypt: test 4 (288 bit key, 1024 byte blocks): 1 operation in 3349 cycles (1024 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.613289][ T374] tcrypt: test 5 (288 bit key, 1420 byte blocks): 1 operation in 4418 cycles (1420 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.616233][ T374] tcrypt: test 6 (288 bit key, 4096 byte blocks): 1 operation in 21600 cycles (4096 bytes)
May 10 13:36:46 test-p10 kernel: [ 59.620221][ T374] tcrypt: test 7 (288 bit key, 8192 byte blocks): 1 operation in 20013 cycles (8192 bytes)
qemu P10 CPU:
May 10 13:40:56 test-p10 kernel: [ 91.672877][ T392] tcrypt: testing speed of multibuffer rfc7539esp(chacha20,poly1305) (rfc7539esp(chacha20-powerpc,poly1305-generic)) encryption
May 10 13:40:56 test-p10 kernel: [ 91.674615][ T392] tcrypt: test 0 (288 bit key, 16 byte blocks): 1 operation in 1471 cycles (16 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.680240][ T392] tcrypt: test 1 (288 bit key, 64 byte blocks): 1 operation in 1733 cycles (64 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.682975][ T392] tcrypt: test 2 (288 bit key, 256 byte blocks): 1 operation in 3248 cycles (256 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.684445][ T392] tcrypt: test 3 (288 bit key, 512 byte blocks): 1 operation in 15211 cycles (512 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.687603][ T392] tcrypt: test 4 (288 bit key, 1024 byte blocks): 1 operation in 20500 cycles (1024 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.690926][ T392] tcrypt: test 5 (288 bit key, 1420 byte blocks): 1 operation in 10159 cycles (1420 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.695009][ T392] tcrypt: test 6 (288 bit key, 4096 byte blocks): 1 operation in 25917 cycles (4096 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.701320][ T392] tcrypt: test 7 (288 bit key, 8192 byte blocks): 1 operation in 63352 cycles (8192 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.713863][ T392] tcrypt: testing speed of multibuffer rfc7539esp(chacha20,poly1305) (rfc7539esp(chacha20-powerpc,poly1305-generic)) decryption
May 10 13:40:56 test-p10 kernel: [ 91.714182][ T392] tcrypt: test 0 (288 bit key, 16 byte blocks): 1 operation in 1502 cycles (16 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.714871][ T392] tcrypt: test 1 (288 bit key, 64 byte blocks): 1 operation in 1778 cycles (64 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.715508][ T392] tcrypt: test 2 (288 bit key, 256 byte blocks): 1 operation in 3322 cycles (256 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.716463][ T392] tcrypt: test 3 (288 bit key, 512 byte blocks): 1 operation in 20980 cycles (512 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.720775][ T392] tcrypt: test 4 (288 bit key, 1024 byte blocks): 1 operation in 8000 cycles (1024 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.724348][ T392] tcrypt: test 5 (288 bit key, 1420 byte blocks): 1 operation in 10155 cycles (1420 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.727952][ T392] tcrypt: test 6 (288 bit key, 4096 byte blocks): 1 operation in 27711 cycles (4096 bytes)
May 10 13:40:56 test-p10 kernel: [ 91.735306][ T392] tcrypt: test 7 (288 bit key, 8192 byte blocks): 1 operation in 52874 cycles (8192 bytes)
Did I do something wrong?
Cheers,
--
Email: Herbert Xu <herbert@...dor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Powered by blists - more mailing lists