| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <1467308170.3969.1.camel@megha-Z97X-UD7-TH>
Date: Thu, 30 Jun 2016 10:36:10 -0700
From: Megha Dey <megha.dey@...ux.intel.com>
To: Herbert Xu <herbert@...dor.apana.org.au>
Cc: Krzysztof Kozlowski <k.kozlowski@...sung.com>,
"David S. Miller" <davem@...emloft.net>,
linux-crypto@...r.kernel.org, linux-kernel@...r.kernel.org,
Bartlomiej Zolnierkiewicz <b.zolnierkie@...sung.com>,
Fenghua Yu <fenghua.yu@...el.com>,
Tim Chen <tim.c.chen@...ux.intel.com>
Subject: Re: crypto: tcrypt - Do not bail on EINPROGRESS in multibuffer hash
test
On Thu, 2016-06-30 at 11:00 +0800, Herbert Xu wrote:
> On Wed, Jun 29, 2016 at 10:45:56AM -0700, Megha Dey wrote:
> > I tested the latest cryptodev tree on my haswell machine and this is
> > what I see:
> > [ 40.402834] modprobe tcrypt mode=422
> > [ 40.403105] testing speed of multibuffer sha1 (sha1_mb)
> > [ 40.403108] test 0 ( 16 byte blocks, 16 bytes per update, 1
> > updates): 32271 cycles/operation, 252 cycles/byte
> > [ 40.403118] At least one hashing failed ret=-115
> > [ 43.218712] modprobe tcrypt mode=423
> > [ 43.218712] testing speed of multibuffer sha256 (sha256_mb)
> > [ 43.218715] test 0 ( 16 byte blocks, 16 bytes per update, 1
> > updates): 106965 cycles/operation, 835 cycles/byte
> > [ 43.218747] At least one hashing failed ret=-115
> > [ 45.346657] modprobe tcrypt mode=424
> > [ 45.346657] testing speed of multibuffer sha512 (sha512_mb)
> > [ 45.346660] test 0 ( 16 byte blocks, 16 bytes per update, 1
> > updates): 43179 cycles/operation, 337 cycles/byte
> > [ 45.346673] At least one hashing failed ret=-115
> >
> > Don't think this is expected, is it?
>
> No that's not right. Does this patch fix it?
>
yes, it fixes it. This is my output:
testing speed of multibuffer sha1 (sha1_mb)
[ 31.342538] test 0 ( 16 byte blocks, 16 bytes per update, 1
updates):
31614 cycles/operation, 246 cycles/byte
[ 31.342549] test 2 ( 64 byte blocks, 64 bytes per update, 1
updates):
10893 cycles/operation, 21 cycles/byte
[ 31.342553] test 5 ( 256 byte blocks, 256 bytes per update, 1
updates):
11043 cycles/operation, 5 cycles/byte
[ 31.342557] test 8 ( 1024 byte blocks, 1024 bytes per update, 1
updates):
17355 cycles/operation, 2 cycles/byte
[ 31.342562] test 12 ( 2048 byte blocks, 2048 bytes per update, 1
updates):
26328 cycles/operation, 1 cycles/byte
[ 31.342570] test 16 ( 4096 byte blocks, 4096 bytes per update, 1
updates):
43833 cycles/operation, 1 cycles/byte
[ 31.342584] test 21 ( 8192 byte blocks, 8192 bytes per update, 1
updates):
81141 cycles/operation, 1 cycles/byte
[ 44.700088]
[ 44.700088] testing speed of multibuffer sha256 (sha256_mb)
[ 44.700091] test 0 ( 16 byte blocks, 16 bytes per update, 1
updates):
116874 cycles/operation, 913 cycles/byte
[ 44.700126] test 2 ( 64 byte blocks, 64 bytes per update, 1
updates):
60894 cycles/operation, 118 cycles/byte
[ 44.700144] test 5 ( 256 byte blocks, 256 bytes per update, 1
updates):
15366 cycles/operation, 7 cycles/byte
[ 44.700149] test 8 ( 1024 byte blocks, 1024 bytes per update, 1
updates):
29130 cycles/operation, 3 cycles/byte
[ 44.700158] test 12 ( 2048 byte blocks, 2048 bytes per update, 1
updates):
48561 cycles/operation, 2 cycles/byte
[ 44.700173] test 16 ( 4096 byte blocks, 4096 bytes per update, 1
updates):
119697 cycles/operation, 3 cycles/byte
[ 44.700207] test 21 ( 8192 byte blocks, 8192 bytes per update, 1
updates):
164163 cycles/operation, 2 cycles/byte
[ 66.309608]
[ 66.309608] testing speed of multibuffer sha512 (sha512_mb)
[ 66.309623] test 0 ( 16 byte blocks, 16 bytes per update, 1
updates):
301050 cycles/operation, 2351 cycles/byte
[ 66.309717] test 2 ( 64 byte blocks, 64 bytes per update, 1
updates):
115416 cycles/operation, 225 cycles/byte
[ 66.309757] test 5 ( 256 byte blocks, 256 bytes per update, 1
updates):
175068 cycles/operation, 85 cycles/byte
[ 66.309814] test 8 ( 1024 byte blocks, 1024 bytes per update, 1
updates):
354834 cycles/operation, 43 cycles/byte
[ 66.309920] test 12 ( 2048 byte blocks, 2048 bytes per update, 1
updates):
599814 cycles/operation, 36 cycles/byte
[ 66.310096] test 16 ( 4096 byte blocks, 4096 bytes per update, 1
updates):
585666 cycles/operation, 17 cycles/byte
[ 66.310263] test 21 ( 8192 byte blocks, 8192 bytes per update, 1
updates):
1057770 cycles/operation, 16 cycles/byte
> Thanks!
>
> ---8<---
> The multibuffer hash speed test is incorrectly bailing because
> of an EINPROGRESS return value. This patch fixes it by setting
> ret to zero if it is equal to -EINPROGRESS.
>
> Reported-by: Megha Dey <megha.dey@...ux.intel.com>
> Signed-off-by: Herbert Xu <herbert@...dor.apana.org.au>
>
> diff --git a/crypto/tcrypt.c b/crypto/tcrypt.c
> index 8a91dc3..202cfa1 100644
> --- a/crypto/tcrypt.c
> +++ b/crypto/tcrypt.c
> @@ -655,8 +486,10 @@ static void test_mb_ahash_speed(const char *algo, unsigned int sec,
>
> for (k = 0; k < 8; k++) {
> ret = crypto_ahash_digest(data[k].req);
> - if (ret == -EINPROGRESS)
> + if (ret == -EINPROGRESS) {
> + ret = 0;
> continue;
> + }
>
> if (ret)
> break;
Powered by blists - more mailing lists