From: Dmitry K. <dm...@dz...> - 2007-02-26 14:44:48
|
Hello, I'm not quite new to devil-linux and have used the distro for several projects in the past. But it's always a pain when you want to rebuild it yourself - even the "stable" 1.2.xx branch. :( Now again I need to compile a smaller version of DL with reduced number of applications. But the build process stops at jailkit claiming I need to have Python installed. Any ideas? Python is enabled in .config file, base system is lfssystem-cvs-20030127-i486-stripped.tar.bz2 -- Best regards, Dmitry Komarov IT Project Manager SIA CITYNET 22/24 - 500 Katolu str. Riga, LV-1003 Latvia |
From: Bruce S. <bw...@ar...> - 2007-02-26 15:05:13
|
> I'm not quite new to devil-linux and have used the distro for several projects > in the past. But it's always a pain when you want to rebuild it yourself - > even the "stable" 1.2.xx branch. :( > > Now again I need to compile a smaller version of DL with reduced number of > applications. But the build process stops at jailkit claiming I need to have > Python installed. > > Any ideas? Python is enabled in .config file, base system is > lfssystem-cvs-20030127-i486-stripped.tar.bz2 It's stopping in the 'build' phase (as opposed to the 'install' phase)? Check the logs for jailkit and Python in /data/build/tmp/LOGS/build (in the LFS chroot). Anything helpful there? - BS |
From: Dmitry K. <dm...@dz...> - 2007-02-26 15:26:38
|
On Monday 26 February 2007 17:05, Bruce Smith wrote: > > I'm not quite new to devil-linux and have used the distro for several > > projects in the past. But it's always a pain when you want to rebuild it > > yourself - even the "stable" 1.2.xx branch. :( > > > > Now again I need to compile a smaller version of DL with reduced number > > of applications. But the build process stops at jailkit claiming I need > > to have Python installed. > > > > Any ideas? Python is enabled in .config file, base system is > > lfssystem-cvs-20030127-i486-stripped.tar.bz2 > > It's stopping in the 'build' phase (as opposed to the 'install' phase)? > > Check the logs for jailkit and Python in /data/build/tmp/LOGS/build > (in the LFS chroot). Anything helpful there? > Yes, it's stopping at build phase. Nothing more helpfull in the log except the statement that I have to have Python installed for jailkit to compile. I even tried to make a fake "done" flag for jailkit and wait for the other packages to compile first (it helped me in the past in some cases). But even if I try to build the jailkit as the last package - the same error there. -- Best regards, Dmitry Komarov IT Project Manager SIA CITYNET 22/24 - 500 Katolu str. Riga, LV-1003 Latvia |
From: Bruce S. <bw...@ar...> - 2007-02-26 15:57:50
|
> > > I'm not quite new to devil-linux and have used the distro for several > > > projects in the past. But it's always a pain when you want to rebuild it > > > yourself - even the "stable" 1.2.xx branch. :( > > > > > > Now again I need to compile a smaller version of DL with reduced number > > > of applications. But the build process stops at jailkit claiming I need > > > to have Python installed. > > > > > > Any ideas? Python is enabled in .config file, base system is > > > lfssystem-cvs-20030127-i486-stripped.tar.bz2 > > > > It's stopping in the 'build' phase (as opposed to the 'install' phase)? > > > > Check the logs for jailkit and Python in /data/build/tmp/LOGS/build > > (in the LFS chroot). Anything helpful there? > > Yes, it's stopping at build phase. Nothing more helpfull in the log except the > statement that I have to have Python installed for jailkit to compile. > > I even tried to make a fake "done" flag for jailkit and wait for the other > packages to compile first (it helped me in the past in some cases). But even > if I try to build the jailkit as the last package - the same error there. Are you SURE that Python built OK? Check it's log file. Inside the chroot, type: python followed by two TAB's. It should list all python binaries in your path: root:/data/build# python python python2 python2.3 -BS |
From: Dmitry K. <dm...@dz...> - 2007-02-26 16:17:37
|
On Monday 26 February 2007 17:57, Bruce Smith wrote: > > > > I'm not quite new to devil-linux and have used the distro for several > > > > projects in the past. But it's always a pain when you want to rebuild > > > > it yourself - even the "stable" 1.2.xx branch. :( > > > > > > > > Now again I need to compile a smaller version of DL with reduced > > > > number of applications. But the build process stops at jailkit > > > > claiming I need to have Python installed. > > > > > > > > Any ideas? Python is enabled in .config file, base system is > > > > lfssystem-cvs-20030127-i486-stripped.tar.bz2 > > > > > > It's stopping in the 'build' phase (as opposed to the 'install' phase)? > > > > > > Check the logs for jailkit and Python in /data/build/tmp/LOGS/build > > > (in the LFS chroot). Anything helpful there? > > > > Yes, it's stopping at build phase. Nothing more helpfull in the log > > except the statement that I have to have Python installed for jailkit to > > compile. > > > > I even tried to make a fake "done" flag for jailkit and wait for the > > other packages to compile first (it helped me in the past in some cases). > > But even if I try to build the jailkit as the last package - the same > > error there. > > Are you SURE that Python built OK? Check it's log file. > > Inside the chroot, type: python > followed by two TAB's. It should list all python binaries in your path: > > root:/data/build# python > python python2 python2.3 > Miracle. After removing everything under /build/tmp and re-running make again - everything compiled OK. What can be the issue?.. -- Best regards, Dmitry Komarov IT Project Manager SIA CITYNET 22/24 - 500 Katolu str. Riga, LV-1003 Latvia |
From: Heiko Z. <he...@zu...> - 2007-02-27 01:35:23
|
On Mon, February 26, 2007 10:18, Dmitry Komarov wrote: > On Monday 26 February 2007 17:57, Bruce Smith wrote: > >>>>> I'm not quite new to devil-linux and have used the distro for >>>>> several projects in the past. But it's always a pain when you want >>>>> to rebuild it yourself - even the "stable" 1.2.xx branch. :( >>>>> >>>>> Now again I need to compile a smaller version of DL with reduced >>>>> number of applications. But the build process stops at jailkit >>>>> claiming I need to have Python installed. >>>>> >>>>> Any ideas? Python is enabled in .config file, base system is >>>>> lfssystem-cvs-20030127-i486-stripped.tar.bz2 >>>> >>>> It's stopping in the 'build' phase (as opposed to the 'install' >>>> phase)? >>>> >>>> Check the logs for jailkit and Python in /data/build/tmp/LOGS/build >>>> (in the LFS chroot). Anything helpful there? >>>> >>> >>> Yes, it's stopping at build phase. Nothing more helpfull in the log >>> except the statement that I have to have Python installed for jailkit >>> to compile. >>> >>> I even tried to make a fake "done" flag for jailkit and wait for the >>> other packages to compile first (it helped me in the past in some >>> cases). But even if I try to build the jailkit as the last package - >>> the same error there. >> >> Are you SURE that Python built OK? Check it's log file. >> >> >> Inside the chroot, type: python >> followed by two TAB's. It should list all python binaries in your path: >> >> >> root:/data/build# python >> python python2 python2.3 >> > > Miracle. > > > After removing everything under /build/tmp and re-running make again - > everything compiled OK. > > What can be the issue?.. Sometimes things get messed up when you're doing testing. If you have strange problems, there are always 2 things you should do: 1) delete the lfssystem folder and unpack to have a fresh and clean build system 2) run 'make mrproper' to clean up build/tmp -- Regards Heiko Zuerker http://www.devil-linux.org |
From: Dmitry K. <dm...@dz...> - 2007-10-30 12:56:13
|
Hello, has anyone tried to compile DL with in-kernel VIA C3/C7 hardware crypto engine support (VIA PadLock) and other required C3/C7 patches for VPN/SSL acceleration? Any success stories or suggestions? -- Regards, Dmit |
From: Heiko Z. <he...@zu...> - 2007-10-30 13:07:03
|
Hey, On Tue, October 30, 2007 07:56, Dmitry Komarov wrote: > has anyone tried to compile DL with in-kernel VIA C3/C7 hardware crypto > engine support (VIA PadLock) and other required C3/C7 patches for VPN/SSL > acceleration? > > Any success stories or suggestions? Is this standard part of the 2.4 or 2.6 series, or does it require additional patches? -- Regards Heiko Zuerker http://www.devil-linux.org |
From: Dmitry K. <dm...@dz...> - 2007-10-30 13:30:32
|
On Tuesday 30 October 2007 15:06:52 Heiko Zuerker wrote: > Hey, > > On Tue, October 30, 2007 07:56, Dmitry Komarov wrote: > > has anyone tried to compile DL with in-kernel VIA C3/C7 hardware crypto > > engine support (VIA PadLock) and other required C3/C7 patches for VPN/SSL > > acceleration? > > > > Any success stories or suggestions? > > Is this standard part of the 2.4 or 2.6 series, or does it require > additional patches? VIA C3 support was included starting from 2.6.11 kernels, and VIA C7 SHA1/SHA256 support was added from 2.6.19. But it may require some patches to OpenSSL libs and some other applications though. Here is the link to VIA PadLock support for Linux page: http://www.logix.cz/michal/devel/padlock/ The main idea behind my question is that you can find a lot unexpensive VIA Mini-ITX based boards and systems which are just a perfect platform for Devil Linux distro, especially if it would have built-in support for their harware crypto engines. For $300 incl. shipping I have bought on eBay a lot of 5 brand new(!) MaxSpeed terminal devices, which all are based on VIA C3 CPU, have 512MB RAM, 512MB CompactFlash card with IDE adaptor and a free PCI slot. You can even put a hard drive there if you need. The system is fanless, absolutely quiet and has very low power consumption. Just a perfect deal for a DL based hardware-accelerated VPN appliance or router for such a low price: http://cgi.ebay.com/Lof-of-5-New-MaxSpeed-8300-XP-512Ram-512-Flash-800Mhz_W0QQitemZ330181384438QQcmdZViewItem?hash=item330181384438 |
From: Serge L. <fi...@in...> - 2007-10-30 18:00:20
|
Dmitry Komarov wrote: > On Tuesday 30 October 2007 15:06:52 Heiko Zuerker wrote: >> Hey, >> >> On Tue, October 30, 2007 07:56, Dmitry Komarov wrote: >>> has anyone tried to compile DL with in-kernel VIA C3/C7 hardware crypto >>> engine support (VIA PadLock) and other required C3/C7 patches for VPN/SSL >>> acceleration? >>> >>> Any success stories or suggestions? >> Is this standard part of the 2.4 or 2.6 series, or does it require >> additional patches? > > VIA C3 support was included starting from 2.6.11 kernels, and VIA C7 > SHA1/SHA256 support was added from 2.6.19. > > But it may require some patches to OpenSSL libs and some other applications > though. Here is the link to VIA PadLock support for Linux page: > > http://www.logix.cz/michal/devel/padlock/ I think we can add that code as a patch (may be only optional, it depends on the real advantages of this patch). > > The main idea behind my question is that you can find a lot unexpensive VIA > Mini-ITX based boards and systems which are just a perfect platform for Devil > Linux distro, especially if it would have built-in support for their harware > crypto engines. Have you got any hardware of this type? I'd like to ask you about small testing of crypto possibilities of the board. Please do the following and share the results with us: # openssl OpenSSL> speed After we add the patch I'll ask you to repeat the test and everybody will be able to check speed up of crypto. Is it ok? It's from my system, CPU: Intel(R) Pentium(R) 4 CPU 2.40GHz stepping 01 type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 1083.86k 2286.95k 3166.46k 3495.30k 3618.13k mdc2 0.00 0.00 0.00 0.00 0.00 md4 13056.59k 46606.95k 147900.22k 342027.30k 544493.12k md5 11080.59k 39827.41k 128398.51k 284474.03k 434088.62k hmac(md5) 17638.54k 63578.20k 181733.97k 339369.98k 453118.63k sha1 11778.76k 39273.37k 106729.05k 189599.21k 247474.86k rmd160 10247.03k 32590.93k 78762.58k 124134.40k 149230.93k rc4 184205.26k 260034.69k 279309.65k 286295.04k 288505.86k des cbc 49821.31k 51994.90k 52212.43k 52670.81k 52568.06k des ede3 18528.82k 18745.07k 18887.34k 18931.71k 18890.75k idea cbc 26481.13k 29136.73k 29953.19k 30030.17k 29878.95k rc2 cbc 20763.43k 22201.94k 22351.27k 22460.53k 22382.39k rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 blowfish cbc 81991.24k 87533.23k 88552.62k 89282.22k 89118.04k cast cbc 45847.92k 54174.72k 56665.86k 57541.63k 57308.50k aes-128 cbc 64413.49k 88553.49k 99803.73k 103181.65k 102530.64k aes-192 cbc 38816.20k 65566.70k 81428.99k 86528.34k 87053.65k aes-256 cbc 36123.11k 59030.44k 71636.91k 75593.39k 75780.08k camellia-128 cbc 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 0.00 0.00 0.00 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 0.00 sha256 7236.61k 16962.65k 30844.07k 39323.65k 42707.63k sha512 4643.79k 18706.35k 32713.73k 48179.54k 56016.90k sign verify sign/s verify/s rsa 512 bits 0.000905s 0.000070s 1105.2 14378.3 rsa 1024 bits 0.004064s 0.000182s 246.1 5484.1 rsa 2048 bits 0.022000s 0.000544s 45.5 1838.6 rsa 4096 bits 0.131711s 0.002079s 7.6 480.9 sign verify sign/s verify/s dsa 512 bits 0.000661s 0.000787s 1513.3 1270.5 dsa 1024 bits 0.001750s 0.002148s 571.3 465.5 dsa 2048 bits 0.005299s 0.006536s 188.7 153.0 -- Sincerely, Serge Leschinsky |
From: Philippe M. <ph...@oz...> - 2007-10-31 06:10:19
|
Hi, I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with 2.6.21.5 kernel compiled for the VIA CPU. I get around 4MB/s of VPN throughput using ipsec-tools which is OK for me as my DSL uplink is only around 1Mb/s and I use it with a WiFi linksys router which does not go much faster than 2-3MB/s. Here is the output of openssl (not patched) speed on this system: OpenSSL 0.9.8e 23 Feb 2007 built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic manual page 2007 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 [sysconf value] timing function used: times The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 445.58k 948.54k 1321.30k 1464.32k 1517.44k mdc2 0.00 0.00 0.00 0.00 0.00 md4 4280.63k 14773.95k 42052.07k 77736.62k 103557.33k md5 3946.47k 13920.19k 40439.58k 78488.40k 107598.35k hmac(md5) 5471.18k 18343.91k 49535.49k 85630.29k 109242.24k sha1 3514.75k 10887.00k 25903.43k 39475.71k 46310.83k rmd160 3080.74k 8926.14k 19499.32k 27568.13k 31437.49k rc4 40253.31k 45109.93k 46369.11k 46791.68k 46952.81k des cbc 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea cbc 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 cbc 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k 24228.46k cast cbc 11002.71k 11694.29k 11943.30k 11968.51k 11953.74k aes-128 cbc 14373.09k 16753.07k 17594.91k 17753.43k 17818.97k aes-192 cbc 12547.46k 14325.70k 14905.43k 15092.87k 15097.47k aes-256 cbc 9892.17k 11976.45k 12751.03k 12893.87k 12966.53k camellia-128 cbc 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 0.00 0.00 0.00 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 0.00 sha256 2111.48k 5018.54k 9106.26k 11497.57k 12397.96k sha512 1468.83k 5865.51k 11094.47k 17036.84k 20446.90k sign verify sign/s verify/s rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 sign verify sign/s verify/s dsa 512 bits 0.001708s 0.002093s 585.5 477.9 dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 I should try to patch when I get some time to compare the results. Philippe |
From: Dmitry K. <dm...@dz...> - 2007-11-01 13:58:21
|
Hello Serge, My eBay ordered boards are still in transit, but I have similar MaxTerm 8500 device for tests, the only reasonable difference - it has 1 GHz VIA C3 CPU, not that 800 MHz ones that I was talking about. So here is the OpenSSL speed test, please note - VIA padlock and padlock-aes modules were enabled in kernel configuration and compiled. I have tried the tests several times WITH and WITHOUT those modules loaded. In both cases results are same, that means that OpenSSL patches are still required to utilize the hardware AES crypto. Some dmesg output: ================================================ [ 0.000000] Detected 1002.311 MHz processor. [ 26.631989] Initializing CPU#0 [ 26.711152] Memory: 483528k/491456k available (2108k kernel code, 7316k reserved, 512k data, 184k init, 0k highmem) [ 26.851970] Calibrating delay using timer specific routine.. 2006.53 BogoMIPS (lpj=10032652) [ 26.852305] CPU: L1 I Cache: 64K (32 bytes/line), D cache 64K (32 bytes/line) [ 26.852314] CPU: L2 Cache: 64K (32 bytes/line) [ 26.852367] CPU: Centaur VIA Nehemiah stepping 08 OpenSSL speed tests: ================================================ The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 427.26k 909.08k 1264.95k 1403.32k 1450.61k mdc2 1081.53k 1224.33k 1261.57k 1277.11k 1281.87k md4 3817.89k 13035.86k 36549.49k 66307.23k 87466.61k md5 3446.69k 12104.06k 35402.50k 67964.00k 93546.65k hmac(md5) 4552.84k 15122.12k 41422.93k 73146.26k 94700.61k sha1 3146.50k 9652.37k 22729.31k 34346.18k 40459.23k rmd160 2735.65k 7883.50k 17057.85k 24027.60k 27352.03k rc4 35043.58k 39253.07k 40256.13k 40701.11k 40834.81k des cbc 10115.70k 10590.26k 10729.89k 10764.93k 10777.51k des ede3 3627.33k 3690.21k 3707.92k 3712.60k 3714.98k idea cbc 6594.41k 6895.35k 6977.91k 6998.92k 7005.38k rc2 cbc 4136.49k 4306.41k 4363.69k 4362.38k 4368.16k rc5-32/12 cbc 32464.64k 36130.49k 37608.95k 37989.04k 38113.21k blowfish cbc 19194.56k 20598.27k 20857.79k 20971.25k 21013.43k cast cbc 9159.76k 9673.53k 9830.74k 9872.58k 9884.83k aes-128 cbc 11937.15k 14327.41k 15190.16k 15424.30k 15494.04k aes-192 cbc 10543.73k 12292.98k 12915.50k 13081.69k 13131.69k aes-256 cbc 9359.60k 10550.16k 11006.55k 11127.92k 11201.19k camellia-128 cbc 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 0.00 0.00 0.00 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 0.00 sha256 2078.33k 4901.00k 8810.48k 11019.06k 11885.20k sha512 204.22k 817.01k 1146.73k 1556.75k 1739.10k sign verify sign/s verify/s rsa 512 bits 0.003140s 0.000246s 318.5 4063.7 rsa 1024 bits 0.014362s 0.000686s 69.6 1456.7 rsa 2048 bits 0.081129s 0.002224s 12.3 449.6 rsa 4096 bits 0.524500s 0.007863s 1.9 127.2 sign verify sign/s verify/s dsa 512 bits 0.002332s 0.002795s 428.9 357.8 dsa 1024 bits 0.006523s 0.007810s 153.3 128.0 dsa 2048 bits 0.021410s 0.025522s 46.7 39.2 On Tuesday 30 October 2007 20:00:04 Serge Leschinsky wrote: > Dmitry Komarov wrote: > > On Tuesday 30 October 2007 15:06:52 Heiko Zuerker wrote: > >> Hey, > >> > >> On Tue, October 30, 2007 07:56, Dmitry Komarov wrote: > >>> has anyone tried to compile DL with in-kernel VIA C3/C7 hardware crypto > >>> engine support (VIA PadLock) and other required C3/C7 patches for > >>> VPN/SSL acceleration? > >>> > >>> Any success stories or suggestions? > >> > >> Is this standard part of the 2.4 or 2.6 series, or does it require > >> additional patches? > > > > VIA C3 support was included starting from 2.6.11 kernels, and VIA C7 > > SHA1/SHA256 support was added from 2.6.19. > > > > But it may require some patches to OpenSSL libs and some other > > applications though. Here is the link to VIA PadLock support for Linux > > page: > > > > http://www.logix.cz/michal/devel/padlock/ > > I think we can add that code as a patch (may be only optional, it depends > on the real advantages of this patch). > > > The main idea behind my question is that you can find a lot unexpensive > > VIA Mini-ITX based boards and systems which are just a perfect platform > > for Devil Linux distro, especially if it would have built-in support for > > their harware crypto engines. > > Have you got any hardware of this type? I'd like to ask you about small > testing of crypto possibilities of the board. Please do the following and > share the results with us: > > # openssl > OpenSSL> speed > > After we add the patch I'll ask you to repeat the test and everybody will > be able to check speed up of crypto. Is it ok? > > It's from my system, CPU: Intel(R) Pentium(R) 4 CPU 2.40GHz stepping 01 > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes md2 1083.86k 2286.95k 3166.46k 3495.30k > 3618.13k mdc2 0.00 0.00 0.00 0.00 > 0.00 md4 13056.59k 46606.95k 147900.22k > 342027.30k 544493.12k md5 11080.59k 39827.41k > 128398.51k 284474.03k 434088.62k hmac(md5) 17638.54k > 63578.20k 181733.97k 339369.98k 453118.63k sha1 11778.76k > 39273.37k 106729.05k 189599.21k 247474.86k rmd160 > 10247.03k 32590.93k 78762.58k 124134.40k 149230.93k rc4 > 184205.26k 260034.69k 279309.65k 286295.04k 288505.86k des cbc > 49821.31k 51994.90k 52212.43k 52670.81k 52568.06k des > ede3 18528.82k 18745.07k 18887.34k 18931.71k 18890.75k > idea cbc 26481.13k 29136.73k 29953.19k 30030.17k > 29878.95k rc2 cbc 20763.43k 22201.94k 22351.27k 22460.53k > 22382.39k rc5-32/12 cbc 0.00 0.00 0.00 > 0.00 0.00 blowfish cbc 81991.24k 87533.23k 88552.62k > 89282.22k 89118.04k cast cbc 45847.92k 54174.72k 56665.86k > 57541.63k 57308.50k aes-128 cbc 64413.49k 88553.49k > 99803.73k 103181.65k 102530.64k aes-192 cbc 38816.20k 65566.70k > 81428.99k 86528.34k 87053.65k aes-256 cbc 36123.11k > 59030.44k 71636.91k 75593.39k 75780.08k camellia-128 cbc > 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc > 0.00 0.00 0.00 0.00 0.00 camellia-256 > cbc 0.00 0.00 0.00 0.00 0.00 sha256 > 7236.61k 16962.65k 30844.07k 39323.65k 42707.63k > sha512 4643.79k 18706.35k 32713.73k 48179.54k > 56016.90k sign verify sign/s verify/s > rsa 512 bits 0.000905s 0.000070s 1105.2 14378.3 > rsa 1024 bits 0.004064s 0.000182s 246.1 5484.1 > rsa 2048 bits 0.022000s 0.000544s 45.5 1838.6 > rsa 4096 bits 0.131711s 0.002079s 7.6 480.9 > sign verify sign/s verify/s > dsa 512 bits 0.000661s 0.000787s 1513.3 1270.5 > dsa 1024 bits 0.001750s 0.002148s 571.3 465.5 > dsa 2048 bits 0.005299s 0.006536s 188.7 153.0 |
From: Dmitry K. <dm...@dz...> - 2007-11-01 15:01:32
|
Hello Serge, some tests taken on another board with VIA C7 1GHz CPU and Gentoo OS: 1) NO hardware crypto enabled: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 byt= es aes-256-ecb =A0=A0=A0=A0=A015138.32k =A0=A0=A015945.93k =A0=A0=A016177.57k = =A0=A0=A016229.37k =A0=A0=A016274.04k sha256 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A02469.35k =A0=A0=A0=A06240.97k =A0= =A0=A011802.20k =A0=A0=A015185.34k =A0=A0=A016643.28k 2) Hardware crypto engine enabled (padlock, padlock-aes, padlock-sha module= s=20 loaded, OpenSSL patches appplied): type =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A016 bytes =A0=A0=A0=A064 bytes =A0= =A0=A0256 bytes =A0=A01024 bytes =A0=A08192 bytes aes-256-ecb =A0=A0=A0=A0107931.79k =A0=A0381123.03k =A0=A0907219.24k =A0138= 5401.38k =A01632910.26k sha256 =A0=A0=A0=A0=A0=A0=A0=A0=A0=A0=A03275.81k =A0=A0=A012607.59k =A0=A0= =A044816.18k =A0=A0124774.06k =A0=A0259879.48k So with that 10x performance difference, including by default VIA PadLock=20 patches into Devil Linux distro is a really good idea. On Tuesday 30 October 2007 20:00:04 Serge Leschinsky wrote: > Dmitry Komarov wrote: > > On Tuesday 30 October 2007 15:06:52 Heiko Zuerker wrote: > >> Hey, > >> > >> On Tue, October 30, 2007 07:56, Dmitry Komarov wrote: > >>> has anyone tried to compile DL with in-kernel VIA C3/C7 hardware cryp= to > >>> engine support (VIA PadLock) and other required C3/C7 patches for > >>> VPN/SSL acceleration? > >>> > >>> Any success stories or suggestions? > >> > >> Is this standard part of the 2.4 or 2.6 series, or does it require > >> additional patches? > > > > VIA C3 support was included starting from 2.6.11 kernels, and VIA C7 > > SHA1/SHA256 support was added from 2.6.19. > > > > But it may require some patches to OpenSSL libs and some other > > applications though. Here is the link to VIA PadLock support for Linux > > page: > > > > http://www.logix.cz/michal/devel/padlock/ > > I think we can add that code as a patch (may be only optional, it depends > on the real advantages of this patch). > > > The main idea behind my question is that you can find a lot unexpensive > > VIA Mini-ITX based boards and systems which are just a perfect platform > > for Devil Linux distro, especially if it would have built-in support for > > their harware crypto engines. > > Have you got any hardware of this type? I'd like to ask you about small > testing of crypto possibilities of the board. Please do the following and > share the results with us: > > # openssl > OpenSSL> speed > > After we add the patch I'll ask you to repeat the test and everybody will > be able to check speed up of crypto. Is it ok? > > It's from my system, CPU: Intel(R) Pentium(R) 4 CPU 2.40GHz stepping 01 > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes md2 1083.86k 2286.95k 3166.46k 3495.30k = =20 > 3618.13k mdc2 0.00 0.00 0.00 0.00= =20 > 0.00 md4 13056.59k 46606.95k 147900.22k =20 > 342027.30k 544493.12k md5 11080.59k 39827.41k =20 > 128398.51k 284474.03k 434088.62k hmac(md5) 17638.54k =20 > 63578.20k 181733.97k 339369.98k 453118.63k sha1 11778.7= 6k > 39273.37k 106729.05k 189599.21k 247474.86k rmd160 =20 > 10247.03k 32590.93k 78762.58k 124134.40k 149230.93k rc4 = =20 > 184205.26k 260034.69k 279309.65k 286295.04k 288505.86k des cbc = =20 > 49821.31k 51994.90k 52212.43k 52670.81k 52568.06k des > ede3 18528.82k 18745.07k 18887.34k 18931.71k 18890.75k > idea cbc 26481.13k 29136.73k 29953.19k 30030.17k =20 > 29878.95k rc2 cbc 20763.43k 22201.94k 22351.27k 22460.5= 3k > 22382.39k rc5-32/12 cbc 0.00 0.00 0.00 =20 > 0.00 0.00 blowfish cbc 81991.24k 87533.23k 88552.62k = =20 > 89282.22k 89118.04k cast cbc 45847.92k 54174.72k 56665.8= 6k > 57541.63k 57308.50k aes-128 cbc 64413.49k 88553.49k =20 > 99803.73k 103181.65k 102530.64k aes-192 cbc 38816.20k 65566.7= 0k > 81428.99k 86528.34k 87053.65k aes-256 cbc 36123.11k =20 > 59030.44k 71636.91k 75593.39k 75780.08k camellia-128 cbc =20 > 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc= =20 > 0.00 0.00 0.00 0.00 0.00 camellia-256 > cbc 0.00 0.00 0.00 0.00 0.00 sha25= 6=20 > 7236.61k 16962.65k 30844.07k 39323.65k 42707.63k > sha512 4643.79k 18706.35k 32713.73k 48179.54k =20 > 56016.90k sign verify sign/s verify/s > rsa 512 bits 0.000905s 0.000070s 1105.2 14378.3 > rsa 1024 bits 0.004064s 0.000182s 246.1 5484.1 > rsa 2048 bits 0.022000s 0.000544s 45.5 1838.6 > rsa 4096 bits 0.131711s 0.002079s 7.6 480.9 > sign verify sign/s verify/s > dsa 512 bits 0.000661s 0.000787s 1513.3 1270.5 > dsa 1024 bits 0.001750s 0.002148s 571.3 465.5 > dsa 2048 bits 0.005299s 0.006536s 188.7 153.0 |
From: Dmitry K. <dm...@dz...> - 2007-11-01 12:30:46
|
Hello, Is it a VIA C3 or C7 CPU? And at which ferquency? You said that you compiled your kernel with VIA support, but did you enable Hardware Crypto option in the kernel with appropriate chipset selection? Did you load those padlock crypto modules before testing openssl speed? On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > Hi, > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with > 2.6.21.5 kernel compiled for the VIA CPU. > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK for > me as my DSL uplink is only around 1Mb/s and I use it with a WiFi > linksys router which does not go much faster than 2-3MB/s. > > Here is the output of openssl (not patched) speed on this system: > > OpenSSL 0.9.8e 23 Feb 2007 > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic manual > page 2007 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) > aes(partial) idea(int) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC > -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN > -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall > -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM > -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 > [sysconf value] > timing function used: times > The 'numbers' are in 1000s of bytes per second processed. > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes md2 445.58k 948.54k 1321.30k 1464.32k > 1517.44k mdc2 0.00 0.00 0.00 0.00 > 0.00 md4 4280.63k 14773.95k 42052.07k > 77736.62k 103557.33k md5 3946.47k 13920.19k 40439.58k > 78488.40k 107598.35k hmac(md5) 5471.18k 18343.91k > 49535.49k 85630.29k 109242.24k sha1 3514.75k 10887.00k > 25903.43k 39475.71k 46310.83k rmd160 3080.74k > 8926.14k 19499.32k 27568.13k 31437.49k rc4 40253.31k > 45109.93k 46369.11k 46791.68k 46952.81k des cbc > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 > 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea cbc > 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 cbc > 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k > rc5-32/12 cbc 0.00 0.00 0.00 0.00 > 0.00 blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k > 24228.46k cast cbc 11002.71k 11694.29k 11943.30k 11968.51k > 11953.74k aes-128 cbc 14373.09k 16753.07k 17594.91k > 17753.43k 17818.97k aes-192 cbc 12547.46k 14325.70k 14905.43k > 15092.87k 15097.47k aes-256 cbc 9892.17k 11976.45k > 12751.03k 12893.87k 12966.53k camellia-128 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-256 cbc > 0.00 0.00 0.00 0.00 0.00 sha256 > 2111.48k 5018.54k 9106.26k 11497.57k 12397.96k sha512 > 1468.83k 5865.51k 11094.47k 17036.84k 20446.90k sign > verify sign/s verify/s > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > sign verify sign/s verify/s > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > I should try to patch when I get some time to compare the results. > > Philippe > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Devil-linux-discuss mailing list > Dev...@li... > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss |
From: Philippe M. <ph...@oz...> - 2007-11-01 16:37:09
|
On Thu, Nov 01, 2007 at 02:30:22PM +0200, Dmitry Komarov wrote: > Hello, > > Is it a VIA C3 or C7 CPU? And at which ferquency? This is a C7 1.2Ghz > > You said that you compiled your kernel with VIA support, but did you enable > Hardware Crypto option in the kernel with appropriate chipset selection? Did > you load those padlock crypto modules before testing openssl speed? They are available as modules: CONFIG_CRYPTO_DEV_PADLOCK=m CONFIG_CRYPTO_DEV_PADLOCK_AES=m CONFIG_CRYPTO_DEV_PADLOCK_SHA=m However they were not loaded (duh!) dmesg gives: padlock: Using VIA PadLock ACE for AES algorithm. padlock: Using VIA PadLock ACE for SHA1/SHA256 algorithms. padlock: 3 drivers are available. I rerun with -engine padlock Here are the results: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes md2 448.75k 956.13k 1323.52k 1465.69k 1514.70k mdc2 0.00 0.00 0.00 0.00 0.00 md4 4282.78k 14819.72k 41912.49k 77754.03k 103565.57k md5 3959.09k 13873.69k 40602.79k 78527.44k 107205.61k hmac(md5) 5490.57k 18332.86k 49374.12k 85938.94k 109242.24k sha1 3519.95k 10891.07k 25894.19k 39226.00k 46636.48k rmd160 3082.02k 8920.64k 19502.49k 27558.57k 31507.69k rc4 40262.22k 45080.70k 46477.19k 46987.46k 46951.08k des cbc 12101.70k 12627.39k 12721.15k 12756.70k 12763.14k des ede3 4266.98k 4347.14k 4341.67k 4342.17k 4361.76k idea cbc 7960.33k 8435.11k 8569.66k 8570.67k 8585.22k rc2 cbc 4671.14k 4805.63k 4846.34k 4858.85k 4860.59k rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 blowfish cbc 21997.17k 23662.88k 23998.63k 24124.13k 24160.94k cast cbc 11005.77k 11734.43k 11903.83k 11972.21k 11990.36k aes-128 cbc 14383.44k 16745.86k 17592.85k 17700.13k 17879.93k aes-192 cbc 12539.86k 14336.00k 14902.53k 15099.00k 15076.01k aes-256 cbc 11135.32k 12442.28k 12781.63k 12979.29k 12997.61k camellia-128 cbc 0.00 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 0.00 0.00 0.00 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 0.00 sha256 2103.55k 5018.65k 9140.31k 11461.97k 12395.21k sha512 1468.30k 5881.13k 11065.86k 17101.14k 20367.29k sign verify sign/s verify/s rsa 512 bits 0.002456s 0.000181s 407.2 5516.4 rsa 1024 bits 0.010418s 0.000461s 96.0 2171.3 rsa 2048 bits 0.053763s 0.001345s 18.6 743.7 rsa 4096 bits 0.318437s 0.004390s 3.1 227.8 sign verify sign/s verify/s dsa 512 bits 0.001709s 0.002074s 585.3 482.1 dsa 1024 bits 0.004340s 0.005220s 230.4 191.6 dsa 2048 bits 0.013038s 0.015497s 76.7 64.5 It doesn't seem that much different :-| Philippe > > On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > > Hi, > > > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with > > 2.6.21.5 kernel compiled for the VIA CPU. > > > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK for > > me as my DSL uplink is only around 1Mb/s and I use it with a WiFi > > linksys router which does not go much faster than 2-3MB/s. > > > > Here is the output of openssl (not patched) speed on this system: > > > > OpenSSL 0.9.8e 23 Feb 2007 > > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic manual > > page 2007 options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) > > aes(partial) idea(int) blowfish(idx) compiler: gcc -fPIC -DOPENSSL_PIC > > -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN > > -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall > > -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM > > -DRMD160_ASM -DAES_ASM available timing options: TIMES TIMEB HZ=100 > > [sysconf value] > > timing function used: times > > The 'numbers' are in 1000s of bytes per second processed. > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes md2 445.58k 948.54k 1321.30k 1464.32k > > 1517.44k mdc2 0.00 0.00 0.00 0.00 > > 0.00 md4 4280.63k 14773.95k 42052.07k > > 77736.62k 103557.33k md5 3946.47k 13920.19k 40439.58k > > 78488.40k 107598.35k hmac(md5) 5471.18k 18343.91k > > 49535.49k 85630.29k 109242.24k sha1 3514.75k 10887.00k > > 25903.43k 39475.71k 46310.83k rmd160 3080.74k > > 8926.14k 19499.32k 27568.13k 31437.49k rc4 40253.31k > > 45109.93k 46369.11k 46791.68k 46952.81k des cbc > > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 > > 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea cbc > > 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 cbc > > 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 > > 0.00 blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k > > 24228.46k cast cbc 11002.71k 11694.29k 11943.30k 11968.51k > > 11953.74k aes-128 cbc 14373.09k 16753.07k 17594.91k > > 17753.43k 17818.97k aes-192 cbc 12547.46k 14325.70k 14905.43k > > 15092.87k 15097.47k aes-256 cbc 9892.17k 11976.45k > > 12751.03k 12893.87k 12966.53k camellia-128 cbc 0.00 > > 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 > > 0.00 0.00 0.00 0.00 camellia-256 cbc > > 0.00 0.00 0.00 0.00 0.00 sha256 > > 2111.48k 5018.54k 9106.26k 11497.57k 12397.96k sha512 > > 1468.83k 5865.51k 11094.47k 17036.84k 20446.90k sign > > verify sign/s verify/s > > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > > sign verify sign/s verify/s > > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > > > I should try to patch when I get some time to compare the results. > > > > Philippe > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Splunk Inc. > > Still grepping through log files to find problems? Stop. > > Now Search log events and configuration files using AJAX and a browser. > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > _______________________________________________ > > Devil-linux-discuss mailing list > > Dev...@li... > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Devil-linux-discuss mailing list > Dev...@li... > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > |
From: Dmitry K. <dm...@dz...> - 2007-11-01 17:07:59
|
Hi, so that really confirms the fact we need this OpenSSL patchset, as you can see from my previous message, similar C7 board gives 10 times higher performance with Gentoo and OpenSSL padlock patches applied... On Thursday 01 November 2007 18:36:55 Philippe Marzouk wrote: > On Thu, Nov 01, 2007 at 02:30:22PM +0200, Dmitry Komarov wrote: > > Hello, > > > > Is it a VIA C3 or C7 CPU? And at which ferquency? > > This is a C7 1.2Ghz > > > You said that you compiled your kernel with VIA support, but did you > > enable Hardware Crypto option in the kernel with appropriate chipset > > selection? Did you load those padlock crypto modules before testing > > openssl speed? > > They are available as modules: > > CONFIG_CRYPTO_DEV_PADLOCK=m > CONFIG_CRYPTO_DEV_PADLOCK_AES=m > CONFIG_CRYPTO_DEV_PADLOCK_SHA=m > > However they were not loaded (duh!) > > dmesg gives: > padlock: Using VIA PadLock ACE for AES algorithm. > padlock: Using VIA PadLock ACE for SHA1/SHA256 algorithms. > padlock: 3 drivers are available. > > I rerun with -engine padlock > > Here are the results: > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes md2 448.75k 956.13k 1323.52k 1465.69k > 1514.70k mdc2 0.00 0.00 0.00 0.00 > 0.00 md4 4282.78k 14819.72k 41912.49k > 77754.03k 103565.57k md5 3959.09k 13873.69k 40602.79k > 78527.44k 107205.61k hmac(md5) 5490.57k 18332.86k > 49374.12k 85938.94k 109242.24k sha1 3519.95k 10891.07k > 25894.19k 39226.00k 46636.48k rmd160 3082.02k > 8920.64k 19502.49k 27558.57k 31507.69k rc4 40262.22k > 45080.70k 46477.19k 46987.46k 46951.08k des cbc > 12101.70k 12627.39k 12721.15k 12756.70k 12763.14k des ede3 > 4266.98k 4347.14k 4341.67k 4342.17k 4361.76k idea cbc > 7960.33k 8435.11k 8569.66k 8570.67k 8585.22k rc2 cbc > 4671.14k 4805.63k 4846.34k 4858.85k 4860.59k > rc5-32/12 cbc 0.00 0.00 0.00 0.00 > 0.00 blowfish cbc 21997.17k 23662.88k 23998.63k 24124.13k > 24160.94k cast cbc 11005.77k 11734.43k 11903.83k 11972.21k > 11990.36k aes-128 cbc 14383.44k 16745.86k 17592.85k > 17700.13k 17879.93k aes-192 cbc 12539.86k 14336.00k 14902.53k > 15099.00k 15076.01k aes-256 cbc 11135.32k 12442.28k > 12781.63k 12979.29k 12997.61k camellia-128 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-256 cbc > 0.00 0.00 0.00 0.00 0.00 sha256 > 2103.55k 5018.65k 9140.31k 11461.97k 12395.21k sha512 > 1468.30k 5881.13k 11065.86k 17101.14k 20367.29k sign > verify sign/s verify/s > rsa 512 bits 0.002456s 0.000181s 407.2 5516.4 > rsa 1024 bits 0.010418s 0.000461s 96.0 2171.3 > rsa 2048 bits 0.053763s 0.001345s 18.6 743.7 > rsa 4096 bits 0.318437s 0.004390s 3.1 227.8 > sign verify sign/s verify/s > dsa 512 bits 0.001709s 0.002074s 585.3 482.1 > dsa 1024 bits 0.004340s 0.005220s 230.4 191.6 > dsa 2048 bits 0.013038s 0.015497s 76.7 64.5 > > It doesn't seem that much different :-| > > Philippe > > > On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > > > Hi, > > > > > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with > > > 2.6.21.5 kernel compiled for the VIA CPU. > > > > > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK for > > > me as my DSL uplink is only around 1Mb/s and I use it with a WiFi > > > linksys router which does not go much faster than 2-3MB/s. > > > > > > Here is the output of openssl (not patched) speed on this system: > > > > > > OpenSSL 0.9.8e 23 Feb 2007 > > > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic > > > manual page 2007 options:bn(64,32) md2(int) rc4(idx,int) > > > des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: > > > gcc -fPIC -DOPENSSL_PIC -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS > > > -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 > > > -fomit-frame-pointer -Wall -DOPENSSL_BN_ASM_PART_WORDS > > > -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM > > > available timing options: TIMES TIMEB HZ=100 [sysconf value] > > > timing function used: times > > > The 'numbers' are in 1000s of bytes per second processed. > > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > > bytes md2 445.58k 948.54k 1321.30k 1464.32k > > > 1517.44k mdc2 0.00 0.00 0.00 > > > 0.00 0.00 md4 4280.63k 14773.95k 42052.07k > > > 77736.62k 103557.33k md5 3946.47k 13920.19k > > > 40439.58k 78488.40k 107598.35k hmac(md5) 5471.18k > > > 18343.91k 49535.49k 85630.29k 109242.24k sha1 > > > 3514.75k 10887.00k 25903.43k 39475.71k 46310.83k rmd160 > > > 3080.74k 8926.14k 19499.32k 27568.13k 31437.49k rc4 > > > 40253.31k 45109.93k 46369.11k 46791.68k 46952.81k des > > > cbc > > > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 > > > 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea > > > cbc 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 > > > cbc 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k > > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 > > > blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k > > > 24228.46k cast cbc 11002.71k 11694.29k 11943.30k > > > 11968.51k 11953.74k aes-128 cbc 14373.09k 16753.07k > > > 17594.91k 17753.43k 17818.97k aes-192 cbc 12547.46k > > > 14325.70k 14905.43k 15092.87k 15097.47k aes-256 cbc > > > 9892.17k 11976.45k 12751.03k 12893.87k 12966.53k camellia-128 > > > cbc 0.00 0.00 0.00 0.00 0.00 > > > camellia-192 cbc 0.00 0.00 0.00 0.00 > > > 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 > > > 0.00 sha256 2111.48k 5018.54k 9106.26k 11497.57k > > > 12397.96k sha512 1468.83k 5865.51k 11094.47k 17036.84k > > > 20446.90k sign verify sign/s verify/s > > > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > > > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > > > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > > > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > > > sign verify sign/s verify/s > > > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > > > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > > > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > > > > > I should try to patch when I get some time to compare the results. > > > > > > Philippe > > > > > > ----------------------------------------------------------------------- > > >-- This SF.net email is sponsored by: Splunk Inc. > > > Still grepping through log files to find problems? Stop. > > > Now Search log events and configuration files using AJAX and a browser. > > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > > _______________________________________________ > > > Devil-linux-discuss mailing list > > > Dev...@li... > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Splunk Inc. > > Still grepping through log files to find problems? Stop. > > Now Search log events and configuration files using AJAX and a browser. > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > _______________________________________________ > > Devil-linux-discuss mailing list > > Dev...@li... > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Devil-linux-discuss mailing list > Dev...@li... > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss |
From: Serge L. <fi...@in...> - 2007-11-01 19:43:42
Attachments:
openssl-0.9.8b-sha.diff.bz2
|
Hi, Dmitry Komarov wrote: > Hi, > > so that really confirms the fact we need this OpenSSL patchset, as you can see > from my previous message, similar C7 board gives 10 times higher performance > with Gentoo and OpenSSL padlock patches applied... Well, is it enough to apply the following patch? I mean I looked through the mentioned page but haven't understood clearly how we should compile padlock engine - static or shared. So I'm going to apply only one patch - it's in the attachment because it was shrunk a bit. If the patch doesn't break the openssl functionality I'll check in it in a day (after minimal preliminary checking). Index: openssl =================================================================== RCS file: /cvsroot/devil-linux/build/scripts/openssl,v retrieving revision 1.25 diff -u -r1.25 openssl --- openssl 19 Apr 2007 15:39:21 -0000 1.25 +++ openssl 1 Nov 2007 19:12:07 -0000 @@ -27,6 +27,10 @@ case $1 in build ) + if [ ! -e .patches_done ]; then + bzcat $SRCDIR/openssl-0.9.8b-sha.diff.bz2 | patch -p1 || exit 1 + touch .patches_done + fi ./Configure linux-elf shared threads zlib-dynamic --prefix=/usr --openssldir=/etc/ssl || exit 1 make depend || exit 1 =================================================================== -- Sincerely Serge Leschinsky |
From: Dmitry K. <dm...@dz...> - 2007-11-01 17:50:49
|
Philippe, what was the exact command line for your test? I did some tests on the cheap MaxTerm 8500 device where I have remote access (VIA C3 1GHz, 512 RAM), it runs ImediaLinux distribution. Results are just terrific for such unexpensive device, wish I have this with Devil: [root@crypto]# openssl speed -evp aes-128-cbc -engine padlock engine "padlock" set. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-128-cbc 35042.17k 124617.57k 331076.12k 560053.16k 701632.55k [root@crypto]# openssl speed -evp aes-192-cbc -engine padlock engine "padlock" set. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-192-cbc 35613.74k 121805.44k 307919.11k 492034.72k 597090.66k [root@crypto]# openssl speed -evp aes-256-cbc -engine padlock engine "padlock" set. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-256-cbc 35539.32k 118421.56k 271725.80k 430060.27k 517467.68k Compare to the same board but without padlock support: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-128-cbc 10915.99k 13937.99k 15080.02k 15395.04k 15491.32k aes-192-cbc 9679.55k 11995.28k 12832.40k 13059.91k 13126.25k aes-256-cbc 8705.40k 10333.62k 10947.27k 11106.15k 11161.26k On Thursday 01 November 2007 18:36:55 Philippe Marzouk wrote: > On Thu, Nov 01, 2007 at 02:30:22PM +0200, Dmitry Komarov wrote: > > Hello, > > > > Is it a VIA C3 or C7 CPU? And at which ferquency? > > This is a C7 1.2Ghz > > > You said that you compiled your kernel with VIA support, but did you > > enable Hardware Crypto option in the kernel with appropriate chipset > > selection? Did you load those padlock crypto modules before testing > > openssl speed? > > They are available as modules: > > CONFIG_CRYPTO_DEV_PADLOCK=m > CONFIG_CRYPTO_DEV_PADLOCK_AES=m > CONFIG_CRYPTO_DEV_PADLOCK_SHA=m > > However they were not loaded (duh!) > > dmesg gives: > padlock: Using VIA PadLock ACE for AES algorithm. > padlock: Using VIA PadLock ACE for SHA1/SHA256 algorithms. > padlock: 3 drivers are available. > > I rerun with -engine padlock > > Here are the results: > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes md2 448.75k 956.13k 1323.52k 1465.69k > 1514.70k mdc2 0.00 0.00 0.00 0.00 > 0.00 md4 4282.78k 14819.72k 41912.49k > 77754.03k 103565.57k md5 3959.09k 13873.69k 40602.79k > 78527.44k 107205.61k hmac(md5) 5490.57k 18332.86k > 49374.12k 85938.94k 109242.24k sha1 3519.95k 10891.07k > 25894.19k 39226.00k 46636.48k rmd160 3082.02k > 8920.64k 19502.49k 27558.57k 31507.69k rc4 40262.22k > 45080.70k 46477.19k 46987.46k 46951.08k des cbc > 12101.70k 12627.39k 12721.15k 12756.70k 12763.14k des ede3 > 4266.98k 4347.14k 4341.67k 4342.17k 4361.76k idea cbc > 7960.33k 8435.11k 8569.66k 8570.67k 8585.22k rc2 cbc > 4671.14k 4805.63k 4846.34k 4858.85k 4860.59k > rc5-32/12 cbc 0.00 0.00 0.00 0.00 > 0.00 blowfish cbc 21997.17k 23662.88k 23998.63k 24124.13k > 24160.94k cast cbc 11005.77k 11734.43k 11903.83k 11972.21k > 11990.36k aes-128 cbc 14383.44k 16745.86k 17592.85k > 17700.13k 17879.93k aes-192 cbc 12539.86k 14336.00k 14902.53k > 15099.00k 15076.01k aes-256 cbc 11135.32k 12442.28k > 12781.63k 12979.29k 12997.61k camellia-128 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 > 0.00 0.00 0.00 0.00 camellia-256 cbc > 0.00 0.00 0.00 0.00 0.00 sha256 > 2103.55k 5018.65k 9140.31k 11461.97k 12395.21k sha512 > 1468.30k 5881.13k 11065.86k 17101.14k 20367.29k sign > verify sign/s verify/s > rsa 512 bits 0.002456s 0.000181s 407.2 5516.4 > rsa 1024 bits 0.010418s 0.000461s 96.0 2171.3 > rsa 2048 bits 0.053763s 0.001345s 18.6 743.7 > rsa 4096 bits 0.318437s 0.004390s 3.1 227.8 > sign verify sign/s verify/s > dsa 512 bits 0.001709s 0.002074s 585.3 482.1 > dsa 1024 bits 0.004340s 0.005220s 230.4 191.6 > dsa 2048 bits 0.013038s 0.015497s 76.7 64.5 > > It doesn't seem that much different :-| > > Philippe > > > On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > > > Hi, > > > > > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with > > > 2.6.21.5 kernel compiled for the VIA CPU. > > > > > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK for > > > me as my DSL uplink is only around 1Mb/s and I use it with a WiFi > > > linksys router which does not go much faster than 2-3MB/s. > > > > > > Here is the output of openssl (not patched) speed on this system: > > > > > > OpenSSL 0.9.8e 23 Feb 2007 > > > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic > > > manual page 2007 options:bn(64,32) md2(int) rc4(idx,int) > > > des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: > > > gcc -fPIC -DOPENSSL_PIC -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS > > > -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 > > > -fomit-frame-pointer -Wall -DOPENSSL_BN_ASM_PART_WORDS > > > -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM > > > available timing options: TIMES TIMEB HZ=100 [sysconf value] > > > timing function used: times > > > The 'numbers' are in 1000s of bytes per second processed. > > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > > bytes md2 445.58k 948.54k 1321.30k 1464.32k > > > 1517.44k mdc2 0.00 0.00 0.00 > > > 0.00 0.00 md4 4280.63k 14773.95k 42052.07k > > > 77736.62k 103557.33k md5 3946.47k 13920.19k > > > 40439.58k 78488.40k 107598.35k hmac(md5) 5471.18k > > > 18343.91k 49535.49k 85630.29k 109242.24k sha1 > > > 3514.75k 10887.00k 25903.43k 39475.71k 46310.83k rmd160 > > > 3080.74k 8926.14k 19499.32k 27568.13k 31437.49k rc4 > > > 40253.31k 45109.93k 46369.11k 46791.68k 46952.81k des > > > cbc > > > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 > > > 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea > > > cbc 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 > > > cbc 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k > > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 > > > blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k > > > 24228.46k cast cbc 11002.71k 11694.29k 11943.30k > > > 11968.51k 11953.74k aes-128 cbc 14373.09k 16753.07k > > > 17594.91k 17753.43k 17818.97k aes-192 cbc 12547.46k > > > 14325.70k 14905.43k 15092.87k 15097.47k aes-256 cbc > > > 9892.17k 11976.45k 12751.03k 12893.87k 12966.53k camellia-128 > > > cbc 0.00 0.00 0.00 0.00 0.00 > > > camellia-192 cbc 0.00 0.00 0.00 0.00 > > > 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 > > > 0.00 sha256 2111.48k 5018.54k 9106.26k 11497.57k > > > 12397.96k sha512 1468.83k 5865.51k 11094.47k 17036.84k > > > 20446.90k sign verify sign/s verify/s > > > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > > > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > > > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > > > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > > > sign verify sign/s verify/s > > > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > > > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > > > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > > > > > I should try to patch when I get some time to compare the results. > > > > > > Philippe > > > > > > ----------------------------------------------------------------------- > > >-- This SF.net email is sponsored by: Splunk Inc. > > > Still grepping through log files to find problems? Stop. > > > Now Search log events and configuration files using AJAX and a browser. > > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > > _______________________________________________ > > > Devil-linux-discuss mailing list > > > Dev...@li... > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > > > ------------------------------------------------------------------------- > > This SF.net email is sponsored by: Splunk Inc. > > Still grepping through log files to find problems? Stop. > > Now Search log events and configuration files using AJAX and a browser. > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > _______________________________________________ > > Devil-linux-discuss mailing list > > Dev...@li... > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Devil-linux-discuss mailing list > Dev...@li... > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss |
From: Heiko Z. <he...@zu...> - 2007-11-01 18:10:50
|
On Thu, November 1, 2007 12:50, Dmitry Komarov wrote: > Philippe, what was the exact command line for your test? > > > I did some tests on the cheap MaxTerm 8500 device where I have remote > access (VIA C3 1GHz, 512 RAM), it runs ImediaLinux distribution. Results > are just terrific for such unexpensive device, wish I have this with > Devil: Serge said he's going to add it, please write any checks in his name (may speed up the process). ;-) -- Regards Heiko Zuerker http://www.devil-linux.org |
From: Philippe M. <ph...@oz...> - 2007-11-01 18:33:19
|
On Thu, Nov 01, 2007 at 07:50:40PM +0200, Dmitry Komarov wrote: > Philippe, what was the exact command line for your test? I used: openssl speed -engine padlock > > I did some tests on the cheap MaxTerm 8500 device where I have remote access > (VIA C3 1GHz, 512 RAM), it runs ImediaLinux distribution. Results are just > terrific for such unexpensive device, wish I have this with Devil: > > [root@crypto]# openssl speed -evp aes-128-cbc -engine padlock > engine "padlock" set. > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes > aes-128-cbc 35042.17k 124617.57k 331076.12k 560053.16k 701632.55k > I tried with the command line above: type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes aes-128-cbc 40436.20k 106822.84k 160375.95k 207363.79k 246063.72k *much* better, this is weird Philippe > [root@crypto]# openssl speed -evp aes-192-cbc -engine padlock > engine "padlock" set. > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes > aes-192-cbc 35613.74k 121805.44k 307919.11k 492034.72k 597090.66k > > [root@crypto]# openssl speed -evp aes-256-cbc -engine padlock > engine "padlock" set. > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes > aes-256-cbc 35539.32k 118421.56k 271725.80k 430060.27k 517467.68k > > Compare to the same board but without padlock support: > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes > aes-128-cbc 10915.99k 13937.99k 15080.02k 15395.04k 15491.32k > aes-192-cbc 9679.55k 11995.28k 12832.40k 13059.91k 13126.25k > aes-256-cbc 8705.40k 10333.62k 10947.27k 11106.15k 11161.26k > > > On Thursday 01 November 2007 18:36:55 Philippe Marzouk wrote: > > On Thu, Nov 01, 2007 at 02:30:22PM +0200, Dmitry Komarov wrote: > > > Hello, > > > > > > Is it a VIA C3 or C7 CPU? And at which ferquency? > > > > This is a C7 1.2Ghz > > > > > You said that you compiled your kernel with VIA support, but did you > > > enable Hardware Crypto option in the kernel with appropriate chipset > > > selection? Did you load those padlock crypto modules before testing > > > openssl speed? > > > > They are available as modules: > > > > CONFIG_CRYPTO_DEV_PADLOCK=m > > CONFIG_CRYPTO_DEV_PADLOCK_AES=m > > CONFIG_CRYPTO_DEV_PADLOCK_SHA=m > > > > However they were not loaded (duh!) > > > > dmesg gives: > > padlock: Using VIA PadLock ACE for AES algorithm. > > padlock: Using VIA PadLock ACE for SHA1/SHA256 algorithms. > > padlock: 3 drivers are available. > > > > I rerun with -engine padlock > > > > Here are the results: > > > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes md2 448.75k 956.13k 1323.52k 1465.69k > > 1514.70k mdc2 0.00 0.00 0.00 0.00 > > 0.00 md4 4282.78k 14819.72k 41912.49k > > 77754.03k 103565.57k md5 3959.09k 13873.69k 40602.79k > > 78527.44k 107205.61k hmac(md5) 5490.57k 18332.86k > > 49374.12k 85938.94k 109242.24k sha1 3519.95k 10891.07k > > 25894.19k 39226.00k 46636.48k rmd160 3082.02k > > 8920.64k 19502.49k 27558.57k 31507.69k rc4 40262.22k > > 45080.70k 46477.19k 46987.46k 46951.08k des cbc > > 12101.70k 12627.39k 12721.15k 12756.70k 12763.14k des ede3 > > 4266.98k 4347.14k 4341.67k 4342.17k 4361.76k idea cbc > > 7960.33k 8435.11k 8569.66k 8570.67k 8585.22k rc2 cbc > > 4671.14k 4805.63k 4846.34k 4858.85k 4860.59k > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 > > 0.00 blowfish cbc 21997.17k 23662.88k 23998.63k 24124.13k > > 24160.94k cast cbc 11005.77k 11734.43k 11903.83k 11972.21k > > 11990.36k aes-128 cbc 14383.44k 16745.86k 17592.85k > > 17700.13k 17879.93k aes-192 cbc 12539.86k 14336.00k 14902.53k > > 15099.00k 15076.01k aes-256 cbc 11135.32k 12442.28k > > 12781.63k 12979.29k 12997.61k camellia-128 cbc 0.00 > > 0.00 0.00 0.00 0.00 camellia-192 cbc 0.00 > > 0.00 0.00 0.00 0.00 camellia-256 cbc > > 0.00 0.00 0.00 0.00 0.00 sha256 > > 2103.55k 5018.65k 9140.31k 11461.97k 12395.21k sha512 > > 1468.30k 5881.13k 11065.86k 17101.14k 20367.29k sign > > verify sign/s verify/s > > rsa 512 bits 0.002456s 0.000181s 407.2 5516.4 > > rsa 1024 bits 0.010418s 0.000461s 96.0 2171.3 > > rsa 2048 bits 0.053763s 0.001345s 18.6 743.7 > > rsa 4096 bits 0.318437s 0.004390s 3.1 227.8 > > sign verify sign/s verify/s > > dsa 512 bits 0.001709s 0.002074s 585.3 482.1 > > dsa 1024 bits 0.004340s 0.005220s 230.4 191.6 > > dsa 2048 bits 0.013038s 0.015497s 76.7 64.5 > > > > It doesn't seem that much different :-| > > > > Philippe > > > > > On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > > > > Hi, > > > > > > > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july with > > > > 2.6.21.5 kernel compiled for the VIA CPU. > > > > > > > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK for > > > > me as my DSL uplink is only around 1Mb/s and I use it with a WiFi > > > > linksys router which does not go much faster than 2-3MB/s. > > > > > > > > Here is the output of openssl (not patched) speed on this system: > > > > > > > > OpenSSL 0.9.8e 23 Feb 2007 > > > > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic > > > > manual page 2007 options:bn(64,32) md2(int) rc4(idx,int) > > > > des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: > > > > gcc -fPIC -DOPENSSL_PIC -DZLIB_SHARED -DZLIB -DOPENSSL_THREADS > > > > -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DL_ENDIAN -DTERMIO -O3 > > > > -fomit-frame-pointer -Wall -DOPENSSL_BN_ASM_PART_WORDS > > > > -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM > > > > available timing options: TIMES TIMEB HZ=100 [sysconf value] > > > > timing function used: times > > > > The 'numbers' are in 1000s of bytes per second processed. > > > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > > > bytes md2 445.58k 948.54k 1321.30k 1464.32k > > > > 1517.44k mdc2 0.00 0.00 0.00 > > > > 0.00 0.00 md4 4280.63k 14773.95k 42052.07k > > > > 77736.62k 103557.33k md5 3946.47k 13920.19k > > > > 40439.58k 78488.40k 107598.35k hmac(md5) 5471.18k > > > > 18343.91k 49535.49k 85630.29k 109242.24k sha1 > > > > 3514.75k 10887.00k 25903.43k 39475.71k 46310.83k rmd160 > > > > 3080.74k 8926.14k 19499.32k 27568.13k 31437.49k rc4 > > > > 40253.31k 45109.93k 46369.11k 46791.68k 46952.81k des > > > > cbc > > > > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des ede3 > > > > 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k idea > > > > cbc 7974.25k 8436.37k 8567.78k 8565.42k 8585.11k rc2 > > > > cbc 4655.45k 4821.64k 4845.40k 4854.78k 4862.97k > > > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 > > > > blowfish cbc 22059.53k 23586.54k 24061.43k 24029.64k > > > > 24228.46k cast cbc 11002.71k 11694.29k 11943.30k > > > > 11968.51k 11953.74k aes-128 cbc 14373.09k 16753.07k > > > > 17594.91k 17753.43k 17818.97k aes-192 cbc 12547.46k > > > > 14325.70k 14905.43k 15092.87k 15097.47k aes-256 cbc > > > > 9892.17k 11976.45k 12751.03k 12893.87k 12966.53k camellia-128 > > > > cbc 0.00 0.00 0.00 0.00 0.00 > > > > camellia-192 cbc 0.00 0.00 0.00 0.00 > > > > 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 > > > > 0.00 sha256 2111.48k 5018.54k 9106.26k 11497.57k > > > > 12397.96k sha512 1468.83k 5865.51k 11094.47k 17036.84k > > > > 20446.90k sign verify sign/s verify/s > > > > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > > > > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > > > > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > > > > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > > > > sign verify sign/s verify/s > > > > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > > > > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > > > > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > > > > > > > I should try to patch when I get some time to compare the results. > > > > > > > > Philippe > > > > > > > > ----------------------------------------------------------------------- > > > >-- This SF.net email is sponsored by: Splunk Inc. > > > > Still grepping through log files to find problems? Stop. > > > > Now Search log events and configuration files using AJAX and a browser. > > > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > > > _______________________________________________ > > > > Devil-linux-discuss mailing list > > > > Dev...@li... > > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > > > > > ------------------------------------------------------------------------- > > > This SF.net email is sponsored by: Splunk Inc. > > > Still grepping through log files to find problems? Stop. > > > Now Search log events and configuration files using AJAX and a browser. > > > Download your FREE copy of Splunk now >> http://get.splunk.com/ > > > _______________________________________________ > > > Devil-linux-discuss mailing list > > > Dev...@li... > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > |
From: Dmitry K. <dm...@dz...> - 2007-11-01 19:07:52
|
On Thursday 01 November 2007 20:33:13 Philippe Marzouk wrote: > I used: > > openssl speed -engine padlock Sounds stupid, but with this command I also get usual low results: aes-128 cbc 10348.65k 13694.81k 15005.60k 15375.65k 15488.60k aes-192 cbc 9252.99k 11821.35k 12781.20k 13046.30k 13126.25k aes-256 cbc 8341.65k 10197.77k 10907.98k 11100.02k 11161.26k Would someone explain this? > > > I did some tests on the cheap MaxTerm 8500 device where I have remote > > access (VIA C3 1GHz, 512 RAM), it runs ImediaLinux distribution. Results > > are just terrific for such unexpensive device, wish I have this with > > Devil: > > > > [root@crypto]# openssl speed -evp aes-128-cbc -engine padlock > > engine "padlock" set. > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes aes-128-cbc 35042.17k 124617.57k 331076.12k 560053.16k > > 701632.55k > > I tried with the command line above: > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > bytes aes-128-cbc 40436.20k 106822.84k 160375.95k 207363.79k > 246063.72k > > *much* better, this is weird > > Philippe > > > [root@crypto]# openssl speed -evp aes-192-cbc -engine padlock > > engine "padlock" set. > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes aes-192-cbc 35613.74k 121805.44k 307919.11k 492034.72k > > 597090.66k > > > > [root@crypto]# openssl speed -evp aes-256-cbc -engine padlock > > engine "padlock" set. > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes aes-256-cbc 35539.32k 118421.56k 271725.80k 430060.27k > > 517467.68k > > > > Compare to the same board but without padlock support: > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > bytes aes-128-cbc 10915.99k 13937.99k 15080.02k 15395.04k > > 15491.32k aes-192-cbc 9679.55k 11995.28k 12832.40k > > 13059.91k 13126.25k aes-256-cbc 8705.40k 10333.62k > > 10947.27k 11106.15k 11161.26k > > > > On Thursday 01 November 2007 18:36:55 Philippe Marzouk wrote: > > > On Thu, Nov 01, 2007 at 02:30:22PM +0200, Dmitry Komarov wrote: > > > > Hello, > > > > > > > > Is it a VIA C3 or C7 CPU? And at which ferquency? > > > > > > This is a C7 1.2Ghz > > > > > > > You said that you compiled your kernel with VIA support, but did you > > > > enable Hardware Crypto option in the kernel with appropriate chipset > > > > selection? Did you load those padlock crypto modules before testing > > > > openssl speed? > > > > > > They are available as modules: > > > > > > CONFIG_CRYPTO_DEV_PADLOCK=m > > > CONFIG_CRYPTO_DEV_PADLOCK_AES=m > > > CONFIG_CRYPTO_DEV_PADLOCK_SHA=m > > > > > > However they were not loaded (duh!) > > > > > > dmesg gives: > > > padlock: Using VIA PadLock ACE for AES algorithm. > > > padlock: Using VIA PadLock ACE for SHA1/SHA256 algorithms. > > > padlock: 3 drivers are available. > > > > > > I rerun with -engine padlock > > > > > > Here are the results: > > > > > > type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 > > > bytes md2 448.75k 956.13k 1323.52k 1465.69k > > > 1514.70k mdc2 0.00 0.00 0.00 > > > 0.00 0.00 md4 4282.78k 14819.72k 41912.49k > > > 77754.03k 103565.57k md5 3959.09k 13873.69k > > > 40602.79k 78527.44k 107205.61k hmac(md5) 5490.57k > > > 18332.86k 49374.12k 85938.94k 109242.24k sha1 > > > 3519.95k 10891.07k 25894.19k 39226.00k 46636.48k rmd160 > > > 3082.02k 8920.64k 19502.49k 27558.57k 31507.69k rc4 > > > 40262.22k 45080.70k 46477.19k 46987.46k 46951.08k des > > > cbc > > > 12101.70k 12627.39k 12721.15k 12756.70k 12763.14k des ede3 > > > 4266.98k 4347.14k 4341.67k 4342.17k 4361.76k idea > > > cbc 7960.33k 8435.11k 8569.66k 8570.67k 8585.22k rc2 > > > cbc 4671.14k 4805.63k 4846.34k 4858.85k 4860.59k > > > rc5-32/12 cbc 0.00 0.00 0.00 0.00 0.00 > > > blowfish cbc 21997.17k 23662.88k 23998.63k 24124.13k > > > 24160.94k cast cbc 11005.77k 11734.43k 11903.83k > > > 11972.21k 11990.36k aes-128 cbc 14383.44k 16745.86k > > > 17592.85k 17700.13k 17879.93k aes-192 cbc 12539.86k > > > 14336.00k 14902.53k 15099.00k 15076.01k aes-256 cbc > > > 11135.32k 12442.28k 12781.63k 12979.29k 12997.61k camellia-128 > > > cbc 0.00 0.00 0.00 0.00 0.00 > > > camellia-192 cbc 0.00 0.00 0.00 0.00 > > > 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 > > > 0.00 sha256 2103.55k 5018.65k 9140.31k 11461.97k > > > 12395.21k sha512 1468.30k 5881.13k 11065.86k 17101.14k > > > 20367.29k sign verify sign/s verify/s > > > rsa 512 bits 0.002456s 0.000181s 407.2 5516.4 > > > rsa 1024 bits 0.010418s 0.000461s 96.0 2171.3 > > > rsa 2048 bits 0.053763s 0.001345s 18.6 743.7 > > > rsa 4096 bits 0.318437s 0.004390s 3.1 227.8 > > > sign verify sign/s verify/s > > > dsa 512 bits 0.001709s 0.002074s 585.3 482.1 > > > dsa 1024 bits 0.004340s 0.005220s 230.4 191.6 > > > dsa 2048 bits 0.013038s 0.015497s 76.7 64.5 > > > > > > It doesn't seem that much different :-| > > > > > > Philippe > > > > > > > On Wednesday 31 October 2007 08:10:13 Philippe Marzouk wrote: > > > > > Hi, > > > > > > > > > > I haven a Via EN12000 with DL 1.3.4 I compiled at the end of july > > > > > with 2.6.21.5 kernel compiled for the VIA CPU. > > > > > > > > > > I get around 4MB/s of VPN throughput using ipsec-tools which is OK > > > > > for me as my DSL uplink is only around 1Mb/s and I use it with a > > > > > WiFi linksys router which does not go much faster than 2-3MB/s. > > > > > > > > > > Here is the output of openssl (not patched) speed on this system: > > > > > > > > > > OpenSSL 0.9.8e 23 Feb 2007 > > > > > built on: Sat Jul 28 19:14:16 Local time zone must be set--see zic > > > > > manual page 2007 options:bn(64,32) md2(int) rc4(idx,int) > > > > > des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) > > > > > compiler: gcc -fPIC -DOPENSSL_PIC -DZLIB_SHARED -DZLIB > > > > > -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H > > > > > -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -Wall > > > > > -DOPENSSL_BN_ASM_PART_WORDS > > > > > -DOPENSSL_IA32_SSE2 -DSHA1_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM > > > > > available timing options: TIMES TIMEB HZ=100 [sysconf value] > > > > > timing function used: times > > > > > The 'numbers' are in 1000s of bytes per second processed. > > > > > type 16 bytes 64 bytes 256 bytes 1024 bytes > > > > > 8192 bytes md2 445.58k 948.54k 1321.30k > > > > > 1464.32k 1517.44k mdc2 0.00 0.00 > > > > > 0.00 0.00 0.00 md4 4280.63k 14773.95k 42052.07k > > > > > 77736.62k 103557.33k md5 3946.47k 13920.19k > > > > > 40439.58k 78488.40k 107598.35k hmac(md5) 5471.18k > > > > > 18343.91k 49535.49k 85630.29k 109242.24k sha1 > > > > > 3514.75k 10887.00k 25903.43k 39475.71k 46310.83k rmd160 > > > > > 3080.74k 8926.14k 19499.32k 27568.13k 31437.49k rc4 > > > > > 40253.31k 45109.93k 46369.11k 46791.68k 46952.81k > > > > > des cbc > > > > > 12141.82k 12626.84k 12675.15k 12774.66k 12769.07k des > > > > > ede3 4262.84k 4348.23k 4346.88k 4352.00k 4359.90k > > > > > idea cbc 7974.25k 8436.37k 8567.78k 8565.42k > > > > > 8585.11k rc2 cbc 4655.45k 4821.64k 4845.40k 4854.78k > > > > > 4862.97k rc5-32/12 cbc 0.00 0.00 0.00 > > > > > 0.00 0.00 blowfish cbc 22059.53k 23586.54k 24061.43k > > > > > 24029.64k 24228.46k cast cbc 11002.71k 11694.29k > > > > > 11943.30k 11968.51k 11953.74k aes-128 cbc 14373.09k > > > > > 16753.07k 17594.91k 17753.43k 17818.97k aes-192 cbc > > > > > 12547.46k 14325.70k 14905.43k 15092.87k 15097.47k aes-256 cbc > > > > > 9892.17k 11976.45k 12751.03k 12893.87k 12966.53k > > > > > camellia-128 cbc 0.00 0.00 0.00 0.00 > > > > > 0.00 camellia-192 cbc 0.00 0.00 0.00 0.00 > > > > > 0.00 camellia-256 cbc 0.00 0.00 0.00 0.00 > > > > > 0.00 sha256 2111.48k 5018.54k 9106.26k 11497.57k > > > > > 12397.96k sha512 1468.83k 5865.51k 11094.47k 17036.84k > > > > > 20446.90k sign verify sign/s verify/s > > > > > rsa 512 bits 0.002456s 0.000180s 407.2 5546.3 > > > > > rsa 1024 bits 0.010418s 0.000459s 96.0 2177.4 > > > > > rsa 2048 bits 0.053710s 0.001352s 18.6 739.7 > > > > > rsa 4096 bits 0.318437s 0.004379s 3.1 228.4 > > > > > sign verify sign/s verify/s > > > > > dsa 512 bits 0.001708s 0.002093s 585.5 477.9 > > > > > dsa 1024 bits 0.004340s 0.005343s 230.4 187.2 > > > > > dsa 2048 bits 0.013025s 0.015876s 76.8 63.0 > > > > > > > > > > I should try to patch when I get some time to compare the results. > > > > > > > > > > Philippe > > > > > > > > > > ------------------------------------------------------------------- > > > > >---- -- This SF.net email is sponsored by: Splunk Inc. > > > > > Still grepping through log files to find problems? Stop. > > > > > Now Search log events and configuration files using AJAX and a > > > > > browser. Download your FREE copy of Splunk now >> > > > > > http://get.splunk.com/ > > > > > _______________________________________________ > > > > > Devil-linux-discuss mailing list > > > > > Dev...@li... > > > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > > > > > > > --------------------------------------------------------------------- > > > >---- This SF.net email is sponsored by: Splunk Inc. > > > > Still grepping through log files to find problems? Stop. > > > > Now Search log events and configuration files using AJAX and a > > > > browser. Download your FREE copy of Splunk now >> > > > > http://get.splunk.com/ > > > > _______________________________________________ > > > > Devil-linux-discuss mailing list > > > > Dev...@li... > > > > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss > > ------------------------------------------------------------------------- > This SF.net email is sponsored by: Splunk Inc. > Still grepping through log files to find problems? Stop. > Now Search log events and configuration files using AJAX and a browser. > Download your FREE copy of Splunk now >> http://get.splunk.com/ > _______________________________________________ > Devil-linux-discuss mailing list > Dev...@li... > https://lists.sourceforge.net/lists/listinfo/devil-linux-discuss |
From: Dmitry K. <dm...@dz...> - 2007-11-01 20:17:56
|
On Thursday 01 November 2007 21:43:24 Serge Leschinsky wrote: > Hi, > > Dmitry Komarov wrote: > > Hi, > > > > so that really confirms the fact we need this OpenSSL patchset, as you > > can see from my previous message, similar C7 board gives 10 times higher > > performance with Gentoo and OpenSSL padlock patches applied... > > Well, is it enough to apply the following patch? I mean I looked through > the mentioned page but haven't understood clearly how we should compile > padlock engine =C2=A0- static or shared. So I'm going to apply only one p= atch - > it's in the attachment because it was shrunk a bit. If the patch doesn't > break the openssl functionality I'll check in it in a day (after minimal > preliminary checking). The page also says: =3D=3D cut =3D=3D Once you get bored with patching heaps of client programs have a look at th= is=20 patch from Cecilia: openssl-0.9.8e-engine.diff, 2007-07-20 01:38 "The openssl-0.9.8e patch will make the ssl-library to load the padlock=20 engine. This means, if you apply the openssl-0.9.8e patch, you do not have = to=20 apply any other patches or modifications, since every time the ssl-library = is=20 called, the padlock-engine is initialized by the ssl-library." In other words - Patch for OpenSSL to always load PadLock engine. =3D=3D cut =3D=3D But I'm not sure if this will not break openssl functionality on systems=20 without padlock.... Should be tested. |
From: Serge L. <fi...@in...> - 2007-11-02 20:17:29
|
Dmitry Komarov wrote: > > But I'm not sure if this will not break openssl functionality on systems > without padlock.... Should be tested. The patch was checked in. My checks have assured me that openssl functionality is not broken - ssh and openvpn were sensors. Heiko, I believe the padlock support should be added to 1.2 after some defense period to avoid problem with crypto in production system. Correct? -- Sincerely, Serge Leschinsky |
From: Heiko Z. <he...@zu...> - 2007-11-02 20:34:37
|
On Fri, November 2, 2007 15:16, Serge Leschinsky wrote: > Dmitry Komarov wrote: > > >> >> But I'm not sure if this will not break openssl functionality on >> systems without padlock.... Should be tested. > The patch was checked in. My checks have assured me that openssl > functionality is not broken - ssh and openvpn were sensors. > > Heiko, I believe the padlock support should be added to 1.2 after some > defense period to avoid problem with crypto in production system. Correct? If it doesn't cause any other troubles I agree that it would be a good addition. -- Regards Heiko Zuerker http://www.devil-linux.org |