Re: [openssl-dev] Multithreading performance patches
Hi! > The reason is that so far it was > reserved for occasion like this, i.e. when somebody requests it and puts > effort into testing it in real-life application:-) Can we do that? I did it. Added to crypto/cryptlib.c: # ifndef OPENSSL_DISABLE_ATOMIC_ADD_INTERNAL int OPENSSL_atomic_add(int *, int, int, const char *, int); # endif and insert at the end of void OPENSSL_cpuid_setup(void) # ifndef OPENSSL_DISABLE_ATOMIC_ADD_INTERNAL CRYPTO_set_add_lock_callback(OPENSSL_atomic_add); # endif So one could disable this feature by defining OPENSSL_DISABLE_ATOMIC_ADD_INTERNAL macro at compile time. Tested on Linux x86 and x64 during last 4 years on different high-load SSL-servers. Tested on Linux x86 and x64, Windows x86 and x64 in different SSL-client applications. No issues observed. Cheers, Andrey. On 12 October 2012 at 18:19, Andy Polyakovwrote: > > I'm working on fast multithreaded TLS server for node.js, and I've > > encountered few problems with openssl threading support which I would > > like to address with attached patches. > > > > Please ask me if you have any questions or comments. > > As for 1st patch. Note that the section in question is guarded by 'if > (add_lock_callback)'. add_lock_callback can be set by application and > perform operation to its liking. In addition, on platforms with > OPENSSL_CPUID_OBJ defined there is OPENSSL_atomic_add that can be used > as callback, so that all you need to do then is > CRYPTO_set_add_lock_callback(OPENSSL_atomic_add) in application code. > Well, you'll surely get into trouble compiling this line, because > OPENSSL_atomic_add is not declared. The reason is that so far it was > reserved for occasion like this, i.e. when somebody requests it and puts > effort into testing it in real-life application:-) Can we do that? > Original idea behind OPENSSL_atomic_add was even to enable it > automatically. On related note one can also add pointer manipulation > subroutines so that one can maintain linked lists in lock-free manner... > > I'll have closer look at second patch and comment later on. > > __ > OpenSSL Project http://www.openssl.org > Development Mailing List openssl-dev@openssl.org > Automated List Manager majord...@openssl.org > -- openssl-dev mailing list To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Re: [openssl-dev] [openssl.org #4073] Segfault in engine processing
On 12 October 2015 at 12:08, Matt Caswell via RTwrote: > Are you using your new GOST > engine or the one currently in master? > Sorry to come in in the middle, but where to get that new GOST engine, that is not on master now? Is it on some other branch? ___ openssl-dev mailing list To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Re: [openssl-dev] [openssl.org #4073] Segfault in engine processing
On 12 October 2015 at 12:08, Matt Caswell via RTwrote: > Are you using your new GOST > engine or the one currently in master? > Sorry to come in in the middle, but where to get that new GOST engine, that is not on master now? Is it on some other branch? ___ openssl-dev mailing list To unsubscribe: https://mta.openssl.org/mailman/listinfo/openssl-dev
Re: [openssl-dev] License compatibility: OpenSSL and Apache v2
This is something I would like to be contributed to OpenSSL. Sure, I'm talking about new engine and new files. And I agree that using the same license as OpenSSL itself would be the best solution. But, if for some reasons licensing of it will be possible under Apache v2 only - what issues could it cause? On 10 December 2014 at 05:54, Salz, Rich rs...@akamai.com wrote: Let's imagine someone develop extension module to OpenSSL, and release it under Apache v2 license. Do you see any possible issues with using this extension module as a part of OpenSSL? Are you writing an extension that you are going to distribute, or is it something you want to contribute to OpenSSL? You cannot change the license on openssl files, and any work derived from them such as modifications to those source files, must have the same license. And if you want to to contribute it to the project, it is easiest for us if the license is the same. But if it is your own code, and you are distributing it, then you can do anything you want :) ___ openssl-dev mailing list openssl-dev@openssl.org https://mta.opensslfoundation.net/mailman/listinfo/openssl-dev ___ openssl-dev mailing list openssl-dev@openssl.org https://mta.opensslfoundation.net/mailman/listinfo/openssl-dev
[openssl-dev] License compatibility: OpenSSL and Apache v2
Hello, Let's imagine someone develop extension module to OpenSSL, and release it under Apache v2 license. Do you see any possible issues with using this extension module as a part of OpenSSL? ___ openssl-dev mailing list openssl-dev@openssl.org https://mta.opensslfoundation.net/mailman/listinfo/openssl-dev
[openssl.org #3613] Patch: Fix warning in Nginx logs on every connect when GOST TLS used.
Hello, As noted here: https://groups.google.com/forum/#!topic/mailing.openssl.dev/Ng3xI7-xZ0E all version of OpenSSL produce scary warning in nginx logs on every connect, if GOST TLS used: [alert] 25532#0: *28 ignoring stale global SSL error (SSL: error:140DD112:SSL routines:SSL_CERT_DUP:library bug) while SSL handshaking All trivial fixes are in ssl_cert_dup() function. Please find patch attached. It can be applied using git apply command. 0001-Update-ssl_cert_dup-for-GOST-key.-Nginx-warning-in-l.patch Description: Binary data
Re: [openssl.org #3533] [PATCH] Ensures that EVP encryption decryption operations check the encrypt flag on the context.
Patch visible. Thanks for contribution. As to actual content - it is up to OpenSSL team. But as to the style: comments of // type is not in OpenSSL code conventions. If you redo them to /**/ way - it could speed-up processing. On 13 November 2014 20:54, Alok Menghrajani via RT r...@openssl.org wrote: Hi, I didn't get any response to my previous email regarding checking the encrypt flag when doing EVP encryption or decryption operations, nor did I get any comments on the patch. I have the feeling I incorrectly formatted my patch, so I'm re-doing it. Alok
Fwd: [openssl.org #3538] 1.0.1h make test fails on test_verify - Debian x64
Strange, but now on the same machine everything works fine. Seems it was fluctuations of world ether... On 21 September 2014 15:08, Andrey Kulikov via RT r...@openssl.org wrote: # uname -a Linux deb7 3.2.0-4-amd64 #1 SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux # gcc --version gcc-4.7.real (Debian 4.7.2-5) 4.7.2 ./config make make test fails with following: ... make[1]: *** [test_verify] Error 2 make: *** [tests] Error 2 make[1]: Leaving directory `openssl-1.0.1h/test' All other versions perform make test Ok, including 1.0.1i __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #3538] 1.0.1h make test fails on test_verify - Debian x64
Strange, but now on the same machine everything works fine. Seems it was fluctuations of world ether... On 21 September 2014 15:08, Andrey Kulikov via RT r...@openssl.org wrote: # uname -a Linux deb7 3.2.0-4-amd64 #1 SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux # gcc --version gcc-4.7.real (Debian 4.7.2-5) 4.7.2 ./config make make test fails with following: ... make[1]: *** [test_verify] Error 2 make: *** [tests] Error 2 make[1]: Leaving directory `openssl-1.0.1h/test' All other versions perform make test Ok, including 1.0.1i
[openssl.org #3538] 1.0.1h make test fails on test_verify - Debian x64
# uname -a Linux deb7 3.2.0-4-amd64 #1 SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux # gcc --version gcc-4.7.real (Debian 4.7.2-5) 4.7.2 ./config make make test fails with following: The following command should have some OK's and some failures There are definitly a few expired certificates ../util/shlib_wrap.sh ../apps/openssl verify -CApath ../certs/demo ../certs/demo/*.pem 47536097586952:error:0B06E06B:x509 certificate routines:X509_get_pubkey_parameters:unable to find parameters in chain:x509_vfy.c:1814: ../certs/demo/ca-cert.pem: C = AU, ST = Queensland, O = CryptSoft Pty Ltd, CN = Test CA (1024 bit) error 20 at 0 depth lookup:unable to get local issuer certificate ../certs/demo/dsa-ca.pem: C = AU, ST = Some-State, O = Internet Widgits Pty Ltd, CN = CA error 20 at 0 depth lookup:unable to get local issuer certificate ../certs/demo/dsa-pca.pem: C = AU, ST = Some-State, O = Internet Widgits Pty Ltd, CN = PCA error 18 at 0 depth lookup:self signed certificate C = AU, ST = Some-State, O = Internet Widgits Pty Ltd, CN = PCA error 10 at 0 depth lookup:certificate has expired OK ../certs/demo/pca-cert.pem: C = AU, ST = Queensland, O = CryptSoft Pty Ltd, CN = Test PCA (1024 bit) error 18 at 0 depth lookup:self signed certificate C = AU, ST = Queensland, O = CryptSoft Pty Ltd, CN = Test PCA (1024 bit) error 10 at 0 depth lookup:certificate has expired OK make[1]: *** [test_verify] Error 2 make: *** [tests] Error 2 make[1]: Leaving directory `openssl-1.0.1h/test' All other versions perform make test Ok, including 1.0.1i __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2938] [PATCH] Severe resource leak in tls_P_hash() (v1.0.1 and up)
Indeed, Improved version of the patch are in: [openssl.org #2937] Handshake performance degradation in 1.0.1 and up. On 9 September 2014 21:16, Rich Salz via RT r...@openssl.org wrote: From an internal review of the patch: Contexts are meant to be reused and (for example) reusing the same context and digest avoids having to reallocate a buffer for the internal context structure. I think this might actually be a problem in the ENGINE not handling reuse properly, or the EVP API not having a way to signal reuse properly. Either way I don't think this is the correct fix. -- Rich Salz, OpenSSL dev team; rs...@openssl.org __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2938] [PATCH] Severe resource leak in tls_P_hash() (v1.0.1 and up)
Indeed, Improved version of the patch are in: [openssl.org #2937] Handshake performance degradation in 1.0.1 and up. On 9 September 2014 21:16, Rich Salz via RT r...@openssl.org wrote: From an internal review of the patch: Contexts are meant to be reused and (for example) reusing the same context and digest avoids having to reallocate a buffer for the internal context structure. I think this might actually be a problem in the ENGINE not handling reuse properly, or the EVP API not having a way to signal reuse properly. Either way I don't think this is the correct fix. -- Rich Salz, OpenSSL dev team; rs...@openssl.org
Re: [openssl.org #1372] default config should diable idea
The last patents for IDEA expired in 2012. Now it's free to use. On 3 July 2014 21:24, Rich Salz via RT r...@openssl.org wrote: As the changelog says for 0.9.8, (IDEA remains enabled despite being patented. This is because IDEA is frequently required for interoperability, and there is no license fee for non-commercial use. As before, no-idea can be used to avoid this algorithm.) __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #3430] ssltest fails when using -DTEMP_GOST_TLS
when I compile using the -DTEMP_GOST_TLS flag What the reason to do it? GOST TLS (at least that one what works) do not require this to be defined.
Re: [openssl.org #3328] [PATCH] Support for GOST R 34.10-2012 digital signature algorithm
Дмитрий, а есть ли у вас планы по внедрению TLS, основанного на новых ГОСТах, в OpenSSL ? Сам собирался занятся этим в начале лета, после отпуска. С вашей помощью, теперь, это совсем тривиально должно получиться. Можно скооперироваться как-нибудь. Если вы, конечно, всё сами не сделаете до этого. :-)
Re: [openssl.org #3322] [PATCH] ccgost to use configured params for 28147-89 in CNT and IMIT mode
Dmitriy, Thanks for noticing! I do not see it either - correcting myself. :-) You are right - according to http://tools.ietf.org/html/draft-chudov-cryptopro-cptls-04 CryptoProParamSetA is required in GOST TLS. But only for content encryption. Premaster secret encryption could use any other parameters. Also, OpenSSL usage is not limited to TLS only - for some other protocols (IPSec, for instance) other parameters can be used. This patch is for them. On 22 April 2014 12:24, Dmitry Belyavsky beld...@gmail.com wrote: Hello Andrey, Thank you for your work, but I do not see the patch :-( I should say that in practice the CNT mode is used in TLS where usage of the Gost28147_CryptoProParamSetA is required. ccgost_CNT_use_params.patch Description: Binary data
[openssl.org #3322] [PATCH] ccgost to use configured params for 28147-89 in CNT and IMIT mode
Currently ccgost engine use configured params (s-boxes) when it works in CFB mode only. For CNT and IMITO parameters are hardcoded to Gost28147_CryptoProParamSetA Supplied patch allow ccgost engine to really use parameters, specified either in config file, or via engine API. When nothing is configured, parameters selection fails back to default - i.e. Gost28147_CryptoProParamSetA. So regression behavior persists. *WARNING: *Some interoperability issues possible, with SSL for example (uses 28147-CNT by default), if some previously configured parameters start to be used, while previously they where ignored. Patch created using this command: \diff -rupN openssl-1.0.1g/engines/ccgost/ openssl-1.0.1g-debug/engines/ccgost/ ccgost_CNT_use_params.patch To apply patch use following command in current OpenSSL root dev. directory: patch -p1 -l -u -b -i ccgost_CNT_use_params.patch Patch created against 1.0.1g. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #3311] [PATCH] Introduce GOST R 34.11-2012 hash function
Hello Dmitry, Thanks a lot for your commitment! It would be good idea to add test cases for new functionality as well. On 14 April 2014 23:52, Dmitry Olshansky via RT r...@openssl.org wrote: It's been a bit over 2 years since the new Russian cryptography standard is out. RFCs 6986 and 7091 been out there for a while[1,2]. Other toolkits are adding support, e.g. Libgcrypt introduced GOST 34.11-2012 in 1.6.0 [3]. Keeping in mind that OpenSSL provides GOST reference engine it seems natural to revise it in the light of the new standard. To summarize the full set of changes for the new standard: - New hash function GOST R 34.11-2012 (Stribog) takes place of GOST R 34.11-94. After 2018 usage of 34.11-94 is basically prohibited. - Digital signature algorithm GOST 34.10-2001 is extended to support both 256bit and 512bit keys. The hash function to use is GOST R 34.11-2012. 256bit version is exactly the same modulo the hashing. - 2 new parameter sets (curves) for 512bit GOST 34.10-2012 are listed as recommended. Older ones stay as is for 256-bit version. - Symmetric cipher stays the same, one new S-box set was defined. - Key Exchange (VKO) is the same algorithm but with different hash function (HMAC of GOST 34.11-2012). This patch adds support for hash algorithm GOST R 34.11-2012 _only_ . The source code was initially tested on x86, PowerPC and ARMv4. New digests have short names md_gost12_256 and md_gost12_512 respectively. See attached patch or browse it on github: https://github.com/openssl/openssl/pull/68 Next steps towards full support are far less involved and consist mostly of minor changes such as adding paramsets and/or removing artificial limitations. 1. GOST R 34.11-2012: Hash Function http://tools.ietf.org/html/rfc6986 2. GOST R 34.10-2012: Digital Signature Algorithm http://tools.ietf.org/html/rfc7091 3. http://lists.gnupg.org/pipermail/gnupg-announce/2013q4/000336.html -- Dmitry Olshansky
Re: [openssl.org #3308] Re: Return missed NULL-check in CMS_add0_cert back
Well... With this check 'make test' fails with: CMS = PKCS#7 compatibility tests signed content DER format, RSA key: generation error make[1]: *** [test_cms] Error 1 Can't reproduce that here. Anyone else seeing this? I saw it on Debian 7 x64, gcc 4.7.2 But anyway, as *pcerts can be NULL legitimately, this seems to be not important. Sorry for wasting your time. :( __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #3310] Can't execute make command
What platform you are using, and what is the config parameters? Can happens due to accidental editing of source file?
[openssl.org #3307] Return missed NULL-check in CMS_add0_cert back
In 1.0.1g duplicated check for (!pcerts) where removed. Had an impression that second appearance was check for (!*pcerts) (as in all other functions). Return it back. Patch applied. 0001-Check-pcerts-for-NULL.patch Description: Binary data
[openssl.org #3308] Re: Return missed NULL-check in CMS_add0_cert back
Well... With this check 'make test' fails with: CMS = PKCS#7 compatibility tests signed content DER format, RSA key: generation error make[1]: *** [test_cms] Error 1 On 14 April 2014 00:16, Andrey Kulikov amde...@gmail.com wrote: In 1.0.1g duplicated check for (!pcerts) where removed. Had an impression that second appearance was check for (!*pcerts) (as in all other functions). Return it back. Patch applied. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #3235] Re: AES256-SHA256 test failed on OpenSSL-1.0.1f ARM64( compiled with -O3 optimization)
Got the same error on Linux, both x86 and x64, when tried to add following lines to ./ssl/t1_enc.c@tls1_change_cipher_state: ssl3_cleanup_key_block(s); OPENSSL_cleanse(s-s3-read_mac_secret, sizeof(s-s3-read_mac_secret)); OPENSSL_cleanse(s-s3-write_mac_secret, sizeof(s-s3-write_mac_secret)); at the very end. Just before return(1); err: SSLerr(SSL_F_TLS1_CHANGE_CIPHER_STATE,ERR_R_MALLOC_FAILURE); Error appears both for TLS 1.2 AES128-SHA256 and AES256-SHA256 ciphersuites. And do NOT appears for any other ciphersuites. On 13 January 2014 20:26, Shawn via RT r...@openssl.org wrote: hey OpenSSL community maintainers, Compiled OpenSSL with GCC -O3 optimization on ARM64 might cause AES256-SHA256 testing fails: - [ 3022s] Testing AES256-SHA256 [ 3024s] Available compression methods: [ 3024s] NONE [ 3024s] ERROR in SERVER [ 3024s] 274890758944:error:1408F119: SSL routines:SSL3_GET_RECORD:decryption failed or bad record mac:s3_pkt.c:484: [ 3024s] TLSv1.2, cipher TLSv1/SSLv3 AES256-SHA256, 2048 bit RSA [ 3024s] 1 handshakes of 256 bytes done [ 3024s] Failed AES256-SHA256 [ 3024s] make[1]: *** [test_ssl] Error 1 [ 3024s] make[1]: Leaving directory `/home/abuild/rpmbuild/BUILD/openssl-1.0.1f/test' --- I'm not sure if this is an issue of OpenSSL. -O3 works for previous release( 1.0.1e). Plz check the fully log which was attached! Thanks! -- GNU powered it... GPL protect it... God blessing it... regards Shawn
Re: [openssl.org #3201] EVP_DigestUpdate crashes because of a NULL pointer
EVP_MD_CTX_copy() also can fail due to external engine usage (including HW engine). In this case underlying work can be more complicated that memory allocation. Descriptors shortage, amount of simultaneous contexts constraints, etc. Or just bugs in engine code. On 27 December 2013 21:39, Misaki.Miyashita via RT r...@openssl.org wrote: I agree that the return value should be checked but I'd like to know the underlying cause. EVP_MD_CTX_copy() shouldn't normally fail unless something is seriously wrong, e.g. memory allocation failures.
Re: [openssl.org #3117] [PATCH] A fast vectorized implementation of binary elliptic curves on x86-64 processors
Dear Manuel, Exciting news! While your paper still unpublished, could you please advice, it there anything even nearly similar possible for curves over primary fields? (e.g. curves secp* ) Best regards, Andrey On 28 August 2013 09:06, Manuel Bluhm via RT r...@openssl.org wrote: Hello all, This patch is a contribution to OpenSSL. It offers an efficient and constant-time implementation of the elliptic curve point multiplication, for the following standard NIST/SECG binary elliptic curves: sect163k1, sect163r1, sect163r2, sect193r1, sect193r2, sect233k1, sect233r1, sect239k1, sect283k1, sect283r1, sect409k1, sect409r1, sect571k1, and sect571r1. The patch implements several improvements at the algorithmic and the coding levels (using SSE/AVX and PCLMULQDQ instructions). Depending on the curve and architecture, this patch offers a speedup of between 4x to 10x for ECDH and ECDSA, compared to the current implementation of OpenSSL 1.0.1e. Additionally, it adds side channel protection to avoid (cache) timing attacks using a number of mechanisms. The code is written in C and uses compiler intrinsics, for simplicity and portability. The following results were obtained with gcc 4.8.1. For detailed explanations of the rationale and algorithms of this code refer to [1]. ECDH performance -- The performance was measured by using openssl speed utility as follows: $ openssl speed ecdh The results for a Core i7-4770 CPU @ 3.40GHz (Haswell) in ECDH op/s: Curve || OpenSSL 1.0.1e || This patch || Speedup || ||||-||--|| |||| || || (nistk163) ||6586.9 || 67029.6|| 10.18 || (nistk233) ||5121.9 || 39441.3|| 7.70 || (nistk283) ||2825.7 || 27718.5|| 9.81 || (nistk409) ||1745.8 || 11634.2|| 6.66 || (nistk571) || 763.2 || 5930.9|| 7.77 || (nistb163) ||6382.5 || 60729.6|| 9.52 || (nistb233) ||4881.9 || 35230.4|| 7.22 || (nistb283) ||2651.6 || 24456.4|| 9.22 || (nistb409) ||1640.3 || 10228.6|| 6.24 || (nistb571) || 693.8 || 5172.1|| 7.45 || |||| || || ||||-||--|| The results for a Core i5-3210M @ 2.50 GHz (Ivy Bridge) in ECDH op/s: Curve || OpenSSL 1.0.1e || This patch || Speedup || ||||-||--|| |||| || || (nistk163) ||3271.5 || 28087.3|| 8.59 || (nistk233) ||2504.9 || 15106.0|| 6.03 || (nistk283) ||1317.0 || 9030.5|| 6.86 || (nistk409) || 772.1 || 3880.8|| 5.03 || (nistk571) || 327.3 || 1821.1|| 5.56 || (nistb163) ||3067.9 || 24357.1|| 7.94 || (nistb233) ||2424.9 || 3147.3|| 5.42 || (nistb283) ||1227.0 || 7765.1|| 6.33 || (nistb409) || 709.7 || 3319.9|| 4.68 || (nistb571) || 296.2 || 1563.9|| 5.28 || |||| || || ||||-||--|| ECDSA performance -- The performance was measured by using openssl speed utility as follows: $ openssl speed ecdsa The results for a Core i7-4770 CPU @ 3.40GHz (Haswell): Curve || OpenSSL 1.0.1e ||This patch || Speedup || ---||-||---||-|| || sign/s verify/s || sign/s verify/s || sign/s verify/s || ||-||---||-|| (nistk163) || 6,465.3 3,159.5 || 36,872.6 26,508.4 || 5.708.39 || (nistk233) || 3,259.2 2,419.8 || 22,998.4 15,557.1 || 7.066.43 || (nistk283) || 2,204.7 1,355.7 || 16,884.9 11,003.2 || 7.668.12 || (nistk409) || 977.0 839.1 || 8,150.0 4,845.0 || 8.345.77 || (nistk571) || 466.4 368.3 || 4,424.1 2,533.6 || 9.496.88 || (nistb163) || 6,487.3 3,043.9 || 35,110.0 24,904.8 || 5.418.18 || (nistb233) || 3,279.2 2,348.0 || 21,468.8 14,095.6 || 6.556.00 || (nistb283) || 2,196.4 1,283.5 || 15,602.7 9,888.5 || 7.107.70 || (nistb409) || 976.3 786.9 || 7,423.1 4,361.9 || 7.605.54 || (nistb571) || 466.6 341.0 || 3,977.0 2,251.6 || 8.526.60 || || || || || ---||-||---||-|| The results for a Core i5-3210M CPU @ 2.50 GHz (Ivy Bridge): Curve
[openssl.org #2953] s_server to show connection duration and transfer speed
Please find attached patch, introducing two new options for s_server: one specify maximum number of connections s_server will accept. It will exit clearly after completing last connection the other tells s_server to show each completed connection duration and data transfer rate. These options exceptionally usable for profiling, debugging and test automation. Patch correctly applied to openssl-1.0.2-stable-SNAP-20130108 In order to apply patch you may use command: patch -p0 -l -u -b -i s_server_limit_show_speed_long.patch and patch -p1 -l -u -b -i s_server_limit_show_speed_doc.patch Please let me know if you have any questions. Andrey. s_server_limit_show_speed_long.patch Description: Binary data s_server_limit_show_speed_doc.patch Description: Binary data
Re: [openssl.org #2937] Handshake performance degradation in 1.0.1 and up.
On 11 December 2012 04:00, Stephen Henson via RT r...@openssl.org wrote: I also notice that even the original HMAC version initialises two HMAC contexts with the same key. That could be improved by initialising one and copying the context across. This kind of optimization can be also applied P_hash implementation via to EVP_DigestSign*. HMAC contexts being re-initialized in inner loop every time it executed. It is possible to improve it here by initializing only once in the beginning of function, save to some temporary variables, and restore restore from them when needed. It would solve original issue with re-initialilization excessive overhead. BUT: 1. It gives only miserable performance benefits in normal case (i.e. software-only) ( 1%). Hashing of few more bytes is almost nothing in comparison to BN-manipulations. 2. In case of external hardware usage it may lead to the same result: in my case cost of MAC copying call is the same as cost of MAC calculation call. And only way to improve performance is to reduce total amount of remote calls. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2937] Handshake performance degradation in 1.0.1 and up.
Please find attached two patches, together implementing proper HMAC context re-initialization instead of full re-creation. In comparison to openssl-1.0.1c it gives ~10% handshake performance improvements when some engine-specific MAC are used. In order to apply patches use command patch -p1 -i filename Patches checked to applied to: openssl-1.0.1c openssl-1.0.2-stable-SNAP-20130108 make test report Ok. for both versions (Linux, x86[_64]) Please let me know if you have any questions. TLS_P_hash-HMAC_reinit.patch Description: Binary data TLS_P_hash-HMAC_reinit2.patch Description: Binary data
Re: ssl3_get_cert_verify performance
I'm affraid flamegraphs for two different servers with two different OpenSSL libraries without information about type of load it was collected with and without any information other than smth. started spending much more time than it was can not give much information about root cause for the issue. You may start with: 1. Describing you test procedure. (what are you measuring? Handshakes? Data transfer? Multithreaded?) 2. Describing test environment (type of certificates, key length, used chiphersuites, etc) 3. Measure performance differences in absolute values (ms, handshakes per second, etc) 4. Measure nginx perfromance with two different OpenSSL versions. 5. Measure your server performance with two different OpenSSL versions. 6. Share your findings. On 5 January 2013 13:14, Fedor Indutny fe...@indutny.com wrote: Hello devs, Right now I'm doing a lot of benchmarks, trying to figure out how to make my https server as fast as are others (for example, nginx). I've found that somewhere between 0.9.8 and 1.0.1c ssl3_get_cert_verify has started spending much more time than it was. I wonder if you're aware of it, or if this thing can depend on some SSL_CTX mode/flag. Here are flamegraphs for you to make it more clearer what I'm talking about: * My server (openssl1.0.1c) - http://blog.indutny.com/f/tlsnappy-x64.svg * Nginx (openssl0.9.8) - http://blog.indutny.com/f/nginx.svg And here are sources of my server, just in case if you need them to figure something out: https://github.com/indutny/tlsnappy/blob/master/src/tlsnappy.cc Thank you, Fedor.
Re: [openssl.org #2951] [PATCH] Fix speed.c to fork N threads
- for(n=0 ; n multi ; ++n) + for(n=0 ; n = multi ; ++n) Isn't |[0..n)| = n and |[0..n]| = n +1 ? I.e this patch will do one more iteration than expected? Assume multi = 1, for example.
Re: [openssl.org #2937] Handshake performance degradation in 1.0.1 and up.
In my case, handshake rate drops down to 5-6% on the same hardware in 1.0.1c in comparison to 1.0.0i. I was wrong. Handshake performance degradation is about 10%. First guilty function is EVP_DigestSignFinal what is perform copying of supplied context. When I replaces in tls1_P_hash() EVP_DigestSignFinal() by it's dull equvivalent, not performing context copying i got these numbers: OpenSSL 1.0.1c - EVP_DigestSignFinalNoCopy Digest init called 105 times. Digest copy called 70 times. Digest cleanup called 174 times. And get 5% performance improvement in comparison to original OpenSSL 1.0.1c (Still, 5% worth than 1.0.0) Second suspect is re-initialization of MAC context in tls1_P_hash()'s loop. While it was real re-initialization when there was HMAC_Init_ex(), what performed initialization of internal i_ctx and o_ctx only when necessary, with EVP_DigestSignInit it is full re-creating, and internal i_ctx and o_ctx is always initialized. This is why we see 3 times more 'digest init' calls. Also, there could be other reasons for what is still hidden from me. Eliminating EVP_DigestSignFinal overhead in tls1_P_hash() by replasing it with calls, what do not perform context copying is trivial. But how can we properly perfrom MAC true re-initialization instead of creation from very beggining? As a drawback, keyblock setup for a chiphersuites with 256-bit encryption and MAC key require about 3 times more intensive usage of hash objects. For example, in order to perform one handshake, in OpenSSL 1.0.0i Digest init called 30 times. Digest copy called 69 times. Digest cleanup called 98 times. OpenSSL 1.0.1c Digest init called 105 times. Digest copy called 160 times. Digest cleanup called 264 times. ~3 times more intensive hashes objects usage definitely not good for performance. In my case, handshake rate drops down to 5-6% on the same hardware in 1.0.1c in comparison to 1.0.0i. Is there any way to reduce hash objects usage, while keeping TLS 1.1/1.2 features?
[openssl.org #2937] Handshake performance degradation in 1.0.1 and up.
In comparison to 1.0.0, in 1.0.1 the implementation of PRF have been changed. In order to supporf TLS 1.1/1.2 features, in file ssl/t1_enc.c, in function tls_P_hash() calls to HMAC_Init/HMAC_Update/HMAC_Final where replaced by EVP_DigestSignInit/EVP_DigestSignUpdate/EVP_DigestSignFinal. As a drawback, keyblock setup for a chiphersuites with 256-bit encryption and MAC key require about 3 times more intensive usage of hash objects. For example, in order to perform one handshake, in OpenSSL 1.0.0i Digest init called 30 times. Digest copy called 69 times. Digest cleanup called 98 times. OpenSSL 1.0.1c Digest init called 105 times. Digest copy called 160 times. Digest cleanup called 264 times. ~3 times more intensive hashes objects usage definitely not good for performance. In my case, handshake rate drops down to 5-6% on the same hardware in 1.0.1c in comparison to 1.0.0i. Also, more intense malloc usage leads to faster head fragmentation. But I didn't able to measure impact of that factor yet. Is there any way to reduce hash objects usage, while keeping TLS 1.1/1.2 features? In comparison to 1.0.0, in 1.0.1 the implementation of PRF have been changed.In order to supporf TLS 1.1/1.2 features, in file ssl/t1_enc.c, in function tls_P_hash() calls to HMAC_Init/HMAC_Update/HMAC_Final where replaced by EVP_DigestSignInit/EVP_DigestSignUpdate/EVP_DigestSignFinal. As a drawback, keyblock setup for a chiphersuites with 256-bit encryption and MAC key require about 3 times more intensive usage of hash objects.For example, in order to perform one handshake, in OpenSSL 1.0.0iDigest init called 30 times.Digest copy called 69 times.Digest cleanup called 98 times. OpenSSL 1.0.1c Digest init called 105 times. Digest copy called 160 times.Digest cleanup called 264 times. ~3 times more intensive hashes objects usage definitely not good for performance.In my case, handshake rate drops down to 5-6% on the same hardware in 1.0.1c in comparison to 1.0.0i. Also, more intense malloc usage leads to faster head fragmentation. But I didnt able to measure impact of that factor yet.Is there any way to reduce hash objects usage, while keeping TLS 1.1/1.2 features?
[openssl.org #2938] [PATCH] Severe resource leak in tls_P_hash() (v1.0.1 and up)
Hello, In v1.0.1 tls_P_hash() has been changed in comparison to early OpenSSL versions. Used hash objects is re-initializing during P_hash calculation. It looks harmless, but only until we come to hash objects, holding references to external objects. E.g. engine-specific hashes, allocating handles to hardware device or some external library. Then re-initing these MACs does not result in memory loss, but handlers to external objects, stored in that memory being lost. (this is why I call it resource leak, not memory). In my case each handshake leads to 87 unfreed handles, and library, interacting with hardware device starts to fail due to out of memory after about 6000 handshakes. Please find attached patch, correcting discovered issue. It can be applied to 1.0.1, 1.0.2 and to latest source tarball. make test report success for 1.0.1c with this patch applied. If anyone see any issues in supplied patch - please let me know. Best wishes, Andrey Kulikov. Hello,In v1.0.1 tls_P_hash() has been changed in comparison to early OpenSSL versions.Used hash objects is re-initializing during P_hash calculation.It looks harmless, but only until we come to hash objects, holding references to external objects. E.g. engine-specific hashes, allocating handles to hardware device or some external library.Then re-initing these MACs does not result in memory loss, but? handlers to external objects, stored in that memory being lost. (this is why I call it resource leak, not memory).In my case each handshake leads to 87 unfreed handles, and library, interacting with hardware device starts to fail due to out of memory after about 6000 handshakes. Please find attached patch, correcting discovered issue.It can be applied to 1.0.1, 1.0.2 and to latest source tarball.make testreport success for 1.0.1c with this patch applied.If anyone see any issues in supplied patch - please let me know. Best wishes,Andrey Kulikov. tls1_P_hash_resource_leak.patch Description: Binary data
[openssl.org #2939] Re: [FIX] 1.0.0d: All platforms: GOST server MUST check correctness of shared UKM
Just noticing the wrong goto label in case of EVP_PKEY_CTX_ctrl() failue. Please find attached corrected patch (gost_server_to_check_ukm_v2.patch) On 17 April 2011 17:54, Andrey Kulikov amde...@gmail.com wrote: According to this document: http://tools.ietf.org/html/draft-chudov-cryptopro-cptls-04#section-3.6 Server, implementing GOST algost MUST check correctness of shared UKM, send by client. == Server MUST verify, that keyBlob.transportParameters.ukm is equal to GOSTR3411(client_random|server_random)[0..7], before decrypting the premaster_secret. == There is no such checks in 1.0.0d at all. Attachecd patch implements missing functionality. To apply patch use following command in root of current OpenSSL development tree: patch -p1 -l -u -b -i gost_server_to_check_ukm.patch Andrey. P.S. Checked to works fine with two CSP from different vendors, as well as openssl itself. Just noticing the wrong goto label in case of EVP_PKEY_CTX_ctrl() failue.Please find attached corrected patch (gost_server_to_check_ukm_v2.patch)On 17 April 2011 17:54, Andrey Kulikov amde...@gmail.com wrote: According to this document:http://tools.ietf.org/html/draft-chudov-cryptopro-cptls-04#section-3.6 Server, implementing GOST algost MUST check correctness of shared UKM, send by client. == Server MUST verify, that keyBlob.transportParameters.ukm is equal to GOSTR3411(client_random|server_random)[0..7], before decrypting the premaster_secret. ==There is no such checks in 1.0.0d at all.Attachecd patch implements missing functionality.To apply patch use following command in root of current OpenSSL development tree: patch -p1 -l -u -b -i gost_server_to_check_ukm.patchAndrey.P.S. Checked to works fine with two CSP from different vendors, as well as openssl itself. gost_server_to_check_ukm_v2.patch Description: Binary data
Re: Multithreading performance patches
Hi Fedor, Thanks for valuable contribution! About your second patch: could you please advice, what the intended way to use BIO_TYPE_NO_EX_DATA in real-life applications? For example TSL-server. Do you have any numbers on performance gain when using this approach?
Re: OpenSSL and CRIME
Hi, OpenSSL enables zlib by default. Could you please advice for what version and platform this is true? openssl-1.0.1c for linux-elf has no-zlib configured by default.
Re: Why x509 displays ECDSA pub. key size two times less than actual?
By convention the key size for ECC is given as the number of bits in the order. E.g. see table 3 in SEC 1. Could you please provide a reference to document, defining this convention? Unfortunatelly table 3 in section B.2.1 of SEC 1 (both v1 and v2) shows only comparison of security level of diffefent algorithmsm and not related to the questions.
Re: Why x509 displays ECDSA pub. key size two times less than actual?
For ECC the key size is stated to be size of n in bits (where n is the order). 'n' is an order of base point 'G' on EC - i.e. size of private key (what should be in range [1; n-1]), not public. Thus, I understand that it is a size of EC private key 'd', what shown in table 3 in SEC 1.
Re: Why x509 displays ECDSA pub. key size two times less than actual?
Talking about the bit-length of the public key data is not particularly helpful because it depends on whether it is in compressed format or not. Sorry, but size of public key does not depends on size of it's representation. It can be compressed, Base64 encoded, etc., but it does not change size of original data. x509 command output was mentioned only for example. The aim was is to understand, why OpenSSL (and not only) consider size of something as twice as less of actual size, while displaying data itself on next line of proper size. Title for this data is 'Public key', not 'Size of private key, public key in this certificate related to'. What trows my programmer's ming into cognitive dissonance. Is it bug, feature, or [yet] hidden convention?
Re: Why x509 displays ECDSA pub. key size two times less than actual?
As I said before it is convention when talking about a _key size_ to mean the number of bits in the order (i.e. the size of the private key). If we talking about just 'key size', then yes, we assume private key size. But I've never heard of convention to assume that for ECC public key size is equal to private key size. If this convention defined, or at least mentioned somewhere, it would be wonderful to have a reference to this document.
Why x509 displays ECDSA pub. key size two times less than actual?
Hello, Let's generate certificate with ECDSA key over 256 bit field: openssl ecparam -out key.pem -name prime256v1 -genkey openssl req -newkey ec:key.pem -x509 -nodes -days 365 -keyout pkey.pem -out cert.pem Then part of output of openssl x509 -text -in cert.pem will be: Subject Public Key Info: Public Key Algorithm: id-ecPublicKey Public-Key: (256 bit) pub: 04:89:9e:f7:b0:fa:69:b7:e3:1b:62:e8:a5:97:56: 68:4b:36:74:fe:df:e7:f8:fe:a8:a2:0b:a8:2b:a3: 2c:80:d1:6f:4e:00:b8:70:92:b1:2b:20:e9:ff:c4: f2:10:76:72:4c:34:80:30:a3:37:b5:60:e1:a4:a7: 07:e0:f8:2a:9e ASN1 OID: prime256v1 Why it states that pub. key is only 256 bits long? By definition public key consists of two coordinates on EC curve, each 256 bits (for selected curve) long. I.e. 512 bits in total. Moreover, pub. key dump contains 512 bits too... (04 in first byte indicates that uncompressed points representation used) Code in ec_ameth.c::do_EC_KEY_print() calculates public key size as order of curve used. What gives us half of real ECDSA pub. key size. Does this behavior a bug, or somewhere documented convention? I've studied FIPS 180-3, SP 800-57 and SEC 1: Elliptic Curve Cryptography but didn't find any indications of such conventions. Maybe I overlooked something? P.S. Interesting, but Windows displays size of pub. key in this certificate also as 256 bits long.
Re: [Q] OpenSSL memory usage per connection
Using SSL_MODE_RELEASE_BUFFERS option can drop memory footprint from 50K to around 5K per connection. It is useful when we'll have big number of connections. Also it may prevent excessive swapping. On 7 April 2012 08:22, eric wang eric_x_w...@yahoo.com wrote: Dear OpenSSL developers We got a question on OpenSSL (0.9.8 or later) and hope to get help from the community. What is the memory usage per connection for OpenSSL 0.9.8 or 1.0.1? What factors affect the memory usage, e.g. build options, authentication method? Really appreicate it if you can help Thanks
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
On 16 December 2011 02:47, Andy Polyakov via RT r...@openssl.org wrote: Attached patch fix the problem on cygwin. Does it mean that you can in fact confirm that modified speed.c runs without hanging? Yes, modified speed.c, being compiled in the same Cygwin environment as before, runs without hanging for all hashes. Verify http://cvs.openssl.org/chngview?cn=21853. Works! __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
[openssl.org #2657] Win32: fail to compile - 1.0.1 snapshot 20111211
VisualStudio 2010 perl Configure VC-WIN32 ms\do_ms.bat nmake -f ms\nt.mak .. ml /nologo /Cp /coff /c /Cx /Zi /Fotmp32\vpaes-x86.obj tmp32\vpaes-x86.asm Assembling: tmp32\vpaes-x86.asm tmp32\vpaes-x86.asm(491) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(527) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(545) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(572) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(616) : error A2070:invalid instruction operands NMAKE : fatal error U1077: 'C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\BIN\ml.EXE' : return code '0x1' Stop. ml Microsoft (R) Macro Assembler Version 10.00.30319.01 Copyright (C) Microsoft Corporation. All rights reserved. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2657] Win32: fail to compile - 1.0.1 snapshot 20111211
The same for SNAP-20111213 On 14 December 2011 19:54, Andrey Kulikov via RT r...@openssl.org wrote: ml /nologo /Cp /coff /c /Cx /Zi /Fotmp32\vpaes-x86.obj tmp32\vpaes-x86.asm Assembling: tmp32\vpaes-x86.asm tmp32\vpaes-x86.asm(491) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(527) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(545) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(572) : error A2070:invalid instruction operands tmp32\vpaes-x86.asm(616) : error A2070:invalid instruction operands NMAKE : fatal error U1077: 'C:\Program Files (x86)\Microsoft Visual Studio 10.0\VC\BIN\ml.EXE' : return code '0x1' Stop. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Checked with 1.0.0e - the same problem. speed hangs up with md4 md5 mdc2 and sha1, but works well with sha256 sha512 and whirlpool. I downloaded 8G file via s_server from 1.0.1 snapshot 20111213 - works fine. Seems problem in speed' itself, or, as Andy mentioned - in time calculation. As in addition, even if speed' do not hangs up, each group of block calkulation MAY take significantly more than 3 sec... On 14 December 2011 01:45, Andrey Kulikov via RT r...@openssl.org wrote: Tested on two computers. Both native (i.e. non-VM) Windows7 x64 Professional (without SP1). One has E8600 CPU, the other is laptop with i5 mobile CPU. Both has 8G RAM. Both working stable for a monthes. No abnormalities, no BSODs. gcc version 3.4.4 (cygming special, gdc 0.12, using dmd 0.125) Cygwin DLL version is 1.7.9-1 Package list (output of cygcheck -cd )attached. On 13 December 2011 13:28, Andy Polyakov via RT r...@openssl.org wrote: Program behavior appears to be sane, Is it virtual machine? Which OS is it? What is the host? I.e. VMware, Xen, ...? Have you ever observed anomalies in ping output? Anomalies like it's reporting negative round-trip time? __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
downloaded 8G file via s_server from 1.0.1 snapshot 20111213 - works fine. Using RC4-SHA chipthersuite. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Checked with 1.0.0e - the same problem. speed hangs up with md4 md5 mdc2 and sha1, but works well with sha256 sha512 and whirlpool. I downloaded 8G file via s_server from 1.0.1 snapshot 20111213 - works fine. (using RC4-SHA chiphersuite) Seems problem in speed' itself, or, as Andy mentioned - in time calculation. As in addition, even if speed' do not hangs up, each group of block calkulation MAY take significantly more than 3 sec... __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Compiler might be complaining about some small things, use your judgment to ignore warnings or to make it compile. It does complain! :-) Attached patch fix the problem on cygwin. Back-ported to 1.0.0e was not break native Win32 (32 bit, VS10) compilation. spped_cygwin.patch Description: Binary data
Re: [openssl.org #2657] Win32: fail to compile - 1.0.1 snapshot 20111211
Verify http://cvs.openssl.org/chngview?cn=21845. This patch helps! Thanks a lot! ml Microsoft (R) Macro Assembler Version 10.00.30319.01 Copyright (C) Microsoft Corporation. All rights reserved. Microsoft Assembler gets *very* limited testing and the official standpoint is to favor nasm. In other words in case of doubt, switch to nasm. Eh... Missed string in INSTALL.W32: Note that NASM is now the only supported assembler. :-) __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
It does complain! :-) Attached patch fix the problem on cygwin. Does it mean that you can in fact confirm that modified speed.c runs without hanging? Yes, modified speed.c, being compiled in the same Cygwin environment as before, runs without hanging for all hashes. Interesting, why we do not have such effect fo ciphers and for three selected hashes BTW, do you get same KBps values? Before patch: The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes256 bytes 1024 bytes 8192 bytes sha256 32264.54k75889.08k 133579.66k 166999.63k 177915.57k sha512 17319.36k69456.32k 103750.87k 144978.57k 166290.93k whirlpool22616.29k51999.79k91632.37k 108593.83k99194.67k rc4 423843.86k 533050.06k 558939.38k 597673.15k 604000.56k idea cbc 64826.69k65397.38k67318.14k66839.71k66455.04k aes-128 cbc 95172.64k 107472.19k 110012.39k 176664.53k 177304.57k aes-256 cbc 69245.45k77572.27k79714.56k 128283.83k 128298.72k After the patch: The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes sha256 32567.32k76673.55k 130986.27k 162825.57k 178479.48k sha512 17374.95k70081.18k 105646.81k 143291.62k 164904.29k whirlpool 22472.70k52594.12k91494.43k 109679.24k 100349.95k rc4 435291.94k 548119.26k 579422.27k 597697.81k 608022.70k idea cbc63872.42k66033.29k66642.98k68983.62k 67100.26k aes-128 cbc 96134.45k 106319.68k 110688.68k 178169.51k 179018.97k aes-256 cbc 69849.36k79412.17k80094.75k 127889.70k 130016.22k __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Tested on two computers. Both native (i.e. non-VM) Windows7 x64 Professional (without SP1). One has E8600 CPU, the other is laptop with i5 mobile CPU. Both has 8G RAM. Both working stable for a monthes. No abnormalities, no BSODs. gcc version 3.4.4 (cygming special, gdc 0.12, using dmd 0.125) Cygwin DLL version is 1.7.9-1 Package list (output of cygcheck -cd )attached. On 13 December 2011 13:28, Andy Polyakov via RT r...@openssl.org wrote: Program behavior appears to be sane, Is it virtual machine? Which OS is it? What is the host? I.e. VMware, Xen, ...? Have you ever observed anomalies in ping output? Anomalies like it's reporting negative round-trip time? Cygwin Package Information Package Version _update-info-dir 01006-1 alternatives 1.3.30c-10 autoconf 10-1 autoconf2.1 2.13-10 autoconf2.5 2.68-1 automake 4-10 automake1.10 1.10.3-1 automake1.11 1.11.1-1 automake1.4 1.4p6-10 automake1.5 1.5-10 automake1.6 1.6.3-11 automake1.7 1.7.9-10 automake1.8 1.8.5-10 automake1.9 1.9.6-10 base-cygwin 3.0-1 base-files4.0-6 base-passwd 3.1-2 bash 4.1.10-4 bash-completion 1.3-1 binutils 2.22.51-1 bison 2.4.2-1 build-docbook-catalog 1.5-2 bzip2 1.0.6-2 ca-certificates 1.78-1 cmake 2.8.4-1 coreutils 8.14-1 cpio 2.11-1 crypt 1.1-1 csih 0.9.4-1 curl 7.22.0-2 cvs 1.12.13-10 cvsps 2.2b1-1 cygrunsrv 1.34-1 cygutils 1.4.6-1 cygwin1.7.9-1 cygwin-doc1.7-1 dash 0.5.7-1 dbus 1.4.16-1 diffutils 2.9-1 docbook-dsssl 1.79-2 docbook-sgml303.0-1 docbook-sgml313.1-1 docbook-sgml404.0-1 docbook-sgml414.1-1 docbook-sgml424.2-1 docbook-sgml434.3-1 docbook-sgml444.4-1 docbook-sgml454.5-1 docbook-utils 0.6.14-1 docbook-xml4124.1.2-2 docbook-xml42 4.2-4 docbook-xml43 4.3-2 docbook-xml44 4.4-2 docbook-xml45 4.5-1 docbook-xsl 1.76.1-1 dog 1.7-1 dos2unix 5.3.1-1 ed1.0-1 editrights1.01-2 file 5.09-1 findutils 4.5.9-2 flex 2.5.35-1 font-adobe-dpi75 1.0.2-1 font-alias1.0.3-1 font-bh-dpi1001.0.2-1 font-bh-dpi75 1.0.2-1 font-bitstream-dpi100 1.0.2-1 font-bitstream-dpi75 1.0.2-1 font-cursor-misc 1.0.2-1 font-encodings1.0.4-1 font-micro-misc 1.0.2-1 font-misc-cyrillic1.0.2-1 font-misc-misc1.1.1-1 font-screen-cyrillic 1.0.3-1 font-sun-misc 1.0.2-1 font-util 1.2.0-1 font-winitzki-cyrillic1.0.2-1 font-xfree86-type11.0.3-1 fontconfig2.8.0-1 gamin 0.1.10-11 gawk 4.0.0-1 gcc-core 3.4.4-999 gcc-g++ 3.4.4-999 gcc-mingw-core20050522-3 gcc-mingw-g++ 20050522-3 gdb 7.3.50-3 geoip 1.3.14-1 gettext 0.18.1.1-2 git 1.7.5.1-1 git-completion1.7.5.1-1 git-gui 1.7.5.1-1 git-oodiff20080328-2 gitk 1.7.5.1-1 grep 2.6.3-1 groff 1.20.1-2 gsettings-desktop-schemas 3.2.0-1 gzip 1.4-1 hexedit 1.2.12-2 indent2.2.11-1 ipc-utils 1.0-1 jadetex 3.13-1 less 444-1 libapr1 1.4.5-1 libaprutil1 1.3.12-1 libattr1 2.4.43-1 libblkid1 2.17.2-1 libbz2_1 1.0.6-2 libcharset1 1.14-2 libcurl4 7.22.0-2 libdb4.5 4.5.20.2-2 libdbus1_31.4.16-1 libedit0 20090923-1 libexpat1 2.0.1-1 libfam0 0.1.10-11 libffi4 4.5.3-3 libfontconfig12.8.0-1 libfontenc1 1.1.0-1 libfreetype6 2.4.7-1 libgcc1 4.5.3-3 libgcrypt11 1.4.6-1 libgdbm4 1.8.3-20 libGL1
[openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
OpenSSL 1.0.1 snaphot 20111211, compiled on Cygwin, no special config options. gcc version 3.4.4 (cygming special, gdc 0.12, using dmd 0.125) CPU Code2 Duo E8600 make tests shows: ../util/shlib_wrap.sh ./shatest test 1 ok test 2 ok test 3 ok ../util/shlib_wrap.sh ./sha1test test 1 ok test 2 ok test 3 ok ../util/shlib_wrap.sh ./sha256t Testing SHA-256 ... passed. Testing SHA-224 ... passed. ../util/shlib_wrap.sh ./sha512t Testing SHA-512 ... passed. Testing SHA-384 ... passed. Both ./apps/openssl speed sha1 and ./apps/openssl speed -evp sha1 MAY results in hangup on any stage. Sometimes speed command complete successfylly. $ ./apps/openssl speed sha1 Doing sha1 for 3s on 16 size blocks: 2726540 sha1's in 2.51s Doing sha1 for 3s on 64 size blocks: 2573136 sha1's in 2.73s Doing sha1 for 3s on 256 size blocks: 1862928 sha1's in 2.61s Doing sha1 for 3s on 1024 size blocks: 935145 sha1's in 2.64s Doing sha1 for 3s on 8192 size blocks: 158296 sha1's in 2.65s OpenSSL 1.0.1-dev xx XXX built on: Mon Dec 12 01:14:51 RST 2011 options:bn(64,32) rc4(8x,mmx) des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: gcc -DOPENSSL_THREADS -DDSO_DLFCN -DHAVE_DLFCN_H -DTERMIOS -DL_ENDIAN -fomit-frame-pointer -O3 -march=i486 -Wall -DOPENSSL_BN_ASM _PART_WORDS -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DRMD160_ASM -DAE S_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes256 bytes 1024 bytes 8192 bytes sha1 17373.41k60322.60k 183004.44k 363273.32k 488974.67k Sometimes it hang up on block calculation of any size. At this time CPU load 100%, and nothing. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
[openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
The same behaviour is for other hashes: md4 md5 mdc2 and sha. But not for sha256, sha512 and whirlpool - everything is ok with these hashes: $ ./apps/openssl speed sha512 sha256 whirlpool OpenSSL 1.0.1-dev xx XXX built on: Tue Dec 13 08:14:31 RST 2011 options:bn(64,32) rc4(8x,mmx) des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx) compiler: gcc -DOPENSSL_THREADS -DDSO_DLFCN -DHAVE_DLFCN_H -g -DTERMIOS -DL_ENDIAN -fomit-frame-pointer -O3 -march=i486 -Wall -DOPENSSL_BN_ASM_PART_WORDS -DOPENSSL_IA32_SSE2 -DOPENSSL_BN_ASM_MONT -DOPENSSL_BN_ASM_GF2m -DSHA1_ASM -DSHA256_ASM -DSHA512_ASM -DMD5_ASM -DRMD160_ASM -DAES_ASM -DVPAES_ASM -DWHIRLPOOL_ASM -DGHASH_ASM The 'numbers' are in 1000s of bytes per second processed. type 16 bytes 64 bytes256 bytes 1024 bytes 8192 bytes sha256 32216.74k75616.73k 132102.69k 165474.27k 179371.50k sha512 17037.17k67636.17k 103818.78k 148730.56k 160855.87k whirlpool21894.41k51207.20k87122.50k 105701.73k96333.88k On 12 December 2011 19:08, Andy Polyakov via RT r...@openssl.org wrote: OpenSSL 1.0.1 snaphot 20111211, compiled on Cygwin, no special config options. gcc version 3.4.4 (cygming special, gdc 0.12, using dmd 0.125) CPU Code2 Duo E8600 Both ./apps/openssl speed sha1 and ./apps/openssl speed -evp sha1 MAY results in hangup on any stage. Sometimes speed command complete successfylly. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
[openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Can you reproduce it under cygwin gdb? I.e. 'gdb apps/openssl' and then at gdb prompt 'run speed sha1'. If yes, ctrl-C and 'info threads'. Then 'cont', ctrl-C and 'info threads' few more times. Look at thread #1 at all occasions. Where is it caught? Does program counter varies? If it's caught in sha1_block_data_order_ssse3, collect several 'thread apply 1 info reg' outputs. Disgusting gdb under sickening cygwin do not understand Crtl-C! Fortunately workaround available: http://www.mingw.org/wiki/Workaround_for_GDB_Ctrl_C_Interrupt openssl was compiled with ./config -g (gdb) info threads Id Target Id Frame * 12 Thread 8820.0x1fc0 0x77b2000d in ntdll!LdrFindResource_U () from /cygdrive/c/Windows/system32/ntdll.dll 8Thread 8820.0x2280 0x77b2f861 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 7Thread 8820.0x19bc 0x77b2fd21 in ntdll!RtlFindSetBits () from /cygdrive/c/Windows/system32/ntdll.dll 6Thread 8820.0x2bd0 0x77b2fd9d in ntdll!RtlFindSetBits () from /cygdrive/c/Windows/system32/ntdll.dll 5Thread 8820.0x2b40 0x77b2f861 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 4Thread 8820.0x1634 0x77b300ed in ntdll!RtlEnableEarlyCriticalSectionEventCreation () from /cygdrive/c/Windows/system32/ntdll.dll 2Thread 8820.0x1f18 0x77b2f895 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 1Thread 8820.0x272c _sha1_block_data_order_ssse3 () at sha1-586.s:1689 (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0xcc6ee5e214363742 ecx0xff2dbcf7 -13779721 edx0x47b21e88 1202855560 ebx0x3430f0e8 875622632 esp0x287e40 0x287e40 ebp0x41a18787 0x41a18787 esi0x47b31288 1202918024 edi0xb3a52001 -1281023999 eip0x555c3e 0x555c3e _sha1_block_data_order_ssse3+990 eflags 0x203[ CF IF ] cs 0x23 35 ss 0x2b 43 ds 0x2b 43 es 0x2b 43 fs 0x53 83 gs 0x2b 43 (gdb) (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0x24f45c66 619994214 ecx0xcb1d6c20 -887264224 edx0x57307c2c 1462795308 ebx0x11f11a9f 301013663 esp0x287f1c 0x287f1c ebp0x10082080x1008208 esi0xe85d345d -396544931 edi0xf70aaf7d -150294659 eip0x556710 0x556710 _sha1_block_data_order_ssse3+3760 eflags 0x202[ IF ] cs 0x23 35 ss 0x2b 43 ds 0x2b 43 es 0x2b 43 fs 0x53 83 gs 0x2b 43 (gdb) (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0x2cbb7e4d 750485069 ecx0x481a01575603989 edx0xe3e6332c -471452884 ebx0xa7f4213a -1477172934 esp0x287e40 0x287e40 ebp0xa2e42130 0xa2e42130 esi0x99405b26 -1723835610 edi0x280b64d3 671835347 eip0x556063 0x556063 _sha1_block_data_order_ssse3+2051 eflags 0x282[ SF IF ] cs 0x23 35 ss 0x2b 43 ds 0x2b 43 es 0x2b 43 fs 0x53 83 gs 0x2b 43 (gdb) (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0xbc696264 -1133944220 ecx0xafee45bd -1343339075 edx0x54158eaf 1410698927 ebx0x5e827d65 1585610085 esp0x287e40 0x287e40 ebp0x25e34b13 0x25e34b13 esi0xf16c38d8 -244565800 edi0xf72a45d5 -148224555 eip0x556049 0x556049 _sha1_block_data_order_ssse3+2025 eflags 0x206[ PF IF ] cs 0x23 35 ss 0x2b 43 ds 0x2b 43 es 0x2b 43 fs 0x53 83 gs 0x2b 43 (gdb) (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0x348d035c 881656668 ecx0xac8d1c04 -1400038396 edx0x6d76965e 1836488286 ebx0xeed891a4 -287796828 esp0x287e40 0x287e40 ebp0x8b021ae9 0x8b021ae9 esi0xee991124 -291958492 edi0x60435d31 1615027505 eip0x555cb5 0x555cb5 _sha1_block_data_order_ssse3+1109 eflags 0x203[ CF IF ] cs 0x23 35 ss 0x2b 43 ds 0x2b 43 es 0x2b 43 fs 0x53 83 gs 0x2b 43 (gdb) (gdb) thread apply 1 info reg Thread 1 (Thread 8820.0x272c): eax0xea86e973 -360257165 ecx0x409b7f15 1083932437 edx
Re: [openssl.org #2655] speed sha1 hang up - 1.0.1 snapshot 20111211 - Cygwin
Where is it caught? Addentum: programm caught in different functions: I saw OPENSSL_Cleanse(), EVP_DigestUpdate(), sha1_something and others. I provided 'thread apply 1 info reg' only if it get caugth in sha1_block_data_order_ssse3 (). openssl was compiled with ./config -g (gdb) info threads Id Target Id Frame * 12 Thread 8820.0x1fc0 0x77b2000d in ntdll!LdrFindResource_U () from /cygdrive/c/Windows/system32/ntdll.dll 8 Thread 8820.0x2280 0x77b2f861 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 7 Thread 8820.0x19bc 0x77b2fd21 in ntdll!RtlFindSetBits () from /cygdrive/c/Windows/system32/ntdll.dll 6 Thread 8820.0x2bd0 0x77b2fd9d in ntdll!RtlFindSetBits () from /cygdrive/c/Windows/system32/ntdll.dll 5 Thread 8820.0x2b40 0x77b2f861 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 4 Thread 8820.0x1634 0x77b300ed in ntdll!RtlEnableEarlyCriticalSectionEventCreation () from /cygdrive/c/Windows/system32/ntdll.dll 2 Thread 8820.0x1f18 0x77b2f895 in ntdll!RtlUpdateClonedSRWLock () from /cygdrive/c/Windows/system32/ntdll.dll 1 Thread 8820.0x272c _sha1_block_data_order_ssse3 () at sha1-586.s:1689 (gdb) thread apply 1 info reg __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: [openssl.org #2650] major ssl read/ write performance improvement - updated
Hello Michael, I have tested youe patch. It is working stable at least with ccgost engine (and without any engine too, of cource). Thanks for contribution! Could you please describe, what was your test environmnet and test methodology? How did you measure that doubling read/write speed? What tool/profiler do you use? How it depends from SSL record size? What the overall speed improvement if we'll count OS IO? I'm asking because I'm trying to measure performance improvement your changes can give with my crypto-accelerator, and my results not even close to doube read/write speed. But my test resources are limited for the moment, and it is possible it is due to these limitations. In any case, I guess comunity will be grateful if your share your expirience. WBR, Andrey On 5 December 2011 14:33, Deng Michael via RT r...@openssl.org wrote: Hi, I have changed the mac code which gives substantial improvement for both read and write (not handshake) The saving is fairly major, on cpu with cryto acceleration, the change can more than double the overall ssl read /write speed for 1K record excluding OS IO time. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
[openssl.org #2365] Limitations of ENGINE interface hamper performance on modern hardware
3) An accellerator device directly supports TLS/SSL record encryption/decryption and the handshake operation itself. We do many bus transactions to the accellerator (and possibly system calls into the OS kernel) where we could do one, doing every single basic cryptographic operation individually when we could actually amortize the cost over the entire record or handshake operation. This is the case for most modern accellerators used with general-purpose CPUs. Application of such technique does not limited to hardware acselerator. Yet another example of such devices is services, allowing to pass the whole record plus encryption and MAC keys, and process it in single call. It is used when for some (security) reasons all cryptography-manipulations performed in separate process/driver/VM, and client operates only with handlers to keys. I saw how it was implemented in extension to MS CryptoAPI. Even without such extensions CryptEncrypt function is able to encrypt and hash data at the same time. Extension I'm taking about does add abbility to pass there pointer to header, body and place where to put tail - i.e. MAC value. Inabbility to process TLS record in single call results to necessity to pass the same data over IPC twice. Andrey. __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: major ssl read/ write performance improvement
Hello, Thanks for interesting contribution! Unfortunately when I apply the patch s_server failed with SEGFAULT, when using ccgost engine (and possibly others) here: EVP_DigestSignFinal if (sctx) r = md_ctx_ptr-pctx-pmeth-signctx(md_ctx_ptr-pctx, sigret, siglen, md_ctx_ptr); else because of pmeth-signctx == 0x08 (or something like this) When I use RSA certificate segfault didn't occur, as pmeth-signctx points to some valid place. Stacktrace is: EVP_DigestSignFinal (ctx=0x87802b0, sigret=0xbfd5f6dc \\\b\\002x\\b\\001\, siglen=0xbfd5f698) tls1_mac (ssl=0x877a088, md=0xbfd5f6dc \\\b\\002x\\b\\001\, send=0) ssl3_get_record (s=0x877a088) ssl3_read_bytes (s=0x877a088, type=22, buf=0x8788d50 \\\020\, len=4, peek=0) ssl3_get_message (s=0x877a088, st1=8608, stn=8609, mt=-1, max=514, ok=0xbfd5f8b8) ssl3_get_cert_verify (s=0x877a088) ssl3_accept (s=0x877a088) ssl3_read_bytes (s=0x877a088, type=23, buf=0x877e7e8 \\, len=4096, peek=0) ssl3_read_internal (s=0x877a088, buf=0x877e7e8, len=4096, peek=0) ssl3_read (s=0x877a088, buf=0x877e7e8, len=4096) SSL_read (s=0x877a088, buf=0x877e7e8, num=4096) ssl_read (b=0x8779370, out=0x877e7e8 \\, outl=4096) BIO_read (b=0x8779370, out=0x877e7e8, outl=4096) buffer_gets (b=0x8777e00, buf=0x877a7e0 \\, size=16382) BIO_gets (b=0x8777e00, in=0x877a7e0 \\, inl=16383) www_body (hostname=0x0, s=6, context=0x0) do_server (port=443, type=1, ret=0x8248ac8, cb=0x8072d24 www_body, context=0x0) s_server_main (argc=0, argv=0xbfd602b8) do_cmd (prog=0x8770868, argc=16, argv=0xbfd60278) main (Argc=16, Argv=0xbfd60278) Could you please advice, what going wrong with your code??? Go check it you need: 1. Adjust your openssl.cnf file, bu adding there: openssl_conf = openssl_def [openssl_def] engines = engine_section [engine_section] gost = gost_section [gost_section] engine_id = gost default_algorithms = ALL somewhhere before [ new_oids ] (if we talking about sample config file from OpenSSL distribution). 2. Generate private key: ./apps/openssl genpkey -engine gost -algorithm gost2001 -pkeyopt paramset:A -out botkey.p8 3. Create self-sign certificate ./apps/openssl req -x509 -days 1095 -subj '/C=US/CN=ccgost_srv/O=sam...@mail.com' -engine gost -new -key botkey.p8 -out botcert.cer 4. Run s_server ./apps/openssl s_server -engine gost -tls1 -www -accept 443 -state -cert botcert.cer -key botkey.p8 -cipher aGOST01 5. Run s_client ./apps/openssl s_client -tls1 -connect 192.168.10.103:443 -msg Well Here s_client will crash with segfault... But if you'll connect via browser - s_server will crash. Please let me know if you'll have any questions. Andrey. On 30 November 2011 05:56, Deng Michael mdeng...@yahoo.com wrote: Thanks Steve for the comment. I guess there are other ways to do similar things, since I was not sure about the intentions of the original code I was trying to make the change in a way such that when checkpoint is not call it should behave like before. Adding a new field for me is less likely to interfere with other code. It seems to me the three evp_md_ctxs contained within the hmac_md_ctx has the data for restoring the state but I was not sure. Also the new field serves as a flag to tell if it has checkpoint data (I could have used an existing flag). My patch also contains some hacking I would think. anyway the real saving comes from redo of state preserving of the evp_md_ctx that contains evp_pkey_ctx which in turn contains hmac_ctx which again contains three evp_md_ctx's. the dup of these are called in tls1_mac() similar place for ssl3 and EVP_DigestSignFinal() these two are the super expensive ones (real super) the copy of ctx in HMAC_Final() --- this one is not too bad can be simplified. I would think the saving is so much that is worth changing maybe in future releases. regards, Michael - Original Message - From: Dr. Stephen Henson st...@openssl.org To: openssl-dev@openssl.org Cc: Sent: Tuesday, November 29, 2011 1:21 PM Subject: Re: major ssl read/ write performance improvement On Mon, Nov 28, 2011, Deng Michael wrote: Hi, I have changed the mac code which gives substantial improvement for both read and write (not handshake) The saving is fairly major, on cpu with cryto acceleration, the change can more than double the overall ssl read /write speed for 1K record excluding OS IO time. this implies the change removed majority of the code overhead for read and write. The basic idea is to remove all the EVP_MD_CTX duplications (which is very cpu intensive) during read and write. the original code involves numerous memory allocations and frees for each read or write all due to the ctx's deep copy. the new way of keeping the ctx is to make it do state checkpoint and restore instead of deep copy, after this change there is NO memory operation for read and write. The changes
Re: [openssl.org #2551] [PATCH] All platforms: Option to disable sending renegotiation_info extension.
First of all - it is not a Microsoft bug. This is bug in some products, who act as a TLS clients. Second: what should I describe to customers? :-) Their programms works with other server, by not with mine. Is the question still valid? :-) On 1 July 2011 12:20, Ben Laurie via RT r...@openssl.org wrote: Why are you modifying OpenSSL for this? This is a Microsoft bug - have you reported it to Microsoft? -- http://www.apache-ssl.org/ben.html http://www.links.org/ There is no limit to what a man can do or how far he can go if he doesn't mind who gets the credit. - Robert Woodruff First of all - it is not a Microsoft bug.This is bug in some products, who act as a TLS clients.Second: what should I describe to customers? :-)Their programms works with other server, by not with mine. Is the question still valid?:-)On 1 July 2011 12:20, Ben Laurie via RT r...@openssl.org wrote: Why are you modifying OpenSSL for this? This is a Microsoft bug - have you reported it to Microsoft? -- http://www.apache-ssl.org/ben.html ? ? ? ? ? http://www.links.org/ There is no limit to what a man can do or how far he can go if he doesnt mind who gets the credit. - Robert Woodruff
What the resons to use SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION may be?
Hello, There is an option available: SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION Descroption laconicaly states: When performing renegotiation as a server, always start a new session (i.e., session resumption requests are only accepted in the initial handshake). This option is not needed for clients. But I can't find any information WHY and WHEN I should (or should not) use this option. Could anyone explain, what the purpose of this option? I.e. what the drawback of allowing session resumption during renegotiation? Is it for compatibility with buggy clients? Or is it some security countermeasure? Would be really appreciated for any comments. Best wishes, Andrey
Re: What the resons to use SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION may be?
Sorry, Some explanations available in sources ssl/t1_lib.c: == - Applications must use SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION if they provide for changing an explicit servername context for the session, i.e. when the session has been established with a servername extension. == So this option must be used if server use server_name extension. But what will happens if server do use server_name extension, but not set SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION? What the security consequences will be? I understand there is some, but can't figure out what exactly... The only thing I was able to find is this: http://mail-archives.apache.org/mod_mbox/httpd-bugs/200409.mbox/%3c20041001143012.6555.qm...@nagoya.betaversion.org%3E If two virtual servers configured to use different ciphersuites, there is a possiblility for malicious client to establish session with one server with it's ciphersuite, and then renegotiate connection with other server. And ciphersuie remains unchanged. I can't say for sure is it security or compatibility issue. Would be really appreciated for any comments. Best wishes, Andrey On 18 June 2011 15:58, Andrey Kulikov amde...@gmail.com wrote: Hello, There is an option available: SSL_OP_NO_SESSION_RESUMPTION_ON_RENEGOTIATION Descroption laconicaly states: When performing renegotiation as a server, always start a new session (i.e., session resumption requests are only accepted in the initial handshake). This option is not needed for clients. But I can't find any information WHY and WHEN I should (or should not) use this option. Could anyone explain, what the purpose of this option? I.e. what the drawback of allowing session resumption during renegotiation? Is it for compatibility with buggy clients? Or is it some security countermeasure? Would be really appreciated for any comments. Best wishes, Andrey
Re: s_client -reconnect with DTLS
Hello, I'm sure you'll get help faster, if you describe: 1. What are you doing exactly. 2. What do you see. 3. What do you expect to see. This is absolutelly necessary steps, as all telepathist is on vacation now. On 22 April 2011 15:50, N. J. nadh...@yahoo.com wrote: Hi again, I am not sure if someone can help confirming that the -reconnect option is broken with the dtls implementation? Please refer to my email below. Looking forward for your support. Regards, Nadhem -- ** Hi there, I have been trying to get the s_client -reconnect option working with my s_server but had no luck when using DTLS, -dtls1. I could not find any information why it is not working so I wonder if this is broken in openssl 1.0.0a. If so, is there any fix? Thanks in advance, Nadhem
Re: [openssl.org #2496] [PATCH] Fix compile problems when various ciphers are disabled
Tim, Seems that you miss to attach a file with actual patch... On 12 April 2011 11:13, Tim Jackson via RT r...@openssl.org wrote: When turning off various ciphers, OpenSSL doesn't always compile. The attached patches fix various issues with OpenSSL when ciphers (particularly RSA or DSA) are turned off. These patches were requested in the attached thread
[openssl.org #2463] [PATCH]: OpenSSL 1.0.0d: Add abbility to load server certificate by ENGINE.
I prefer more generic method similar to ENGINE_load_ssl_client_cert, i.e. I need EVP keys, corresponding certificates and the certificate chain. Additional methods has server in it's names for the same reason why ENGINE_load_ssl_client_cert has client in it. ENGINE_load_server_certificate() intended to work in pair with ENGINE_load_private_key() to allow engine to load certificate correspondent to it's private key. And present obtained certificate to client during SSL handshake. That it. You can always get required EVP_PKEY keys from loaded X509 structure. But I can't get why you may need certificate chain for your certificate if you are server. Andrey I prefer more generic method similar to ENGINE_load_ssl_client_cert, i.e. I need EVP keys, corresponding certificates and the certificate chain. Additional methods has server in its names for the same reason why ENGINE_load_ssl_client_cert has client in it. ENGINE_load_server_certificate() intended to work in pair with ENGINE_load_private_key() to allow engine to load certificate correspondent to its private key.And present obtained certificate to client during SSL handshake. That it. You can always get required EVP_PKEY keys from loaded X509 structure.But I cant get why you may need certificate chain for your certificate if you are server. Andrey
[openssl.org #2463] [PATCH]: OpenSSL 1.0.0d: Add abbility to load server certificate by ENGINE.
Hello, Please find file attached: server_cert_from_engine4.patch This is a patch to allow loading server SSL certificate by ENGINE. Currently OpenSSL allows loading certificate only from a file. Loading by specific engine is required for hardware-based engines, which used their own certificate storages, as well as any Microsoft CAPI-based CSP. Although there is no such engines implemented for OpenSSL yet, sooner it will. Affected files: apps/apps.c apps/s_server.c crypto/engine/engine.h crypto/engine/eng_int.h crypto/engine/eng_pkey.c Patch created using this command: diff -rupN openssl-1.0.0d/ openssl-1.0.0d-engine-srv-cert/ server_cert_from_engine4.patch To apply pach use follwing command in current OpenSSL root dev. directory: patch -p1 -l -u -b -i server_cert_from_engine4.patch After applying this patch s_server will accept -certform ENGINE option. This patch supplied by Stonesoft Corporation, who give me permission to supply it to OpenSSL. Feel free to contact with me with any related questions. Andrey. Hello,Please find file attached: server_cert_from_engine4.patch This is a patch to allow loading server SSL certificate by ENGINE.Currently OpenSSL allows loading certificate only from a file.Loading by specific engine is required for hardware-based engines, which used their own certificate storages, as well as any Microsoft CAPI-based CSP. Although there is no such engines implemented for OpenSSL yet, sooner it will.Affected files:apps/apps.capps/s_server.ccrypto/engine/engine.hcrypto/engine/eng_int.hcrypto/engine/eng_pkey.c Patch created using this command:diff -rupN openssl-1.0.0d/ openssl-1.0.0d-engine-srv-cert/ server_cert_from_engine4.patchTo apply pach use follwing command in current OpenSSL root dev. directory: patch -p1 -l -u -b -i server_cert_from_engine4.patchAfter applying this patch s_server will accept -certform ENGINE option.This patch supplied by Stonesoft Corporation, who give me permission to supply it to OpenSSL. Feel free to contact with me with any related questions.Andrey. server_cert_from_engine4.patch Description: Binary data
HowTo: Engine-specific TLS master secret generation?
Hello, Now OpenSSL generates master secret and read/write keys inside the library, left only premaster secret decryption to the engine. In case of hardware-based TLS engine it could be not an option, as there may be no possibility to set read/write keys from outside (or it may be restricted according to some rules of such hardware usage). If someone would needed to implement such device support in OpenSSL: 1. How would you estimate required effort? 2. What is the best base OpenSSL version to start with? 3. What is necessary to take into account, but may be not visible from the beginning? Would be really appreciated for any answer. Andrey.
Re: adding new encryption algorithm to OpenSSL
This explanation may help you to add new algorithms to OpenSSL: http://www.mail-archive.com/openssl-users@openssl.org/msg59630.html On 18 January 2011 08:59, M. V. bored_to_deat...@yahoo.com wrote: hi, for a project, i have to add my own encryption algorithm to racoon (part of ipsec-tools) in freebsd. i realized racoon uses openssl libraries for encryption algorithms. so, i wanted to know , does anyone have any experience in doing such thing? does anyone have any suggestions for me? 1- what's the best and easiest way? adding my algorithm to racoon, openssl or else (as a dynamic library, etc)? 2- is there any example, code template or documentation which could help me doing this? thank you.
What the purpose of EVP_PKEY_derive() function?
Hello, I'm exploring how to implement custom engine, and can't undestand the purpose of EVP_PKEY_derive() function. It is possible to set pointer to it's implementation using EVP_PKEY_meth_set_derive() call. But it used only in *pkeyutl* command. It is not used in SSL handshake. The only engine which set it is ccgost, but as I understand in not used at all. According to pkeyutl command documentation it is used for key derivation (agreement) operations. Could please anyone explain, what it is designed for, and according to which standard it should be implemented in new engine? Is it required at all? Andrey.
Re: What is the REALLY proper way to use an ENGINE?
On 4 January 2011 16:54, Dr. Stephen Henson st...@openssl.org wrote: I can't see how that would happen: I find a way to reproduce it: just call ENGINE_load_builtin_engines() twice. Or call ENGINE_load_builtin_engines() after OPENSSL_config() which calls it too. It happens because internal ccgost AMETH structures are static (which, obviously, does matter). First call of ENGINE_load_builtin_engines() creates new ENGINE structure and store it in some list, etc. Second call to ENGINE_load_builtin_engines() creates new engine structure AGAIN, but AMETH part of it reffers to the same static variable (ameth_GostR3410_2001). But ENGINE_add() in second call fails, because engine with the same ID already in the list. So, reference count not increased, and just created engine released in the very next line. AMETH structures being destoryed, and it makes them invalid for originally created engine too (because there is only one such structure). Ways to avoid it: 1. Make ccgost engine allocates AMETH structures dynamically. As there is no other engines, used this functionality (calls to ENGINE_set_pkey_asn1_meths() ), I can't compare. But from my point of view this is a bad idea, because it introduces unnecessary complexity. And, there is no clear reason to create them more that once. Of cource we can imagine some engine which provides different NID_id's depend on configuration. But ccgost is not the one. 2. Make ENGINE_load_builtin_engines() single-callable. Just like OPENSSL_config(). I can't see any reason why it have to be called more that once. Would be really appreciated if guru comment on that. Andrey.
What is the REALLY proper way to use an ENGINE?
If we take a look at any ENGINE_load_XXX function, we find that they all has similar structure: ENGINE *toadd = engine_XXX(); if(!toadd) return; ENGINE_add(toadd); ENGINE_free(toadd); ERR_clear_error(); My question is: why we need call ENGINE_free(toadd) ?? Somewhere inside it calls EVP_PKEY_asn1_free(), which besides everything else desroy all AMETH structures, created during engine initialization via EVP_PKEY_asn1_set_* . So, let's consider following example: CRYPTO_malloc_init(); ERR_load_crypto_strings(); ENGINE_load_builtin_engines(); e = ENGINE_by_id(XXX); ENGINE_set_default(e, ENGINE_METHOD_ALL); OpenSSL_add_all_algorithms(); OpenSSL_add_ssl_algorithms(); EVP_PKEY *signing_key = ENGINE_load_private_key(e, NULL, NULL, NULL); // bla-bla-bla... And, if in engine load priv func we write something like EVP_PKEY *pkey = EVP_PKEY_new(); pkey-ameth = ENGINE_get_pkey_asn1_meth(eng, NID_id_SOME_NID); it's not gonna work, because all ASN1 structures, carefully created by engine via calls to EVP_PKEY_asn1_set_* are invalid and possibly corrupted at this moment. Is it a design issue, or provided example is invalid (but it taken from openssl sources, hm...)?
Re: What is the REALLY proper way to use an ENGINE?
Update: adding ENGINE_init(e) after e = ENGINE_by_id(XXX); doesn't make any difference, as in my case functional reference count is 8(???) at the moment of ENGINE_init(e) call, so engine is not re-initialised. :( On 4 January 2011 04:12, Andrey Kulikov amde...@gmail.com wrote: If we take a look at any ENGINE_load_XXX function, we find that they all has similar structure: ENGINE *toadd = engine_XXX(); if(!toadd) return; ENGINE_add(toadd); ENGINE_free(toadd); ERR_clear_error(); My question is: why we need call ENGINE_free(toadd) ?? Somewhere inside it calls EVP_PKEY_asn1_free(), which besides everything else desroy all AMETH structures, created during engine initialization via EVP_PKEY_asn1_set_* . So, let's consider following example: CRYPTO_malloc_init(); ERR_load_crypto_strings(); ENGINE_load_builtin_engines(); e = ENGINE_by_id(XXX); ENGINE_set_default(e, ENGINE_METHOD_ALL); OpenSSL_add_all_algorithms(); OpenSSL_add_ssl_algorithms(); EVP_PKEY *signing_key = ENGINE_load_private_key(e, NULL, NULL, NULL); // bla-bla-bla... And, if in engine load priv func we write something like EVP_PKEY *pkey = EVP_PKEY_new(); pkey-ameth = ENGINE_get_pkey_asn1_meth(eng, NID_id_SOME_NID); it's not gonna work, because all ASN1 structures, carefully created by engine via calls to EVP_PKEY_asn1_set_* are invalid and possibly corrupted at this moment. Is it a design issue, or provided example is invalid (but it taken from openssl sources, hm...)?
Re: What is the REALLY proper way to use an ENGINE?
Thanks for a explanations. Let's consider following main, using ccgost engine: main(){ OPENSSL_config(NULL); ENGINE *e = ENGINE_by_id(gost); ENGINE_init(e); ENGINE_free(e); ENGINE_set_default(e, ENGINE_METHOD_ALL); OpenSSL_add_all_algorithms(); // emulating ENGINE_load_private_key() EVP_PKEY *pkey = EVP_PKEY_new(); pkey-ameth = ENGINE_get_pkey_asn1_meth(e, NID_id_GostR3410_2001); bla-bla-bla... //end emulating EVP_PKEY_CTX *ctx = EVP_PKEY_CTX_new(pkey, e); } Result ctx is always NULL. It happens because ENGINE_get_pkey_asn1_meth() returns pointer to corrupted internall ccgost structure ameth_GostR3410_2001. It IS initialized, then being freed. And contains garbage. Specifically pkey-ameth-pkey_id contains some random value instead of 811 (NID_id_GostR3410_2001). Is this code contain some error or invalid engine API usage? On 4 January 2011 06:23, Dr. Stephen Henson st...@openssl.org wrote: On Tue, Jan 04, 2011, Andrey Kulikov wrote: If we take a look at any ENGINE_load_XXX function, we find that they all has similar structure: ENGINE *toadd = engine_XXX(); if(!toadd) return; ENGINE_add(toadd); ENGINE_free(toadd); ERR_clear_error(); My question is: why we need call ENGINE_free(toadd) ?? To avoid a memory leak. When you call engine_XXX() you get a reference to the ENGINE. When it is added to the ENGINE list the reference count is incremented. When you call ENGINE_free() the count is decremented and the ENGINE is freed only if the reference count is zero. Steve.
Re: s_client - debug arg make the difference
I had the same behaviour (different results in debug/not debug). In my case answer was obvious: in debug environment I had different value for OPENSSL_CONF env. variable, and that why in debug nothing works. On 22 December 2010 16:58, Benjamin benja...@gentilkiwi.net wrote: Hello, Maybe a mistake, but on OpenSSL 1.0.0c (2 Dec 2010) the debug arg make a real difference in s_client usage, a difference that I can't explain ?? Maybe a timing problem ?? Edited for text visualisation :
Re: [openssl.org #2398] [PATCH] gost code cleanup
On 16 December 2010 16:31, Mark Andrews via RT r...@openssl.org wrote: encountered a NULL pointer dereference in the gost code due to a undetected EVP_PKEY_assign() failure in pkey_gost01_paramgen(). Thnaks for report! Could you please share a secret knowelege: why EVP_PKEY_assign() fails in pkey_gost01_paramgen() in your case? I eyeball their code, but can't find out why it can fails. (Except passing incorrect pointers.) Is it reproducible from command-line, or it happens in your specific code? Andrey __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Is it really possible to generate gost-mac using ccgost engine?
Hello, I've got a problem with calculating gost-mac using Openssl 1.0.0a May be problem with cmd options, but I was unable to find out how to get it work Trying to generate gost-mac. Example from documentation (engines/ccgost/README.gost) Calculation of GOST 28147 MAC openssl dgst -mac gost-mac -macopt key:32 bytes of key datafile Note absense of an option that specifies digest algorithm. gost-mac algorithm supports only one digest (which is actually part of implementation of this mac) and OpenSSL is clever enough to find out this. # ./apps/openssl dgst -mac gost-mac -macopt key:FF openssl.doxy Algorithm gost-mac not found Well, do it like this: # ./apps/openssl dgst -gost-mac -macopt key:ff openssl.doxy Read Error in openssl.doxy 3076327052:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: The reason is that key for this mac is not set. The only place where it can be set is control function gost_imit_ctrl engines/ccgost/gost_crypt.c:595 But this function int gost_imit_ctrl(EVP_MD_CTX, int, int, void *) never called. So length of key option is not a root cause. As I understand, control functions for dgst called only if parameter -mac is specified. See apps/dgst.c:228 else if (!strcmp(*argv,-mac)) and apps/dgst.c:362 if (mac_name) But since there is no such parameter it not supposed to work. But if we do specify it - openssl can't find an algorithms (see first command line example). So the question is: is it a bug or it is possible to specify some valid parameter for -mac option in this case? Or may be there is other way to get it work? __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: Source level debug config for 1.0.0a?
As a quick hack I added to root Makefile following lines: == debug: CFLAG+= -ggdb3 -O0 -DDEBUG debug: all == just before all: Makefile build_all openssl.pc libssl.pc libcrypto.pc Then # make clean # make debug After that NetBeans can walk through source using gdb. Linux i368. It consumes about 3G of disk space. And it's a good idea to exclude test from DIRS= crypto ssl engines apps test tools in Makefile if you do not need them, as it's require enormous time to build. Scott Cherf Fri, 15 Oct 2010 15:22:45 -0700 Hello - I'm trying to compile with source level debugging enabled using the 1.0.0a distribution and I'm not having any luck at all. I would appreciate a clue if there is one. ... I'm unable to build a version that gives me source level debug using gdb and I can't figure out why. Anyone? __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: Is it really possible to generate gost-mac using ccgost engine?
On 12 November 2010 19:26, Valery Blazhnov vblazh...@lissi.ru wrote: This command works with ccgost engine: openssl dgst -mac gost-mac -binary -macopt hexkey:3132333435363738393031323334353637383930313233343536373839303132 -out mac.bin data.bin What openssl version do you use? I use 1.0.0a, and this command output is just Algorithm gost-mac not found. But this command works: openssl dgst -engine gost -mac gost-mac -macopt key: openssl.doxy __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: 1.0.0a s_server with GOST certificate can't compute sha1 digest???
On 12 November 2010 15:20, Dr. Stephen Henson st...@openssl.org wrote: On Fri, Nov 12, 2010, Andrey Kulikov wrote: Hello, I'm trying to make s_server and s_client work with GOST encryption using ccgost engine and certificates with GOST algos. But it unable to work, complaining to bad mac computing. (If I use RSA-based certificates, everything works just fine.) Please try a recent 1.0.0 snapshot and see if you still have this problem. Steve. I check with ftp://ftp.openssl.org/snapshot/openssl-1.0.1-stable-SNAP-20101112.tar.gz Visible behaviour is the same. Commands output is the same. Test procedure the same as described in original message: Modify openssl.cnf export OPENSSL_CONF=path to /openssl.cnf ./apps/openssl genpkey -engine gost -algorithm gost2001 -pkeyopt paramset:A -out botkey.p8 ./apps/openssl req -config ./apps/openssl.cnf -x509 -days 1095 -subj '/C=RU/CN=csp_srv/o=lapu...@mail.ru' -engine gost -new -key botkey.p8 -out botcert.pem ./apps/openssl s_server -www -engine gost -accept 4333 -state -cert botcert.pem -key botkey.p8 ./apps/openssl s_client -engine gost -ssl3 -connect localhost:4333 s_client output (truncated): == ~/comt_area/openssl-1.0.1-stable-SNAP-20101112# ./apps/openssl s_client -engine gost -ssl3 -connect localhost:4333 engine gost set. CONNECTED(0003) depth=0 C = RU, CN = csp_srv, O = lapu...@mail.ru verify error:num=18:self signed certificate verify return:1 depth=0 C = RU, CN = csp_srv, O = lapu...@mail.ru verify return:1 3076486796:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: 3076486796:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: 3076486796:error:140943FC:SSL routines:SSL3_READ_BYTES:sslv3 alert bad record mac:s3_pkt.c:1212:SSL alert number 20 3076486796:error:1409E0E5:SSL routines:SSL3_WRITE_BYTES:ssl handshake failure:s3_pkt.c:591: --- Certificate chain 0 s:/C=RU/CN=csp_srv/o=lapu...@mail.ru i:/C=RU/CN=csp_srv/o=lapu...@mail.ru --- == s_server output (truncated) == :~/comt_area/openssl-1.0.1-stable-SNAP-20101112# ./apps/openssl s_server -www -engine gost -accept 4333 -state -cert botcert.pem -key botkey.p8 engine gost set. Using default temp DH parameters Using default temp ECDH parameters ACCEPT SSL_accept:before/accept initialization SSL_accept:SSLv3 read client hello A SSL_accept:SSLv3 write server hello A SSL_accept:SSLv3 write certificate A SSL_accept:SSLv3 write server done A SSL_accept:SSLv3 flush data SSL_accept:SSLv3 read client key exchange A SSL3 alert write:fatal:bad record mac SSL_accept:error in SSLv3 read certificate verify A SSL_accept:error in SSLv3 read certificate verify A 3076945548:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: 3076945548:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:8808C074:lib(136):GOST_IMIT_FINAL:mac key not set:gost_crypt.c:564: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076945548:error:8808C074:lib(136):GOST_IMIT_FINAL:mac key not set:gost_crypt.c:564: 3076945548:error:1408F119:SSL routines:SSL3_GET_RECORD:decryption failed or bad record mac:s3_pkt.c:479: __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: 1.0.0a s_server with GOST certificate can't compute sha1 digest???
Sorry, previous email is about 1.0.1 latest snapshot. I just checked with 1.0.1 ftp://ftp.openssl.org/snapshot/openssl-1.0.0-stable-SNAP-20101112.tar.gz Results exactly the same. If you'll need any details - please let me know. Andrey On 13 November 2010 00:43, Andrey Kulikov amde...@gmail.com wrote: On 12 November 2010 15:20, Dr. Stephen Henson st...@openssl.org wrote: On Fri, Nov 12, 2010, Andrey Kulikov wrote: Hello, I'm trying to make s_server and s_client work with GOST encryption using ccgost engine and certificates with GOST algos. But it unable to work, complaining to bad mac computing. (If I use RSA-based certificates, everything works just fine.) Please try a recent 1.0.0 snapshot and see if you still have this problem. Steve. I check with ftp://ftp.openssl.org/snapshot/openssl-1.0.1-stable-SNAP-20101112.tar.gz Visible behaviour is the same. Commands output is the same. Test procedure the same as described in original message: __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: 1.0.0a s_server with GOST certificate can't compute sha1 digest???
Hello, On 13 November 2010 03:33, Dr. Stephen Henson st...@openssl.org wrote: I've just tried 1.0.1 and it does have a problem with GOST and TLS v1.1 which is the default for OpenSSL 1.0.1. If you include -no_tls1_1 in the command line it should work or if you try a recent 1.0.0 snapshot (OpenSSL 1.0.0 doesn't support TLS v1.1). I'll look into the TLS v1.1 issue. Steve. Sorry to confuse you - I've mixed up versions and snapshot filenames. I just re-check these snapshots of 1.0.0 ftp://ftp.openssl.org/snapshot/openssl-1.0.0-stable-SNAP-20101112.tar.gz and 1.0.1 ftp://ftp.openssl.org/snapshot/openssl-1.0.1-stable-SNAP-20101112.tar.gz I disable TSL in there: ./config no-tls I add -ssl3 parameter to s_server and s_client commands. ./apps/openssl s_server -ssl3 -www -engine gost -accept 4333 -state -cert botcert.pem -key botkey.p8 ./apps/openssl s_client -ssl3 -engine gost -connect localhost:4333 And both of these versions are not working in the same way. s_server = SSL3 alert write:fatal:bad record mac SSL_accept:error in SSLv3 read certificate verify A 3076736652:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: = s_client = verify return:1 3076413068:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:668: = So, the problem not in TLS1.1 BTW: 1.0.1 s_server doesn't accept -no_tls1_1 option, while have it in help option list. s_client does accept this option. WBR, Andrey __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
1.0.0a s_server with GOST certificate can't compute sha1 digest???
Hello, I'm trying to make s_server and s_client work with GOST encryption using ccgost engine and certificates with GOST algos. But it unable to work, complaining to bad mac computing. (If I use RSA-based certificates, everything works just fine.) Openssl 1.0.0a, Linux i386 I have ccgost statically linked in openssl. I've add following lines to openssl config file: = openssl_conf = openssl_def [openssl_def] engines = engine_section [engine_section] gost = gost_section [gost_section] engine_id = gost default_algorithms = ALL CRYPT_PARAMS = id-Gost28147-89-CryptoPro-A-ParamSet = Then execute export OPENSSL_CONF=path to /openssl.cnf Generate key and self-signed certificate using GOST algos. #Generate private key ./apps/openssl genpkey -engine gost -algorithm gost2001 -pkeyopt paramset:A -out botkey.p8 # Create self-sign certificate ./apps/openssl req -config ./apps/openssl.cnf -x509 -days 1095 -subj '/C=RU/CN=csp_srv/o=lapu...@mail.ru' -engine gost -new -key botkey.p8 -out botcert.pem # Dump ASN format files to check that it what we expect openssl asn1parse -in botkey.p8 openssl asn1parse -in botcert.pem When start s_server ./apps/openssl s_server -www -engine gost -accept 4333 -state -cert botcert.pem -key botkey.p8 ./apps/openssl s_client -engine gost -ssl3 -connect localhost:4333 Client output (truncated): . 3076867724:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:669: 3076867724:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:669: 3076867724:error:140943FC:SSL routines:SSL3_READ_BYTES:sslv3 alert bad record mac:s3_pkt.c:1193:SSL alert number 20 3076867724:error:1409E0E5:SSL routines:SSL3_WRITE_BYTES:ssl handshake failure:s3_pkt.c:590: --- .. --- New, TLSv1/SSLv3, Cipher is GOST2001-GOST89-GOST89 Server public key is 256 bit Secure Renegotiation IS supported Compression: NONE Expansion: NONE SSL-Session: Protocol : SSLv3 Cipher: GOST2001-GOST89-GOST89 Session-ID: 214A3B31604763646B6E6A1DC0B793BC5979C454C25AE2ED2EE3947AA08078B0 Session-ID-ctx: Master-Key: 93B1DB1A2B5A40CFB2E3328280FB200B044CF446E73353B735328253D690F88A01AEB9E004BC06085E4C81770269EBD7 Key-Arg : None PSK identity: None PSK identity hint: None Start Time: 1289516998 Timeout : 7200 (sec) Verify return code: 18 (self signed certificate) Server output (truncated): ... SSL_accept:SSLv3 write server done A SSL_accept:SSLv3 flush data SSL_accept:SSLv3 read client key exchange A SSL3 alert write:fatal:bad record mac SSL_accept:error in SSLv3 read certificate verify A SSL_accept:error in SSLv3 read certificate verify A 3076998796:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:669: 3076998796:error:1411D144:SSL routines:ssl3_handshake_mac:digest requred for handshake isn't computed:s3_enc.c:669: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:8808C074:lib(136):GOST_IMIT_FINAL:mac key not set:gost_crypt.c:564: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: 3076998796:error:8808C074:lib(136):GOST_IMIT_FINAL:mac key not set:gost_crypt.c:564: 3076998796:error:1408F119:SSL routines:SSL3_GET_RECORD:decryption failed or bad record mac:s3_pkt.c:478: ACCEPT The problem is in ssl3_final_finish_mac function (ssl/s3_enc.c:636), called somewhere from ssl3_do_change_cipher_spec(). It try to compute NID_md5 and NID_sha1 digests, but in SSL-s3-handshake_dgst[] there is only one digest - NID_id_GostR3411_94. So we got that errors. Could please someone advice, what the root cause of this situation, and how to deal with it? How to get s_server and s_client working with GOST certificates? Best wishes, Andrey __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: Source level debug config for 1.0.0a?
Thanks for pointing to this feature! In context of original question ./Configure debug-rse looks better than others, as it add -ggdb3 switch, while just -g (like in all others debugging targets) was not work well (for me). On 10 November 2010 16:33, Martin Kaiser li...@kaiser.cx wrote: Hello, there's a number of debugging targets in Configure. For OpenSSL-1.0.0a on Linux i386, you can simply do ./Configure debug-linux-elf-noefence make clean make and you have debug infos enabled Martin __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
How to generate gost-mac using ccgost engine?
Hello, I've got a problem with calculating gost-mac using Openssl 1.0.0a May be problem with cmd options, but I was unable to find out how to get it work Trying to generate gost-mac. Example from documentation (engines/ccgost/README.gost) Calculation of GOST 28147 MAC openssl dgst -mac gost-mac -macopt key:32 bytes of key datafile Note absense of an option that specifies digest algorithm. gost-mac algorithm supports only one digest (which is actually part of implementation of this mac) and OpenSSL is clever enough to find out this. # ./apps/openssl dgst -mac gost-mac -macopt key:FF openssl.doxy Algorithm gost-mac not found Well, do it like this: # ./apps/openssl dgst -gost-mac -macopt key:ff openssl.doxy Read Error in openssl.doxy 3076327052:error:88073074:lib(136):GOST_IMIT_UPDATE:mac key not set:gost_crypt.c:527: The reason is that key for this mac is not set. The only place where it can be set is control function gost_imit_ctrl engines/ccgost/gost_crypt.c:595 But this function int gost_imit_ctrl(EVP_MD_CTX, int, int, void *) never called. So length of key option is not a root cause. As I understand, control functions for dgst called only if parameter -mac is specified. See apps/dgst.c:228 else if (!strcmp(*argv,-mac)) and apps/dgst.c:362 if (mac_name) But since there is no such parameter it not supposed to work. But if we do specify it - openssl can't find an algorithms (see first command line example). So the question is: is it a bug or it is possible to specify some valid parameter for -mac option in this case? Or may be there is other way to get it work? __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org
Re: Source level debug config for 1.0.0a?
As a quick hack I added to root Makefile following lines: == debug: CFLAG+= -ggdb3 -O0 -DDEBUG debug: all == just before all: Makefile build_all openssl.pc libssl.pc libcrypto.pc Then # make clean # make debug After that NetBeans can walk through source using gdb. Linux i368. It consumes about 3G of disk space. And it's a good idea to exclude test from DIRS= crypto ssl engines apps test tools in Makefile if you do not need them, as it's require enormous time to build. Scott Cherf Fri, 15 Oct 2010 15:22:45 -0700 Hello - I'm trying to compile with source level debugging enabled using the 1.0.0a distribution and I'm not having any luck at all. I would appreciate a clue if there is one. ... I'm unable to build a version that gives me source level debug using gdb and I can't figure out why. Anyone? __ OpenSSL Project http://www.openssl.org Development Mailing List openssl-dev@openssl.org Automated List Manager majord...@openssl.org