Bug #45498 for Crypt-OpenSSL-RSA: OpenSSL error: padding check failed at t/rsa.t line 149

Mon Apr 27 18:49:07 2009 MARKLE [...] cpan.org - Ticket created

Subject:

OpenSSL error: padding check failed at t/rsa.t line 149

I expanded the test a little to warn the actual $@ if "ok(!$@)" fails. PERL_DL_NONLAZY=1 /usr/bin/perl "-MExtUtils::Command::MM" "-e" "test_harness(0, 'blib/lib', 'blib/arch')" t/*.t t/bignum.t .. skipped: (no reason given) t/format.t .. ok t/rsa.t ..... 1/63 # Failed test 63 in t/rsa.t at line 151 # t/rsa.t line 151 is: ok(!$err); # error: RSA.xs:301: OpenSSL error: padding check failed at t/rsa.t line 149. t/rsa.t ..... Failed 1/63 subtests

Mon Jun 01 13:56:07 2009 jason [...] long.name - Correspondence added

I also get this failure, on 64-bit RHEL 5.3. The failure does NOT occur on 32-bit Fedora 9. This is the part of the t/rsa.t test script that fails... Show quoted text

># check subclassing > >eval { Crypt::OpenSSL::RSA::Subpackage->generate_key(256); }; >ok(!$@);

I find that if I change the key length in this line to 512, the test script passes. Not sure why though. Jason Long

Mon Jun 01 13:56:08 2009 The RT System itself - Status changed from 'new' to 'open'

Mon Jun 01 14:03:09 2009 MARKLE [...] cpan.org - Correspondence added

The error comes from openssl. Maybe RHell/CentOS modify openssl libs to require 512 bit keys?

Tue Jun 02 04:36:35 2009 justincase [...] yopmail.com - Correspondence added

On Mon Jun 01 14:03:09 2009, MARKLE wrote: Show quoted text

> The error comes from openssl. Maybe RHell/CentOS modify openssl libs

to Show quoted text

> require 512 bit keys?

I don't think so. I run two different RHell/CentOS distros, one 32bit and the other 64bit. The 32bit machine passes this test, the 64bit machine fails this test. The 32bit machine is: [root@wilber ~]# cat /etc/release ClarkConnect Community Edition release 4.2 [root@wilber ~]# cat /etc/redhat-release CentOS release 4.2 (Final) [root@wilber ~]# By memory ClarkConnect also use some RHEL4 rpms as well. The 64bit machine is: [root@server ~]# cat /etc/redhat-release CentOS release 5.3 (Final) [root@server ~]# The 32bit machine openssl libs: [root@wilber ~]# rpm -qa | grep -i ssl | sort cc-ssl-4.2-10 docbook-style-dsssl-1.78-4 mod_ssl-2.0.52-22.cc openssl096b-0.9.6b-22.46 openssl-0.9.7a-43.17.el4_7.2 openssl-devel-0.9.7a-43.17.el4_7.2 openssl-perl-0.9.7a-43.17.el4_7.2 perl-Crypt-SSLeay-0.51-5 perl-IO-Socket-SSL-0.97-1.cc perl-Net-SSLeay-1.26-2.cc [root@wilber ~]# The 64bit machine openssl libs: [root@server ~]# rpm -qa | grep -i ssl | sort mod_ssl-2.2.3-22.el5.centos openssl097a-0.9.7a-9.el5_2.1 openssl-0.9.8e-7.el5 openssl-0.9.8e-7.el5 openssl-devel-0.9.8e-7.el5 openssl-devel-0.9.8e-7.el5 openssl-perl-0.9.8e-7.el5 perl-Crypt-SSLeay-0.51-11.el5 perl-IO-Socket-SSL-1.01-1.fc6 perl-Net-SSLeay-1.30-4.fc6 pyOpenSSL-0.6-1.p24.7.2.2 xmlsec1-openssl-1.2.9-8.1 xmlsec1-openssl-1.2.9-8.1 xmlsec1-openssl-devel-1.2.9-8.1 xmlsec1-openssl-devel-1.2.9-8.1 [root@server ~]#

Tue Nov 17 06:35:36 2009 MSTEVENS [...] cpan.org - Correspondence added

I'm seeing a similar problem on my redhat 5.3 system: not ok 63 # Failed test 63 in t/rsa.t at line 150 # t/rsa.t line 150 is: ok(!$@);

Tue Nov 17 23:12:52 2009 iroberts+perl [...] red-bean.com - Correspondence added

I'm not seeing this behavior on my suse 11.1 64bit box with openssl-0.9.8h (also 64 bit).. Moreover, I can bump the keysize on the test in question as low as 31 bits; after that, the error is "OpenSSL error: bad signature", not the "padding check failed" error. One person mentioned that bumping the key size from 256 to 512 made the problem go away. Are others seeing the same behavior? Those who are seeing this; does the problem go away if you change line 149 from generate_key(256) to generate_key(512)? Also, what happens if you run: openssl genrsa 256 from the command line?

Tue Nov 17 23:22:53 2009 MARKLE [...] cpan.org - Correspondence added

On Tue Nov 17 23:12:52 2009, IROBERTS wrote: Show quoted text

> Also, what happens if you run: > > openssl genrsa 256 > > from the command line?

Yes, in CentOS with perl 5.8.8 changing the test to 512 makes the test succeed. hedges@anubis:~$ openssl genrsa 256 Generating RSA private key, 256 bit long modulus ...+++++++++++++++++++++++++++ ..................+++++++++++++++++++++++++++ 1584:error:0408C070:rsa routines:FIPS_RSA_SIGN:digest too big for rsa key:fips_rsa_sign.c:280: 1584:error:2D06D075:FIPS routines:fips_pkey_signature_test:test failure:fips.c:672: 1584:error:2D06A06B:FIPS routines:FIPS_CHECK_RSA:pairwise test failed:fips_rsa_gen.c:132: 1584:error:04081003:rsa routines:RSA_BUILTIN_KEYGEN:BN lib:fips_rsa_gen.c:298: hedges@anubis:~$ openssl genrsa 512 Generating RSA private key, 512 bit long modulus ...++++++++++++ .....++++++++++++ unable to write 'random state' e is 65537 (0x10001) Show quoted text

Tue Nov 17 23:39:34 2009 iroberts+perl [...] red-bean.com - Correspondence added

Another thing to try: after running make test (and having it fail), try running just this: PERL_DL_NONLAZY=1 /usr/bin/perl -e '$dir = `pwd`; chop $dir; unshift @INC, "$dir/blib/lib", "$dir/blib/arch"; eval "use Crypt::OpenSSL::RSA"; Crypt::OpenSSL::RSA->generate_key(256);' I'm wondering if there are some unexpected interactions between the tests at play here; running the above can help either eliminate or support that possibility.

Tue Nov 17 23:41:30 2009 iroberts+perl [...] red-bean.com - Correspondence added

On Tue Nov 17 23:22:53 2009, MARKLE wrote: Show quoted text

> On Tue Nov 17 23:12:52 2009, IROBERTS wrote:

> > Also, what happens if you run: > > > > openssl genrsa 256 > > > > from the command line?

> > Yes, in CentOS with perl 5.8.8 changing the test to 512 makes the test > succeed. > > hedges@anubis:~$ openssl genrsa 256 > Generating RSA private key, 256 bit long modulus > ...+++++++++++++++++++++++++++ > ..................+++++++++++++++++++++++++++ > 1584:error:0408C070:rsa routines:FIPS_RSA_SIGN:digest too big for rsa > key:fips_rsa_sign.c:280: > 1584:error:2D06D075:FIPS routines:fips_pkey_signature_test:test > failure:fips.c:672: > 1584:error:2D06A06B:FIPS routines:FIPS_CHECK_RSA:pairwise test > failed:fips_rsa_gen.c:132: > 1584:error:04081003:rsa routines:RSA_BUILTIN_KEYGEN:BN > lib:fips_rsa_gen.c:298: > > hedges@anubis:~$ openssl genrsa 512 > Generating RSA private key, 512 bit long modulus > ...++++++++++++ > .....++++++++++++ > unable to write 'random state' > e is 65537 (0x10001) > -----BEGIN RSA PRIVATE KEY----- > MIIBOwIBAAJBANTqQfnWtnj5ZPdyzdq+MuvmGfPctQMOZ0NokaOlfsnugY1u6oSn > kY+9n5Yi1uTW7IeyfQTZpX5ElOOwmR3Bv5sCAwEAAQJAL3BKZuoqGeYnUuLnbtd1 > 7Q2Ftsn0IswwwdooqPeUulTbgetAVVBL04uJ1aF2bcjhDPWepFTb/Y8r2Ki/nvj1 > oQIhAPt1jBvbH/+Rg/LyM6tpqtQLGdZS6klEWC+p1N2oJOlJAiEA2MKHE4R0yo23 > kFvJFvJvnRm0drGX1ezMSZlKkE0IpcMCIQDAp5+v2+OZJ8DDSuNhafTSvqqN9TrA > 5Zb29ZQ8Os4E2QIhALCRXUjHx/9Li7D7I4YevE6jpr5dEgr/1rVJOlTre1TTAiAJ > E9nuA4eRvnXMoRug0a5GwSX7RXrCQNTsdkK4QPKYDA== > -----END RSA PRIVATE KEY----- > > Thanks for checking this out! I would love to have another item on my > checklist of why RedHate sucks. --mark--

Sorry - I didn't see your response before my second response. I'll work on getting out a new release with a 512 bit key instead of 256 bit in the next few days. - Ian

Wed Nov 18 15:04:32 2009 MARKLE [...] cpan.org - Correspondence added

'use blib' accomplishes the same thing and incidentally should be used in all the test scripts. Your suggestion resulted in this: hedges@anubis:~/lame/Crypt-OpenSSL-RSA-0.25$ PERL_DL_NONLAZY=1 /usr/bin/perl -e '$dir = `pwd`; chop $dir; unshift @INC, "$dir/blib/lib", "$dir/blib/arch"; eval "use Crypt::OpenSSL::RSA"; Crypt::OpenSSL::RSA->generate_key(256);' RSA.xs:301: OpenSSL error: digest too big for rsa key at -e line 1. Why isn't a 256-bit key valid? It works on other platforms. openssl is hacked by RedHate to disallow it? Mark

Sun Nov 22 15:46:09 2009 iroberts+perl [...] red-bean.com - Correspondence added

Fixed in 0.26, now available at https://pause.perl.org/pub/PAUSE/authors/id/I/IR/IROBERTS/ and doubtlessly working its way through the distribution channels.

Sun Nov 22 15:46:33 2009 iroberts+perl [...] red-bean.com - Correspondence added

On Sun Nov 22 15:46:09 2009, IROBERTS wrote: Show quoted text

> Fixed in 0.26, now available at > https://pause.perl.org/pub/PAUSE/authors/id/I/IR/IROBERTS/ and > doubtlessly working its way through the distribution channels.

Closing

Sun Nov 22 15:46:34 2009 iroberts+perl [...] red-bean.com - Status changed from 'open' to 'resolved'