segmentation fault on AIX with code working on Linux & Mac OS X

[Harry-64]

Hello alltogether,

I'm trying to run the tools pemtrans (in order to import openSSL's RSA  
private key and certificate into a cryptlib certificate store) on AIX.  
The URL for the tool is http://toroid.org/ams/pemtrans (if anyone  
looks for it)
The tool, linked against cryptlib 3.3.2 under Linux and/or Mac OS X  
works fine, but under AIX 5.3 with GCC 4.2.0, the program segfaults  
right at the call of the function "cryptSetAttributeString( pKey,  
CRYPT_CTXINFO_KEY_COMPONENTS, &rsa, sizeof( CRYPT_PKCINFO_RSA ) );".  
My first thought was it's all about the endianess of the openSSL  
bignum values, but they seem to be OK, I've printed out all the  
corresponding bits and bytes: they seem correct. I've traced down  
cryptlib to find a problem, but I'm not involved so deeply (at some  
point I was in some kernel and message queueing stuff of cryptlib  
which I don't understand).
Does anyone have any clue why this program fails on AIX (which is  
running on a RS/6000 CPU, perhaps it's really about the endianess) but  
works on Linux and Mac OS X (both Intel platform)?

Regards,
Harry


_______________________________________________
Cryptlib mailing list
Cryptlib@... via Mail: cryptlib-request@...
Archive: ftp://ftp.franken.de/pub/crypt/cryptlib/archives/
http://news.gmane.org/gmane.comp.encryption.cryptlib
Posts from non-subscribed addresses are blocked to prevent spam, please
subscribe in order to post messages.

[Scott Neugroschl]

> Hello alltogether,
>
> I'm trying to run the tools pemtrans (in order to import openSSL's RSA
> private key and certificate into a cryptlib certificate store) on AIX.
> The URL for the tool is http://toroid.org/ams/pemtrans (if anyone
> looks for it)
> The tool, linked against cryptlib 3.3.2 under Linux and/or Mac OS X
> works fine, but under AIX 5.3 with GCC 4.2.0, the program segfaults
> right at the call of the function "cryptSetAttributeString( pKey,
> CRYPT_CTXINFO_KEY_COMPONENTS, &rsa, sizeof( CRYPT_PKCINFO_RSA ) );".
> My first thought was it's all about the endianess of the openSSL
> bignum values, but they seem to be OK, I've printed out all the
> corresponding bits and bytes: they seem correct. I've traced down
> cryptlib to find a problem, but I'm not involved so deeply (at some
> point I was in some kernel and message queueing stuff of cryptlib
> which I don't understand).
> Does anyone have any clue why this program fails on AIX (which is
> running on a RS/6000 CPU, perhaps it's really about the endianess) but
> works on Linux and Mac OS X (both Intel platform)?
>

[[SAN]]
What error status are you getting as the return value from
cryptSetAttributeString?
And what is your key size?  I believe that 3.3.2 doesn't support 512-bit
moduli.



_______________________________________________
Cryptlib mailing list
Cryptlib@... via Mail: cryptlib-request@...
Archive: ftp://ftp.franken.de/pub/crypt/cryptlib/archives/
http://news.gmane.org/gmane.comp.encryption.cryptlib
Posts from non-subscribed addresses are blocked to prevent spam, please
subscribe in order to post messages.

[Harry-64]

Hi!

Am 06.07.2009 um 23:10 schrieb Scott Neugroschl:

>> I'm trying to run the tools pemtrans (in order to import openSSL's  
>> RSA
>> private key and certificate into a cryptlib certificate store) on  
>> AIX.
>> The URL for the tool is http://toroid.org/ams/pemtrans (if anyone
>> looks for it)
>> The tool, linked against cryptlib 3.3.2 under Linux and/or Mac OS X
>> works fine, but under AIX 5.3 with GCC 4.2.0, the program segfaults
>> right at the call of the function "cryptSetAttributeString( pKey,
>> CRYPT_CTXINFO_KEY_COMPONENTS, &rsa, sizeof( CRYPT_PKCINFO_RSA ) );".
>> My first thought was it's all about the endianess of the openSSL
>> bignum values, but they seem to be OK, I've printed out all the
>> corresponding bits and bytes: they seem correct. I've traced down
>> cryptlib to find a problem, but I'm not involved so deeply (at some
>> point I was in some kernel and message queueing stuff of cryptlib
>> which I don't understand).
>> Does anyone have any clue why this program fails on AIX (which is
>> running on a RS/6000 CPU, perhaps it's really about the endianess)  
>> but
>> works on Linux and Mac OS X (both Intel platform)?
> [[SAN]]
> What error status are you getting as the return value from
> cryptSetAttributeString?
> And what is your key size?  I believe that 3.3.2 doesn't support 512-
> bit
> moduli.

There is no return value since the function doesn't return, it just  
throws a segmentation fault (and coredumps). This is an indication  
that a pointer isn't used correctly.
The key size is 1024 bits, but since the code works under Linux and  
Mac OS X, I assume this isn't a cryptlib limitation.

Regards,
Harry

_______________________________________________
Cryptlib mailing list
Cryptlib@... via Mail: cryptlib-request@...
Archive: ftp://ftp.franken.de/pub/crypt/cryptlib/archives/
http://news.gmane.org/gmane.comp.encryption.cryptlib
Posts from non-subscribed addresses are blocked to prevent spam, please
subscribe in order to post messages.