Feature #8667: Unable to set OpenSSL GCM iv_length in Ruby

• Status: Closed
• Assignee: -
• Target version: -

Hello,

In OpenSSL you are allowed to change the iv_length on an AES-BCM cipher. (
http://www.openssl.org/docs/crypto/EVP_EncryptInit.html#GCM_Mode) However
this was not implemented in the ruby-wrapper. Since I am a novice in C and
OpenSSL I think by no means my supplied patch is complete, it is a start
however. Maybe this missing function can be added to Ruby 2.0?

You can now set the iv_length using:

cipher = OpenSSL::Cipher.new('aes-128-gcm').encrypt
cipher.iv_len = 16

An issue I already spotted is that OpenSSL sets the ivlen on the
cipher_data (snippet from OpenSSL crypto/evp/e_aes.c):
EVP_AES_GCM_CTX *gctx = c->cipher_data;
gctx->ivlen = arg;

and not the c->cipher->iv_len. So querying for the iv_len in ruby by using
cipher.iv_len will still report the default which is 12. Encryption however
is done correctly using the new iv-length. I tested it by comparing it to
results from other programming languages (Java and C#).

Regards Andres

[drbrain (Eric Hodel)]

• Category set to ext/openssl
• Status changed from Open to Assigned
• Assignee set to MartinBosslet (Martin Bosslet)
• ruby -v set to -

[dwaite (David Waite)]

• File 0001-Alternate-implementation-of-gcm-IV-length-support.patch 0001-Alternate-implementation-of-gcm-IV-length-support.patch added

Alternate patch which makes iv length changes implicit in iv= when using GCM.

This issue currently requires me to require a patched ruby for my software to run.

[zzak (zzak _)]

• Assignee changed from MartinBosslet (Martin Bosslet) to 7150

Without knowing what HAVE_AUTHENTICATED_ENCRYPTION is, I'm not really sure.. also we need a test for this.

[rhenium (Kazuki Yamaguchi)]

• Related to Feature #10420: Cannot set IV length to OpenSSL::Cipher added

[rhenium (Kazuki Yamaguchi)]

• Tracker changed from Bug to Feature
• Status changed from Assigned to Closed

I think this is good to have. I thought about changing Cipher#iv= first, too, but I didn't because of the current (2.3) behavior of Cipher#iv=. Please see:

https://github.com/ruby/openssl/pull/58