Table of contents

• Table of contents
• Installation
• Configuration
• Using Engine_cudamrg with the openssl command
• Using Engine_cudamrg with the openssl config file
• Using Engine_cudamrg with the OpenSSL autoloading
• Engine_cudamrg Options

Installation

Before installing the engine_cudamrg you MUST install the NVIDIA CUDA Toolkit as described in the official CUDA documentation and install the OpenSSL toolkit.

Installing engine_cudamrg is quite simple:

wget http://engine-cuda.googlecode.com/files/engine_cudamrg-v_0.1.1.tar.gz
tar xvzf engine_cudamrg-v_0.1.1.tar.gz 
cd engine_cudamrg/
./configure 
make 
make install

If you install a driver different from the one provided by the NVIDIA web site you MUST specify the path to the libcuda.so library on the configure command-line.

For example: if you have the library into the directory /usr/lib64/nvidia-current you MUST add to the command-line something like this LDFLAGS=-L/usr/lib64/nvidia-current, in other case the script will fail.

If your distribution has a version of gcc above 4.4 and a version of the CUDA Toolkit below 3.1 you MUST specify on the configure command-line the path to a version of gcc below 4.4 throutgh the --with-ccbin option.

For example: if you have gcc-4.3 installed into /usr/bin you MUST add to the command-line something like this --with-ccbin=/usr/bin/gcc-4.3, in other case the script will fail.

By default the engine_cudamrg installs into /usr/local/lib/engines, if you want to install the engine into a different directory you must specify the path through the --prefix option on the configure command-line.

For example: if you want to install the engine into the directory /opt you must specify on the command-line something like this --prefix=/opt.

Alternatevely you can download a more recent version from the svn repository using the svn command instead of wget:

svn checkout http://engine-cuda.googlecode.com/svn/trunk/ engine-cudamrg

Configuration

Using Engine_cudamrg with the openssl command

You can run the OpenSSL command shell, load the engine and then run any command using the engine.

Here is an example:

$ openssl
OpenSSL> engine -t dynamic -pre SO_PATH:/usr/local/lib/engines/libcudamrg.so -pre ID:cudamrg -pre LIST_ADD:1 -pre LOAD
OpenSSL> enc -engine cudamrg -e -aes-256-ecb -k $YOUR_KEY -in $YOUR_INPUT_FILE -out $YOUR_OUTPUT_FILE -bufsize $BUFFER_SIZE

In this example the engine_cudamrg is loaded.

The second command encrypt a file named $YOUR_INPUT_FILE into $YOUR_OUTPUT_FILE using $YOUR_KEY and a buffer of size $BUFFER_SIZE.

Using Engine_cudamrg with the openssl config file

You can create or edit an openssl config file, so you don't need to type in or paste the above commands every the time.

Here is an example for OpenSSL 0.9.8:

openssl_conf = openssl_init

[openssl_init]
engines = engine_section

[engine_section]
cudamrg = cudamrg_section

[cudamrg_section]
engine_id = cudamrg
dynamic_path = /usr/local/lib/engines/libcudamrg.so
default_algorithms = ALL
init = 1

With such a config file you can directly call openssl to use that engine...

openssl enc -engine cudamrg -e -aes-256-ecb -k $YOUR_KEY -in $YOUR_INPUT_FILE -out $YOUR_OUTPUT_FILE -bufsize $BUFFER_SIZE

...or...

openssl -e -aes-256-ecb -k $YOUR_KEY -in $YOUR_INPUT_FILE -out $YOUR_OUTPUT_FILE -bufsize $BUFFER_SIZE

Using Engine_cudamrg with the OpenSSL autoloading

OpenSSL 0.9.8+ can automaticaly load engines. If you want to use this feature, add a symlink from libcudamrg.so in the lib/engines/ directory or install it directly into that directory.

After that you can directly call openssl to use this engine...

openssl enc -engine cudamrg -e -aes-256-ecb -k $YOUR_KEY -in $YOUR_INPUT_FILE -out $YOUR_OUTPUT_FILE -bufsize $BUFFER_SIZE

Engine_cudamrg Options

Options you can use with engine_cudamrg:

• SO_PATH: Specifies the path to the 'cuda-engine' shared library
• VERBOSE: Print additional details
• QUIET: Remove additional details
• BUFFER_SIZE: Specifies the size of the buffer between central memory and GPU memory in kilobytes (default: 8MB)

WARNING: if you use these command with the -post option in an OpenSSL command shell or into the configuration file you will get an error like this: Error: you cannot set command 201 when the engine is already initialized.[Failure]: VERBOSE.