DRM Library integration¶

The Accelize DRM can be integrated into an application using the library API.

The DRM library supports the following languages:

C : C API
C++ : C++ API
Python : Python API
Cython : Python API, Cython headers section (.pxd files)

Note

No examples are provided in the Cython language, see the DRM library Python source code for an example of Cython integration.

Install DRM API¶

DRM Library installation

Include DRM API¶

In C¶

#include "accelize/drmc.h"

In C++¶

#include "accelize/drm.h"
namespace drm = Accelize::DRM;

In Python¶

from accelize_drm import DrmManager

Use DRM API in your code¶

Create a DrmManager object¶

To instantiate a DRM Manager object, the software developer needs to provide some inputs.

First, the DRM manager requires the paths to the conf.json and the cred.json files. Those files are used to configure the DRM (i.e licensing mode) and authenticate the user. See Configuration for more information about configuration files.

The DRM API needs to access the DRM controller registers in the design both in read and write directions. The means to access these registers depend on the design and the driver. So it is the user’s responsibility to implement two callback functions (read and write functions) to perform those registers access operations.

Note

The read and write register functions will be used asynchronously from the rest of the application. Adequate protections against concurrent accesses is left to attention of the user.

Finally, the DRM Manager needs a third callback function to manage asynchronous errors that the DRM thread might report. The most basic function is a call that simply displays the message.

In the following examples, The read and write registers function are named fpga_read_register and fpga_write_register and are imported from a FPGA C library driver.

In C¶

#include <stdio.h>

// Define functions to read and write FPGA registers to use them as
// callbacks in DrmManager.

int read_register( uint32_t offset, uint32_t* value, void* user_p ) {
    return fpga_read_register( drm_controller_base_addr + offset, value );
}

int write_register( uint32_t offset, uint32_t value, void* user_p ) {
    return fpga_write_register( drm_controller_base_addr + offset, value );
}

// Define asynchronous error callback
void asynch_error( const char* err_msg, void* user_p ) {
    fprintf( stderr, "%s", err_msg );
}

// Instantiate DrmManager with previously defined functions and
// configuration files

DrmManager* drm_manager = NULL;
int ctx = 0;

if ( DrmManager_alloc(
        &drm_manager,

        // Configuration files paths
        "./conf.json",
        "./cred.json",

        // Read/write register functions callbacks
        read_register,
        write_register,

        // Asynchronous error callback
        asynch_error,

        // Contextual pointer that the user can use at his/her convenience
        &ctx ) ) {
    // In the C case, the last error message is stored inside the
    // "DrmManager"
    fprintf( stderr, "%s", drm_manager->error_message );
}

In C++¶

#include <iostream>
#include <string>

// Define functions to read and write FPGA registers and use them as
// callbacks to instantiate the DrmManager.
// Note: This example use C++11 Lambda function.

drm::DrmManager drm_manager(
    // Configuration files paths
    "./conf.json",
    "./cred.json",

    // Read/write register functions callbacks
    [&](uint32_t offset, uint32_t* value) {
        return fpga_read_register( drm_controller_base_addr + offset, value );
    },

    [&](uint32_t offset, uint32_t value) {
        return fpga_write_register( drm_controller_base_addr + offset, value );
    },

    // Asynchronous error callback
    [&](const std::string &err_msg) {
        std::cerr << err_msg << std::endl;
        GlobalDrmError = true; // Any global machanism to warn the other threads of a DRM error
    }
);

In Python¶

# The FPGA driver is most likely written in C and read and write
# functions can be imported using the "ctypes" module like in following
# example
import ctypes

libfpga = ctypes.cdll.LoadLibrary("libfpga.so")

fpga_read_register = libfpga.fpga_read_register
fpga_read_register.restype = ctypes.c_int  # return code
fpga_read_register.argtypes = (
    ctypes.c_uint32,  # offset
    ctypes.POINTER(ctypes.c_uint32)  # value
)

fpga_write_register = libfpga.fpga_write_register
fpga_write_register.restype = ctypes.c_int  # return code
fpga_write_register.argtypes = (
    ctypes.c_uint32,  # offset
    ctypes.c_uint32  # value
)

drm_manager = DrmManager(
    # Configuration files paths
    "./conf.json" ,
    "./cred.json" ,

    # Read/write register functions callbacks
    # Python lambda function are used to simplify code
    lambda offset, value: fpga_read_register(
        drm_controller_base_addr + offset, value),
    lambda offset, value: fpga_write_register(
        drm_controller_base_addr + offset, value),

    # Python API provides a default asynchronous error callback that
    # raise Python exceptions on error. It is possible to override it if
    # needed
    )

Activate the protected hardware¶

Once the DrmManager object allocated, you need to activate the protected IPs to use your design. This is performed with the activate method.

When the function returns and the license is valid, the protected IPs are guaranteed to be unlocked and usable.

In case of metering and floating licensing, this function spawns a thread to keep the design unlocked and periodically send metering information to the Accelize Web Service.

Warning

The activate method call may take few seconds. Especially on the first call due to internet and network delay.

C¶

if ( DrmManager_activate( drm_manager, false ) )
    fprintf( stderr, "%s", drm_manager->error_message );

C++¶

try {
    drm_manager.activate();
} catch( const drm::Exception& e ) {
    cerr << "DRM error: " << e.what() << endl;
}

Python¶

drm_manager.activate()

Deactivate the protected hardware¶

At the end of the application execution, when the protected IP is no longer required, the DRM must be deactivated.

When this function returns, the protected IPs are guaranteed to be locked (Except with the node-locked licensing mode where the IP is kept unlocked).

Note

Calling this function at the end of your application allows you to reuse the same design in different licensing modes without the need to recompile.

Warning

In Floating or Metering mode, the deactivate method call may take some seconds due to internet or network delay.

C¶

if ( DrmManager_deactivate( drm_manager, false ) )
    fprintf( stderr, "%s", drm_manager->error_message );

// In the C case, the "DrmManager" needs also to be freed to deallocate
// associated resources.
if ( DrmManager_free( &drm_manager ) )
    fprintf( stderr, "%s", drm_manager->error_message );

C++¶

try {
    drm_manager.deactivate();
} catch( const drm::Exception& e ) {
    cerr << "DRM error: " << e.what() << endl;
}

Python¶

drm_manager.deactivate()

Compile your application¶

Your application needs to be linked against the adequate C or C++ DRM library and have the thread support enabled.

C with GCC compiler¶

gcc source.c -pthread -laccelize_drmc -o application

C++ with GCC compiler¶

g++ source.cpp -pthread -laccelize_drm -o application

Warning

The library requires JsonCpp that is included as json/json.h for portability.

On some OS like Ubuntu or CentOS, JsonCpp include is in a jsoncpp subdirectory of the system include directory (resulting in header file path jsoncpp/json/json.h).

Make sure to specify the correct JsonCpp include path when compiling.

GCC: JsonCpp include with an absolute path specification¶

gcc source.c -pthread -laccelize_drmc -o application -I/usr/include/jsoncpp

CMake: JsonCpp include path auto-detection¶

find_package(PkgConfig REQUIRED)
pkg_check_modules(JSONCPP jsoncpp)
include_directories(${JSONCPP_INCLUDEDIR})

Advanced usage¶

By default, the activate call creates a new DRM session and the deactivate call closes it.

When a session is open, a call to the Accelize Web Service is performed to:

get a new license key to unlock the hardware design
get a new session ID

When a session is closed, the metering counter is reset and a call to the Accelize Web Service is performed to:

send the last metering data
close the current session

In some cases, it might be useful to pause and resume a session instead of closing it, in particular to keep the same session ID. It may also reduce the activate/deactivate call delay by skipping the license request to the Accelize Web Service if the license is still valid.

Warning

Pausing a session at application level maintains the session open on Accelize Web Service. The session must be closed explicitly by the application before it terminates.

This strategy is usually not recommended but it can be useful in the following cases:

As a general way, when there is no possibility to have a background service,
When the call to the hardware function is short but frequent (less than the license expiration period).

Pausing/resuming a DRM session¶

Following code snippets show how to implement pause/resume functionality:

C¶

// Activate the DRM and resume the existing session if any
if ( DrmManager_activate( drm_manager, true ) )
    fprintf( stderr, "%s", drm_manager->error_message );

// [...]

// Deactivate the DRM, but pause the session instead of closing it
if ( DrmManager_deactivate( drm_manager, true ) )
    fprintf( stderr, "%s", drm_manager->error_message );

C++¶

// Activate the DRM and resume the existing session if any
try {
    drm_manager.activate( true );

    // [...]

    // Deactivate the DRM, but pause the session instead of closing it
    drm_manager.deactivate( true );

} catch( const drm::Exception& e ) {
    cerr << "DRM error: " << e.what() << endl;
}

Python¶

# Activate the DRM and resume the existing session if any
drm_manager.activate(True)

# [...]

# Deactivate the DRM, but pause the session instead of closing it
drm_manager.deactivate(True)

Full API documentation¶

For more information about the API in your favorite language, refer to DRM Library APIs.