Copyright (c) University of Luxembourg 2022.

Created by Enrico VIGANO, [email protected], SnT, 2022.

DAMAt example

Introduction

This is an example of how to apply the DAMAt tool and procedure to improve an existing test suite. We recommend executing this procedure on Ubuntu 20.04.4 LTS, but it should work in any Linux environment. To apply the DAMAt procedure we need a Software Under Test and the relative test suite to evaluate. We decided to use libCSP, a C library implementing the Cubesat Space Protocol (CSP), a small protocol stack written to facilitate communication between the components of a CubeSat. Additional information on libCSP can be found at this link https://github.com/libcsp/libcsp. LibCSP does not come with a test suite, but it includes some examples; starting from one of these examples, we devised a small, intentionally flawed example of an integration test suite. The test suite contains a single test case that emulates a server/client configuration over loop-back and verifies that a minimum number of packets is transmitted by the client and correctly received by the server in a given time.

Installation procedure

Inside this repository you will find three sub-folders:

• damat-config: this folder contains all the files needed to apply DAMAt to libCSP.
• libcsp: this folder contains the source code for libCSP.
• test_suite: this folder contains the test suite under test that we are going to try and improve.

Compiling LibCSP

To compile libCSP and its test cases you will need to install the following dependencies:

sudo apt-get python gcc pkg-config libsocketcan-dev libzmq3-dev

This version of libCSP has been slightly modified with the insertion of DAMAt's mutation probes (see Step 3: Inserting the mutation probes). To compile it, you will need to go to the libcsp folder and run the following command.

./waf distclean configure build --mutation-opt -1 --singleton TRUE \
--with-os=posix --enable-rdp --enable-promisc --enable-hmac --enable-dedup \
--enable-can-socketcan --with-driver-usart=linux --enable-if-zmqhub \
--enable-examples

The flag --mutation-opt -1 will set the mutation probes to an inactive state.

To execute the original test case, go to the test_suite folder and run the following commands:

make clean
make test_01

The test should pass.

Installing DAMAt

DAMAt is not included in this repository. You can clone its source code from this repo https://github.com/SNTSVV/DAMAT. DAMAt is contained in the damat-pipeline folder. Is in that folder that every step of the pipeline will take place.

From now on, when we write commands, we will refer to the folder damat-pipeline in the DAMAt repository as $DAMAT_PIPELINE. If you export an environment variable called $DAMAT_PIPELINE containing the path to damat-pipeline you should be able to follow the commands written in the following.

You can do it using this command, replacing of course path/to/the/damat/repo with the path where you cloned the DAMAt repo.

export DAMAT_PIPELINE="/path/to/the/damat/repo/DAMAT/damat-pipeline"

The scripts contained in damat-config are configuration files for DAMAT and must be copied in the damat-pipeline folder in place of the placeholders that are present when you clone the repository. You can do that using the following commands:

cp damat-config/DAMAt_run_tests.sh $DAMAT_PIPELINE/DAMAt_run_tests.sh
cp damat-config/DAMAt_compile.sh $DAMAT_PIPELINE/DAMAt_compile.sh

Executing DAMAt

Step 1: Fault Model Specification

The first step for applying the DAMAt procedure is specifying a fault model. The fault model must be written in the csv format supported by DAMAt. You can find it here:

damat-config/fault_model_libcsp.csv

This csv file contains two Fault Models implementing a total of 39 Mutation Operators, which will generate the mutants. Another csv file in the same folder, test.csv should contain the following:

test_01,10000

The first column will contain the names of the test cases and the second the normal maximum execution time.

You should copy these two csv files in the $DAMAT_PIPELINE folder, in place of the examples.

cp damat-config/fault_model_libcsp.csv $DAMAT_PIPELINE/fault_model.csv
cp damat-config/tests.csv $DAMAT_PIPELINE/tests.csv

The following variable declarations, defined in $DAMAT_PIPELINE/DAMAt_configure.sh should be modified to contain the following values:

# the datatype of the elements of the target buffer
buffer_type="int"

# padding: can be used to skip the first n bit of a buffer, normally set to 0
padding=0

# singleton: can set to true to load the fault model into a singleton variable, normally set to "TRUE", can also  be set to "FALSE"
singleton="TRUE"

Step 2: Generating the mutation API

To generate the mutation API, run the following commands:

cd $DAMAT_PIPELINE
./DAMAt_probe_generation.sh

You should see that a new file is generated in the $DAMAT_PIPELINE folder: FAQAS_dataDrivenMutator.h. This file will contain the implementation of the mutation probes that will be used to modify the SUT's buffer data and generate the mutants. Another newly generated file, function_calls.out contains the definitions of the functions implementing the mutation probes. There is a probe for every different fault model.

Step 3: Insert the mutation probes

The mutation probes were inserted in a specific libCSP source file:

libcsp/src/csp_io.c

This file contains the definitions of the two functions that we will target with the probe insertion strategy:

• csp_send
• csp_read

these functions take as input the data structure csp_conn_t, which we are going to mutate. By looking at the content of the file you can see how the probes were embedded in the code. This is an example:

csp_packet_t * csp_read(csp_conn_t * conn, uint32_t timeout) {

 /* mutation probe */

 int damat_buffer_read[7];

 damat_buffer_read[0] = 0;
 damat_buffer_read[1] = conn->idin.pri;
 damat_buffer_read[2] = conn->idin.src;
 damat_buffer_read[3] = conn->idin.dst;
 damat_buffer_read[4] = conn->idin.dport;
 damat_buffer_read[5] = conn->idin.sport;
 damat_buffer_read[6] = conn->idin.flags;

 mutate_FM_Read(damat_buffer_read);

 conn->idin.pri = damat_buffer_read[1];
 conn->idin.src = damat_buffer_read[2];
 conn->idin.dst = damat_buffer_read[3];
 conn->idin.dport = damat_buffer_read[4];
 conn->idin.sport = damat_buffer_read[5];
 conn->idin.flags = damat_buffer_read[6];

 /* end of the probe */

 csp_packet_t * packet = NULL;

 if ((conn == NULL) || (conn->state != CONN_OPEN)) {
  return NULL;
 }

#if (CSP_USE_RDP)
 /* RDP: timeout can either be 0 (for no hang poll/check)
  or minimum the "connection timeout" */

 if (timeout && (conn->idin.flags & CSP_FRDP) && (timeout < conn->rdp.conn_timeout)) {
  timeout = conn->rdp.conn_timeout;
 }
#endif

 if (csp_queue_dequeue(conn->rx_queue, &packet, timeout) != CSP_QUEUE_OK) {
  return NULL;
 }

#if (CSP_USE_RDP)
 /* Packet read could trigger ACK transmission */
 if ((conn->idin.flags & CSP_FRDP) && conn->rdp.delayed_acks) {
  csp_rdp_check_ack(conn);
 }
#endif

 return packet;
}

The extra lines of code are there because, in its current form, DAMAt can only mutate arrays. The file also contains the following inclusion:

#include "FAQAS_dataDrivenMutator.h"

Step 4: Compile mutants

To enable this step, which will be automatically performed by the DAMAt pipeline, the user will need to modify this script:

$DAMAT_PIPELINE/DAMAt_compile.sh

to include the commands to compile the SUT. By taking a look at it, you will see how it has been done for this particular case. You just need to assign the absolute path to the root of this tutorial's repo to the variable REPO_ROOT.

# here the user must invoke the compilation of the SUT

# REPO_ROOT must contain the path to the root of this tutorial's repo
REPO_ROOT=
LIBCSP_ROOT="$REPO_ROOT/libcsp"

pushd $LIBCSP_ROOT

echo "$deco"
echo "compiling test"
echo "$deco"

./waf distclean configure build --mutation-opt $mutant_id $EXTRA_FLAGS_SINGL --with-os=posix --enable-rdp --enable-promisc --enable-hmac --enable-dedup --enable-can-socketcan --with-driver-usart=linux --enable-if-zmqhub #--enable-examples

echo "$deco"
if [ $? -eq 0 ]; then
    echo $x "$(tput setaf 2) compilation OK $(tput sgr0)"
else
    echo $x "$(tput setaf 1) compilation FAILED $(tput sgr0)"
fi
echo "$deco"
popd

Step 5: Execute the test suite

As with Step 4, Step 5 will be performed automatically by the pipeline, but the user must take a few preparatory steps. First, the user must provide the csv file containing the name of the test case and a normal execution time (useful to set a timeout in case a mutant should cause an infinite loop); You can find the one used in this example here:

damat-config/tests.csv

Then, the commands to run the test cases must be included in the proper space of the script $DAMAT_PIPELINE/DAMAt_run_tests.sh. In this case, they have been already added so you just need to assign the absolute path to the root of this tutorial's repo to the variable REPO_ROOT, as you did for the previous script. In this case, this is the code that has been added:

###############################################################################
# here the user shall call the execution of the current test case,
# REPO_ROOT must contain the path to the root of this tutorial's repo

  REPO_ROOT=

  TEST_SUITE="$REPO_ROOT/test_suite"

  pushd $TEST_SUITE

  echo "$deco" 2>&1 | tee -a $execution_log
  echo "$(tput setaf 1) RUNNING THE TEST NOW! $(tput sgr0)" 2>&1 | tee -a $execution_log
  echo "$deco" 2>&1 | tee -a $execution_log

  # timeout 30 ./build/examples/csp_server_client -t
  make clean
  make $tst

  EXEC_RET_CODE=$?

  echo "$deco" 2>&1 | tee -a $execution_log
  echo "$(tput setaf 2) FINITO! $(tput sgr0)" 2>&1 | tee -a $execution_log
  echo "$deco" 2>&1 | tee -a $execution_log

  popd

###############################################################################

The $tst variable is read from the first column of the csv file. When new test cases are added they must be included in that file to be executed during the procedure.

Step 6: Generate the results

At this point to execute the test suite against the mutants and gather the results you just need to run the pipeline by typing:

cd $DAMAT_PIPELINE
./DAMAt_mutants_launcher.sh

When the execution is complete you will see a message with the evaluation provided by DAMAt.

Improving on the existing test suite

After the first execution of DAMAt, the metrics describing the performance of the test suite should be the following:

• Fault Model Coverage 100%
• Mutation Operation Coverage 98%
• Mutation Score 52%

At a first glance we can see that not all the input partitions have been covered, since the MOC is < 100%, and some oracles are missing or incomplete since the MS is < 100%.

The file

results/detailed_results.csv

contains information on the status of every single mutant. This file will allow us to see what fault models, data items, and input partitions, in particular, are not well tested due to the test suite's shortcomings. For example, the DataItem column contains information on which data item in the buffer is targeted by the mutant. In particular:

• 1 = conn-><idin/idout>.pri
• 2 = conn-><idin/idout>.src
• 3 = conn-><idin/idout>.dst
• 4 = conn-><idin/idout>.dport
• 5 = conn-><idin/idout>.sport
• 6 = conn-><idin/idout>.flags

We will add test cases based on the data gathered by DAMAt to improve the test suite. A summary of this process, including metrics and tables is contained in the file

summary_of_the_results.xlsx

Improving the Mutation Operation Coverage

We can improve the Mutation Operation Coverage by adding new test cases that exercise partitions not covered by the test suite such as the ones targeted by mutants that were NOT_APPLIED.

By looking at the mutants that were NOT_APPLIED we can identify input partitions not covered by the test suite:

• there is no test case that covers a value of conn->idin.pri and conn->idout.pri > 3;

Improving the Mutation Score

We can improve the Mutation Score by adding new test cases that contain oracles on the values modified by LIVE mutants. In this case, this being an integration test suite, our primary focus is to check whether the different components (server and client) interact correctly and if the connection data contained in the structure csp_conn_t is correctly handled and preserved through these interactions.

By looking at the mutants that were APPLIED but not KILLED by the test suite, we notice that they belong to some specific members of the csp_conn_t structure:

• conn->idin.pri and conn->idout.pri, which define the priority of the connection;
• conn->idin.src, conn->idout.src, conn->idin.dst, and conn->idout.src, which represent the source and destination;
• conn->idin.sport, which represents the source port.

Test 02: Priority (pri)

A test case containing an oracle that checks if the conn->idin.pri and conn->idout.pri coincide between server and client should detect eventual mismanagement of the priority in the connection interfaces, and kill the mutants emulating these kinds of faults, that were previously APPLIED but not KILLED. In the test the client will send 5 packages with the four different priorities defined by libCSP:

• CSP_PRIO_CRITICAL (0)
• CSP_PRIO_HIGH (1)
• CSP_PRIO_NORM (2)
• CSP_PRIO_LOW (3)

Then it will test what happens if the priority is not defined by libCSP, for example, if it is equal to 6, thus hopefully improving the Mutation Operation Coverage, leading to a more extensive test suite. The content of conn->idin.pri as received by the server will be checked against the priority established by the client when connecting. The test case is implemented in the file

test_suite/test_02/test_02.c

and can be executed by compiling libCSP, moving to the test_suite folder, and executing the following command:

make test_02

Then you can add the test to the $DAMAT_PIPELINE/tests.csv so that it will be recognized and executed by DAMAt: The file should now contain these lines:

test_01,10000
test_02,10000

After adding test_02 to the test suite and re-executing DAMAt, the metrics should become the following:

• Fault Model Coverage 100%
• Mutation Operation Coverage 100%
• Mutation Score 61%

Test 03: Source (src) and Destination (dst)

A test case containing an oracle that checks the content of conn->idin.dst, conn->idin.src, conn->idout.dst, conn->idout.src should detect eventual alteration of these values between server and client, and kill the mutants emulating these kinds of faults, that were previously APPLIED but not KILLED.

The test consists of the client sending 5 messages to the server. For every message, the content of the aforementioned members of the csp_conn_t data structure shall be checked for discrepancies between server and client.

The test case is implemented in the file

test_suite/test_03/test_03.c

and can be executed by compiling libCSP, moving to the test_suite folder, and executing the following command:

make test_03

Then you can add the test to the $DAMAT_PIPELINE/tests.csv so that it will be recognized and executed by DAMAt: The file should now contain these lines:

test_01,10000
test_02,10000
test_03,10000

After adding test_03 to the test suite and re-executing DAMAt, the metrics will become the following:

• Fault Model Coverage 100%
• Mutation Operation Coverage 100%
• Mutation Score 67%

Test 04: Source Port (sport) and Destination Port (dport)

A test case containing an oracle that checks the content of conn->idin.sport and conn->idin.dport should detect eventual alteration of these values between server and client, and kill the mutants emulating these kinds of faults, that were previously APPLIED but not KILLED.

The test consists of the client sending 4 messages to the server. For every message, the content of the aforementioned members of the csp_conn_t data structure shall be checked for discrepancies between server and client. In particular, the server-side dport shall coincide with the client-side sport and vice-versa.

The test case is implemented in the file

test_suite/test_04/test_04.c

and can be executed by compiling libCSP, moving to the test_suite folder, and executing the following command:

make test_04

Then you can add the test to the $DAMAT_PIPELINE/tests.csv so that it will be recognized and executed by DAMAt: The file should now contain these lines:

test_01,10000
test_02,10000
test_03,10000
test_04,10000

After adding test_04 to the test suite and re-executing DAMAt, the metrics will become the following:

• Fault Model Coverage 100%
• Mutation Operation Coverage 100%
• Mutation Score 81%

Test 05: Flags (flags)

A test case containing an oracle that checks the content of conn->idin.flags and conn->idin.flags should detect eventual alteration of these values between server and client, and kill the mutants emulating these kinds of faults, that were previously APPLIED but not KILLED.

The test case is implemented in the file

test_suite/test_04/test_04.c

and can be executed by compiling libCSP, moving to the test_suite folder, and executing the following command:

make test_04

Then you can add the test to the $DAMAT_PIPELINE/tests.csv so that it will be recognized and executed by DAMAt: The file should now contain these lines:

test_01,10000
test_02,10000
test_03,10000
test_04,10000

After adding test_05 to the test suite and re-executing DAMAt, the metrics will become the following:

• Fault Model Coverage 100%
• Mutation Operation Coverage 100%
• Mutation Score 95%

Conclusion

This example is intended to show that, with the help of DAMAt, a user can obtain valuable indications on how to improve a test suite. The new test suite for libCSP, while intentionally still very limited and simple, should be more capable of identifying problems in the SUT than it was at the start.