Skip to content

WorksButNotTested/stalker-coverage

Repository files navigation

stalker-coverage

Stalker coverage is a small TypeScript module designed to be used with FRIDA to generate coverage information in DynamoRio DRCOV format. This information can then be loaded into IDA using the lighthouse plugin or Ghidra using the Dragondance plugin.

Example

Source Code

The following C code will be used as our example:

#include <stdio.h>
#include <stdlib.h>

long square(long x)
{
    return x * x;
}

int main(int argc, char *argv[])
{
    if (argc < 2) {
        printf("Give me an argument!\n");
        return 1;
    }

    long x = strtol(argv[1], NULL, 0);
    long sq_x = square(x);
    printf("%ld squared is %ld\n", x, sq_x);
    return 0;
}

We first compile the code as follows (note the use of the -rdynamic flag to allow FRIDA to more readily find the function):

$ gcc -rdynamic -o test test.c

Project

We can then start using the frida-agent-example as a template project and install the stalker-coverage module:

$ npm install
$ npm install --save ./stalker-coverage

Using Stalker-Coverage

We can then interact with this module using the following typescript code (put this in frida-agent-example/agent/index.ts):

import { Coverage } from "@worksbutnottested/stalker-coverage/dist/coverage";
/*
 * This sample replaces the 'main' function of the target application with one which starts
 * collecting coverage information, before calling the original 'main' function. Once the
 * original 'main' function returns, coverage collection is stopped. This coverage
 * information is written into a file which can then be directly loaded into IDA lighthouse
 * or Ghidra Dragondance.
 */

/*
 * The address of the symbol 'main' which is to be used as the start and finish point to
 * collect coverage information.
 */
const mainAddress = DebugSymbol.fromName("main").address;

/**
 * The main module for the program for which we will collect coverage information (we will
 * not collect coverage information for any library dependencies).
 */
const mainModule = Process.enumerateModules()[0];

/*
 * A NativeFunction type for the 'main' function which will be used to call the original
 * function.
 */
const mainFunctionPointer = new NativeFunction(
    mainAddress,
    "int",
    ["int", "pointer"],
    { traps : "all"});

/*
 * A function to be used to replace the 'main' function. This function will start collecting
 * coverage information before calling the original 'main' function. Once this function
 * returns, the coverage collection will be stopped and flushed. Note that we cannot use
 * Interceptor.attach here, since this interferes with Stalker (which is used to provide the
 * coverage data).
 */
const mainReplacement = new NativeCallback(
    (argc, argv) => {
        const coverageFileName = `${mainModule.path}.dat`;
        const coverageFile = new File(coverageFileName, "wb+");

        const coverage = Coverage.start({
            moduleFilter: (m) => Coverage.mainModule(m),
            onCoverage: (coverageData) => {
                coverageFile.write(coverageData);
            },
            threadFilter: (t) => Coverage.currentThread(t),
        });

        const ret = mainFunctionPointer(argc, argv) as number;

        coverage.stop();
        coverageFile.close();

        return ret;
    },
    "int",
    ["int", "pointer"]);

/*
 * Replace the 'main' function with our replacement function defined above.
 */
Interceptor.replace(mainAddress, mainReplacement);

We can build our project using the following command:

$ npm run build

Running

First we must download the FRIDA gadget from here (be sure to rename it frida-gadget.so) and create a configuration file called frida-gadget.config

{
  "interaction": {
    "type": "script",
    "path": "./_agent.js"
  }
}

Now we can run our application and collect the coverage information as follows:

$ LD_PRELOAD=./frida-gadget.so ./test 123
123 squared is 15129

Coverage data should be stored along-side our test application in test.dat. This file can then de directly loaded into ghidra using the Dragondance plugin.

Screenshots

We can then see the resulting coverage data inside of ghidra using the Dragondance plugin. Coverage1.png Coverage2.png