Speed up communication between Ghidra and cwe_checker (#145)

... also disabled CWE 78 check for standard runs until performance problems of the check are solved.

caller/Cargo.toml

@@ -8,4 +8,5 @@ edition = "2018"
 structopt = "0.3"
 cwe_checker_rs = { path = "../cwe_checker_rs" }
 serde_json = "1.0"
-directories = "3.0"
\ No newline at end of file
+directories = "3.0"
+nix = "0.19.1"
\ No newline at end of file

caller/src/main.rs

@@ -3,9 +3,11 @@ use cwe_checker_rs::utils::log::print_all_messages;
 use cwe_checker_rs::utils::{get_ghidra_plugin_path, read_config_file};
 use cwe_checker_rs::AnalysisResults;
 use cwe_checker_rs::{intermediate_representation::Project, utils::log::LogMessage};
+use nix::{sys::stat, unistd};
 use std::collections::HashSet;
 use std::path::{Path, PathBuf};
 use std::process::Command;
+use std::thread;
 use structopt::StructOpt;
 
 #[derive(Debug, StructOpt)]
@@ -124,6 +126,13 @@ fn run_with_ghidra(args: CmdlineArgs) {
     // Filter the modules to be executed if the `--partial` parameter is set.
     if let Some(ref partial_module_list) = args.partial {
         filter_modules_for_partial_run(&mut modules, partial_module_list);
+    } else {
+        // TODO: CWE78 is disabled on a standard run for now,
+        // because it uses up huge amounts of RAM and computation time on some binaries.
+        modules = modules
+            .into_iter()
+            .filter(|module| module.name != "CWE78")
+            .collect();
     }
 
     let binary_file_path = PathBuf::from(args.binary.unwrap());
@@ -241,49 +250,66 @@ fn get_project_from_ghidra(file_path: &Path, binary: &[u8], quiet_flag: bool) ->
     let filename = file_path
         .file_name()
         .expect("Invalid file name")
-        .to_string_lossy();
-    let output_filename = format!("{}_{}.json", filename, timestamp_suffix);
-    let output_path = tmp_folder.join(output_filename);
+        .to_string_lossy()
+        .to_string();
     let ghidra_plugin_path = get_ghidra_plugin_path("p_code_extractor");
-    // Execute Ghidra
-    let output = match Command::new(&headless_path)
-        .arg(&tmp_folder) // The folder where temporary files should be stored
-        .arg(format!("PcodeExtractor_{}_{}", filename, timestamp_suffix)) // The name of the temporary Ghidra Project.
-        .arg("-import") // Import a file into the Ghidra project
-        .arg(file_path) // File import path
-        .arg("-postScript") // Execute a script after standard analysis by Ghidra finished
-        .arg(ghidra_plugin_path.join("PcodeExtractor.java")) // Path to the PcodeExtractor.java
-        .arg(&output_path) // Output file path
-        .arg("-scriptPath") // Add a folder containing additional script files to the Ghidra script file search paths
-        .arg(ghidra_plugin_path) // Path to the folder containing the PcodeExtractor.java (so that the other java files can be found.)
-        .arg("-deleteProject") // Delete the temporary project after the script finished
-        .arg("-analysisTimeoutPerFile") // Set a timeout for how long the standard analysis can run before getting aborted
-        .arg("3600") // Timeout of one hour (=3600 seconds) // TODO: The post-script can detect that the timeout fired and react accordingly.
-        .output() // Execute the command and catch its output.
-    {
-        Ok(output) => output,
-        Err(err) => {
-            eprintln!("Error: Ghidra could not be executed:\n{}", err);
-            std::process::exit(101);
-        }
-    };
-    if !output.status.success() {
-        match output.status.code() {
-            Some(code) => {
-                eprintln!("{}", String::from_utf8(output.stdout).unwrap());
-                eprintln!("{}", String::from_utf8(output.stderr).unwrap());
-                eprintln!("Execution of Ghidra plugin failed with exit code {}", code);
+
+    // Create a unique name for the pipe
+    let fifo_path = tmp_folder.join(format!("pcode_{}.pipe", timestamp_suffix));
+
+    // Create a new fifo and give read and write rights to the owner
+    if let Err(err) = unistd::mkfifo(&fifo_path, stat::Mode::from_bits(0o600).unwrap()) {
+        eprintln!("Error creating FIFO pipe: {}", err);
+        std::process::exit(101);
+    }
+
+    let thread_fifo_path = fifo_path.clone();
+    let thread_file_path = file_path.to_path_buf();
+    let thread_tmp_folder = tmp_folder.to_path_buf();
+    // Execute Ghidra in a new thread and return a Join Handle, so that the thread is only joined
+    // after the output has been read into the cwe_checker
+    let ghidra_subprocess = thread::spawn(move || {
+        let output = match Command::new(&headless_path)
+            .arg(&thread_tmp_folder) // The folder where temporary files should be stored
+            .arg(format!("PcodeExtractor_{}_{}", filename, timestamp_suffix)) // The name of the temporary Ghidra Project.
+            .arg("-import") // Import a file into the Ghidra project
+            .arg(thread_file_path) // File import path
+            .arg("-postScript") // Execute a script after standard analysis by Ghidra finished
+            .arg(ghidra_plugin_path.join("PcodeExtractor.java")) // Path to the PcodeExtractor.java
+            .arg(thread_fifo_path) // The path to the named pipe (fifo)
+            .arg("-scriptPath") // Add a folder containing additional script files to the Ghidra script file search paths
+            .arg(ghidra_plugin_path) // Path to the folder containing the PcodeExtractor.java (so that the other java files can be found.)
+            .arg("-deleteProject") // Delete the temporary project after the script finished
+            .arg("-analysisTimeoutPerFile") // Set a timeout for how long the standard analysis can run before getting aborted
+            .arg("3600") // Timeout of one hour (=3600 seconds) // TODO: The post-script can detect that the timeout fired and react accordingly.
+            .output() // Execute the command and catch its output.
+        {
+            Ok(output) => output,
+            Err(err) => {
+                eprintln!("Error: Ghidra could not be executed:\n{}", err);
                 std::process::exit(101);
             }
-            None => {
-                eprintln!("Execution of Ghidra plugin failed: Process was terminated.");
-                std::process::exit(101);
+        };
+
+        if !output.status.success() {
+            match output.status.code() {
+                Some(code) => {
+                    eprintln!("{}", String::from_utf8(output.stdout).unwrap());
+                    eprintln!("{}", String::from_utf8(output.stderr).unwrap());
+                    eprintln!("Execution of Ghidra plugin failed with exit code {}", code);
+                    std::process::exit(101);
+                }
+                None => {
+                    eprintln!("Execution of Ghidra plugin failed: Process was terminated.");
+                    std::process::exit(101);
+                }
             }
         }
-    }
-    // Read the results from the Ghidra script
-    let file =
-        std::fs::File::open(&output_path).expect("Could not read results of the Ghidra script");
+    });
+
+    // Open the FIFO
+    let file = std::fs::File::open(fifo_path.clone()).expect("Could not open FIFO.");
+
     let mut project_pcode: cwe_checker_rs::pcode::Project =
         serde_json::from_reader(std::io::BufReader::new(file)).unwrap();
     project_pcode.normalize();
@@ -301,7 +327,12 @@ fn get_project_from_ghidra(file_path: &Path, binary: &[u8], quiet_flag: bool) ->
             project
         }
     };
-    // delete the temporary file again.
-    std::fs::remove_file(output_path).unwrap();
+
+    ghidra_subprocess
+        .join()
+        .expect("The Ghidra thread to be joined has panicked!");
+
+    std::fs::remove_file(fifo_path).unwrap();
+
     project
 }

ghidra/p_code_extractor/PcodeExtractor.java

@@ -2,7 +2,6 @@ import java.io.FileNotFoundException;
 import java.util.ArrayList;
 import java.util.HashMap;
 import java.util.stream.StreamSupport;
-import java.util.concurrent.TimeUnit;
 
 import bil.*;
 import term.*;
@@ -58,7 +57,6 @@ public class PcodeExtractor extends GhidraScript {
         String jsonPath = getScriptArgs()[0];
         Serializer ser = new Serializer(project, jsonPath);
         ser.serializeProject();
-        TimeUnit.SECONDS.sleep(3);
 
     }