Support report accumulation for casr-libfuzzer and casr-afl (#223)

casr/Cargo.toml

@@ -37,6 +37,7 @@ tokio = { version = "1", features = ["rt", "macros"], optional = true }
 toml = { version = "0.8", optional = true }
 wait-timeout = "0.2"
 which = "6.0"
+copy_dir = "0.1"
 
 libcasr = { path = "../libcasr", version = "2.12.0", features = ["serde", "exploitable"] }
 
@@ -52,4 +53,3 @@ required-features = ["dojo"]
 
 [dev-dependencies]
 lazy_static = "1.4"
-copy_dir = "0.1.3"

casr/src/bin/casr-afl.rs

@@ -75,6 +75,15 @@ fn main() -> Result<()> {
                 .value_parser(clap::value_parser!(PathBuf))
                 .help("Output directory with triaged reports")
         )
+        .arg(
+            Arg::new("join")
+                .long("join")
+                .env("CASR_PREV_CLUSTERS_DIR")
+                .action(ArgAction::Set)
+                .value_parser(clap::value_parser!(PathBuf))
+                .value_name("PREV_CLUSTERS_DIR")
+                .help("Use directory with previously triaged reports for new reports accumulation")
+        )
         .arg(
             Arg::new("force-remove")
                 .short('f')

casr/src/bin/casr-cluster.rs

@@ -762,11 +762,11 @@ fn main() -> Result<()> {
             println!("Number of new clusters: {result}");
         }
         // Print crashline dedup summary
-        if before != 0 {
+        if dedup_crashlines {
             println!("Number of reports before crashline deduplication in new clusters: {before}");
-        }
-        if before != after {
             println!("Number of reports after crashline deduplication in new clusters: {after}");
+        } else {
+            println!("Number of reports in new clusters: {after}");
         }
         let sil = calc_avg_sil(paths[1], jobs)?;
         println!("Cluster silhouette score: {sil}");

casr/src/bin/casr-libfuzzer.rs

@@ -74,6 +74,15 @@ fn main() -> Result<()> {
                 .value_name("OUTPUT_DIR")
                 .help("Output directory with triaged reports")
         )
+        .arg(
+            Arg::new("join")
+                .long("join")
+                .env("CASR_PREV_CLUSTERS_DIR")
+                .action(ArgAction::Set)
+                .value_parser(clap::value_parser!(PathBuf))
+                .value_name("PREV_CLUSTERS_DIR")
+                .help("Use directory with previously triaged reports for new reports accumulation")
+        )
         .arg(
             Arg::new("force-remove")
                 .short('f')

casr/src/triage.rs

@@ -159,8 +159,16 @@ pub fn fuzzing_crash_triage_pipeline(
         bail!("No crashes found");
     }
 
+    let accum_mode = matches.contains_id("join");
+
     let output_dir = initialize_dirs(matches)?;
 
+    let casrep_dir = if accum_mode {
+        output_dir.join("casrep")
+    } else {
+        output_dir.to_path_buf()
+    };
+
     // Get timeout
     let timeout = *matches.get_one::<u64>("timeout").unwrap();
 
@@ -189,7 +197,7 @@ pub fn fuzzing_crash_triage_pipeline(
         .join(
             || {
                 crashes.par_iter().try_for_each(|(_, crash)| {
-                    if let Err(e) = crash.run_casr(output_dir.as_path(), timeout) {
+                    if let Err(e) = crash.run_casr(casrep_dir.as_path(), timeout) {
                         // Disable util::log_progress
                         *counter.write().unwrap() = total;
                         bail!(e);
@@ -210,7 +218,7 @@ pub fn fuzzing_crash_triage_pipeline(
     info!("Deduplicating CASR reports...");
     let casr_cluster_d = Command::new(&casr_cluster)
         .arg("-d")
-        .arg(output_dir.clone().into_os_string())
+        .arg(casrep_dir.clone().into_os_string())
         .output()
         .with_context(|| format!("Couldn't launch {casr_cluster:?}"))?;
 
@@ -230,41 +238,66 @@ pub fn fuzzing_crash_triage_pipeline(
     }
 
     if !matches.get_flag("no-cluster") {
-        if output_dir
-            .read_dir()?
-            .flatten()
-            .map(|e| e.path())
-            .filter(|e| e.extension().is_some() && e.extension().unwrap() == "casrep")
-            .count()
-            < 2
-        {
-            info!("There are less than 2 CASR reports, nothing to cluster.");
-            return summarize_results(matches, crashes, gdb_args);
-        }
-        info!("Clustering CASR reports...");
-        let casr_cluster_c = Command::new(&casr_cluster)
-            .arg("-c")
-            .arg(output_dir.clone().into_os_string())
-            .output()
-            .with_context(|| format!("Couldn't launch {casr_cluster:?}"))?;
+        if accum_mode {
+            info!("Accumulating CASR reports...");
+            let casr_cluster_u = Command::new(&casr_cluster)
+                .arg("-u")
+                .arg(casrep_dir.clone().into_os_string())
+                .arg(output_dir.clone().into_os_string())
+                .output()
+                .with_context(|| format!("Couldn't launch {casr_cluster:?}"))?;
+
+            if casr_cluster_u.status.success() {
+                info!(
+                    "{}",
+                    String::from_utf8_lossy(&casr_cluster_u.stdout).trim_end()
+                );
+            } else {
+                error!(
+                    "{}",
+                    String::from_utf8_lossy(&casr_cluster_u.stderr).trim_end()
+                );
+            }
 
-        if casr_cluster_c.status.success() {
-            info!(
-                "{}",
-                String::from_utf8_lossy(&casr_cluster_c.stdout).trim_end()
-            );
+            // Remove reports from deduplication phase. They are in clusters now.
+            fs::remove_dir_all(casrep_dir)?;
         } else {
-            error!(
-                "{}",
-                String::from_utf8_lossy(&casr_cluster_c.stderr).trim_end()
-            );
-        }
+            if casrep_dir
+                .read_dir()?
+                .flatten()
+                .map(|e| e.path())
+                .filter(|e| e.extension().is_some() && e.extension().unwrap() == "casrep")
+                .count()
+                < 2
+            {
+                info!("There are less than 2 CASR reports, nothing to cluster.");
+                return summarize_results(matches, crashes, gdb_args);
+            }
+            info!("Clustering CASR reports...");
+            let casr_cluster_c = Command::new(&casr_cluster)
+                .arg("-c")
+                .arg(output_dir.clone().into_os_string())
+                .output()
+                .with_context(|| format!("Couldn't launch {casr_cluster:?}"))?;
+
+            if casr_cluster_c.status.success() {
+                info!(
+                    "{}",
+                    String::from_utf8_lossy(&casr_cluster_c.stdout).trim_end()
+                );
+            } else {
+                error!(
+                    "{}",
+                    String::from_utf8_lossy(&casr_cluster_c.stderr).trim_end()
+                );
+            }
 
-        // Remove reports from deduplication phase. They are in clusters now.
-        for casrep in fs::read_dir(output_dir)?.flatten().map(|e| e.path()) {
-            if let Some(ext) = casrep.extension() {
-                if ext == "casrep" {
-                    let _ = fs::remove_file(casrep);
+            // Remove reports from deduplication phase. They are in clusters now.
+            for casrep in fs::read_dir(casrep_dir)?.flatten().map(|e| e.path()) {
+                if let Some(ext) = casrep.extension() {
+                    if ext == "casrep" {
+                        let _ = fs::remove_file(casrep);
+                    }
                 }
             }
         }

casr/src/util.rs

@@ -9,6 +9,7 @@ use libcasr::stacktrace::{
 
 use anyhow::{bail, Context, Result};
 use clap::ArgMatches;
+use copy_dir::copy_dir;
 use gdb_command::stacktrace::StacktraceExt;
 use is_executable::IsExecutable;
 use log::{info, warn};
@@ -316,28 +317,34 @@ pub fn get_atheris_lib() -> Result<String> {
 pub fn initialize_dirs(matches: &clap::ArgMatches) -> Result<&PathBuf> {
     // Get output dir
     let output_dir = matches.get_one::<PathBuf>("output").unwrap();
-    if !output_dir.exists() {
-        fs::create_dir_all(output_dir).with_context(|| {
-            format!("Couldn't create output directory {}", output_dir.display())
-        })?;
-    } else if output_dir.read_dir()?.next().is_some() {
+    if output_dir.exists() && output_dir.read_dir()?.next().is_some() {
         if matches.get_flag("force-remove") {
             fs::remove_dir_all(output_dir)?;
-            fs::create_dir_all(output_dir).with_context(|| {
-                format!("Couldn't create output directory {}", output_dir.display())
-            })?;
         } else {
             bail!("Output directory is not empty.");
         }
     }
+
+    if let Some(join_dir) = matches.get_one::<PathBuf>("join") {
+        copy_dir(join_dir, output_dir)
+            .with_context(|| format!("Couldn't copy join directory {}", join_dir.display()))?;
+        // Get casrep dir
+        let casrep_dir = output_dir.join("casrep");
+        if !casrep_dir.exists() && fs::create_dir_all(&casrep_dir).is_err() {
+            bail!("Failed to create dir {}", &casrep_dir.to_str().unwrap());
+        }
+    } else if !output_dir.exists() && fs::create_dir_all(output_dir).is_err() {
+        format!("Couldn't create output directory {}", output_dir.display());
+    }
+
     // Get oom dir
     let oom_dir = output_dir.join("oom");
-    if fs::create_dir_all(&oom_dir).is_err() {
+    if !oom_dir.exists() && fs::create_dir_all(&oom_dir).is_err() {
         bail!("Failed to create dir {}", &oom_dir.to_str().unwrap());
     }
     // Get timeout dir
     let timeout_dir = output_dir.join("timeout");
-    if fs::create_dir_all(&timeout_dir).is_err() {
+    if !timeout_dir.exists() && fs::create_dir_all(&timeout_dir).is_err() {
         bail!("Failed to create dir {}", &timeout_dir.to_str().unwrap());
     }