From b73e441d5a6c1988986b8c42eef27b8319ba937e Mon Sep 17 00:00:00 2001
From: Philippe Antoine <pantoine@oisf.net>
Date: Mon, 27 Jul 2026 20:58:38 +0200
Subject: [PATCH] fuzz: script to generate corpus for datasets target

Ticket: 3590
---
 scripts/fuzz_gen_corpus_datasets.py | 123 ++++++++++++++++++++++++++++
 1 file changed, 123 insertions(+)
 create mode 100644 scripts/fuzz_gen_corpus_datasets.py

diff --git a/scripts/fuzz_gen_corpus_datasets.py b/scripts/fuzz_gen_corpus_datasets.py
new file mode 100644
index 0000000000..e3e02686a7
--- /dev/null
+++ b/scripts/fuzz_gen_corpus_datasets.py
@@ -0,0 +1,123 @@
+#!/usr/bin/env python3
+
+import argparse
+import re
+import sys
+from pathlib import Path
+
+
+KEYWORD_RE = re.compile(r"\b(datarep|dataset)\s*:\s*(.*?);", re.IGNORECASE | re.DOTALL)
+DATASET_FILE_RE = re.compile(r"\b(load|state|save)\s+([^,;\s]+)", re.IGNORECASE)
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(
+        description=(
+            "Generate corpus elements for fuzz_dataset from suricata-verify tests "
+            "using datarep/dataset keywords."
+        )
+    )
+    parser.add_argument(
+        "output",
+        type=Path,
+        help="Output directory where generated corpus files are written.",
+    )
+    parser.add_argument(
+        "--sv-root",
+        type=Path,
+        default=(Path(__file__).resolve().parent.parent / "suricata-verify"),
+        help="Path to suricata-verify root (default: ../suricata-verify).",
+    )
+    return parser.parse_args()
+
+
+def sanitize_name(name: str) -> str:
+    sanitized = re.sub(r"[^A-Za-z0-9_.-]", "_", name)
+    return sanitized or "unnamed_test"
+
+
+def normalize_argument(raw: str) -> str:
+    return re.sub(r"\s+", " ", raw).strip() + ";"
+
+
+def maybe_load_dataset_bytes(rules_file: Path, argument: str) -> bytes:
+    match = DATASET_FILE_RE.search(argument)
+    if not match:
+        return b""
+
+    dataset_name = match.group(2)
+    dataset_path = (rules_file.parent / dataset_name).resolve()
+    try:
+        return dataset_path.read_bytes()
+    except OSError:
+        # Some tests use paths that are not intended to exist. Keep corpus valid.
+        return b""
+
+
+def replace_dataset_filename(argument: str) -> str:
+    def _repl(match: re.Match) -> str:
+        return f"{match.group(1)} dataset.fuzz"
+
+    return DATASET_FILE_RE.sub(_repl, argument)
+
+
+def iter_test_rule_files(sv_root: Path):
+    tests_root = sv_root / "tests"
+    if not tests_root.is_dir():
+        raise FileNotFoundError(f"suricata-verify tests directory not found: {tests_root}")
+
+    by_test_dir = {}
+    for rules_file in sorted(tests_root.rglob("*.rules")):
+        by_test_dir.setdefault(rules_file.parent, []).append(rules_file)
+
+    for test_dir in sorted(by_test_dir):
+        yield test_dir, sorted(by_test_dir[test_dir])
+
+
+def main() -> int:
+    args = parse_args()
+
+    output_dir = args.output
+    sv_root = args.sv_root.resolve()
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+
+    generated = 0
+    for test_dir, rule_files in iter_test_rule_files(sv_root):
+        entries = []
+        for rules_file in rule_files:
+            try:
+                rules_text = rules_file.read_text(encoding="utf-8", errors="replace")
+            except OSError:
+                continue
+
+            for keyword, raw_argument in KEYWORD_RE.findall(rules_text):
+                argument = normalize_argument(raw_argument)
+                argument = replace_dataset_filename(argument)
+                prefix = "A" if keyword.lower() == "datarep" else "B"
+                dataset_bytes = maybe_load_dataset_bytes(rules_file, raw_argument)
+
+                payload = prefix.encode("ascii") + argument.encode("utf-8") + dataset_bytes
+                entries.append(payload)
+
+        if not entries:
+            continue
+
+        base_name = sanitize_name(test_dir.name)
+        if len(entries) == 1:
+            out_path = output_dir / base_name
+            out_path.write_bytes(entries[0])
+            generated += 1
+            continue
+
+        for idx, payload in enumerate(entries, start=1):
+            out_path = output_dir / f"{base_name}.{idx}"
+            out_path.write_bytes(payload)
+            generated += 1
+
+    print(f"Generated {generated} corpus files in {output_dir}")
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())
\ No newline at end of file