🛡️ Sentinel: [CRITICAL/HIGH] Fix path traversal in module resolution

.jules/sentinel.md

@@ -0,0 +1,4 @@
+## 2024-06-06 - Path Traversal in TS Extractor
+**Vulnerability:** A path traversal vulnerability in `crates/flow/src/incremental/extractors/typescript.rs` during manual module path resolution where popping `std::path::Component::ParentDir` could allow an attacker to traverse to restricted directories or files when canonicalize fails.
+**Learning:** During manual canonicalization of path components, ignoring `..` entirely when popping isn't sufficient. If the path navigates up from the root or prefix, popping an empty stack leads to bypassing of security boundaries. Moreover, relative paths like `../../a` get wrongly reduced.
+**Prevention:** Explicitly block `Component::ParentDir` from popping `Component::RootDir` or `Component::Prefix`. If the components list is empty or its last element is `Component::ParentDir`, the new `Component::ParentDir` must be pushed instead of ignored, to correctly preserve valid relative paths.

crates/ast-engine/src/tree_sitter/mod.rs

@@ -553,9 +553,8 @@ impl ContentExt for String {
         let mut bytes = std::mem::take(self).into_bytes();
         let original_len = bytes.len();
         bytes.splice(safe_start..safe_end, full_inserted);
-        *self = Self::from_utf8(bytes).unwrap_or_else(|e| {
-            Self::from_utf8_lossy(&e.into_bytes()).into_owned()
-        });
+        *self = Self::from_utf8(bytes)
+            .unwrap_or_else(|e| Self::from_utf8_lossy(&e.into_bytes()).into_owned());
 
         // We calculate new_end_byte using the difference in the new overall string length
         // to correctly align the end offset, taking any potential replacement bytes from
@@ -791,7 +790,10 @@ mod test {
 
         let tree2 = parse_lang(|p| p.parse(&src, Some(&tree)), &Tsx.get_ts_language())?;
         let fresh_tree = parse(&src)?;
-        assert_eq!(tree2.root_node().to_sexp(), fresh_tree.root_node().to_sexp());
+        assert_eq!(
+            tree2.root_node().to_sexp(),
+            fresh_tree.root_node().to_sexp()
+        );
         Ok(())
     }
 }

crates/flow/src/incremental/extractors/typescript.rs

@@ -808,7 +808,20 @@ impl TypeScriptDependencyExtractor {
                 for component in resolved.components() {
                     match component {
                         std::path::Component::ParentDir => {
-                            components.pop();
+                            // SECURITY: Prevent path traversal vulnerabilities by checking the last component
+                            // before popping. We shouldn't pop RootDir or Prefix components.
+                            // If the component list is empty or the last component is already ParentDir,
+                            // we must push ParentDir to preserve relative paths like "../../a".
+                            match components.last() {
+                                Some(std::path::Component::RootDir)
+                                | Some(std::path::Component::Prefix(_)) => {}
+                                Some(std::path::Component::ParentDir) | None => {
+                                    components.push(component)
+                                }
+                                _ => {
+                                    components.pop();
+                                }
+                            }
                         }
                         std::path::Component::CurDir => {}
                         _ => components.push(component),

crates/rule-engine/src/check_var.rs

@@ -104,9 +104,9 @@ fn get_vars_from_rules<'r>(rule: &'r Rule, utils: &'r RuleRegistration) -> Rapid
     vars
 }
 
-fn check_var_in_constraints<'r>(
+fn check_var_in_constraints(
     mut vars: RapidSet<String>,
-    constraints: &'r RapidMap<thread_ast_engine::meta_var::MetaVariableID, Rule>,
+    constraints: &RapidMap<thread_ast_engine::meta_var::MetaVariableID, Rule>,
 ) -> RResult<RapidSet<String>> {
     for rule in constraints.values() {
         for var in rule.defined_vars() {
@@ -125,9 +125,9 @@ fn check_var_in_constraints<'r>(
     Ok(vars)
 }
 
-fn check_var_in_transform<'r>(
+fn check_var_in_transform(
     mut vars: RapidSet<String>,
-    transform: &'r Option<Transform>,
+    transform: &Option<Transform>,
 ) -> RResult<RapidSet<String>> {
     let Some(transform) = transform else {
         return Ok(vars);

crates/rule-engine/src/rule/mod.rs

@@ -246,7 +246,11 @@ impl Rule {
 
     pub fn defined_vars(&self) -> RapidSet<String> {
         match self {
-            Rule::Pattern(p) => p.defined_vars().into_iter().map(|s| s.to_string()).collect(),
+            Rule::Pattern(p) => p
+                .defined_vars()
+                .into_iter()
+                .map(|s| s.to_string())
+                .collect(),
             Rule::Kind(_) => RapidSet::default(),
             Rule::Regex(_) => RapidSet::default(),
             Rule::NthChild(n) => n.defined_vars(),

crates/rule-engine/src/rule/referent_rule.rs

@@ -27,10 +27,7 @@ impl<R> Clone for Registration<R> {
 
 impl<R> Registration<R> {
     fn read(&self) -> Arc<RapidMap<String, R>> {
-        self.0
-            .read()
-            .unwrap_or_else(|e| e.into_inner())
-            .clone()
+        self.0.read().unwrap_or_else(|e| e.into_inner()).clone()
     }
     pub(crate) fn contains_key(&self, key: &str) -> bool {
         self.read().contains_key(key)