writer: Reject empty xattr names

Came out of a related discussion in ostree ostreedev/ostree#3346 (comment) Today we faithfully write this, and `fsck.erofs` doesn't seem to complain...but such a thing makes no sense so we shouldn't support it. Actually digging in more here...in practice the way xattrs work in Linux at least is they're restricted to starting with one of the standard prefixes like `security.` or `user.`; but enforcing/validating that is a whole other thing. For now let's just deny this edge case. Signed-off-by: Colin Walters <walters@verbum.org>
containers · Dec 2, 2024 · 02077e8 · 02077e8
1 parent d62aec4
commit 02077e8
Show file tree

Hide file tree

Showing 3 changed files with 4 additions and 1 deletion.
diff --git a/libcomposefs/lcfs-writer.c b/libcomposefs/lcfs-writer.c
@@ -1691,7 +1691,7 @@ int lcfs_node_set_xattr_internal(struct lcfs_node_s *node, const char *name,
 	char *k, *v;
 
 	const size_t namelen = strlen(name);
-	if (namelen > XATTR_NAME_MAX) {
+	if (namelen == 0 || namelen > XATTR_NAME_MAX) {
 		errno = ERANGE;
 		return -1;
 	}

diff --git a/tests/assets/should-fail-empty-xattr-key.dump b/tests/assets/should-fail-empty-xattr-key.dump
@@ -0,0 +1,2 @@
+/ 4096 40555 2 0 0 0 1633950376.0 - - -
+/noxattr-key 0 100644 1 0 0 0 0.0 - - - =somevalue
diff --git a/tests/meson.build b/tests/meson.build
@@ -21,6 +21,7 @@ test_assets_small_extra = [
 
 test_assets_should_fail = [
   'should-fail-long-link.dump',
+  'should-fail-empty-xattr-key.dump',
   'should-fail-long-xattr-key.dump',
   'should-fail-long-xattr-value.dump',
   'should-fail-honggfuzz-long-xattr.dump',
Original file line number	Diff line number	Diff line change
		@@ -0,0 +1,2 @@
		/ 4096 40555 2 0 0 0 1633950376.0 - - -
		/noxattr-key 0 100644 1 0 0 0 0.0 - - - =somevalue