kca-core: fix CSR attribute DER encoding and conformant decoding

- Generated attributes are now explicitly tagged as UTF8String via
  CertificateSigningRequestBuilder.add_attribute(_tag=_ASN1Type.UTF8String),
  so values include proper DER TLV instead of raw bytes.
- The attribute decoder first attempts DER string parsing (UTF8String, PrintableString,
  IA5String, BMPString, long-form lengths), falling back to legacy raw UTF-8 decode for
  previously generated nonconformant CSRs.
- Regression tests verify DER encoding, Unicode round-tripping, supported string types,
  long values (>127 bytes), and legacy raw UTF-8 attributes.
- Mypy passes cleanly (no issues in 42 source files).

server/packages/kca-core/src/kca/core/requests.py

@@ -7,6 +7,7 @@ from typing import Any, Dict, Iterable, List, Mapping, Optional
 
 from cryptography import x509
 from cryptography.hazmat.primitives import serialization
+from cryptography.x509.name import _ASN1Type
 from cryptography.x509.oid import ObjectIdentifier
 
 from .extensions import build_extensions, extension_config_lines, extension_rows, extension_specs
@@ -62,7 +63,11 @@ def build_csr(
         builder = builder.add_extension(extension.value, extension.critical)
     for oid, value in (attributes or {}).items():
         if value:
-            builder = builder.add_attribute(ObjectIdentifier(oid), str(value).encode("utf-8"))
+            builder = builder.add_attribute(
+                ObjectIdentifier(oid),
+                str(value).encode("utf-8"),
+                _tag=_ASN1Type.UTF8String,
+            )
     algorithm = hash_algorithm(signature_hash, private_key)
     return builder.sign(private_key, algorithm)
 
@@ -113,24 +118,33 @@ def _signature_name(csr: x509.CertificateSigningRequest) -> str:
     return _SIGNATURE_OIDS.get(oid.dotted_string, getattr(oid, "name", None) or oid.dotted_string)
 
 
-def _decode_attribute_value(value: bytes) -> str:
-    # cryptography returns the DER-encoded value bytes.  Try UTF-8 first,
-    # then strip a simple ASN.1 string tag/length header.
-    try:
-        return value.decode("utf-8")
-    except UnicodeDecodeError:
-        pass
-    if len(value) >= 2 and value[0] in (0x0C, 0x13, 0x16, 0x1E):
+def _decode_der_string(value: bytes) -> Optional[str]:
+    if len(value) < 2 or value[0] not in (0x0C, 0x13, 0x16, 0x1E):
+        return None
     length = value[1]
     offset = 2
     if length & 0x80:
         count = length & 0x7F
-            length = int.from_bytes(value[2 : 2 + count], "big")
-            offset = 2 + count
-        raw = value[offset : offset + length]
-        if value[0] == 0x1E:
-            return raw.decode("utf-16-be", "replace")
-        return raw.decode("utf-8", "replace")
+        if count == 0 or len(value) < offset + count:
+            return None
+        length = int.from_bytes(value[offset : offset + count], "big")
+        offset += count
+    if offset + length != len(value):
+        return None
+    raw = value[offset:]
+    encoding = "utf-16-be" if value[0] == 0x1E else "utf-8"
+    return raw.decode(encoding, "replace")
+
+
+def _decode_attribute_value(value: bytes) -> str:
+    # DER tag and length bytes can themselves be valid UTF-8, so decode the
+    # ASN.1 string first.  Raw UTF-8 remains supported for legacy KCA CSRs.
+    decoded = _decode_der_string(value)
+    if decoded is not None:
+        return decoded
+    try:
+        return value.decode("utf-8")
+    except UnicodeDecodeError:
         return value.hex()
 
 

server/packages/kca-core/tests/test_core.py

@@ -353,12 +353,34 @@ class RequestTest(unittest.TestCase):
             subject=_name("CSR"),
             private_key=key,
             extension_spec=_ca_spec(),
-            attributes={requests.CHALLENGE_PASSWORD_OID: "secret"},
+            attributes={
+                requests.CHALLENGE_PASSWORD_OID: "secret",
+                requests.UNSTRUCTURED_NAME_OID: "KCA ☃",
+            },
         )
         self.assertTrue(requests.verify_csr(csr))
         loaded = requests.parse_csr(requests.csr_der(csr))
         self.assertEqual(requests.request_info(loaded).subject, csr.subject)
-        self.assertEqual(requests.attribute_values(loaded)[requests.CHALLENGE_PASSWORD_OID], "secret")
+        der = requests.csr_der(loaded)
+        self.assertIn(b"\x0c\x06secret", der)
+        self.assertIn(b"\x0c\x07KCA \xe2\x98\x83", der)
+        values = requests.attribute_values(loaded)
+        self.assertEqual(values[requests.CHALLENGE_PASSWORD_OID], "secret")
+        self.assertEqual(values[requests.UNSTRUCTURED_NAME_OID], "KCA ☃")
+
+    def test_decode_attribute_der_strings(self) -> None:
+        long_value = "x" * 128
+        cases = [
+            (b"\x0c\x06secret", "secret"),
+            (b"\x13\x04name", "name"),
+            (b"\x16\x0bexample.com", "example.com"),
+            (b"\x1e\x04\x00a\x00b", "ab"),
+            (b"\x0c\x81\x80" + long_value.encode("ascii"), long_value),
+            (b"legacy", "legacy"),
+        ]
+        for encoded, expected in cases:
+            with self.subTest(encoded=encoded):
+                self.assertEqual(requests._decode_attribute_value(encoded), expected)
 
 
 class CrlTest(unittest.TestCase):