uuid: relax UUID value validation

This patch completely relaxes UUID checks and accepts an arbitrary
128-bit sequence as an UUID for binary data. String representations
still should match the grammars in RFC 4122, Section 3 [1] and RFC 9562,
Section 4 [2].

[1]: https://datatracker.ietf.org/doc/html/rfc4122#section-3
[2]: https://datatracker.ietf.org/doc/html/rfc9562#name-uuid-format

Closes tarantool#5444

@TarantoolBot document
Title: uuid: relaxed UUID validation

[The UUID module documentation][1] mentions that Tarantool generates
UUIDs following the rules for RFC 4122,[version 4, variant 1][2]. It is
worth mentioning that the user can store an arbitrary 128-bit sequence
as an UUID for binary data. String representations still should match
the grammars in RFC 4122, [Section 3][3], and RFC 9562, [Section 4][4].

[1]: https://www.tarantool.io/en/doc/latest/reference/reference_lua/uuid/
[2]: https://en.wikipedia.org/wiki/Universally_unique_identifier#Version_4_(random)
[3]: https://datatracker.ietf.org/doc/html/rfc4122#section-3
[4]: https://datatracker.ietf.org/doc/html/rfc9562#name-uuid-format

changelogs/unreleased/gh-5444-relax-uuid-validation.md

@@ -0,0 +1,4 @@
+## feature/core
+
+* Any 128-byte sequence is considered as a valid UUID value to support all
+  RFC 9562 UUID versions (gh-5444).

src/box/sql/mem.c

@@ -812,8 +812,7 @@ static inline int
 bin_to_uuid(struct Mem *mem)
 {
 	assert(mem->type == MEM_TYPE_BIN);
-	if (mem->n != UUID_LEN ||
-	    tt_uuid_validate((struct tt_uuid *)mem->z) != 0)
+	if (mem->n != UUID_LEN)
 		return -1;
 	mem_set_uuid(mem, (struct tt_uuid *)mem->z);
 	return 0;

src/lib/core/mp_uuid.c

@@ -55,7 +55,6 @@ uuid_pack(char *data, const struct tt_uuid *uuid)
 struct tt_uuid *
 uuid_unpack(const char **data, uint32_t len, struct tt_uuid *uuid)
 {
-	const char *const svp = *data;
 	if (len != UUID_PACKED_LEN)
 		return NULL;
 	uuid->time_low = mp_load_u32(data);
@@ -66,10 +65,6 @@ uuid_unpack(const char **data, uint32_t len, struct tt_uuid *uuid)
 	for (int i = 0; i < 6; i++)
 		uuid->node[i] = mp_load_u8(data);
 
-	if (tt_uuid_validate(uuid) != 0) {
-		*data = svp;
-		return NULL;
-	}
 	return uuid;
 }
 

src/lib/core/tt_uuid.c

@@ -65,9 +65,6 @@ tt_uuid_create(struct tt_uuid *uu)
 }
 #endif
 
-extern inline int
-tt_uuid_validate(struct tt_uuid *uu);
-
 extern inline int
 tt_uuid_from_string(const char *in, struct tt_uuid *uu);
 

src/lib/core/tt_uuid.h

@@ -62,16 +62,6 @@ struct tt_uuid {
 void
 tt_uuid_create(struct tt_uuid *uu);
 
-inline int
-tt_uuid_validate(struct tt_uuid *uu)
-{
-	/* Check variant (NCS, RFC4122, MSFT) */
-	uint8_t n = uu->clock_seq_hi_and_reserved;
-	if ((n & 0x80) != 0x00 && (n & 0xc0) != 0x80 &&	(n & 0xe0) != 0xc0)
-		return 1;
-	return 0;
-}
-
 /**
  * \brief Parse UUID from string.
  * \param in string
@@ -88,8 +78,7 @@ tt_uuid_from_string(const char *in, struct tt_uuid *uu)
 		   &uu->node[0], &uu->node[1], &uu->node[2], &uu->node[3],
 		   &uu->node[4], &uu->node[5]) != 11)
 		return 1;
-
-	return tt_uuid_validate(uu);
+	return 0;
 }
 
 /**

test/app-luatest/gh_5444_relax_uuid_validation_test.lua

@@ -0,0 +1,46 @@
+local uuid = require('uuid')
+local msgpack = require('msgpack')
+local t = require('luatest')
+
+local group_versions = t.group('group_versions', {
+    -- UUID examples taken from the RFC 9562 edition:
+    -- https://datatracker.ietf.org/doc/html/rfc9562#name-test-vectors
+
+    -- The v2 implementation is not specified. Thus, it is
+    -- omitted here.
+    {version = 'v1', uuid = 'c232ab00-9414-11ec-b3c8-9f6bdeced846'},
+    {version = 'v3', uuid = '5df41881-3aed-3515-88a7-2f4a814cf09e'},
+    {version = 'v4', uuid = '919108f7-52d1-4320-9bac-f847db4148a8'},
+    {version = 'v5', uuid = '2ed6657d-e927-568b-95e1-2665a8aea6a2'},
+    {version = 'v6', uuid = '1ec9414c-232a-6b00-b3c8-9f6bdeced846'},
+    {version = 'v7', uuid = '017f22e2-79b0-7cc3-98c4-dc0c0c07398f'},
+    {version = 'v8_time_based', uuid = '2489e9ad-2ee2-8e00-8ec9-32d5f69181c0'},
+    {version = 'v8_name_based', uuid = '5c146b14-3c52-8afd-938a-375d0df1fbf6'},
+})
+
+-- Base test.
+group_versions.test_version = function(cg)
+    local version = cg.params.version
+    local u = cg.params.uuid
+    t.assert(uuid.fromstr(u),
+             ('UUID %s value (%s) failed to parse'):format(version, u))
+end
+
+local group_gh_5444 = t.group('group_gh_5444', {
+    {uuid = 'bea80698-e07d-11ea-fe85-00155d373b0c'},
+    {uuid = 'bee815b2-e07d-11ea-fe85-00155d373b0c'},
+    {uuid = '4429d312-18d4-11eb-94f6-77e22d44915a'},
+    {uuid = '64ecacb4-18d4-11eb-94f6-cbb989f20a7e'},
+    {uuid = '64ecacb5-18d4-11eb-94f6-cf778123fd70'},
+    {uuid = '00000000-0000-0000-e000-000000000000'},
+    {uuid = '98c0cfc3-2b03-461d-b662-9a869bf46c75'},
+})
+
+-- Test all examples from gh-5444.
+group_gh_5444.test_all_examples = function(cg)
+    local u = cg.params.uuid
+    local uuid_cdata = uuid.fromstr(u)
+    t.assert(uuid_cdata, ('UUID value (%s) failed to parse'):format(u))
+    t.assert(msgpack.decode(msgpack.encode(uuid_cdata)),
+             ('UUID value (%s) failed to be decoded in msgpack'):format(u))
+end

tools/tarantool-gdb.py

@@ -544,8 +544,6 @@ def decode_uuid(cls, data, len): # uuid_unpack
             clock_seq_low = data.read_u8(),
             node = [ data.read_u8() for _ in range(0, 6) ],
         ))
-        if not uuid.is_valid():
-            return uuid, "uuid_unpack: invalid uuid"
         return uuid, None
 
     @classmethod
@@ -998,12 +996,6 @@ class Uuid:
         def __init__(self, val):
             self.val = val
 
-        def is_valid(self): # tt_uuid_validate
-            n = self.val['clock_seq_hi_and_reserved']
-            if (n & 0x80) != 0x00 and (n & 0xc0) != 0x80 and (n & 0xe0) != 0xc0:
-                return False
-            return True
-
         def __str__(self): # tt_uuid_to_string
             return '{:08x}-{:04x}-{:04x}-{:02x}{:02x}-{:02x}{:02x}{:02x}{:02x}{:02x}{:02x}'.format(
                 self.val['time_low'], self.val['time_mid'], self.val['time_hi_and_version'],