Add pg_encoding_set_invalid()

There are cases where we cannot / do not want to error out for invalidly
encoded input. In such cases it can be useful to replace e.g. an incomplete
multi-byte characters with bytes that will trigger an error when getting
validated as part of a larger string.

Unfortunately, until now, for some encoding no such sequence existed. For
those encodings this commit removes one previously accepted input combination
- we consider that to be ok, as the chosen bytes are outside of the valid
ranges for the encodings, we just previously failed to detect that.

As we cannot add a new field to pg_wchar_table without breaking ABI, this is
implemented "in-line" in the newly added function.

Author: Noah Misch <noah@leadboat.com>
Reviewed-by: Andres Freund <andres@anarazel.de>
Backpatch-through: 13
Security: CVE-2025-1094

src/common/wchar.c

@@ -16,6 +16,25 @@
 #include "utils/ascii.h"
 
 
+/*
+ * In today's multibyte encodings other than UTF8, this two-byte sequence
+ * ensures pg_encoding_mblen() == 2 && pg_encoding_verifymbstr() == 0.
+ *
+ * For historical reasons, several verifychar implementations opt to reject
+ * this pair specifically.  Byte pair range constraints, in encoding
+ * originator documentation, always excluded this pair.  No core conversion
+ * could translate it.  However, longstanding verifychar implementations
+ * accepted any non-NUL byte.  big5_to_euc_tw and big5_to_mic even translate
+ * pairs not valid per encoding originator documentation.  To avoid tightening
+ * core or non-core conversions in a security patch, we sought this one pair.
+ *
+ * PQescapeString() historically used spaces for BYTE1; many other values
+ * could suffice for BYTE1.
+ */
+#define NONUTF8_INVALID_BYTE0 (0x8d)
+#define NONUTF8_INVALID_BYTE1 (' ')
+
+
 /*
  * Operations on multi-byte encodings are driven by a table of helper
  * functions.
@@ -1526,6 +1545,11 @@ pg_big5_verifychar(const unsigned char *s, int len)
 	if (len < l)
 		return -1;
 
+	if (l == 2 &&
+		s[0] == NONUTF8_INVALID_BYTE0 &&
+		s[1] == NONUTF8_INVALID_BYTE1)
+		return -1;
+
 	while (--l > 0)
 	{
 		if (*++s == '\0')
@@ -1575,6 +1599,11 @@ pg_gbk_verifychar(const unsigned char *s, int len)
 	if (len < l)
 		return -1;
 
+	if (l == 2 &&
+		s[0] == NONUTF8_INVALID_BYTE0 &&
+		s[1] == NONUTF8_INVALID_BYTE1)
+		return -1;
+
 	while (--l > 0)
 	{
 		if (*++s == '\0')
@@ -1624,6 +1653,11 @@ pg_uhc_verifychar(const unsigned char *s, int len)
 	if (len < l)
 		return -1;
 
+	if (l == 2 &&
+		s[0] == NONUTF8_INVALID_BYTE0 &&
+		s[1] == NONUTF8_INVALID_BYTE1)
+		return -1;
+
 	while (--l > 0)
 	{
 		if (*++s == '\0')
@@ -2067,6 +2101,19 @@ pg_utf8_islegal(const unsigned char *source, int length)
 }
 
 
+/*
+ * Fills the provided buffer with two bytes such that:
+ *   pg_encoding_mblen(dst) == 2 && pg_encoding_verifymbstr(dst) == 0
+ */
+void
+pg_encoding_set_invalid(int encoding, char *dst)
+{
+	Assert(pg_encoding_max_length(encoding) > 1);
+
+	dst[0] = (encoding == PG_UTF8 ? 0xc0 : NONUTF8_INVALID_BYTE0);
+	dst[1] = NONUTF8_INVALID_BYTE1;
+}
+
 /*
  *-------------------------------------------------------------------
  * encoding info table
@@ -2189,5 +2236,11 @@ pg_encoding_max_length(int encoding)
 {
 	Assert(PG_VALID_ENCODING(encoding));
 
-	return pg_wchar_table[encoding].maxmblen;
+	/*
+	 * Check for the encoding despite the assert, due to some mingw versions
+	 * otherwise issuing bogus warnings.
+	 */
+	return PG_VALID_ENCODING(encoding) ?
+		pg_wchar_table[encoding].maxmblen :
+		pg_wchar_table[PG_SQL_ASCII].maxmblen;
 }

src/include/mb/pg_wchar.h

@@ -573,6 +573,7 @@ extern int	pg_valid_server_encoding_id(int encoding);
  * (in addition to the ones just above).  The constant tables declared
  * earlier in this file are also available from libpgcommon.
  */
+extern void pg_encoding_set_invalid(int encoding, char *dst);
 extern int	pg_encoding_mblen(int encoding, const char *mbstr);
 extern int	pg_encoding_mblen_bounded(int encoding, const char *mbstr);
 extern int	pg_encoding_dsplen(int encoding, const char *mbstr);

src/test/regress/expected/conversion.out

@@ -5,6 +5,13 @@
 \getenv libdir PG_LIBDIR
 \getenv dlsuffix PG_DLSUFFIX
 \set regresslib :libdir '/regress' :dlsuffix
+CREATE FUNCTION test_enc_setup() RETURNS void
+    AS :'regresslib', 'test_enc_setup'
+    LANGUAGE C STRICT;
+SELECT FROM test_enc_setup();
+--
+(1 row)
+
 CREATE FUNCTION test_enc_conversion(bytea, name, name, bool, validlen OUT int, result OUT bytea)
     AS :'regresslib', 'test_enc_conversion'
     LANGUAGE C STRICT;

src/test/regress/regress.c

@@ -1099,6 +1099,56 @@ test_opclass_options_func(PG_FUNCTION_ARGS)
 	PG_RETURN_NULL();
 }
 
+/* one-time tests for encoding infrastructure */
+PG_FUNCTION_INFO_V1(test_enc_setup);
+Datum
+test_enc_setup(PG_FUNCTION_ARGS)
+{
+	/* Test pg_encoding_set_invalid() */
+	for (int i = 0; i < _PG_LAST_ENCODING_; i++)
+	{
+		char		buf[2],
+					bigbuf[16];
+		int			len,
+					mblen,
+					valid;
+
+		if (pg_encoding_max_length(i) == 1)
+			continue;
+		pg_encoding_set_invalid(i, buf);
+		len = strnlen(buf, 2);
+		if (len != 2)
+			elog(WARNING,
+				 "official invalid string for encoding \"%s\" has length %d",
+				 pg_enc2name_tbl[i].name, len);
+		mblen = pg_encoding_mblen(i, buf);
+		if (mblen != 2)
+			elog(WARNING,
+				 "official invalid string for encoding \"%s\" has mblen %d",
+				 pg_enc2name_tbl[i].name, mblen);
+		valid = pg_encoding_verifymbstr(i, buf, len);
+		if (valid != 0)
+			elog(WARNING,
+				 "official invalid string for encoding \"%s\" has valid prefix of length %d",
+				 pg_enc2name_tbl[i].name, valid);
+		valid = pg_encoding_verifymbstr(i, buf, 1);
+		if (valid != 0)
+			elog(WARNING,
+				 "first byte of official invalid string for encoding \"%s\" has valid prefix of length %d",
+				 pg_enc2name_tbl[i].name, valid);
+		memset(bigbuf, ' ', sizeof(bigbuf));
+		bigbuf[0] = buf[0];
+		bigbuf[1] = buf[1];
+		valid = pg_encoding_verifymbstr(i, bigbuf, sizeof(bigbuf));
+		if (valid != 0)
+			elog(WARNING,
+				 "trailing data changed official invalid string for encoding \"%s\" to have valid prefix of length %d",
+				 pg_enc2name_tbl[i].name, valid);
+	}
+
+	PG_RETURN_VOID();
+}
+
 /*
  * Call an encoding conversion or verification function.
  *

src/test/regress/sql/conversion.sql

@@ -8,6 +8,11 @@
 
 \set regresslib :libdir '/regress' :dlsuffix
 
+CREATE FUNCTION test_enc_setup() RETURNS void
+    AS :'regresslib', 'test_enc_setup'
+    LANGUAGE C STRICT;
+SELECT FROM test_enc_setup();
+
 CREATE FUNCTION test_enc_conversion(bytea, name, name, bool, validlen OUT int, result OUT bytea)
     AS :'regresslib', 'test_enc_conversion'
     LANGUAGE C STRICT;