libpq: reject extraneous data after SSL or GSS encryption handshake.

libpq collects up to a bufferload of data whenever it reads data from
the socket.  When SSL or GSS encryption is requested during startup,
any additional data received with the server's yes-or-no reply
remained in the buffer, and would be treated as already-decrypted data
once the encryption handshake completed.  Thus, a man-in-the-middle
with the ability to inject data into the TCP connection could stuff
some cleartext data into the start of a supposedly encryption-protected
database session.

This could probably be abused to inject faked responses to the
client's first few queries, although other details of libpq's behavior
make that harder than it sounds.  A different line of attack is to
exfiltrate the client's password, or other sensitive data that might
be sent early in the session.  That has been shown to be possible with
a server vulnerable to CVE-2021-23214.

To fix, throw a protocol-violation error if the internal buffer
is not empty after the encryption handshake.

Our thanks to Jacob Champion for reporting this problem.

Security: CVE-2021-23222

doc/src/sgml/protocol.sgml

@@ -1477,6 +1477,20 @@ SELCT 1/0;<!-- this typo is intentional -->
     and proceed without requesting <acronym>SSL</acronym>.
    </para>
 
+   <para>
+    When <acronym>SSL</acronym> encryption can be performed, the server
+    is expected to send only the single <literal>S</literal> byte and then
+    wait for the frontend to initiate an <acronym>SSL</acronym> handshake.
+    If additional bytes are available to read at this point, it likely
+    means that a man-in-the-middle is attempting to perform a
+    buffer-stuffing attack
+    (<ulink url="https://www.postgresql.org/support/security/CVE-2021-23222/">CVE-2021-23222</ulink>).
+    Frontends should be coded either to read exactly one byte from the
+    socket before turning the socket over to their SSL library, or to
+    treat it as a protocol violation if they find they have read additional
+    bytes.
+   </para>
+
    <para>
     An initial SSLRequest can also be used in a connection that is being
     opened to send a CancelRequest message.
@@ -1539,6 +1553,20 @@ SELCT 1/0;<!-- this typo is intentional -->
     encryption.
    </para>
 
+   <para>
+    When <acronym>GSSAPI</acronym> encryption can be performed, the server
+    is expected to send only the single <literal>G</literal> byte and then
+    wait for the frontend to initiate a <acronym>GSSAPI</acronym> handshake.
+    If additional bytes are available to read at this point, it likely
+    means that a man-in-the-middle is attempting to perform a
+    buffer-stuffing attack
+    (<ulink url="https://www.postgresql.org/support/security/CVE-2021-23222/">CVE-2021-23222</ulink>).
+    Frontends should be coded either to read exactly one byte from the
+    socket before turning the socket over to their GSSAPI library, or to
+    treat it as a protocol violation if they find they have read additional
+    bytes.
+   </para>
+
    <para>
     An initial GSSENCRequest can also be used in a connection that is being
     opened to send a CancelRequest message.

src/interfaces/libpq/fe-connect.c

@@ -3097,6 +3097,19 @@ keep_going:						/* We will come back to here until there is
 				pollres = pqsecure_open_client(conn);
 				if (pollres == PGRES_POLLING_OK)
 				{
+					/*
+					 * At this point we should have no data already buffered.
+					 * If we do, it was received before we performed the SSL
+					 * handshake, so it wasn't encrypted and indeed may have
+					 * been injected by a man-in-the-middle.
+					 */
+					if (conn->inCursor != conn->inEnd)
+					{
+						appendPQExpBufferStr(&conn->errorMessage,
+											 libpq_gettext("received unencrypted data after SSL response\n"));
+						goto error_return;
+					}
+
 					/* SSL handshake done, ready to send startup packet */
 					conn->status = CONNECTION_MADE;
 					return PGRES_POLLING_WRITING;
@@ -3196,6 +3209,19 @@ keep_going:						/* We will come back to here until there is
 				pollres = pqsecure_open_gss(conn);
 				if (pollres == PGRES_POLLING_OK)
 				{
+					/*
+					 * At this point we should have no data already buffered.
+					 * If we do, it was received before we performed the GSS
+					 * handshake, so it wasn't encrypted and indeed may have
+					 * been injected by a man-in-the-middle.
+					 */
+					if (conn->inCursor != conn->inEnd)
+					{
+						appendPQExpBufferStr(&conn->errorMessage,
+											 libpq_gettext("received unencrypted data after GSSAPI encryption response\n"));
+						goto error_return;
+					}
+
 					/* All set for startup packet */
 					conn->status = CONNECTION_MADE;
 					return PGRES_POLLING_WRITING;