PROXY protocol support

Started by Magnus Haganderalmost 5 years ago48 messages
#1Magnus Hagander
magnus@hagander.net
1 attachment(s)

PFA a simple patch that implements support for the PROXY protocol.

This is a protocol common and very light weight in proxies and load
balancers (haproxy is one common example, but also for example the AWS
cloud load balancers). Basically this protocol prefixes the normal
connection with a header and a specification of what the original host
was, allowing the server to unwrap that and get the correct client
address instead of just the proxy ip address. It is a one-way protocol
in that there is no response from the server, it's just purely a
prefix of the IP information.

Using this when PostgreSQL is behind a proxy allows us to keep using
pg_hba.conf rules based on the original ip address, as well as track
the original address in log messages and pg_stat_activity etc.

The implementation adds a parameter named proxy_servers which lists
the ips or ip+cidr mask to be trusted. Since a proxy can decide what
the origin is, and this is used for security decisions, it's very
important to not just trust any server, only those that are
intentionally used. By default, no servers are listed, and thus the
protocol is disabled.

When specified, and the connection on the normal port has the proxy
prefix on it, and the connection comes in from one of the addresses
listed as valid proxy servers, we will replace the actual IP address
of the client with the one specified in the proxy packet.

Currently there is no information about the proxy server in the
pg_stat_activity view, it's only available as a log message. But maybe
it should go in pg_stat_activity as well? Or in a separate
pg_stat_proxy view?

(In passing, I note that pq_discardbytes were in pqcomm.h, yet listed
as static in pqcomm.c -- but now made non-static)

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..d4f6fad5b0 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-servers"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index b5718fc136..fc7de25378 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,30 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names or cidr specifications
+        of proxy servers to trust. If a connection using the PROXY protocol is made
+        from one of these IP addresses, <productname>PostgreSQL</productname> will
+        read the client IP address from the PROXY header and consider that the
+        address of the client, instead of listing all connections as coming from
+        the proxy server.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.  This parameter can only
+        be set at server start.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 27a298f110..9163761cc2 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -53,6 +53,7 @@
  *		pq_getmessage	- get a message with length word from connection
  *		pq_getbyte		- get next byte from connection
  *		pq_peekbyte		- peek at next byte from connection
+ *		pq_peekbytes    - peek at a known number of bytes from connection
  *		pq_putbytes		- send bytes to connection (not flushed until pq_flush)
  *		pq_flush		- flush pending output
  *		pq_flush_if_writable - flush pending output if writable without blocking
@@ -1039,6 +1040,27 @@ pq_peekbyte(void)
 	return (unsigned char) PqRecvBuffer[PqRecvPointer];
 }
 
+
+/* --------------------------------
+ *		pq_peekbytes		- peek at a known number of bytes from connection.
+ *							  Note! Does NOT wait for more data to arrive.
+ *
+ *		returns 0 if OK, EOF if trouble
+ * --------------------------------
+ */
+int
+pq_peekbytes(char *s, size_t len)
+{
+	Assert(PqCommReadingMsg);
+
+	if (PqRecvLength - PqRecvPointer < len)
+		return EOF;
+
+	memcpy(s, PqRecvBuffer + PqRecvPointer, len);
+
+	return 0;
+}
+
 /* --------------------------------
  *		pq_getbyte_if_available - get a single byte from connection,
  *			if available
@@ -1135,7 +1157,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 3f1ce135a8..0473129cb4 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -204,6 +205,10 @@ char	   *Unix_socket_directories;
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -1911,6 +1916,203 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		allowed = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, port->raddr.salen);
+
+	pq_startmsgread();
+
+	/* Peek at the very first byte just to trigger a read */
+	if (pq_peekbyte() == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete startup packet")));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * PROXY requests always start with: \x0D \x0A \x0D \x0A \x00 \x0D \x0A
+	 * \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_peekbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		/*
+		 * Not enough bytes to be a proxy header, so fall through to normal
+		 * processing
+		 */
+		pq_endmsgread();
+		return STATUS_OK;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		/*
+		 * Data is there but it wasn't a proxy header. Also fall through to
+		 * normal processing
+		 */
+		pq_endmsgread();
+		return STATUS_OK;
+	}
+
+	/* Header is valid. Verify that the proxy is actually authorized! */
+	for (i = 0; i < *((int *) TrustedProxyServers); i += 2)
+	{
+		if (raddr_save.addr.ss_family == TrustedProxyServers[i + 1].ss_family &&
+			pg_range_sockaddr(&raddr_save.addr,
+							  &TrustedProxyServers[i + 1],
+							  &TrustedProxyServers[i + 2]))
+		{
+			allowed = true;
+			break;
+		}
+	}
+	if (!allowed)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("proxy connection from unauthorized server")));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * This is a valid proxy header, so unwrap it. First, skip past the header
+	 * itself
+	 */
+	pq_discardbytes(sizeof(proxyheader));
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("oversized proxy packet")));
+		return STATUS_ERROR;
+	}
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+		pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));
+
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -4344,6 +4546,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (TrustedProxyServers)
+	{
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+	}
+
+	disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4395,28 +4624,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index d626731723..381067b737 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -4241,6 +4245,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12228,4 +12243,108 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index ee06528bb0..aa7ac35f67 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,8 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_servers = ''			# what IP/netmasks of proxy servers to trust
+					# (change requires restart)
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index e4e5c21565..6125b93e86 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -74,6 +74,8 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_peekbytes(char *s, size_t len);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putbytes(const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..38e7644371 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -21,6 +21,8 @@ extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
#2Arthur Nascimento
tureba@gmail.com
In reply to: Magnus Hagander (#1)
Re: PROXY protocol support

Hi,

On Tue, 2 Mar 2021 at 14:43, Magnus Hagander <magnus@hagander.net> wrote:

PFA a simple patch that implements support for the PROXY protocol.

Nice. I didn't know I needed this. But in hindsight, I would've used
it quite a few times in the past if I could have.

The implementation adds a parameter named proxy_servers which lists
the ips or ip+cidr mask to be trusted. Since a proxy can decide what
the origin is, and this is used for security decisions, it's very
important to not just trust any server, only those that are
intentionally used. By default, no servers are listed, and thus the
protocol is disabled.

Might make sense to add special cases for 'samehost' and 'samenet', as
in hba rules, as proxy servers are commonly on the same machine or
share one of the same internal networks.

Despite the security issues, I'm sure people will soon try and set
proxy_servers='*' or 'all' if they think this setting works as
listen_addresses or as pg_hba. But I don't think I'd make these use
cases easier.

Tureba - Arthur Nascimento

#3Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#1)
Re: PROXY protocol support

On Tue, 2021-03-02 at 18:43 +0100, Magnus Hagander wrote:

PFA a simple patch that implements support for the PROXY protocol.

I'm not all the way through the patch yet, but this part jumped out at
me:

+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		/*
+		 * Data is there but it wasn't a proxy header. Also fall through to
+		 * normal processing
+		 */
+		pq_endmsgread();
+		return STATUS_OK;

From my reading, the spec explicitly disallows this sort of fallback
behavior:

The receiver MUST be configured to only receive the protocol described in this
specification and MUST not try to guess whether the protocol header is present
or not. This means that the protocol explicitly prevents port sharing between
public and private access.

You might say, "if we already trust the proxy server, why should we
care?" but I think the point is that you want to catch
misconfigurations where the middlebox is forwarding bare TCP without
adding a PROXY header of its own, which will "work" for innocent
clients but in reality is a ticking timebomb. If you've decided to
trust an intermediary to use PROXY connections, then you must _only_
accept PROXY connections from that intermediary. Does that seem like a
reasonable interpretation?

--Jacob

#4Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#3)
1 attachment(s)
Re: PROXY protocol support

On Wed, Mar 3, 2021 at 1:50 AM Jacob Champion <pchampion@vmware.com> wrote:

On Tue, 2021-03-02 at 18:43 +0100, Magnus Hagander wrote:

PFA a simple patch that implements support for the PROXY protocol.

I'm not all the way through the patch yet, but this part jumped out at
me:

+     if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+     {
+             /*
+              * Data is there but it wasn't a proxy header. Also fall through to
+              * normal processing
+              */
+             pq_endmsgread();
+             return STATUS_OK;

From my reading, the spec explicitly disallows this sort of fallback
behavior:

The receiver MUST be configured to only receive the protocol described in this
specification and MUST not try to guess whether the protocol header is present
or not. This means that the protocol explicitly prevents port sharing between
public and private access.

You might say, "if we already trust the proxy server, why should we
care?" but I think the point is that you want to catch
misconfigurations where the middlebox is forwarding bare TCP without
adding a PROXY header of its own, which will "work" for innocent
clients but in reality is a ticking timebomb. If you've decided to
trust an intermediary to use PROXY connections, then you must _only_
accept PROXY connections from that intermediary. Does that seem like a
reasonable interpretation?

I definitely missed that part of the spec. Ugh.

That said, I'm not sure it's *actually* an issue in the case of
PostgreSQL. Given that doing what you're suggesting, accidentally
passing connections without PROXY, will get caught in pg_hba.conf.

That said, I agree with your interpretation, and it's pretty easy to
change it to that. Basically we just have to do the IP check *before*
doing the PROXY protocol check. It makes testing a bit more difficult
though, but maybe worth it?

I've attached a POC that does that. Note that I have *not* updated the docs!

Another option would of course be to listen on a separate port for it,
which seems to be the "haproxy way". That would be slightly more code
(we'd still want to keep the code for validating the list of trusted
proxies I'd say), but maybe worth doing?

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_only.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_only.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..d4f6fad5b0 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-servers"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index b5718fc136..fc7de25378 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,30 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names or cidr specifications
+        of proxy servers to trust. If a connection using the PROXY protocol is made
+        from one of these IP addresses, <productname>PostgreSQL</productname> will
+        read the client IP address from the PROXY header and consider that the
+        address of the client, instead of listing all connections as coming from
+        the proxy server.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.  This parameter can only
+        be set at server start.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 27a298f110..9163761cc2 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -53,6 +53,7 @@
  *		pq_getmessage	- get a message with length word from connection
  *		pq_getbyte		- get next byte from connection
  *		pq_peekbyte		- peek at next byte from connection
+ *		pq_peekbytes    - peek at a known number of bytes from connection
  *		pq_putbytes		- send bytes to connection (not flushed until pq_flush)
  *		pq_flush		- flush pending output
  *		pq_flush_if_writable - flush pending output if writable without blocking
@@ -1039,6 +1040,27 @@ pq_peekbyte(void)
 	return (unsigned char) PqRecvBuffer[PqRecvPointer];
 }
 
+
+/* --------------------------------
+ *		pq_peekbytes		- peek at a known number of bytes from connection.
+ *							  Note! Does NOT wait for more data to arrive.
+ *
+ *		returns 0 if OK, EOF if trouble
+ * --------------------------------
+ */
+int
+pq_peekbytes(char *s, size_t len)
+{
+	Assert(PqCommReadingMsg);
+
+	if (PqRecvLength - PqRecvPointer < len)
+		return EOF;
+
+	memcpy(s, PqRecvBuffer + PqRecvPointer, len);
+
+	return 0;
+}
+
 /* --------------------------------
  *		pq_getbyte_if_available - get a single byte from connection,
  *			if available
@@ -1135,7 +1157,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 3f1ce135a8..514ab72d40 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -204,6 +205,10 @@ char	   *Unix_socket_directories;
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -1911,6 +1916,184 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	for (i = 0; i < *((int *) TrustedProxyServers); i += 2)
+	{
+		if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family &&
+			pg_range_sockaddr(&port->raddr.addr,
+							  &TrustedProxyServers[i + 1],
+							  &TrustedProxyServers[i + 2]))
+		{
+			useproxy = true;
+			break;
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		return STATUS_OK;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, port->raddr.salen);
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with: \x0D \x0A \x0D \x0A \x00 \x0D \x0A
+	 * \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("oversized proxy packet")));
+		return STATUS_ERROR;
+	}
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+		pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));
+
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -4344,6 +4527,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (TrustedProxyServers)
+	{
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+	}
+
+	disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4395,28 +4605,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index d626731723..381067b737 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -4241,6 +4245,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12228,4 +12243,108 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index ee06528bb0..aa7ac35f67 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,8 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_servers = ''			# what IP/netmasks of proxy servers to trust
+					# (change requires restart)
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index e4e5c21565..6125b93e86 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -74,6 +74,8 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_peekbytes(char *s, size_t len);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putbytes(const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..38e7644371 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -21,6 +21,8 @@ extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
#5Magnus Hagander
magnus@hagander.net
In reply to: Magnus Hagander (#4)
1 attachment(s)
Re: PROXY protocol support

On Wed, Mar 3, 2021 at 10:00 AM Magnus Hagander <magnus@hagander.net> wrote:

On Wed, Mar 3, 2021 at 1:50 AM Jacob Champion <pchampion@vmware.com> wrote:

On Tue, 2021-03-02 at 18:43 +0100, Magnus Hagander wrote:

PFA a simple patch that implements support for the PROXY protocol.

I'm not all the way through the patch yet, but this part jumped out at
me:

+     if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+     {
+             /*
+              * Data is there but it wasn't a proxy header. Also fall through to
+              * normal processing
+              */
+             pq_endmsgread();
+             return STATUS_OK;

From my reading, the spec explicitly disallows this sort of fallback
behavior:

The receiver MUST be configured to only receive the protocol described in this
specification and MUST not try to guess whether the protocol header is present
or not. This means that the protocol explicitly prevents port sharing between
public and private access.

You might say, "if we already trust the proxy server, why should we
care?" but I think the point is that you want to catch
misconfigurations where the middlebox is forwarding bare TCP without
adding a PROXY header of its own, which will "work" for innocent
clients but in reality is a ticking timebomb. If you've decided to
trust an intermediary to use PROXY connections, then you must _only_
accept PROXY connections from that intermediary. Does that seem like a
reasonable interpretation?

I definitely missed that part of the spec. Ugh.

That said, I'm not sure it's *actually* an issue in the case of
PostgreSQL. Given that doing what you're suggesting, accidentally
passing connections without PROXY, will get caught in pg_hba.conf.

That said, I agree with your interpretation, and it's pretty easy to
change it to that. Basically we just have to do the IP check *before*
doing the PROXY protocol check. It makes testing a bit more difficult
though, but maybe worth it?

I've attached a POC that does that. Note that I have *not* updated the docs!

Another option would of course be to listen on a separate port for it,
which seems to be the "haproxy way". That would be slightly more code
(we'd still want to keep the code for validating the list of trusted
proxies I'd say), but maybe worth doing?

In order to figure that out, I hacked up a poc on that. Once again
without updates to the docs, but shows approximately how much code
complexity it adds (not much).

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_separate_port.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_separate_port.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..d4f6fad5b0 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-servers"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index b5718fc136..fc7de25378 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,30 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names or cidr specifications
+        of proxy servers to trust. If a connection using the PROXY protocol is made
+        from one of these IP addresses, <productname>PostgreSQL</productname> will
+        read the client IP address from the PROXY header and consider that the
+        address of the client, instead of listing all connections as coming from
+        the proxy server.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.  This parameter can only
+        be set at server start.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 27a298f110..401f2d2464 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -53,6 +53,7 @@
  *		pq_getmessage	- get a message with length word from connection
  *		pq_getbyte		- get next byte from connection
  *		pq_peekbyte		- peek at next byte from connection
+ *		pq_peekbytes    - peek at a known number of bytes from connection
  *		pq_putbytes		- send bytes to connection (not flushed until pq_flush)
  *		pq_flush		- flush pending output
  *		pq_flush_if_writable - flush pending output if writable without blocking
@@ -336,7 +337,7 @@ socket_close(int code, Datum arg)
 int
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 pgsocket ListenSocket[], bool ProxyList[], bool isProxy, int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -602,6 +603,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							familyDesc, addrDesc, (int) portNumber)));
 
 		ListenSocket[listen_index] = fd;
+		ProxyList[listen_index] = isProxy;
 		added++;
 	}
 
@@ -1039,6 +1041,27 @@ pq_peekbyte(void)
 	return (unsigned char) PqRecvBuffer[PqRecvPointer];
 }
 
+
+/* --------------------------------
+ *		pq_peekbytes		- peek at a known number of bytes from connection.
+ *							  Note! Does NOT wait for more data to arrive.
+ *
+ *		returns 0 if OK, EOF if trouble
+ * --------------------------------
+ */
+int
+pq_peekbytes(char *s, size_t len)
+{
+	Assert(PqCommReadingMsg);
+
+	if (PqRecvLength - PqRecvPointer < len)
+		return EOF;
+
+	memcpy(s, PqRecvBuffer + PqRecvPointer, len);
+
+	return 0;
+}
+
 /* --------------------------------
  *		pq_getbyte_if_available - get a single byte from connection,
  *			if available
@@ -1135,7 +1158,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 3f1ce135a8..5b1e4f5592 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -195,15 +196,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,6 +226,7 @@ int			ReservedBackends;
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
 static pgsocket ListenSocket[MAXLISTEN];
+static bool ListenSocketIsProxy[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -1124,7 +1133,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
+	{
 		ListenSocket[i] = PGINVALID_SOCKET;
+		ListenSocketIsProxy[i] = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1156,12 +1168,14 @@ PostmasterMain(int argc, char *argv[])
 				status = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
-										  ListenSocket, MAXLISTEN);
+										  ListenSocket, ListenSocketIsProxy,
+										  false, MAXLISTEN);
 			else
 				status = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
-										  ListenSocket, MAXLISTEN);
+										  ListenSocket, ListenSocketIsProxy,
+										  false, MAXLISTEN);
 
 			if (status == STATUS_OK)
 			{
@@ -1177,6 +1191,27 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					status = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, ListenSocketIsProxy,
+											  true, MAXLISTEN);
+				else
+					status = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, ListenSocketIsProxy,
+											  true, MAXLISTEN);
+				if (status != STATUS_OK)
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
 		if (!success && elemlist != NIL)
@@ -1254,7 +1289,8 @@ PostmasterMain(int argc, char *argv[])
 			status = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
-									  ListenSocket, MAXLISTEN);
+									  ListenSocket, ListenSocketIsProxy,
+									  false, MAXLISTEN);
 
 			if (status == STATUS_OK)
 			{
@@ -1731,6 +1767,8 @@ ServerLoop(void)
 					port = ConnCreate(ListenSocket[i]);
 					if (port)
 					{
+						port->isProxy = ListenSocketIsProxy[i];
+
 						BackendStartup(port);
 
 						/*
@@ -1911,6 +1949,190 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers); i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family &&
+				pg_range_sockaddr(&port->raddr.addr,
+								  &TrustedProxyServers[i + 1],
+								  &TrustedProxyServers[i + 2]))
+			{
+				useproxy = true;
+				break;
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, port->raddr.salen);
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with: \x0D \x0A \x0D \x0A \x00 \x0D \x0A
+	 * \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("oversized proxy packet")));
+		return STATUS_ERROR;
+	}
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+		pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));
+
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -4344,6 +4566,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4395,28 +4644,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index d626731723..9be50b1532 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2290,6 +2294,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4241,6 +4255,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12228,4 +12253,108 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index ee06528bb0..aa7ac35f67 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,8 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_servers = ''			# what IP/netmasks of proxy servers to trust
+					# (change requires restart)
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 7be1a67d69..57edda122a 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -121,6 +121,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index e4e5c21565..549e2d86a7 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -60,7 +60,7 @@ extern WaitEventSet *FeBeWaitSet;
 
 extern int	StreamServerPort(int family, const char *hostName,
 							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+							 pgsocket ListenSocket[], bool ProxyList[], bool isProxy, int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -74,6 +74,8 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_peekbytes(char *s, size_t len);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putbytes(const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..9ed219dfda 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
#6Bruno Lavoie
bl@brunol.com
In reply to: Magnus Hagander (#1)
Re: PROXY protocol support

+10 on this one!

Hosting a farm of read replicas and r/w endpoint behind an HAproxy makes
the powerful pg_hba purpose by hiding the real source address... which is
bad for some environments with strict conformance and audit requirements

Le mar. 2 mars 2021 à 12:43, Magnus Hagander <magnus@hagander.net> a écrit :

Show quoted text

PFA a simple patch that implements support for the PROXY protocol.

This is a protocol common and very light weight in proxies and load
balancers (haproxy is one common example, but also for example the AWS
cloud load balancers). Basically this protocol prefixes the normal
connection with a header and a specification of what the original host
was, allowing the server to unwrap that and get the correct client
address instead of just the proxy ip address. It is a one-way protocol
in that there is no response from the server, it's just purely a
prefix of the IP information.

Using this when PostgreSQL is behind a proxy allows us to keep using
pg_hba.conf rules based on the original ip address, as well as track
the original address in log messages and pg_stat_activity etc.

The implementation adds a parameter named proxy_servers which lists
the ips or ip+cidr mask to be trusted. Since a proxy can decide what
the origin is, and this is used for security decisions, it's very
important to not just trust any server, only those that are
intentionally used. By default, no servers are listed, and thus the
protocol is disabled.

When specified, and the connection on the normal port has the proxy
prefix on it, and the connection comes in from one of the addresses
listed as valid proxy servers, we will replace the actual IP address
of the client with the one specified in the proxy packet.

Currently there is no information about the proxy server in the
pg_stat_activity view, it's only available as a log message. But maybe
it should go in pg_stat_activity as well? Or in a separate
pg_stat_proxy view?

(In passing, I note that pq_discardbytes were in pqcomm.h, yet listed
as static in pqcomm.c -- but now made non-static)

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#7Tatsuo Ishii
ishii@sraoss.co.jp
In reply to: Magnus Hagander (#1)
Re: PROXY protocol support

PFA a simple patch that implements support for the PROXY protocol.

This is a protocol common and very light weight in proxies and load
balancers (haproxy is one common example, but also for example the AWS
cloud load balancers). Basically this protocol prefixes the normal
connection with a header and a specification of what the original host
was, allowing the server to unwrap that and get the correct client
address instead of just the proxy ip address. It is a one-way protocol
in that there is no response from the server, it's just purely a
prefix of the IP information.

Is there any formal specification for the "a protocol common and very
light weight in proxies"? I am asking because I was expecting that is
explained in your patch (hopefully in "Frontend/Backend Protocol"
chapter) but I couldn't find it in your patch.

Also we need a regression test for this feature.

Best regards,
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp

#8Jacob Champion
pchampion@vmware.com
In reply to: Tatsuo Ishii (#7)
Re: PROXY protocol support

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

which is maintained by HAProxy Technologies.

--Jacob

#9Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#5)
Re: PROXY protocol support

On Wed, 2021-03-03 at 10:39 +0100, Magnus Hagander wrote:

On Wed, Mar 3, 2021 at 10:00 AM Magnus Hagander <magnus@hagander.net> wrote:

Another option would of course be to listen on a separate port for it,
which seems to be the "haproxy way". That would be slightly more code
(we'd still want to keep the code for validating the list of trusted
proxies I'd say), but maybe worth doing?

In order to figure that out, I hacked up a poc on that. Once again
without updates to the docs, but shows approximately how much code
complexity it adds (not much).

From a configuration perspective, I like that the separate-port
approach can shift the burden of verifying trust to an external
firewall, and that it seems to match the behavior of other major server
software. But I don't have any insight into the relative security of
the two options in practice; hopefully someone else can chime in.

memset((char *) &hints, 0, sizeof(hints));
hints.ai_flags = AI_NUMERICHOST;
hints.ai_family = AF_UNSPEC;

ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

--Jacob

#10Jan Wieck
jan@wi3ck.info
In reply to: Jacob Champion (#8)
Re: PROXY protocol support

On 3/4/21 2:45 PM, Jacob Champion wrote:

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

which is maintained by HAProxy Technologies.

--Jacob

This looks like it would only need a few extra protocol messages to be
understood by the backend. It might be possible to implement that with
the loadable wire protocol extensions proposed here:

https://commitfest.postgresql.org/32/3018/

Regards, Jan

--
Jan Wieck
Principle Database Engineer
Amazon Web Services

#11Magnus Hagander
magnus@hagander.net
In reply to: Jan Wieck (#10)
Re: PROXY protocol support

On Thu, Mar 4, 2021 at 9:29 PM Jan Wieck <jan@wi3ck.info> wrote:

On 3/4/21 2:45 PM, Jacob Champion wrote:

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

which is maintained by HAProxy Technologies.

--Jacob

This looks like it would only need a few extra protocol messages to be
understood by the backend. It might be possible to implement that with
the loadable wire protocol extensions proposed here:

https://commitfest.postgresql.org/32/3018/

Actually the whole point of it is that it *doesn't* need any new
protocol messages. And that it *wraps* whatever is there, definitely
doesn't replace it. It should equally be wrapping whatever an
extension uses.

So while the base topic is not unrelated, I don't think there is any
overlap between these.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#12Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#9)
Re: PROXY protocol support

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

On Wed, 2021-03-03 at 10:39 +0100, Magnus Hagander wrote:

On Wed, Mar 3, 2021 at 10:00 AM Magnus Hagander <magnus@hagander.net> wrote:

Another option would of course be to listen on a separate port for it,
which seems to be the "haproxy way". That would be slightly more code
(we'd still want to keep the code for validating the list of trusted
proxies I'd say), but maybe worth doing?

In order to figure that out, I hacked up a poc on that. Once again
without updates to the docs, but shows approximately how much code
complexity it adds (not much).

From a configuration perspective, I like that the separate-port
approach can shift the burden of verifying trust to an external
firewall, and that it seems to match the behavior of other major server
software. But I don't have any insight into the relative security of
the two options in practice; hopefully someone else can chime in.

Yeah I think that and the argument that the spec explicitly says it
should be on it's own port is the advantage. The disadvantage is,
well, more ports and more configuration. But it does definitely make a
more clean separation of concerns.

memset((char *) &hints, 0, sizeof(hints));
hints.ai_flags = AI_NUMERICHOST;
hints.ai_family = AF_UNSPEC;

ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Iid think about the other end, if you had a proxy on a different
machine accepting unix connections and passing them on over
PROXY-over-tcp. But I doubt it's useful to know it was unix in that
case (since it still couldn't do peer or such for the auth) --
instead, that seems like an argument where it'd be better to proxy
without using PROXY and just letting the IP address be.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#13Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#8)
Re: PROXY protocol support

On Thu, Mar 4, 2021 at 8:45 PM Jacob Champion <pchampion@vmware.com> wrote:

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

Yeah, it's currently in one of the comments, but should probably be
added to the docs side as well.

And yes tests :) Probably not a regression test, but some level of tap
testing should definitely be added. We'll just have to find a way to
do that without making haproxy a dependency to run the tests :)

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#14Jan Wieck
jan@wi3ck.info
In reply to: Magnus Hagander (#11)
Re: PROXY protocol support

On 3/4/21 3:40 PM, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:29 PM Jan Wieck <jan@wi3ck.info> wrote:

This looks like it would only need a few extra protocol messages to be
understood by the backend. It might be possible to implement that with
the loadable wire protocol extensions proposed here:

https://commitfest.postgresql.org/32/3018/

Actually the whole point of it is that it *doesn't* need any new
protocol messages. And that it *wraps* whatever is there, definitely
doesn't replace it. It should equally be wrapping whatever an
extension uses.

So while the base topic is not unrelated, I don't think there is any
overlap between these.

I might be missing something here, but isn't sending some extra,
informational *header*, which is understood by the backend, in essence a
protocol extension?

Regards, Jan

--
Jan Wieck
Principle Database Engineer
Amazon Web Services

#15Magnus Hagander
magnus@hagander.net
In reply to: Jan Wieck (#14)
Re: PROXY protocol support

On Thu, Mar 4, 2021 at 10:01 PM Jan Wieck <jan@wi3ck.info> wrote:

On 3/4/21 3:40 PM, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:29 PM Jan Wieck <jan@wi3ck.info> wrote:

This looks like it would only need a few extra protocol messages to be
understood by the backend. It might be possible to implement that with
the loadable wire protocol extensions proposed here:

https://commitfest.postgresql.org/32/3018/

Actually the whole point of it is that it *doesn't* need any new
protocol messages. And that it *wraps* whatever is there, definitely
doesn't replace it. It should equally be wrapping whatever an
extension uses.

So while the base topic is not unrelated, I don't think there is any
overlap between these.

I might be missing something here, but isn't sending some extra,
informational *header*, which is understood by the backend, in essence a
protocol extension?

Bad choice of words, I guess.

The points being, there is a single packet sent ahead of the normal
stream. There are no new messages in "the postgresql protocol" or "the
febe protocol" or whatever we call it. And it doesn't change the
properties of any part of that protocol. And, importantly for the
simplicity, there is no negotiation and there are no packets going the
other way.

But sure, you can call it a protocol extension if you want. And yes,
it could probably be built on top of part of the ideas in that other
patch, but most of it would be useless (the abstraction of the listen
functionality into listen_have_free_slot/listen_add_socket would be
the big thing that could be used)

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#16Tatsuo Ishii
ishii@sraoss.co.jp
In reply to: Magnus Hagander (#13)
Re: PROXY protocol support

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

Yeah, it's currently in one of the comments, but should probably be
added to the docs side as well.

It seems the protocol is HAproxy product specific and I think it would
be better to be mentioned in the docs.

And yes tests :) Probably not a regression test, but some level of tap
testing should definitely be added. We'll just have to find a way to
do that without making haproxy a dependency to run the tests :)

Agreed.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese:http://www.sraoss.co.jp

#17Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#12)
Re: PROXY protocol support

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

Iid think about the other end, if you had a proxy on a different
machine accepting unix connections and passing them on over
PROXY-over-tcp. But I doubt it's useful to know it was unix in that
case (since it still couldn't do peer or such for the auth) --
instead, that seems like an argument where it'd be better to proxy
without using PROXY and just letting the IP address be.

You could potentially design a system that lets you proxy a "local all
all trust" setup from a different (trusted) machine, without having to
actually let people onto the machine that's running Postgres. That
would require some additional authentication on the PROXY connection
(i.e. something stronger than host-based auth) to actually be useful.

-- other notes --

A small nitpick on the current separate-port PoC is that I'm forced to
set up a "regular" TCP port, even if I only want the PROXY behavior.

The original-host logging isn't working for me:

WARNING: pg_getnameinfo_all() failed: ai_family not supported
LOG: proxy connection from: host=??? port=???

and I think the culprit is this:

/* Store a copy of the original address, for logging */
memcpy(&raddr_save, &port->raddr, port->raddr.salen);

port->raddr.salen is the length of port->raddr.addr; we want the length
of the copy to be sizeof(port->raddr) here, no?

--Jacob

#18Hannu Krosing
hannuk@google.com
In reply to: Jacob Champion (#17)
Re: PROXY protocol support

The current proposal seems to miss the case of transaction pooling
(and statement pooling) where the same established connection
multiplexes transactions / statements from multiple remote clients.

What we would need for that case would be a functionl

pg_set_remote_client_address( be_key, remote_ip, remote_hostname)

where only be_key and remote_ip are required, but any string (up to a
certain length) would be accepted as hostname.

It would be really nice if we could send this request at protocol level but
if that is hard to do then having a function would get us half way there.

the be_key in the function is the key from PGcancel, which is stored
by libpq when making the connection, and it is there, to make sure
that only the directly connecting proxy can successfully call the function.

Cheers
Hannu

Show quoted text

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

Iid think about the other end, if you had a proxy on a different
machine accepting unix connections and passing them on over
PROXY-over-tcp. But I doubt it's useful to know it was unix in that
case (since it still couldn't do peer or such for the auth) --
instead, that seems like an argument where it'd be better to proxy
without using PROXY and just letting the IP address be.

You could potentially design a system that lets you proxy a "local all
all trust" setup from a different (trusted) machine, without having to
actually let people onto the machine that's running Postgres. That
would require some additional authentication on the PROXY connection
(i.e. something stronger than host-based auth) to actually be useful.

-- other notes --

A small nitpick on the current separate-port PoC is that I'm forced to
set up a "regular" TCP port, even if I only want the PROXY behavior.

The original-host logging isn't working for me:

WARNING: pg_getnameinfo_all() failed: ai_family not supported
LOG: proxy connection from: host=??? port=???

and I think the culprit is this:

/* Store a copy of the original address, for logging */
memcpy(&raddr_save, &port->raddr, port->raddr.salen);

port->raddr.salen is the length of port->raddr.addr; we want the length
of the copy to be sizeof(port->raddr) here, no?

--Jacob

#19Álvaro Hernández
aht@ongres.com
In reply to: Jacob Champion (#17)
Re: PROXY protocol support

On 5/3/21 0:21, Jacob Champion wrote:

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

    Here's a use case: Envoy's Postgres filter (see [1]https://www.envoyproxy.io/docs/envoy/latest/configuration/listeners/network_filters/postgres_proxy_filter, [2]https://www.cncf.io/blog/2020/08/13/envoy-1-15-introduces-a-new-postgres-extension-with-monitoring-support/). Right now
is able to capture protocol-level metrics and send them to a metrics
collector (eg. Prometheus) while proxying the traffic. More capabilities
are being added as of today, and will eventually manage HBA too. It
would greatly benefit from this proposal, since it proxies the traffic
with, obviously, its IP, not the client's. It may be used (we do)
locally fronting Postgres, via UDS (so it can be easily trusted).

    Álvaro

[1]: https://www.envoyproxy.io/docs/envoy/latest/configuration/listeners/network_filters/postgres_proxy_filter
https://www.envoyproxy.io/docs/envoy/latest/configuration/listeners/network_filters/postgres_proxy_filter
[2]: https://www.cncf.io/blog/2020/08/13/envoy-1-15-introduces-a-new-postgres-extension-with-monitoring-support/
https://www.cncf.io/blog/2020/08/13/envoy-1-15-introduces-a-new-postgres-extension-with-monitoring-support/

--

Alvaro Hernandez

-----------
OnGres

#20Magnus Hagander
magnus@hagander.net
In reply to: Hannu Krosing (#18)
Re: PROXY protocol support

On Fri, Mar 5, 2021 at 12:57 AM Hannu Krosing <hannuk@google.com> wrote:

The current proposal seems to miss the case of transaction pooling
(and statement pooling) where the same established connection
multiplexes transactions / statements from multiple remote clients.

Not at all.

The current proposal is there to implement the PROXY protocol. It
doesn't try to do anything with connection pooling at all.

Solving a similar problem for connection poolers would also definitely
be a useful thing, but it is entirely out of scope of this patch, and
is a completely separate implementation.

I'd definitely like to see that one solved as well, but let's look at
it on a different thread so we don't derail this one.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#21Magnus Hagander
magnus@hagander.net
In reply to: Álvaro Hernández (#19)
Re: PROXY protocol support

On Fri, Mar 5, 2021 at 1:33 AM Álvaro Hernández <aht@ongres.com> wrote:

On 5/3/21 0:21, Jacob Champion wrote:

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

Here's a use case: Envoy's Postgres filter (see [1], [2]). Right now
is able to capture protocol-level metrics and send them to a metrics
collector (eg. Prometheus) while proxying the traffic. More capabilities
are being added as of today, and will eventually manage HBA too. It
would greatly benefit from this proposal, since it proxies the traffic
with, obviously, its IP, not the client's. It may be used (we do)
locally fronting Postgres, via UDS (so it can be easily trusted).

Yeah, Envoy is definitely a great example of a usecase for the proxy
protocol in general.

Specifically about the Unix socket though -- doesn't envoy normally
run on a different instance (or in a different container at least),
thus normally uses tcp between envoy and postgres? Or would it be a
reasonable usecase that you ran it locally on the postgres server,
having it speak IP to the clients but unix sockets to the postgres
backend? I guess maybe it is outside of the containerized world?

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#22Magnus Hagander
magnus@hagander.net
In reply to: Tatsuo Ishii (#16)
Re: PROXY protocol support

On Fri, Mar 5, 2021 at 12:08 AM Tatsuo Ishii <ishii@sraoss.co.jp> wrote:

On Thu, 2021-03-04 at 10:42 +0900, Tatsuo Ishii wrote:

Is there any formal specification for the "a protocol common and very
light weight in proxies"?

See

https://www.haproxy.org/download/1.8/doc/proxy-protocol.txt

Yeah, it's currently in one of the comments, but should probably be
added to the docs side as well.

It seems the protocol is HAproxy product specific and I think it would
be better to be mentioned in the docs.

It's definitely not HAProxy specific, it's more or less an industry
standard. It's just maintained by them. That said, yes, it should be
referenced in the docs.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#23Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#17)
Re: PROXY protocol support

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

Iid think about the other end, if you had a proxy on a different
machine accepting unix connections and passing them on over
PROXY-over-tcp. But I doubt it's useful to know it was unix in that
case (since it still couldn't do peer or such for the auth) --
instead, that seems like an argument where it'd be better to proxy
without using PROXY and just letting the IP address be.

You could potentially design a system that lets you proxy a "local all
all trust" setup from a different (trusted) machine, without having to
actually let people onto the machine that's running Postgres. That
would require some additional authentication on the PROXY connection
(i.e. something stronger than host-based auth) to actually be useful.

-- other notes --

A small nitpick on the current separate-port PoC is that I'm forced to
set up a "regular" TCP port, even if I only want the PROXY behavior.

Yeah. I'm not sure there's a good way to avoid that without making
configuations a lot more complex.

The original-host logging isn't working for me:

WARNING: pg_getnameinfo_all() failed: ai_family not supported
LOG: proxy connection from: host=??? port=???

and I think the culprit is this:

/* Store a copy of the original address, for logging */
memcpy(&raddr_save, &port->raddr, port->raddr.salen);

port->raddr.salen is the length of port->raddr.addr; we want the length
of the copy to be sizeof(port->raddr) here, no?

That's interesting -- it works perfectly fine here. What platform are
you testing on?

But yes, you are correct, it should do that. I guess it's a case of
the salen actually ending up being uninitialized in the copy, and thus
failing at a later stage. (I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#24Álvaro Hernández
aht@ongres.com
In reply to: Magnus Hagander (#21)
Re: PROXY protocol support

On 5/3/21 10:03, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 1:33 AM Álvaro Hernández <aht@ongres.com> wrote:

On 5/3/21 0:21, Jacob Champion wrote:

On Thu, 2021-03-04 at 21:45 +0100, Magnus Hagander wrote:

On Thu, Mar 4, 2021 at 9:07 PM Jacob Champion <pchampion@vmware.com> wrote:

Idle thought I had while setting up a local test rig: Are there any
compelling cases for allowing PROXY packets to arrive over Unix
sockets? (By which I mean, the proxy is running on the same machine as
Postgres, and connects to it using the .s.PGSQL socket file instead of
TCP.) Are there cases where you want some other software to interact
with the TCP stack instead of Postgres, but it'd still be nice to have
the original connection information available?

I'm uncertain what that usecase would be for something like haproxy,
tbh. It can't do connection pooling, so adding it on the same machine
as postgres itself wouldn't really add anything, I think?

Yeah, I wasn't thinking HAproxy so much as some unspecified software
appliance that's performing Some Task before allowing a TCP client to
speak to Postgres. But it'd be better to hear from someone that has an
actual use case, instead of me spitballing.

Here's a use case: Envoy's Postgres filter (see [1], [2]). Right now
is able to capture protocol-level metrics and send them to a metrics
collector (eg. Prometheus) while proxying the traffic. More capabilities
are being added as of today, and will eventually manage HBA too. It
would greatly benefit from this proposal, since it proxies the traffic
with, obviously, its IP, not the client's. It may be used (we do)
locally fronting Postgres, via UDS (so it can be easily trusted).

Yeah, Envoy is definitely a great example of a usecase for the proxy
protocol in general.

    Actually Envoy already implements the Proxy protocol:
https://www.envoyproxy.io/docs/envoy/latest/configuration/listeners/listener_filters/proxy_protocol.html
But I believe it would need some further cooperation with the Postgres
filter, unless they can be chained directly. Still, Postgres needs to
understand it, which is what your patch would add (thanks!).

Specifically about the Unix socket though -- doesn't envoy normally
run on a different instance (or in a different container at least),
thus normally uses tcp between envoy and postgres? Or would it be a
reasonable usecase that you ran it locally on the postgres server,
having it speak IP to the clients but unix sockets to the postgres
backend? I guess maybe it is outside of the containerized world?

    This is exactly the architecture we use at StackGres [1]https://stackgres.io[2]https://stackgres.io/doc/latest/intro/architecture/#stackgres-pod-architecture-diagram. We use
Envoy as a sidecar (so it runs on the same pod, server as Postgres) and
connects via UDS. But then exposes the connection to the outside clients
via TCP/IP. So in my opinion it is quite applicable to the container
world :)

    Álvaro

[1]: https://stackgres.io
[2]: https://stackgres.io/doc/latest/intro/architecture/#stackgres-pod-architecture-diagram
https://stackgres.io/doc/latest/intro/architecture/#stackgres-pod-architecture-diagram

--

Alvaro Hernandez

-----------
OnGres

#25Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#23)
Re: PROXY protocol support

On Fri, 2021-03-05 at 10:22 +0100, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

A small nitpick on the current separate-port PoC is that I'm forced to
set up a "regular" TCP port, even if I only want the PROXY behavior.

Yeah. I'm not sure there's a good way to avoid that without making
configuations a lot more complex.

A generic solution would also solve the "I want to listen on more than
one port" problem, but that's probably not something to tackle at the
same time.

The original-host logging isn't working for me:

[...]

That's interesting -- it works perfectly fine here. What platform are
you testing on?

Ubuntu 20.04.

But yes, you are correct, it should do that. I guess it's a case of
the salen actually ending up being uninitialized in the copy, and thus
failing at a later stage.

That seems right; EAI_FAMILY can be returned for a mismatched addrlen.

(I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

Cool. Did you mean to attach a patch?

== More Notes ==

(Stop me if I'm digging too far into a proof of concept patch.)

+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("oversized proxy packet")));
+		return STATUS_ERROR;
+	}

I think this is not quite right -- if there's additional data beyond
the IPv6 header size, that just means there are TLVs tacked onto the
header that we should ignore. (Or, eventually, use.)

Additionally, we need to check for underflow as well. A misbehaving
proxy might not send enough data to fill up the address block for the
address family in use.

+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+		pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));

This looks like dead code, given that we'll error out for the same
check above -- but once it's no longer dead code, the return value of
pq_discardbytes should be checked for EOF.

+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}

I'm trying to reason through the fallout of setting raddr and not
laddr. I understand why we're not setting laddr -- several places in
the code rely on the laddr to actually refer to a machine-local address
-- but the fact that there is no actual connection from raddr to laddr
could cause shenanigans. For example, the ident auth protocol will just
break (and it might be nice to explicitly disable it for PROXY
connections). Are there any other situations where a "faked" raddr
could throw off Postgres internals?

--Jacob

#26Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#25)
1 attachment(s)
Re: PROXY protocol support

On Fri, Mar 5, 2021 at 8:11 PM Jacob Champion <pchampion@vmware.com> wrote:

On Fri, 2021-03-05 at 10:22 +0100, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

The original-host logging isn't working for me:

[...]

That's interesting -- it works perfectly fine here. What platform are
you testing on?

Ubuntu 20.04.

Curious. It doesn't show up on my debian.

But either way -- it was clearly wrong :)

(I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

Cool. Did you mean to attach a patch?

I didn't, I had some other hacks that were broken :) I've attached one
now which includes those changes.

== More Notes ==

(Stop me if I'm digging too far into a proof of concept patch.)

Definitely not -- much appreciated, and just what was needed to take
it from poc to a proper one!

+     proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+     if (proxyaddrlen > sizeof(proxyaddr))
+     {
+             ereport(COMMERROR,
+                             (errcode(ERRCODE_PROTOCOL_VIOLATION),
+                              errmsg("oversized proxy packet")));
+             return STATUS_ERROR;
+     }

I think this is not quite right -- if there's additional data beyond
the IPv6 header size, that just means there are TLVs tacked onto the
header that we should ignore. (Or, eventually, use.)

Yeah, you're right. Fallout of too much moving around. I think inthe
end that code should just be removed, in favor of the discard path as
you mentinoed below.

Additionally, we need to check for underflow as well. A misbehaving
proxy might not send enough data to fill up the address block for the
address family in use.

I used to have that check. I seem to have lost it in restructuring. Added back!

+     /* If there is any more header data present, skip past it */
+     if (proxyaddrlen > sizeof(proxyaddr))
+             pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));

This looks like dead code, given that we'll error out for the same
check above -- but once it's no longer dead code, the return value of
pq_discardbytes should be checked for EOF.

Yup.

+     else if (proxyheader.fam == 0x11)
+     {
+             /* TCPv4 */
+             port->raddr.addr.ss_family = AF_INET;
+             port->raddr.salen = sizeof(struct sockaddr_in);
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+     }

I'm trying to reason through the fallout of setting raddr and not
laddr. I understand why we're not setting laddr -- several places in
the code rely on the laddr to actually refer to a machine-local address
-- but the fact that there is no actual connection from raddr to laddr
could cause shenanigans. For example, the ident auth protocol will just
break (and it might be nice to explicitly disable it for PROXY
connections). Are there any other situations where a "faked" raddr
could throw off Postgres internals?

That's a good point to discuss. I thought about it initially and
figured it'd be even worse to actually copy over laddr since that
woudl then suddenly have the IP address belonging to a different
machine.. And then I forgot to enumerate the other cases.

For ident, disabling the method seems reasonable.

Another thing that shows up with added support for running the proxy
protocol over Unix sockets, is that PostgreSQL refuses to do SSL over
Unix sockets. So that check has to be updated to allow it over proxy
connections. Same for GSSAPI.

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

Attached is an updated, which covers your comments, as well as adds
unix socket support (per your question and Alvaros confirmed usecase).
It allows proxy connections over unix sockets, but I saw no need to
get into unix sockets over the proxy protocol (dealing with paths
between machines etc).

I changed the additional ListenSocket array to instead declare
ListenSocket as an array of structs holding two fields. Seems cleaner,
and especially should there be further extensions needed in the
future.

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly). They probably need some more love but it's a start.

And of course rebased.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_3.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_3.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..5d8fcc3d50 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 967de73596..e1999bfeb5 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 994251e7d9..470e36db58 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1761,6 +1761,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 4c7b1e7bfd..9f92195097 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -308,13 +308,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -359,10 +359,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -385,7 +385,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -402,7 +402,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -579,16 +579,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -1113,7 +1113,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index edab95a19e..0344f907c9 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -195,15 +196,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -217,7 +225,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -581,6 +589,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1124,7 +1133,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1153,17 +1165,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1177,9 +1189,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1189,7 +1222,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1251,12 +1284,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1267,9 +1300,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1281,7 +1328,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1430,10 +1477,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1722,15 +1769,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1898,7 +1947,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1911,6 +1960,213 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2019,7 +2275,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2056,7 +2312,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2555,10 +2811,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4320,6 +4576,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4371,28 +4654,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 3fd1a5fbe2..8b3be40a26 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2291,6 +2295,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4242,6 +4256,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12227,4 +12252,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index ee06528bb0..b002228393 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 7be1a67d69..57edda122a 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -121,6 +121,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index b20deeb555..c06ee29f88 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -33,6 +33,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -54,9 +60,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -69,6 +75,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..9ed219dfda 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..c84d797e4f
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,157 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 35;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+listen_addresses = 'localhost'
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "mha", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub inet_socket
+{
+	my ($port) = @_;
+	return IO::Socket::INET->new(
+		PeerAddr => "127.0.0.1",
+		PeerPort => $port,
+		Proto    => "tcp",
+		Type     => SOCK_STREAM);
+}
+
+sub unix_socket
+{
+	my ($port) = @_;
+	return IO::Socket::UNIX->new(
+		Peer => $node->host() . "/.s.PGSQL." . $port,
+		Type => SOCK_STREAM);
+}
+
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(
+	inet_socket($node->port()), undef,
+	"normal ipv4 connection",   "127.0.0.1");
+test_connection(unix_socket($node->port()),
+	undef, "normal unix connection", "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(inet_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(inet_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(inet_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+test_connection(unix_socket($node->port() + 1),
+	"11.22.33.44", "proxy unix", "11.22.33.44");
+
+test_connection(unix_socket($node->port() + 1),
+    "11.22.33.44", "proxy unix with extra", "11.22.33.44", undef, "abcdef"x100);
#27Magnus Hagander
magnus@hagander.net
In reply to: Magnus Hagander (#26)
1 attachment(s)
Re: PROXY protocol support

On Sat, Mar 6, 2021 at 4:17 PM Magnus Hagander <magnus@hagander.net> wrote:

On Fri, Mar 5, 2021 at 8:11 PM Jacob Champion <pchampion@vmware.com> wrote:

On Fri, 2021-03-05 at 10:22 +0100, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

The original-host logging isn't working for me:

[...]

That's interesting -- it works perfectly fine here. What platform are
you testing on?

Ubuntu 20.04.

Curious. It doesn't show up on my debian.

But either way -- it was clearly wrong :)

(I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

Cool. Did you mean to attach a patch?

I didn't, I had some other hacks that were broken :) I've attached one
now which includes those changes.

== More Notes ==

(Stop me if I'm digging too far into a proof of concept patch.)

Definitely not -- much appreciated, and just what was needed to take
it from poc to a proper one!

+     proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+     if (proxyaddrlen > sizeof(proxyaddr))
+     {
+             ereport(COMMERROR,
+                             (errcode(ERRCODE_PROTOCOL_VIOLATION),
+                              errmsg("oversized proxy packet")));
+             return STATUS_ERROR;
+     }

I think this is not quite right -- if there's additional data beyond
the IPv6 header size, that just means there are TLVs tacked onto the
header that we should ignore. (Or, eventually, use.)

Yeah, you're right. Fallout of too much moving around. I think inthe
end that code should just be removed, in favor of the discard path as
you mentinoed below.

Additionally, we need to check for underflow as well. A misbehaving
proxy might not send enough data to fill up the address block for the
address family in use.

I used to have that check. I seem to have lost it in restructuring. Added back!

+     /* If there is any more header data present, skip past it */
+     if (proxyaddrlen > sizeof(proxyaddr))
+             pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));

This looks like dead code, given that we'll error out for the same
check above -- but once it's no longer dead code, the return value of
pq_discardbytes should be checked for EOF.

Yup.

+     else if (proxyheader.fam == 0x11)
+     {
+             /* TCPv4 */
+             port->raddr.addr.ss_family = AF_INET;
+             port->raddr.salen = sizeof(struct sockaddr_in);
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+     }

I'm trying to reason through the fallout of setting raddr and not
laddr. I understand why we're not setting laddr -- several places in
the code rely on the laddr to actually refer to a machine-local address
-- but the fact that there is no actual connection from raddr to laddr
could cause shenanigans. For example, the ident auth protocol will just
break (and it might be nice to explicitly disable it for PROXY
connections). Are there any other situations where a "faked" raddr
could throw off Postgres internals?

That's a good point to discuss. I thought about it initially and
figured it'd be even worse to actually copy over laddr since that
woudl then suddenly have the IP address belonging to a different
machine.. And then I forgot to enumerate the other cases.

For ident, disabling the method seems reasonable.

Another thing that shows up with added support for running the proxy
protocol over Unix sockets, is that PostgreSQL refuses to do SSL over
Unix sockets. So that check has to be updated to allow it over proxy
connections. Same for GSSAPI.

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

Attached is an updated, which covers your comments, as well as adds
unix socket support (per your question and Alvaros confirmed usecase).
It allows proxy connections over unix sockets, but I saw no need to
get into unix sockets over the proxy protocol (dealing with paths
between machines etc).

I changed the additional ListenSocket array to instead declare
ListenSocket as an array of structs holding two fields. Seems cleaner,
and especially should there be further extensions needed in the
future.

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly). They probably need some more love but it's a start.

And of course rebased.

Pfft, I was hoping for cfbot to pick it up and test it on a different
platform. Of course, for it to do that, I need to include the test
directory in the Makefile. Here's a new one which adds that, no other
changes.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_4.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_4.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..5d8fcc3d50 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 967de73596..e1999bfeb5 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 994251e7d9..470e36db58 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1761,6 +1761,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 4c7b1e7bfd..9f92195097 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -308,13 +308,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -359,10 +359,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -385,7 +385,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -402,7 +402,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -579,16 +579,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -1113,7 +1113,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index edab95a19e..0344f907c9 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -195,15 +196,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -217,7 +225,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -581,6 +589,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1124,7 +1133,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1153,17 +1165,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1177,9 +1189,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1189,7 +1222,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1251,12 +1284,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1267,9 +1300,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1281,7 +1328,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1430,10 +1477,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1722,15 +1769,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1898,7 +1947,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1911,6 +1960,213 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2019,7 +2275,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2056,7 +2312,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2555,10 +2811,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4320,6 +4576,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4371,28 +4654,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 3fd1a5fbe2..8b3be40a26 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2291,6 +2295,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4242,6 +4256,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12227,4 +12252,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index ee06528bb0..b002228393 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 7be1a67d69..57edda122a 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -121,6 +121,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index b20deeb555..c06ee29f88 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -33,6 +33,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -54,9 +60,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -69,6 +75,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..9ed219dfda 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index f7859c2fd5..cfb9a319a2 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -13,7 +13,7 @@ top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
 SUBDIRS = perl regress isolation modules authentication recovery subscription \
-	  locale
+	  locale protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..c84d797e4f
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,157 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 35;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+listen_addresses = 'localhost'
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "mha", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub inet_socket
+{
+	my ($port) = @_;
+	return IO::Socket::INET->new(
+		PeerAddr => "127.0.0.1",
+		PeerPort => $port,
+		Proto    => "tcp",
+		Type     => SOCK_STREAM);
+}
+
+sub unix_socket
+{
+	my ($port) = @_;
+	return IO::Socket::UNIX->new(
+		Peer => $node->host() . "/.s.PGSQL." . $port,
+		Type => SOCK_STREAM);
+}
+
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(
+	inet_socket($node->port()), undef,
+	"normal ipv4 connection",   "127.0.0.1");
+test_connection(unix_socket($node->port()),
+	undef, "normal unix connection", "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(inet_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(inet_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(inet_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+test_connection(unix_socket($node->port() + 1),
+	"11.22.33.44", "proxy unix", "11.22.33.44");
+
+test_connection(unix_socket($node->port() + 1),
+    "11.22.33.44", "proxy unix with extra", "11.22.33.44", undef, "abcdef"x100);
#28Magnus Hagander
magnus@hagander.net
In reply to: Magnus Hagander (#27)
1 attachment(s)
Re: PROXY protocol support

On Sat, Mar 6, 2021 at 5:30 PM Magnus Hagander <magnus@hagander.net> wrote:

On Sat, Mar 6, 2021 at 4:17 PM Magnus Hagander <magnus@hagander.net> wrote:

On Fri, Mar 5, 2021 at 8:11 PM Jacob Champion <pchampion@vmware.com> wrote:

On Fri, 2021-03-05 at 10:22 +0100, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

The original-host logging isn't working for me:

[...]

That's interesting -- it works perfectly fine here. What platform are
you testing on?

Ubuntu 20.04.

Curious. It doesn't show up on my debian.

But either way -- it was clearly wrong :)

(I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

Cool. Did you mean to attach a patch?

I didn't, I had some other hacks that were broken :) I've attached one
now which includes those changes.

== More Notes ==

(Stop me if I'm digging too far into a proof of concept patch.)

Definitely not -- much appreciated, and just what was needed to take
it from poc to a proper one!

+     proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+     if (proxyaddrlen > sizeof(proxyaddr))
+     {
+             ereport(COMMERROR,
+                             (errcode(ERRCODE_PROTOCOL_VIOLATION),
+                              errmsg("oversized proxy packet")));
+             return STATUS_ERROR;
+     }

I think this is not quite right -- if there's additional data beyond
the IPv6 header size, that just means there are TLVs tacked onto the
header that we should ignore. (Or, eventually, use.)

Yeah, you're right. Fallout of too much moving around. I think inthe
end that code should just be removed, in favor of the discard path as
you mentinoed below.

Additionally, we need to check for underflow as well. A misbehaving
proxy might not send enough data to fill up the address block for the
address family in use.

I used to have that check. I seem to have lost it in restructuring. Added back!

+     /* If there is any more header data present, skip past it */
+     if (proxyaddrlen > sizeof(proxyaddr))
+             pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));

This looks like dead code, given that we'll error out for the same
check above -- but once it's no longer dead code, the return value of
pq_discardbytes should be checked for EOF.

Yup.

+     else if (proxyheader.fam == 0x11)
+     {
+             /* TCPv4 */
+             port->raddr.addr.ss_family = AF_INET;
+             port->raddr.salen = sizeof(struct sockaddr_in);
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+     }

I'm trying to reason through the fallout of setting raddr and not
laddr. I understand why we're not setting laddr -- several places in
the code rely on the laddr to actually refer to a machine-local address
-- but the fact that there is no actual connection from raddr to laddr
could cause shenanigans. For example, the ident auth protocol will just
break (and it might be nice to explicitly disable it for PROXY
connections). Are there any other situations where a "faked" raddr
could throw off Postgres internals?

That's a good point to discuss. I thought about it initially and
figured it'd be even worse to actually copy over laddr since that
woudl then suddenly have the IP address belonging to a different
machine.. And then I forgot to enumerate the other cases.

For ident, disabling the method seems reasonable.

Another thing that shows up with added support for running the proxy
protocol over Unix sockets, is that PostgreSQL refuses to do SSL over
Unix sockets. So that check has to be updated to allow it over proxy
connections. Same for GSSAPI.

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

Attached is an updated, which covers your comments, as well as adds
unix socket support (per your question and Alvaros confirmed usecase).
It allows proxy connections over unix sockets, but I saw no need to
get into unix sockets over the proxy protocol (dealing with paths
between machines etc).

I changed the additional ListenSocket array to instead declare
ListenSocket as an array of structs holding two fields. Seems cleaner,
and especially should there be further extensions needed in the
future.

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly). They probably need some more love but it's a start.

And of course rebased.

Pfft, I was hoping for cfbot to pick it up and test it on a different
platform. Of course, for it to do that, I need to include the test
directory in the Makefile. Here's a new one which adds that, no other
changes.

So cfbot didn't like thato ne one bit. Turns out that it's not a great
idea to hardcode the username "mha" in the tests :)

And also changed to only use unix sockets for the tests on linux, and
tcp only on windows. Because that's how our tests are supposed to be.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_5.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_5.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index b420486a0a..5d8fcc3d50 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 529876895b..6c2bd2f4fe 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 994251e7d9..470e36db58 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1761,6 +1761,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 4c7b1e7bfd..9f92195097 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -308,13 +308,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -359,10 +359,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -385,7 +385,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -402,7 +402,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -579,16 +579,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -1113,7 +1113,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index edab95a19e..0344f907c9 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -195,15 +196,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -217,7 +225,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -581,6 +589,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1124,7 +1133,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1153,17 +1165,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1177,9 +1189,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1189,7 +1222,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1251,12 +1284,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1267,9 +1300,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1281,7 +1328,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1430,10 +1477,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1722,15 +1769,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1898,7 +1947,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1911,6 +1960,213 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2019,7 +2275,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2056,7 +2312,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2555,10 +2811,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4320,6 +4576,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4371,28 +4654,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index e337df42cb..67e59664b6 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -46,10 +46,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -227,6 +229,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2300,6 +2304,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4251,6 +4265,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12236,4 +12261,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index c6483fa1ff..a52ca19fa9 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 30fb4e613d..2051ceb6e6 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -121,6 +121,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index b20deeb555..c06ee29f88 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -33,6 +33,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -54,9 +60,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -69,6 +75,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index cfa59c4dc0..9ed219dfda 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index f7859c2fd5..cfb9a319a2 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -13,7 +13,7 @@ top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
 SUBDIRS = perl regress isolation modules authentication recovery subscription \
-	  locale
+	  locale protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..edc032d49c
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgresNode::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgresNode::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#29Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#28)
Re: PROXY protocol support

On Tue, 2021-03-09 at 11:25 +0100, Magnus Hagander wrote:

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly).

Yeah. The tests I'm writing for this and NSS have been the same way;
it's a real problem. I'm basically writing supplemental tests in Python
as the "daily driver", then trying to port whatever is easiest (not
much) into Perl, when I get time.

== More Notes ==

Some additional spec-compliance stuff:

/* Lower 4 bits hold type of connection */
if (proxyheader.fam == 0)
{
/* LOCAL connection, so we ignore the address included */
}

(fam == 0) is the UNSPEC case, which isn't necessarily LOCAL. We have
to do something different for the LOCAL case:

- \x0 : LOCAL : [...] The receiver must accept this connection as
valid and must use the real connection endpoints and discard the
protocol block including the family which is ignored.

So we should ignore the entire "protocol block" (by which I believe
they mean the protocol-and-address-family byte) in the case of LOCAL,
and just accept it with the original address info intact. That seems to
match the sample code in the back of the spec. The current behavior in
the patch will apply the PROXY behavior incorrectly if the sender sends
a LOCAL header with something other than UNSPEC -- which is strange
behavior but not explicitly prohibited as far as I can see.

We also need to reject all connections that aren't either LOCAL or
PROXY commands:

- other values are unassigned and must not be emitted by senders.
Receivers must drop connections presenting unexpected values here.

...and naturally it'd be Nice (tm) if the tests covered those corner
cases.

Over on the struct side:

+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
... snip ...
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{

Given the importance of these hardcoded lengths matching reality, is it
possible to add some static assertions to make sure that sizeof(<ipv4
block>) == 12 and so on? That would also save any poor souls who are
using compilers with nonstandard struct-packing behavior.

--Jacob

#30Magnus Hagander
magnus@hagander.net
In reply to: Magnus Hagander (#28)
1 attachment(s)
Re: PROXY protocol support

On Tue, Mar 9, 2021 at 11:25 AM Magnus Hagander <magnus@hagander.net> wrote:

On Sat, Mar 6, 2021 at 5:30 PM Magnus Hagander <magnus@hagander.net> wrote:

On Sat, Mar 6, 2021 at 4:17 PM Magnus Hagander <magnus@hagander.net> wrote:

On Fri, Mar 5, 2021 at 8:11 PM Jacob Champion <pchampion@vmware.com> wrote:

On Fri, 2021-03-05 at 10:22 +0100, Magnus Hagander wrote:

On Fri, Mar 5, 2021 at 12:21 AM Jacob Champion <pchampion@vmware.com> wrote:

The original-host logging isn't working for me:

[...]

That's interesting -- it works perfectly fine here. What platform are
you testing on?

Ubuntu 20.04.

Curious. It doesn't show up on my debian.

But either way -- it was clearly wrong :)

(I sent for sizeof(SockAddr) to make it
easier to read without having to look things up, but the net result is
the same)

Cool. Did you mean to attach a patch?

I didn't, I had some other hacks that were broken :) I've attached one
now which includes those changes.

== More Notes ==

(Stop me if I'm digging too far into a proof of concept patch.)

Definitely not -- much appreciated, and just what was needed to take
it from poc to a proper one!

+     proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+     if (proxyaddrlen > sizeof(proxyaddr))
+     {
+             ereport(COMMERROR,
+                             (errcode(ERRCODE_PROTOCOL_VIOLATION),
+                              errmsg("oversized proxy packet")));
+             return STATUS_ERROR;
+     }

I think this is not quite right -- if there's additional data beyond
the IPv6 header size, that just means there are TLVs tacked onto the
header that we should ignore. (Or, eventually, use.)

Yeah, you're right. Fallout of too much moving around. I think inthe
end that code should just be removed, in favor of the discard path as
you mentinoed below.

Additionally, we need to check for underflow as well. A misbehaving
proxy might not send enough data to fill up the address block for the
address family in use.

I used to have that check. I seem to have lost it in restructuring. Added back!

+     /* If there is any more header data present, skip past it */
+     if (proxyaddrlen > sizeof(proxyaddr))
+             pq_discardbytes(proxyaddrlen - sizeof(proxyaddr));

This looks like dead code, given that we'll error out for the same
check above -- but once it's no longer dead code, the return value of
pq_discardbytes should be checked for EOF.

Yup.

+     else if (proxyheader.fam == 0x11)
+     {
+             /* TCPv4 */
+             port->raddr.addr.ss_family = AF_INET;
+             port->raddr.salen = sizeof(struct sockaddr_in);
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+             ((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+     }

I'm trying to reason through the fallout of setting raddr and not
laddr. I understand why we're not setting laddr -- several places in
the code rely on the laddr to actually refer to a machine-local address
-- but the fact that there is no actual connection from raddr to laddr
could cause shenanigans. For example, the ident auth protocol will just
break (and it might be nice to explicitly disable it for PROXY
connections). Are there any other situations where a "faked" raddr
could throw off Postgres internals?

That's a good point to discuss. I thought about it initially and
figured it'd be even worse to actually copy over laddr since that
woudl then suddenly have the IP address belonging to a different
machine.. And then I forgot to enumerate the other cases.

For ident, disabling the method seems reasonable.

Another thing that shows up with added support for running the proxy
protocol over Unix sockets, is that PostgreSQL refuses to do SSL over
Unix sockets. So that check has to be updated to allow it over proxy
connections. Same for GSSAPI.

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

Attached is an updated, which covers your comments, as well as adds
unix socket support (per your question and Alvaros confirmed usecase).
It allows proxy connections over unix sockets, but I saw no need to
get into unix sockets over the proxy protocol (dealing with paths
between machines etc).

I changed the additional ListenSocket array to instead declare
ListenSocket as an array of structs holding two fields. Seems cleaner,
and especially should there be further extensions needed in the
future.

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly). They probably need some more love but it's a start.

And of course rebased.

Pfft, I was hoping for cfbot to pick it up and test it on a different
platform. Of course, for it to do that, I need to include the test
directory in the Makefile. Here's a new one which adds that, no other
changes.

So cfbot didn't like thato ne one bit. Turns out that it's not a great
idea to hardcode the username "mha" in the tests :)

And also changed to only use unix sockets for the tests on linux, and
tcp only on windows. Because that's how our tests are supposed to be.

PFA a rebase to make cfbot happy.

There's another set or review notes from Jacob on March 11, that I
will also address, but it's not included in this version.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_6.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_6.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 6098f6b020..c8a7d2a3b7 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 967b5ef73c..6cf6e51708 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1851,6 +1851,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 89a5f901aa..dd63be54e2 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -311,13 +311,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -362,10 +362,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -388,7 +388,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -405,7 +405,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -584,16 +584,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -1118,7 +1118,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 5a050898fe..afa40bcdce 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -196,15 +197,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,7 +226,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -586,6 +594,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1135,7 +1144,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1164,17 +1176,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1188,9 +1200,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1200,7 +1233,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1262,12 +1295,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1278,9 +1311,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1292,7 +1339,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1441,10 +1488,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1733,15 +1780,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1909,7 +1958,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1922,6 +1971,213 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Lower 4 bits hold type of connection */
+	if (proxyheader.fam == 0)
+	{
+		/* LOCAL connection, so we ignore the address included */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2030,7 +2286,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2067,7 +2323,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2579,10 +2835,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4363,6 +4619,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4414,28 +4697,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 480e8cd199..2eac7e7264 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -50,10 +50,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -234,6 +236,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2358,6 +2362,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4331,6 +4345,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12536,4 +12561,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index b696abfe54..2e224fef36 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 02015efe13..471c76fb30 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,6 +126,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index 6c51b2f20f..cdaae030e1 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -63,9 +69,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -78,6 +84,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 extern bool pq_check_connection(void);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 0efdd7c232..2a029ef786 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..edc032d49c
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgresNode::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgresNode::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#31Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#29)
1 attachment(s)
Re: PROXY protocol support

On Thu, Mar 11, 2021 at 12:05 AM Jacob Champion <pchampion@vmware.com> wrote:

On Tue, 2021-03-09 at 11:25 +0100, Magnus Hagander wrote:

I've also added some trivial tests (man that took an ungodly amount of
fighting perl -- it's clearly been a long time since I used perl
properly).

Yeah. The tests I'm writing for this and NSS have been the same way;
it's a real problem. I'm basically writing supplemental tests in Python
as the "daily driver", then trying to port whatever is easiest (not
much) into Perl, when I get time.

== More Notes ==

Some additional spec-compliance stuff:

/* Lower 4 bits hold type of connection */
if (proxyheader.fam == 0)
{
/* LOCAL connection, so we ignore the address included */
}

(fam == 0) is the UNSPEC case, which isn't necessarily LOCAL. We have
to do something different for the LOCAL case:

Oh ugh. yeah, and the comment is wrong too -- it got the "command"
confused with "connection family". Too many copy/paste I think.

- \x0 : LOCAL : [...] The receiver must accept this connection as
valid and must use the real connection endpoints and discard the
protocol block including the family which is ignored.

So we should ignore the entire "protocol block" (by which I believe
they mean the protocol-and-address-family byte) in the case of LOCAL,
and just accept it with the original address info intact. That seems to
match the sample code in the back of the spec. The current behavior in
the patch will apply the PROXY behavior incorrectly if the sender sends
a LOCAL header with something other than UNSPEC -- which is strange
behavior but not explicitly prohibited as far as I can see.

Yeah, I think we do the right thing in the "right usecase".

We also need to reject all connections that aren't either LOCAL or
PROXY commands:

Indeed.

- other values are unassigned and must not be emitted by senders.
Receivers must drop connections presenting unexpected values here.

...and naturally it'd be Nice (tm) if the tests covered those corner
cases.

I think that's covered in the attached update.

Over on the struct side:

+             struct
+             {                                               /* for TCP/UDP over IPv4, len = 12 */
+                     uint32          src_addr;
+                     uint32          dst_addr;
+                     uint16          src_port;
+                     uint16          dst_port;
+             }                       ip4;
... snip ...
+             /* TCPv4 */
+             if (proxyaddrlen < 12)
+             {

Given the importance of these hardcoded lengths matching reality, is it
possible to add some static assertions to make sure that sizeof(<ipv4
block>) == 12 and so on? That would also save any poor souls who are
using compilers with nonstandard struct-packing behavior.

Yeah, probably makes sense. Added.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_7.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_7.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 6098f6b020..c8a7d2a3b7 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 967b5ef73c..6cf6e51708 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1851,6 +1851,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 89a5f901aa..dd63be54e2 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -311,13 +311,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -362,10 +362,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -388,7 +388,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -405,7 +405,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -584,16 +584,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -1118,7 +1118,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 5a050898fe..725d6fac34 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -196,15 +197,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,7 +226,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -586,6 +594,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1135,7 +1144,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1164,17 +1176,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1188,9 +1200,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1200,7 +1233,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1262,12 +1295,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1278,9 +1311,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1292,7 +1339,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1441,10 +1488,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1733,15 +1780,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1909,7 +1958,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1922,6 +1971,256 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/*
+	 * Assert the size of the structs that are part of the protocol,
+	 * to defend against strange compilers.
+	 */
+	StaticAssertStmt(sizeof(proxyheader) == 16, "proxy header struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip4) == 12, "proxy address ipv4 struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip6) == 36, "proxy address ipv6 struct has invalid size");
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * Proxy command is in the low 4 bits of the first byte.
+	 * 0x00 = local, 0x01 = proxy, all others should be rejected
+	 */
+	if ((proxyheader.ver_cmd & 0x0F) == 0x00)
+	{
+		if (proxyheader.fam != 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family %x for local connection", proxyheader.fam)));
+			return STATUS_ERROR;
+		}
+	}
+	else if ((proxyheader.ver_cmd & 0x0F) == 0x01)
+	{
+		if (proxyheader.fam == 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family 0 for non-local connection")));
+			return STATUS_ERROR;
+		}
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol command: %x", (proxyheader.ver_cmd & 0x0f))));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Connection family */
+	if (proxyheader.fam == 0)
+	{
+		/*
+		 * UNSPEC connection over LOCAL (verified above).
+		 * in this case we just ignore the address included.
+		 */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s",
+						remote_host,
+						remote_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2030,7 +2329,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2067,7 +2366,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2579,10 +2878,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4363,6 +4662,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4414,28 +4740,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 480e8cd199..2eac7e7264 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -50,10 +50,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -234,6 +236,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2358,6 +2362,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4331,6 +4345,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12536,4 +12561,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP addrress %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index b696abfe54..2e224fef36 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 02015efe13..471c76fb30 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,6 +126,7 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index 6c51b2f20f..cdaae030e1 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -63,9 +69,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -78,6 +84,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 extern bool pq_check_connection(void);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 0efdd7c232..2a029ef786 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..edc032d49c
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgresNode::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgresNode::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#32Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#31)
Re: PROXY protocol support

Hi Magnus,

I'm only just starting to page this back into my head, so this is by no
means a full review of the v7 changes -- just stuff I've noticed over
the last day or so of poking around.

On Tue, 2021-06-29 at 11:48 +0200, Magnus Hagander wrote:

On Thu, Mar 11, 2021 at 12:05 AM Jacob Champion <pchampion@vmware.com> wrote:

On Tue, 2021-03-09 at 11:25 +0100, Magnus Hagander wrote:

- \x0 : LOCAL : [...] The receiver must accept this connection as
valid and must use the real connection endpoints and discard the
protocol block including the family which is ignored.

So we should ignore the entire "protocol block" (by which I believe
they mean the protocol-and-address-family byte) in the case of LOCAL,
and just accept it with the original address info intact. That seems to
match the sample code in the back of the spec. The current behavior in
the patch will apply the PROXY behavior incorrectly if the sender sends
a LOCAL header with something other than UNSPEC -- which is strange
behavior but not explicitly prohibited as far as I can see.

Yeah, I think we do the right thing in the "right usecase".

The current implementation is, I think, stricter than the spec asks
for. We're supposed to ignore the family for LOCAL cases, and it's not
clear to me whether we're supposed to also ignore the entire "fam"
family-and-protocol byte (the phrase "protocol block" is not actually
defined in the spec).

It's probably not a big deal in practice, but it could mess with
interoperability for lazier proxy implementations. I think I'll ask the
HAProxy folks for some clarification tomorrow.

+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.

The documentation mentions that host names are valid in proxy_servers,
but check_proxy_servers() uses the AI_NUMERICHOST hint with
getaddrinfo(), so host names get rejected.

+ GUC_check_errdetail("Invalid IP addrress %s", tok);

s/addrress/address/

I've been thinking more about your earlier comment:

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

IMO these should return the "virtual" dst_addr/port, instead of
exposing the physical connection information to the client. That way,
if you intend for your proxy to be transparent, you're not advertising
your network internals to connected clients. It also means that clients
can reasonably expect to be able to reconnect to the addr:port that we
give them, and prevents confusion if the proxy is using an address
family that the client doesn't even support (e.g. the client is IPv4-
only but the proxy connects via IPv6).

--Jacob

#33Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#32)
1 attachment(s)
Re: PROXY protocol support

On Fri, Jul 9, 2021 at 1:42 AM Jacob Champion <pchampion@vmware.com> wrote:

Hi Magnus,

I'm only just starting to page this back into my head, so this is by no
means a full review of the v7 changes -- just stuff I've noticed over
the last day or so of poking around.

On Tue, 2021-06-29 at 11:48 +0200, Magnus Hagander wrote:

On Thu, Mar 11, 2021 at 12:05 AM Jacob Champion <pchampion@vmware.com> wrote:

On Tue, 2021-03-09 at 11:25 +0100, Magnus Hagander wrote:

- \x0 : LOCAL : [...] The receiver must accept this connection as
valid and must use the real connection endpoints and discard the
protocol block including the family which is ignored.

So we should ignore the entire "protocol block" (by which I believe
they mean the protocol-and-address-family byte) in the case of LOCAL,
and just accept it with the original address info intact. That seems to
match the sample code in the back of the spec. The current behavior in
the patch will apply the PROXY behavior incorrectly if the sender sends
a LOCAL header with something other than UNSPEC -- which is strange
behavior but not explicitly prohibited as far as I can see.

Yeah, I think we do the right thing in the "right usecase".

The current implementation is, I think, stricter than the spec asks
for. We're supposed to ignore the family for LOCAL cases, and it's not
clear to me whether we're supposed to also ignore the entire "fam"
family-and-protocol byte (the phrase "protocol block" is not actually
defined in the spec).

It's probably not a big deal in practice, but it could mess with
interoperability for lazier proxy implementations. I think I'll ask the
HAProxy folks for some clarification tomorrow.

Thanks!

Yeah, I have no problem being stricter than necessary, unless that
actually causes any interop problems. It's a lot worse to not be
strict enough..

+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more host names, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.

The documentation mentions that host names are valid in proxy_servers,
but check_proxy_servers() uses the AI_NUMERICHOST hint with
getaddrinfo(), so host names get rejected.

Ah, good point. Should say "ip addresses".

+ GUC_check_errdetail("Invalid IP addrress %s", tok);

s/addrress/address/

Oops.

I've been thinking more about your earlier comment:

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

IMO these should return the "virtual" dst_addr/port, instead of
exposing the physical connection information to the client. That way,
if you intend for your proxy to be transparent, you're not advertising
your network internals to connected clients. It also means that clients
can reasonably expect to be able to reconnect to the addr:port that we
give them, and prevents confusion if the proxy is using an address
family that the client doesn't even support (e.g. the client is IPv4-
only but the proxy connects via IPv6).

That reasoning I think makes a lot of sense, especially with the
comment about being able to connect back to it.

The question at that point extends to, would we also add extra
functions to get the data on the proxy connection itself? Maybe add a
inet_proxy_addr()/inet_proxy_port()? Better names?

PFA a patch that fixes the above errors, and changes
inet_server_addr()/inet_server_port(). Does not yet add anything to
receive the actual local port in this case.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_8.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_8.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 381d8636ab..778a20a179 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is done over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more ip addresses, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
index 6388385edc..3884412a8b 100644
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -21872,7 +21872,12 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        IP address used to connect to the proxy server.
+       </para>
+       </entry>
       </row>
 
       <row>
@@ -21888,7 +21893,13 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        port used to connect to the proxy server.
+       </para>
+
+       </entry>
       </row>
 
       <row>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index 8cc23ef7fb..65f8e69131 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1696,6 +1696,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 89a5f901aa..a8d6c5fa4c 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -311,13 +311,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -362,10 +362,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -388,7 +388,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -405,7 +405,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -584,16 +584,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -747,6 +747,9 @@ StreamConnection(pgsocket server_fd, Port *port)
 		return STATUS_ERROR;
 	}
 
+	/* copy over to daddr to make sure it's set for the non-proxy case */
+	memcpy(&port->daddr, &port->laddr, sizeof(port->laddr));
+
 	/* select NODELAY and KEEPALIVE options if it's a TCP connection */
 	if (!IS_AF_UNIX(port->laddr.addr.ss_family))
 	{
@@ -1118,7 +1121,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 5a050898fe..5991fa9f2b 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -102,6 +102,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -196,15 +197,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,7 +226,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -586,6 +594,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1135,7 +1144,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1164,17 +1176,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1188,9 +1200,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any TCP/IP sockets")));
 
@@ -1200,7 +1233,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1262,12 +1295,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1278,9 +1311,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1292,7 +1339,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1441,10 +1488,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1733,15 +1780,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1909,7 +1958,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1922,6 +1971,284 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	SockAddr	laddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/*
+	 * Assert the size of the structs that are part of the protocol,
+	 * to defend against strange compilers.
+	 */
+	StaticAssertStmt(sizeof(proxyheader) == 16, "proxy header struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip4) == 12, "proxy address ipv4 struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip6) == 36, "proxy address ipv6 struct has invalid size");
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+	memcpy(&laddr_save, &port->laddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * Proxy command is in the low 4 bits of the first byte.
+	 * 0x00 = local, 0x01 = proxy, all others should be rejected
+	 */
+	if ((proxyheader.ver_cmd & 0x0F) == 0x00)
+	{
+		if (proxyheader.fam != 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family %x for local connection", proxyheader.fam)));
+			return STATUS_ERROR;
+		}
+	}
+	else if ((proxyheader.ver_cmd & 0x0F) == 0x01)
+	{
+		if (proxyheader.fam == 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family 0 for non-local connection")));
+			return STATUS_ERROR;
+		}
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol command: %x", (proxyheader.ver_cmd & 0x0f))));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Connection family */
+	if (proxyheader.fam == 0)
+	{
+		/*
+		 * UNSPEC connection over LOCAL (verified above).
+		 * in this case we just ignore the address included.
+		 */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+
+		port->daddr.addr.ss_family = AF_INET;
+		port->daddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->daddr.addr)->sin_addr.s_addr = proxyaddr.ip4.dst_addr;
+		((struct sockaddr_in *) &port->daddr.addr)->sin_port = proxyaddr.ip4.dst_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+
+
+		port->daddr.addr.ss_family = AF_INET6;
+		port->daddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->daddr.addr)->sin6_addr, proxyaddr.ip6.dst_addr, 16);
+		((struct sockaddr_in6 *) &port->daddr.addr)->sin6_port = proxyaddr.ip6.dst_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		char		proxy_host[NI_MAXHOST];
+		char		proxy_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		proxy_host[0] = '\0';
+		proxy_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&laddr_save.addr, laddr_save.salen,
+									  proxy_host, sizeof(proxy_host),
+									  proxy_port, sizeof(proxy_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s (proxy host=%s port=%s)",
+						remote_host,
+						remote_port,
+						proxy_host,
+						proxy_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2030,7 +2357,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2067,7 +2394,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
 			GSSok = 'G';
 #endif
 
@@ -2579,10 +2906,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4363,6 +4690,33 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * Note: AuthenticationTimeout is applied here while waiting for the
+	 * startup packet, and then again in InitPostgres for the duration of any
+	 * authentication operations.  So a hostile client could tie up the
+	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 *
+	 * Note: because PostgresMain will call InitializeTimeouts again, the
+	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
+	 * since we never use it again after this function.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4414,28 +4768,11 @@ BackendInitialize(Port *port)
 		strspn(remote_host, "0123456789ABCDEFabcdef:") < strlen(remote_host))
 		port->remote_hostname = strdup(remote_host);
 
-	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
-	 *
-	 * Note: AuthenticationTimeout is applied here while waiting for the
-	 * startup packet, and then again in InitPostgres for the duration of any
-	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
-	 *
-	 * Note: because PostgresMain will call InitializeTimeouts again, the
-	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
-	 * since we never use it again after this function.
-	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
-	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
 	/*
 	 * Receive the startup packet (which might turn out to be a cancel request
 	 * packet).
 	 */
+	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/adt/network.c b/src/backend/utils/adt/network.c
index 0ab54316f8..9198a29f51 100644
--- a/src/backend/utils/adt/network.c
+++ b/src/backend/utils/adt/network.c
@@ -1802,6 +1802,8 @@ inet_client_port(PG_FUNCTION_ARGS)
 
 /*
  * IP address that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_addr(PG_FUNCTION_ARGS)
@@ -1813,7 +1815,7 @@ inet_server_addr(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1826,14 +1828,14 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 	local_host[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 local_host, sizeof(local_host),
 							 NULL, 0,
 							 NI_NUMERICHOST | NI_NUMERICSERV);
 	if (ret != 0)
 		PG_RETURN_NULL();
 
-	clean_ipv6_addr(port->laddr.addr.ss_family, local_host);
+	clean_ipv6_addr(port->daddr.addr.ss_family, local_host);
 
 	PG_RETURN_INET_P(network_in(local_host, false));
 }
@@ -1841,6 +1843,8 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 /*
  * port that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_port(PG_FUNCTION_ARGS)
@@ -1852,7 +1856,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1865,7 +1869,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 
 	local_port[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 NULL, 0,
 							 local_port, sizeof(local_port),
 							 NI_NUMERICHOST | NI_NUMERICSERV);
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 480e8cd199..e2e5af2fd8 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -50,10 +50,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -234,6 +236,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2358,6 +2362,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4331,6 +4345,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12536,4 +12561,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP address %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index b696abfe54..2e224fef36 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -61,6 +61,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 02015efe13..e91a4d7607 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,9 +126,11 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
+	SockAddr    daddr;          /* destination addr (postmaster, or proxy server if proxy protocol used) */
 	char	   *remote_host;	/* name (or ip addr) of remote host */
 	char	   *remote_hostname;	/* name (not ip addr) of remote host, if
 									 * available */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index 6c51b2f20f..cdaae030e1 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -63,9 +69,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -78,6 +84,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 extern bool pq_check_connection(void);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 0efdd7c232..2a029ef786 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..edc032d49c
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = get_new_node('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgresNode::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgresNode::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#34Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#33)
Re: PROXY protocol support

On Mon, 2021-07-12 at 18:28 +0200, Magnus Hagander wrote:

Yeah, I have no problem being stricter than necessary, unless that
actually causes any interop problems. It's a lot worse to not be
strict enough..

Agreed. Haven't heard back from the HAProxy mailing list yet, so
staying strict seems reasonable in the meantime. That could always be
rolled back later.

I've been thinking more about your earlier comment:

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

IMO these should return the "virtual" dst_addr/port, instead of
exposing the physical connection information to the client. That way,
if you intend for your proxy to be transparent, you're not advertising
your network internals to connected clients. It also means that clients
can reasonably expect to be able to reconnect to the addr:port that we
give them, and prevents confusion if the proxy is using an address
family that the client doesn't even support (e.g. the client is IPv4-
only but the proxy connects via IPv6).

That reasoning I think makes a lot of sense, especially with the
comment about being able to connect back to it.

The question at that point extends to, would we also add extra
functions to get the data on the proxy connection itself? Maybe add a
inet_proxy_addr()/inet_proxy_port()? Better names?

What's the intended use case? I have trouble viewing those as anything
but information disclosure vectors, but I'm jaded. :)

If the goal is to give a last-ditch debugging tool to someone whose
proxy isn't behaving properly -- though I'd hope the proxy in question
has its own ways to debug its behavior -- maybe they could be locked
behind one of the pg_monitor roles, so that they're only available to
someone who could get that information anyway?

PFA a patch that fixes the above errors, and changes
inet_server_addr()/inet_server_port(). Does not yet add anything to
receive the actual local port in this case.

Looking good in local testing. I'm going to reread the spec with fresh
eyes and do a full review pass, but this is shaping up nicely IMO.

Something that I haven't thought about very hard yet is proxy
authentication, but I think the simple IP authentication will be enough
for a first version. For the Unix socket case, it looks like anyone
currently relying on peer auth will need to switch to a
unix_socket_group/_permissions model. For now, that sounds like a
reasonable v1 restriction, though I think not being able to set the
proxy socket's permissions separately from the "normal" one might lead
to some complications in more advanced setups.

--Jacob

#35Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#34)
Re: PROXY protocol support

On Wed, Jul 14, 2021 at 8:24 PM Jacob Champion <pchampion@vmware.com> wrote:

On Mon, 2021-07-12 at 18:28 +0200, Magnus Hagander wrote:

Yeah, I have no problem being stricter than necessary, unless that
actually causes any interop problems. It's a lot worse to not be
strict enough..

Agreed. Haven't heard back from the HAProxy mailing list yet, so
staying strict seems reasonable in the meantime. That could always be
rolled back later.

Any further feedback from them now, two months later? :)

(Sorry, I was out on vacation for the end of the last CF, so didn't
get around to this one, but it seemed there'd be plenty of time in
this CF)

I've been thinking more about your earlier comment:

An interesting thing is what to do about
inet_server_addr/inet_server_port. That sort of loops back up to the
original question of where/how to expose the information about the
proxy in general (since right now it just logs). Right now you can
actually use inet_server_port() to see if the connection was proxied
(as long as it was over tcp).

IMO these should return the "virtual" dst_addr/port, instead of
exposing the physical connection information to the client. That way,
if you intend for your proxy to be transparent, you're not advertising
your network internals to connected clients. It also means that clients
can reasonably expect to be able to reconnect to the addr:port that we
give them, and prevents confusion if the proxy is using an address
family that the client doesn't even support (e.g. the client is IPv4-
only but the proxy connects via IPv6).

That reasoning I think makes a lot of sense, especially with the
comment about being able to connect back to it.

The question at that point extends to, would we also add extra
functions to get the data on the proxy connection itself? Maybe add a
inet_proxy_addr()/inet_proxy_port()? Better names?

What's the intended use case? I have trouble viewing those as anything
but information disclosure vectors, but I'm jaded. :)

"Covering all the bases"?

I'm not entirely sure what the point is of the *existing* functions
for that though, so I'm definitely not wedded to including it.

If the goal is to give a last-ditch debugging tool to someone whose
proxy isn't behaving properly -- though I'd hope the proxy in question
has its own ways to debug its behavior -- maybe they could be locked
behind one of the pg_monitor roles, so that they're only available to
someone who could get that information anyway?

Yeah, agreed.

PFA a patch that fixes the above errors, and changes
inet_server_addr()/inet_server_port(). Does not yet add anything to
receive the actual local port in this case.

Looking good in local testing. I'm going to reread the spec with fresh
eyes and do a full review pass, but this is shaping up nicely IMO.

Thanks!

Something that I haven't thought about very hard yet is proxy
authentication, but I think the simple IP authentication will be enough
for a first version. For the Unix socket case, it looks like anyone
currently relying on peer auth will need to switch to a
unix_socket_group/_permissions model. For now, that sounds like a
reasonable v1 restriction, though I think not being able to set the
proxy socket's permissions separately from the "normal" one might lead
to some complications in more advanced setups.

Agreed in principle, but I think those are some quite uncommon
usecases, so definitely something we don't need to cover in a first
feature.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#36Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#35)
Re: PROXY protocol support

On Tue, 2021-09-07 at 12:24 +0200, Magnus Hagander wrote:

On Wed, Jul 14, 2021 at 8:24 PM Jacob Champion <pchampion@vmware.com> wrote:

On Mon, 2021-07-12 at 18:28 +0200, Magnus Hagander wrote:

Yeah, I have no problem being stricter than necessary, unless that
actually causes any interop problems. It's a lot worse to not be
strict enough..

Agreed. Haven't heard back from the HAProxy mailing list yet, so
staying strict seems reasonable in the meantime. That could always be
rolled back later.

Any further feedback from them now, two months later? :)

Not yet :( I've bumped the thread; in the meantime I still think the
stricter operation is fine, since in the worst case you just make it
less strict in the future.

(Sorry, I was out on vacation for the end of the last CF, so didn't
get around to this one, but it seemed there'd be plenty of time in
this CF)

No worries!

The question at that point extends to, would we also add extra
functions to get the data on the proxy connection itself? Maybe add a
inet_proxy_addr()/inet_proxy_port()? Better names?

What's the intended use case? I have trouble viewing those as anything
but information disclosure vectors, but I'm jaded. :)

"Covering all the bases"?

I'm not entirely sure what the point is of the *existing* functions
for that though, so I'm definitely not wedded to including it.

I guess I'm in the same boat. I'm probably not the right person to
weigh in.

Looking good in local testing. I'm going to reread the spec with fresh
eyes and do a full review pass, but this is shaping up nicely IMO.

Thanks!

I still owe you that overall review. Hoping to get to it this week.

Something that I haven't thought about very hard yet is proxy
authentication, but I think the simple IP authentication will be enough
for a first version. For the Unix socket case, it looks like anyone
currently relying on peer auth will need to switch to a
unix_socket_group/_permissions model. For now, that sounds like a
reasonable v1 restriction, though I think not being able to set the
proxy socket's permissions separately from the "normal" one might lead
to some complications in more advanced setups.

Agreed in principle, but I think those are some quite uncommon
usecases, so definitely something we don't need to cover in a first
feature.

Hm. I guess I'm overly optimistic that "properly securing your
database" is not such an uncommon case, but... :)

--Jacob

#37Jacob Champion
pchampion@vmware.com
In reply to: Jacob Champion (#36)
Re: PROXY protocol support

On Wed, 2021-09-08 at 18:51 +0000, Jacob Champion wrote:

I still owe you that overall review. Hoping to get to it this week.

And here it is. I focused on things other than UnwrapProxyConnection()
for this round, since I think that piece is looking solid.

+       if (port->isProxy)
+       {
+               ereport(LOG,
+                               (errcode_for_socket_access(),
+                                errmsg("Ident authentication cannot be used over PROXY connections")));

What are the rules on COMMERROR vs LOG when dealing with authentication
code? I always thought COMMERROR was required, but I see now that LOG
(among others) is suppressed to the client during authentication.

#ifdef USE_SSL
/* No SSL when disabled or on Unix sockets */
-               if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+               if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
SSLok = 'N';
else
SSLok = 'S';            /* Support for SSL */
@@ -2087,7 +2414,7 @@ retry1:
#ifdef ENABLE_GSS
/* No GSSAPI encryption when on Unix socket */
-               if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+               if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
GSSok = 'G';

Now that we have port->daddr, could these checks be simplified to just
IS_AF_UNIX(port->daddr...)? Or is there a corner case I'm missing for
the port->isProxy case?

+    * Note: AuthenticationTimeout is applied here while waiting for the
+    * startup packet, and then again in InitPostgres for the duration of any
+    * authentication operations.  So a hostile client could tie up the
+    * process for nearly twice AuthenticationTimeout before we kick him off.

This comment needs to be adjusted after the move; waiting for the
startup packet comes later, and it looks like we can now tie up 3x the
timeout for the proxy case.

+       /* Check if this is a proxy connection and if so unwrap the proxying */
+       if (port->isProxy)
+       {
+               enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+               if (UnwrapProxyConnection(port) != STATUS_OK)
+                       proc_exit(0);

I think the timeout here could comfortably be substantially less than
the overall authentication timeout, since the proxy should send its
header immediately even if the client takes its time with the startup
packet. The spec suggests allowing 3 seconds minimum to cover a
retransmission. Maybe something to tune in the future?

+                       /* Also listen to the PROXY port on this address, if configured */
+                       if (ProxyPortNumber)
+                       {
+                               if (strcmp(curhost, "*") == 0)
+                                       socket = StreamServerPort(AF_UNSPEC, NULL,
+                                                                                         (unsigned short) ProxyPortNumber,
+                                                                                         NULL,
+                                                                                         ListenSocket, MAXLISTEN);

Sorry if you already talked about this upthread somewhere, but it looks
like another downside of treating "proxy mode" as a server-wide on/off
switch is that it cuts the effective MAXLISTEN in half, from 64 to 32,
since we're opening two sockets for every address. If I've understood
that correctly, it might be worth mentioning in the docs.

-               if (!success && elemlist != NIL)
+               if (socket == NULL && elemlist != NIL)
ereport(FATAL,
(errmsg("could not create any TCP/IP sockets")));

With this change in PostmasterMain, it looks like `success` is no
longer a useful variable. But I'm not convinced that this is the
correct logic -- this is just checking to see if the last socket
creation succeeded, as opposed to seeing if any of them succeeded. Is
that what you intended?

+plan tests => 25;
+
+my $node = get_new_node('node');

The TAP test will need to be rebased over the changes in 201a76183e.

--Jacob

#38Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#37)
1 attachment(s)
Re: PROXY protocol support

On Fri, Sep 10, 2021 at 1:44 AM Jacob Champion <pchampion@vmware.com> wrote:

On Wed, 2021-09-08 at 18:51 +0000, Jacob Champion wrote:

I still owe you that overall review. Hoping to get to it this week.

And here it is. I focused on things other than UnwrapProxyConnection()
for this round, since I think that piece is looking solid.

Thanks!

+       if (port->isProxy)
+       {
+               ereport(LOG,
+                               (errcode_for_socket_access(),
+                                errmsg("Ident authentication cannot be used over PROXY connections")));

What are the rules on COMMERROR vs LOG when dealing with authentication
code? I always thought COMMERROR was required, but I see now that LOG
(among others) is suppressed to the client during authentication.

I honestly don't know :) In this case, LOG is what's used for all the
other message in errors in ident_inet(), so I picked it for
consistency.

#ifdef USE_SSL
/* No SSL when disabled or on Unix sockets */
-               if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+               if (!LoadedSSL || (IS_AF_UNIX(port->laddr.addr.ss_family) && !port->isProxy))
SSLok = 'N';
else
SSLok = 'S';            /* Support for SSL */
@@ -2087,7 +2414,7 @@ retry1:
#ifdef ENABLE_GSS
/* No GSSAPI encryption when on Unix socket */
-               if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+               if (!IS_AF_UNIX(port->laddr.addr.ss_family) || port->isProxy)
GSSok = 'G';

Now that we have port->daddr, could these checks be simplified to just
IS_AF_UNIX(port->daddr...)? Or is there a corner case I'm missing for
the port->isProxy case?

Yeah, I think they could.

+    * Note: AuthenticationTimeout is applied here while waiting for the
+    * startup packet, and then again in InitPostgres for the duration of any
+    * authentication operations.  So a hostile client could tie up the
+    * process for nearly twice AuthenticationTimeout before we kick him off.

This comment needs to be adjusted after the move; waiting for the
startup packet comes later, and it looks like we can now tie up 3x the
timeout for the proxy case.

Good point.

+       /* Check if this is a proxy connection and if so unwrap the proxying */
+       if (port->isProxy)
+       {
+               enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+               if (UnwrapProxyConnection(port) != STATUS_OK)
+                       proc_exit(0);

I think the timeout here could comfortably be substantially less than
the overall authentication timeout, since the proxy should send its
header immediately even if the client takes its time with the startup
packet. The spec suggests allowing 3 seconds minimum to cover a
retransmission. Maybe something to tune in the future?

Maybe. I'll leave it with a new comment for now about us diong it, and
that we may want to consider igt in the future.

+                       /* Also listen to the PROXY port on this address, if configured */
+                       if (ProxyPortNumber)
+                       {
+                               if (strcmp(curhost, "*") == 0)
+                                       socket = StreamServerPort(AF_UNSPEC, NULL,
+                                                                                         (unsigned short) ProxyPortNumber,
+                                                                                         NULL,
+                                                                                         ListenSocket, MAXLISTEN);

Sorry if you already talked about this upthread somewhere, but it looks
like another downside of treating "proxy mode" as a server-wide on/off
switch is that it cuts the effective MAXLISTEN in half, from 64 to 32,
since we're opening two sockets for every address. If I've understood
that correctly, it might be worth mentioning in the docs.

Correct. I don't see a way to avoid that without complicating things
(as long as we want the ports to be separate), but I also don't see it
as something that's reality to be an issue in reality.

I would agree with documenting it, but I can't actually find us
documenting the MAXLISTEN value anywhere. Do we?

-               if (!success && elemlist != NIL)
+               if (socket == NULL && elemlist != NIL)
ereport(FATAL,
(errmsg("could not create any TCP/IP sockets")));

With this change in PostmasterMain, it looks like `success` is no
longer a useful variable. But I'm not convinced that this is the
correct logic -- this is just checking to see if the last socket
creation succeeded, as opposed to seeing if any of them succeeded. Is
that what you intended?

Eh, no, that's clearly a code-moving-bug.

I think the reasonable thing is to succeed if we create either a
regular socket *or* a proxy one, but FATAL out if you configured
either of them but we failed co create any.

+plan tests => 25;
+
+my $node = get_new_node('node');

The TAP test will need to be rebased over the changes in 201a76183e.

Done, and adjustments above according to your comments, along with a
small docs fix "a proxy connection is done" -> "a proxy connection is
made".

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_9.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_9.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 0a8e35c59f..b1b3613fd1 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is made over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more ip addresses, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
index 78812b2dbe..8af4d8b69c 100644
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22213,7 +22213,12 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        IP address used to connect to the proxy server.
+       </para>
+       </entry>
       </row>
 
       <row>
@@ -22229,7 +22234,13 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        port used to connect to the proxy server.
+       </para>
+
+       </entry>
       </row>
 
       <row>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index a317aef1c9..f8c32ad492 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1696,6 +1696,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 89a5f901aa..a8d6c5fa4c 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -311,13 +311,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -362,10 +362,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -388,7 +388,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -405,7 +405,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -584,16 +584,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -747,6 +747,9 @@ StreamConnection(pgsocket server_fd, Port *port)
 		return STATUS_ERROR;
 	}
 
+	/* copy over to daddr to make sure it's set for the non-proxy case */
+	memcpy(&port->daddr, &port->laddr, sizeof(port->laddr));
+
 	/* select NODELAY and KEEPALIVE options if it's a TCP connection */
 	if (!IS_AF_UNIX(port->laddr.addr.ss_family))
 	{
@@ -1118,7 +1121,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index e2a76ba055..797a317a04 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -101,6 +101,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -196,15 +197,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,7 +226,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -586,6 +594,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1185,7 +1194,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1214,17 +1226,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1238,6 +1250,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+				{
+					success++;
+					socket->isProxy = true;
+				}
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
 		if (!success && elemlist != NIL)
@@ -1250,7 +1286,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1312,12 +1348,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1328,9 +1364,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1342,7 +1392,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1497,10 +1547,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1789,15 +1839,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1965,7 +2017,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1978,6 +2030,284 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	SockAddr	laddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/*
+	 * Assert the size of the structs that are part of the protocol,
+	 * to defend against strange compilers.
+	 */
+	StaticAssertStmt(sizeof(proxyheader) == 16, "proxy header struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip4) == 12, "proxy address ipv4 struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip6) == 36, "proxy address ipv6 struct has invalid size");
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+	memcpy(&laddr_save, &port->laddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * Proxy command is in the low 4 bits of the first byte.
+	 * 0x00 = local, 0x01 = proxy, all others should be rejected
+	 */
+	if ((proxyheader.ver_cmd & 0x0F) == 0x00)
+	{
+		if (proxyheader.fam != 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family %x for local connection", proxyheader.fam)));
+			return STATUS_ERROR;
+		}
+	}
+	else if ((proxyheader.ver_cmd & 0x0F) == 0x01)
+	{
+		if (proxyheader.fam == 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family 0 for non-local connection")));
+			return STATUS_ERROR;
+		}
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol command: %x", (proxyheader.ver_cmd & 0x0f))));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Connection family */
+	if (proxyheader.fam == 0)
+	{
+		/*
+		 * UNSPEC connection over LOCAL (verified above).
+		 * in this case we just ignore the address included.
+		 */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+
+		port->daddr.addr.ss_family = AF_INET;
+		port->daddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->daddr.addr)->sin_addr.s_addr = proxyaddr.ip4.dst_addr;
+		((struct sockaddr_in *) &port->daddr.addr)->sin_port = proxyaddr.ip4.dst_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+
+
+		port->daddr.addr.ss_family = AF_INET6;
+		port->daddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->daddr.addr)->sin6_addr, proxyaddr.ip6.dst_addr, 16);
+		((struct sockaddr_in6 *) &port->daddr.addr)->sin6_port = proxyaddr.ip6.dst_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		char		proxy_host[NI_MAXHOST];
+		char		proxy_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		proxy_host[0] = '\0';
+		proxy_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&laddr_save.addr, laddr_save.salen,
+									  proxy_host, sizeof(proxy_host),
+									  proxy_port, sizeof(proxy_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s (proxy host=%s port=%s)",
+						remote_host,
+						remote_port,
+						proxy_host,
+						proxy_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2086,7 +2416,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || IS_AF_UNIX(port->daddr.addr.ss_family))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2123,7 +2453,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->daddr.addr.ss_family))
 			GSSok = 'G';
 #endif
 
@@ -2635,10 +2965,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4422,6 +4752,31 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * If this is a proxy connection, we apply the timeout once while waiting
+	 * for the proxy header. It is then reapplied further down when we process
+	 * the startup packet, which means it can apply multiple times.
+	 *
+	 * For the time being we re-use AuthenticationTimeout for this, but it may
+	 * be considered for a separate tunable in the future.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4474,27 +4829,20 @@ BackendInitialize(Port *port)
 		port->remote_hostname = strdup(remote_host);
 
 	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
+	 * Receive the startup packet (which might turn out to be a cancel request
+	 * packet).
 	 *
 	 * Note: AuthenticationTimeout is applied here while waiting for the
 	 * startup packet, and then again in InitPostgres for the duration of any
 	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 * process for nearly twice (or three times in the case of a proxy connection)
+	 * AuthenticationTimeout before we kick him off.
 	 *
 	 * Note: because PostgresMain will call InitializeTimeouts again, the
 	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
 	 * since we never use it again after this function.
 	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
 	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
-	/*
-	 * Receive the startup packet (which might turn out to be a cancel request
-	 * packet).
-	 */
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/adt/network.c b/src/backend/utils/adt/network.c
index 0ab54316f8..9198a29f51 100644
--- a/src/backend/utils/adt/network.c
+++ b/src/backend/utils/adt/network.c
@@ -1802,6 +1802,8 @@ inet_client_port(PG_FUNCTION_ARGS)
 
 /*
  * IP address that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_addr(PG_FUNCTION_ARGS)
@@ -1813,7 +1815,7 @@ inet_server_addr(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1826,14 +1828,14 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 	local_host[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 local_host, sizeof(local_host),
 							 NULL, 0,
 							 NI_NUMERICHOST | NI_NUMERICSERV);
 	if (ret != 0)
 		PG_RETURN_NULL();
 
-	clean_ipv6_addr(port->laddr.addr.ss_family, local_host);
+	clean_ipv6_addr(port->daddr.addr.ss_family, local_host);
 
 	PG_RETURN_INET_P(network_in(local_host, false));
 }
@@ -1841,6 +1843,8 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 /*
  * port that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_port(PG_FUNCTION_ARGS)
@@ -1852,7 +1856,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1865,7 +1869,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 
 	local_port[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 NULL, 0,
 							 local_port, sizeof(local_port),
 							 NI_NUMERICHOST | NI_NUMERICSERV);
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index d2ce4a8450..dc8a44b0af 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -50,10 +50,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -234,6 +236,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2382,6 +2386,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4355,6 +4369,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12559,4 +12584,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP address %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index 3fe9a53cb3..aec19705ab 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -62,6 +62,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 02015efe13..e91a4d7607 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,9 +126,11 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
+	SockAddr    daddr;          /* destination addr (postmaster, or proxy server if proxy protocol used) */
 	char	   *remote_host;	/* name (or ip addr) of remote host */
 	char	   *remote_hostname;	/* name (not ip addr) of remote host, if
 									 * available */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index 6c51b2f20f..cdaae030e1 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -63,9 +69,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -78,6 +84,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 extern bool pq_check_connection(void);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 0efdd7c232..2a029ef786 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..ad560da6f7
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use TestLib;
+use PostgresNode;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = PostgresNode->new('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgresNode::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgresNode::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#39Daniel Gustafsson
daniel@yesql.se
In reply to: Magnus Hagander (#38)
Re: PROXY protocol support

On 28 Sep 2021, at 15:23, Magnus Hagander <magnus@hagander.net> wrote:
On Fri, Sep 10, 2021 at 1:44 AM Jacob Champion <pchampion@vmware.com> wrote:

The TAP test will need to be rebased over the changes in 201a76183e.

Done

And now the TAP test will need to be rebased over the changes in
b3b4d8e68ae83f432f43f035c7eb481ef93e1583.

--
Daniel Gustafsson https://vmware.com/

#40Magnus Hagander
magnus@hagander.net
In reply to: Daniel Gustafsson (#39)
1 attachment(s)
Re: PROXY protocol support

On Wed, Nov 3, 2021 at 2:36 PM Daniel Gustafsson <daniel@yesql.se> wrote:

On 28 Sep 2021, at 15:23, Magnus Hagander <magnus@hagander.net> wrote:
On Fri, Sep 10, 2021 at 1:44 AM Jacob Champion <pchampion@vmware.com>

wrote:

The TAP test will need to be rebased over the changes in 201a76183e.

Done

And now the TAP test will need to be rebased over the changes in
b3b4d8e68ae83f432f43f035c7eb481ef93e1583.

Thanks for the pointer, PFA a rebase.

--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/&gt;
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/&gt;

Attachments:

proxy_protocol_10.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_10.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index de77f14573..5211c1f7b1 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is made over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more ip addresses, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
index 4b49dff2ff..74f38d4891 100644
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22238,7 +22238,12 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        IP address used to connect to the proxy server.
+       </para>
+       </entry>
       </row>
 
       <row>
@@ -22254,7 +22259,13 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        port used to connect to the proxy server.
+       </para>
+
+       </entry>
       </row>
 
       <row>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index a317aef1c9..f8c32ad492 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1696,6 +1696,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 89a5f901aa..a8d6c5fa4c 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -311,13 +311,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -362,10 +362,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -388,7 +388,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -405,7 +405,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -584,16 +584,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -747,6 +747,9 @@ StreamConnection(pgsocket server_fd, Port *port)
 		return STATUS_ERROR;
 	}
 
+	/* copy over to daddr to make sure it's set for the non-proxy case */
+	memcpy(&port->daddr, &port->laddr, sizeof(port->laddr));
+
 	/* select NODELAY and KEEPALIVE options if it's a TCP connection */
 	if (!IS_AF_UNIX(port->laddr.addr.ss_family))
 	{
@@ -1118,7 +1121,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index e2a76ba055..797a317a04 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -101,6 +101,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -196,15 +197,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -218,7 +226,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -586,6 +594,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1185,7 +1194,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1214,17 +1226,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1238,6 +1250,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+				{
+					success++;
+					socket->isProxy = true;
+				}
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
 		if (!success && elemlist != NIL)
@@ -1250,7 +1286,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1312,12 +1348,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1328,9 +1364,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1342,7 +1392,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1497,10 +1547,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1789,15 +1839,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1965,7 +2017,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1978,6 +2030,284 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	SockAddr	laddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/*
+	 * Assert the size of the structs that are part of the protocol,
+	 * to defend against strange compilers.
+	 */
+	StaticAssertStmt(sizeof(proxyheader) == 16, "proxy header struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip4) == 12, "proxy address ipv4 struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip6) == 36, "proxy address ipv6 struct has invalid size");
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+	memcpy(&laddr_save, &port->laddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * Proxy command is in the low 4 bits of the first byte.
+	 * 0x00 = local, 0x01 = proxy, all others should be rejected
+	 */
+	if ((proxyheader.ver_cmd & 0x0F) == 0x00)
+	{
+		if (proxyheader.fam != 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family %x for local connection", proxyheader.fam)));
+			return STATUS_ERROR;
+		}
+	}
+	else if ((proxyheader.ver_cmd & 0x0F) == 0x01)
+	{
+		if (proxyheader.fam == 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family 0 for non-local connection")));
+			return STATUS_ERROR;
+		}
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol command: %x", (proxyheader.ver_cmd & 0x0f))));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Connection family */
+	if (proxyheader.fam == 0)
+	{
+		/*
+		 * UNSPEC connection over LOCAL (verified above).
+		 * in this case we just ignore the address included.
+		 */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+
+		port->daddr.addr.ss_family = AF_INET;
+		port->daddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->daddr.addr)->sin_addr.s_addr = proxyaddr.ip4.dst_addr;
+		((struct sockaddr_in *) &port->daddr.addr)->sin_port = proxyaddr.ip4.dst_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+
+
+		port->daddr.addr.ss_family = AF_INET6;
+		port->daddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->daddr.addr)->sin6_addr, proxyaddr.ip6.dst_addr, 16);
+		((struct sockaddr_in6 *) &port->daddr.addr)->sin6_port = proxyaddr.ip6.dst_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		char		proxy_host[NI_MAXHOST];
+		char		proxy_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		proxy_host[0] = '\0';
+		proxy_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&laddr_save.addr, laddr_save.salen,
+									  proxy_host, sizeof(proxy_host),
+									  proxy_port, sizeof(proxy_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s (proxy host=%s port=%s)",
+						remote_host,
+						remote_port,
+						proxy_host,
+						proxy_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2086,7 +2416,7 @@ ProcessStartupPacket(Port *port, bool ssl_done, bool gss_done)
 
 #ifdef USE_SSL
 		/* No SSL when disabled or on Unix sockets */
-		if (!LoadedSSL || IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!LoadedSSL || IS_AF_UNIX(port->daddr.addr.ss_family))
 			SSLok = 'N';
 		else
 			SSLok = 'S';		/* Support for SSL */
@@ -2123,7 +2453,7 @@ retry1:
 
 #ifdef ENABLE_GSS
 		/* No GSSAPI encryption when on Unix socket */
-		if (!IS_AF_UNIX(port->laddr.addr.ss_family))
+		if (!IS_AF_UNIX(port->daddr.addr.ss_family))
 			GSSok = 'G';
 #endif
 
@@ -2635,10 +2965,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4422,6 +4752,31 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * If this is a proxy connection, we apply the timeout once while waiting
+	 * for the proxy header. It is then reapplied further down when we process
+	 * the startup packet, which means it can apply multiple times.
+	 *
+	 * For the time being we re-use AuthenticationTimeout for this, but it may
+	 * be considered for a separate tunable in the future.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4474,27 +4829,20 @@ BackendInitialize(Port *port)
 		port->remote_hostname = strdup(remote_host);
 
 	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
+	 * Receive the startup packet (which might turn out to be a cancel request
+	 * packet).
 	 *
 	 * Note: AuthenticationTimeout is applied here while waiting for the
 	 * startup packet, and then again in InitPostgres for the duration of any
 	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 * process for nearly twice (or three times in the case of a proxy connection)
+	 * AuthenticationTimeout before we kick him off.
 	 *
 	 * Note: because PostgresMain will call InitializeTimeouts again, the
 	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
 	 * since we never use it again after this function.
 	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
 	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
-	/*
-	 * Receive the startup packet (which might turn out to be a cancel request
-	 * packet).
-	 */
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/adt/network.c b/src/backend/utils/adt/network.c
index 0ab54316f8..9198a29f51 100644
--- a/src/backend/utils/adt/network.c
+++ b/src/backend/utils/adt/network.c
@@ -1802,6 +1802,8 @@ inet_client_port(PG_FUNCTION_ARGS)
 
 /*
  * IP address that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_addr(PG_FUNCTION_ARGS)
@@ -1813,7 +1815,7 @@ inet_server_addr(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1826,14 +1828,14 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 	local_host[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 local_host, sizeof(local_host),
 							 NULL, 0,
 							 NI_NUMERICHOST | NI_NUMERICSERV);
 	if (ret != 0)
 		PG_RETURN_NULL();
 
-	clean_ipv6_addr(port->laddr.addr.ss_family, local_host);
+	clean_ipv6_addr(port->daddr.addr.ss_family, local_host);
 
 	PG_RETURN_INET_P(network_in(local_host, false));
 }
@@ -1841,6 +1843,8 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 /*
  * port that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_port(PG_FUNCTION_ARGS)
@@ -1852,7 +1856,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1865,7 +1869,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 
 	local_port[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 NULL, 0,
 							 local_port, sizeof(local_port),
 							 NI_NUMERICHOST | NI_NUMERICSERV);
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index e91d5a3cfd..6aa0e7fca6 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -50,10 +50,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -235,6 +237,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2383,6 +2387,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4368,6 +4382,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12572,4 +12597,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP address %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index 1cbc9feeb6..3e67fa6b3c 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -62,6 +62,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index 02015efe13..e91a4d7607 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,9 +126,11 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
+	SockAddr    daddr;          /* destination addr (postmaster, or proxy server if proxy protocol used) */
 	char	   *remote_host;	/* name (or ip addr) of remote host */
 	char	   *remote_hostname;	/* name (not ip addr) of remote host, if
 									 * available */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index 6c51b2f20f..cdaae030e1 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -63,9 +69,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetSocketPos 0
 #define FeBeWaitSetLatchPos 1
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -78,6 +84,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
 extern bool pq_check_connection(void);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 0efdd7c232..2a029ef786 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..64619058c8
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgreSQL::Test::Cluster::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgreSQL::Test::Cluster::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#41Jacob Champion
pchampion@vmware.com
In reply to: Magnus Hagander (#40)
Re: PROXY protocol support

On Thu, 2021-11-04 at 12:03 +0100, Magnus Hagander wrote:

Thanks for the pointer, PFA a rebase.

I think the Unix socket handling needs the same "success" fix that you
applied to the TCP socket handling above it:

@@ -1328,9 +1364,23 @@ PostmasterMain(int argc, char *argv[])
ereport(WARNING,
(errmsg("could not create Unix-domain socket in directory \"%s\"",
socketdir)));
+
+           if (ProxyPortNumber)
+           {
+               socket = StreamServerPort(AF_UNIX, NULL,
+                                         (unsigned short) ProxyPortNumber,
+                                         socketdir,
+                                         ListenSocket, MAXLISTEN);
+               if (socket)
+                   socket->isProxy = true;
+               else
+                   ereport(WARNING,
+                           (errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+                                   socketdir)));
+           }
}
-       if (!success && elemlist != NIL)
+       if (socket == NULL && elemlist != NIL)
ereport(FATAL,
(errmsg("could not create any Unix-domain sockets")));

Other than that, I can find nothing else to improve, and I think this
is ready for more eyes than mine. :)

--

To tie off some loose ends from upthread:

I didn't find any MAXLISTEN documentation either, so I guess it's only
a documentation issue if someone runs into it, heh.

I was not able to find any other cases (besides ident) where using
daddr instead of laddr would break things. I am going a bit snow-blind
on the patch, though, and there's a lot of auth code.

I never did hear back from the PROXY spec maintainer on how strict to
be with LOCAL; another contributor did chime in but only to add that
they didn't know the answer. That conversation is at [1]https://www.mail-archive.com/haproxy@formilux.org/msg40899.html, in case
someone picks it up in the future.

A summary of possible improvements talked about upthread, for a future
v2:

- SQL functions to get the laddr info (scoped to superusers, somehow),
if there's a use case for them

- Setting up PROXY Unix socket permissions separately from the "main"
socket

- Allowing PROXY-only communication (disable the "main" port)

Thanks,
--Jacob

[1]: https://www.mail-archive.com/haproxy@formilux.org/msg40899.html

#42Magnus Hagander
magnus@hagander.net
In reply to: Jacob Champion (#41)
1 attachment(s)
Re: PROXY protocol support

On Tue, Nov 16, 2021 at 12:03 AM Jacob Champion <pchampion@vmware.com> wrote:

On Thu, 2021-11-04 at 12:03 +0100, Magnus Hagander wrote:

Thanks for the pointer, PFA a rebase.

I think the Unix socket handling needs the same "success" fix that you
applied to the TCP socket handling above it:

@@ -1328,9 +1364,23 @@ PostmasterMain(int argc, char *argv[])
ereport(WARNING,
(errmsg("could not create Unix-domain socket in directory \"%s\"",
socketdir)));
+
+           if (ProxyPortNumber)
+           {
+               socket = StreamServerPort(AF_UNIX, NULL,
+                                         (unsigned short) ProxyPortNumber,
+                                         socketdir,
+                                         ListenSocket, MAXLISTEN);
+               if (socket)
+                   socket->isProxy = true;
+               else
+                   ereport(WARNING,
+                           (errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+                                   socketdir)));
+           }
}
-       if (!success && elemlist != NIL)
+       if (socket == NULL && elemlist != NIL)
ereport(FATAL,
(errmsg("could not create any Unix-domain sockets")));

Other than that, I can find nothing else to improve, and I think this
is ready for more eyes than mine. :)

Here's another rebase on top of the AF_UNIX patch.

To tie off some loose ends from upthread:

I didn't find any MAXLISTEN documentation either, so I guess it's only
a documentation issue if someone runs into it, heh.

I was not able to find any other cases (besides ident) where using
daddr instead of laddr would break things. I am going a bit snow-blind
on the patch, though, and there's a lot of auth code.

Yeah, that's definitely a good reason for more eyes on it.

A summary of possible improvements talked about upthread, for a future
v2:

- SQL functions to get the laddr info (scoped to superusers, somehow),
if there's a use case for them

- Setting up PROXY Unix socket permissions separately from the "main"
socket

- Allowing PROXY-only communication (disable the "main" port)

These all seem useful, but I'm liking the idea of putting them in a
v2, to avoid expanding the scope too much.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

Attachments:

proxy_protocol_11.patchtext/x-patch; charset=US-ASCII; name=proxy_protocol_11.patchDownload
diff --git a/doc/src/sgml/client-auth.sgml b/doc/src/sgml/client-auth.sgml
index 02f0489112..a3ff09b3ac 100644
--- a/doc/src/sgml/client-auth.sgml
+++ b/doc/src/sgml/client-auth.sgml
@@ -353,6 +353,15 @@ hostnogssenc  <replaceable>database</replaceable>  <replaceable>user</replaceabl
        the client's host name instead of the IP address in the log.
       </para>
 
+      <para>
+       If <xref linkend="guc-proxy-port"/> is enabled and the
+       connection is made through a proxy server using the PROXY
+       protocol, the actual IP address of the client will be used
+       for matching. If a connection is made through a proxy server
+       not using the PROXY protocol, the IP address of the
+       proxy server will be used.
+      </para>
+
       <para>
        These fields do not apply to <literal>local</literal> records.
       </para>
diff --git a/doc/src/sgml/config.sgml b/doc/src/sgml/config.sgml
index 7ed8c82a9d..e0847b6347 100644
--- a/doc/src/sgml/config.sgml
+++ b/doc/src/sgml/config.sgml
@@ -682,6 +682,56 @@ include_dir 'conf.d'
       </listitem>
      </varlistentry>
 
+     <varlistentry id="guc-proxy-port" xreflabel="proxy_port">
+      <term><varname>proxy_port</varname> (<type>integer</type>)
+      <indexterm>
+       <primary><varname>proxy_port</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        The TCP port the server listens on for PROXY connections, disabled by
+        default. If set to a number, <productname>PostgreSQL</productname>
+        will listen on this port on the same addresses as for regular
+        connections, but expect all connections to use the PROXY protocol to
+        identify the client.  This parameter can only be set at server start.
+       </para>
+       <para>
+        If a proxy connection is made over this port, and the proxy is listed
+        in <xref linkend="guc-proxy-servers" />, the actual client address
+        will be considered as the address of the client, instead of listing
+        all connections as coming from the proxy server.
+       </para>
+       <para>
+         The <ulink url="http://www.haproxy.org/download/1.9/doc/proxy-protocol.txt">PROXY
+         protocol</ulink> is maintained by <productname>HAProxy</productname>,
+         and supported in many proxies and load
+         balancers. <productname>PostgreSQL</productname> supports version 2
+         of the protocol.
+       </para>
+      </listitem>
+     </varlistentry>
+
+     <varlistentry id="guc-proxy-servers" xreflabel="proxy_servers">
+      <term><varname>proxy_servers</varname> (<type>string</type>)
+      <indexterm>
+       <primary><varname>proxy_servers</varname> configuration parameter</primary>
+      </indexterm>
+      </term>
+      <listitem>
+       <para>
+        A comma separated list of one or more ip addresses, cidr specifications or the
+        literal <literal>unix</literal>, indicating which proxy servers to trust when
+        connecting on the port specified in <xref linkend="guc-proxy-port" />.
+       </para>
+       <para>
+        If a proxy connection is made from an IP address not covered by this
+        list, the connection will be rejected. By default no proxy is trusted
+        and all proxy connections will be rejected.
+       </para>
+      </listitem>
+     </varlistentry>
+
      <varlistentry id="guc-max-connections" xreflabel="max_connections">
       <term><varname>max_connections</varname> (<type>integer</type>)
       <indexterm>
diff --git a/doc/src/sgml/func.sgml b/doc/src/sgml/func.sgml
index df3cd5987b..10dce1beca 100644
--- a/doc/src/sgml/func.sgml
+++ b/doc/src/sgml/func.sgml
@@ -22323,7 +22323,12 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        IP address used to connect to the proxy server.
+       </para>
+       </entry>
       </row>
 
       <row>
@@ -22339,7 +22344,13 @@ SELECT * FROM pg_ls_dir('.') WITH ORDINALITY AS t(ls,n);
         connection,
         or <literal>NULL</literal> if the current connection is via a
         Unix-domain socket.
-       </para></entry>
+       </para>
+       <para>
+        If the connection is a PROXY connection, this function returns the
+        port used to connect to the proxy server.
+       </para>
+
+       </entry>
       </row>
 
       <row>
diff --git a/src/backend/libpq/auth.c b/src/backend/libpq/auth.c
index efc53f3135..cdc20455fe 100644
--- a/src/backend/libpq/auth.c
+++ b/src/backend/libpq/auth.c
@@ -1698,6 +1698,14 @@ ident_inet(hbaPort *port)
 			   *la = NULL,
 				hints;
 
+	if (port->isProxy)
+	{
+		ereport(LOG,
+				(errcode_for_socket_access(),
+				 errmsg("Ident authentication cannot be used over PROXY connections")));
+		return STATUS_ERROR;
+	}
+
 	/*
 	 * Might look a little weird to first convert it to text and then back to
 	 * sockaddr, but it's protocol independent.
diff --git a/src/backend/libpq/pqcomm.c b/src/backend/libpq/pqcomm.c
index 47923b9e9d..abe5b4c5ea 100644
--- a/src/backend/libpq/pqcomm.c
+++ b/src/backend/libpq/pqcomm.c
@@ -326,13 +326,13 @@ socket_close(int code, Datum arg)
  * Successfully opened sockets are added to the ListenSocket[] array (of
  * length MaxListen), at the first position that isn't PGINVALID_SOCKET.
  *
- * RETURNS: STATUS_OK or STATUS_ERROR
+ * RETURNS: The PQlistenSocket listening on, or NULL in case of error
  */
 
-int
+PQlistenSocket *
 StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 				 const char *unixSocketDir,
-				 pgsocket ListenSocket[], int MaxListen)
+				 PQlistenSocket ListenSocket[], int MaxListen)
 {
 	pgsocket	fd;
 	int			err;
@@ -378,10 +378,10 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("Unix-domain socket path \"%s\" is too long (maximum %d bytes)",
 							unixSocketPath,
 							(int) (UNIXSOCK_PATH_BUFLEN - 1))));
-			return STATUS_ERROR;
+			return NULL;
 		}
 		if (Lock_AF_UNIX(unixSocketDir, unixSocketPath) != STATUS_OK)
-			return STATUS_ERROR;
+			return NULL;
 		service = unixSocketPath;
 	}
 	else
@@ -404,7 +404,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 							service, gai_strerror(ret))));
 		if (addrs)
 			pg_freeaddrinfo_all(hint.ai_family, addrs);
-		return STATUS_ERROR;
+		return NULL;
 	}
 
 	for (addr = addrs; addr; addr = addr->ai_next)
@@ -421,7 +421,7 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 		/* See if there is still room to add 1 more socket. */
 		for (; listen_index < MaxListen; listen_index++)
 		{
-			if (ListenSocket[listen_index] == PGINVALID_SOCKET)
+			if (ListenSocket[listen_index].socket == PGINVALID_SOCKET)
 				break;
 		}
 		if (listen_index >= MaxListen)
@@ -600,16 +600,16 @@ StreamServerPort(int family, const char *hostName, unsigned short portNumber,
 					(errmsg("listening on %s address \"%s\", port %d",
 							familyDesc, addrDesc, (int) portNumber)));
 
-		ListenSocket[listen_index] = fd;
+		ListenSocket[listen_index].socket = fd;
 		added++;
 	}
 
 	pg_freeaddrinfo_all(hint.ai_family, addrs);
 
 	if (!added)
-		return STATUS_ERROR;
+		return NULL;
 
-	return STATUS_OK;
+	return &ListenSocket[listen_index];
 }
 
 
@@ -763,6 +763,9 @@ StreamConnection(pgsocket server_fd, Port *port)
 		return STATUS_ERROR;
 	}
 
+	/* copy over to daddr to make sure it's set for the non-proxy case */
+	memcpy(&port->daddr, &port->laddr, sizeof(port->laddr));
+
 	/* select NODELAY and KEEPALIVE options if it's a TCP connection */
 	if (port->laddr.addr.ss_family != AF_UNIX)
 	{
@@ -1134,7 +1137,7 @@ pq_getbytes(char *s, size_t len)
  *		returns 0 if OK, EOF if trouble
  * --------------------------------
  */
-static int
+int
 pq_discardbytes(size_t len)
 {
 	size_t		amount;
diff --git a/src/backend/postmaster/postmaster.c b/src/backend/postmaster/postmaster.c
index 80bb269599..caad6d61a0 100644
--- a/src/backend/postmaster/postmaster.c
+++ b/src/backend/postmaster/postmaster.c
@@ -103,6 +103,7 @@
 #include "common/string.h"
 #include "lib/ilist.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "libpq/pqsignal.h"
@@ -198,15 +199,22 @@ BackgroundWorker *MyBgworkerEntry = NULL;
 
 
 
-/* The socket number we are listening for connections on */
+/* The TCP port number we are listening for connections on */
 int			PostPortNumber;
 
+/* The TCP port number we are listening for proxy connections on */
+int			ProxyPortNumber;
+
 /* The directory names for Unix socket(s) */
 char	   *Unix_socket_directories;
 
 /* The TCP listen address(es) */
 char	   *ListenAddresses;
 
+/* Trusted proxy servers */
+char	   *TrustedProxyServersString = NULL;
+struct sockaddr_storage *TrustedProxyServers = NULL;
+
 /*
  * ReservedBackends is the number of backends reserved for superuser use.
  * This number is taken out of the pool size given by MaxConnections so
@@ -220,7 +228,7 @@ int			ReservedBackends;
 
 /* The socket(s) we're listening to. */
 #define MAXLISTEN	64
-static pgsocket ListenSocket[MAXLISTEN];
+static PQlistenSocket ListenSocket[MAXLISTEN];
 
 /*
  * These globals control the behavior of the postmaster in case some
@@ -588,6 +596,7 @@ PostmasterMain(int argc, char *argv[])
 	bool		listen_addr_saved = false;
 	int			i;
 	char	   *output_config_variable = NULL;
+	PQlistenSocket *socket = NULL;
 
 	InitProcessGlobals();
 
@@ -1185,7 +1194,10 @@ PostmasterMain(int argc, char *argv[])
 	 * charged with closing the sockets again at postmaster shutdown.
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
-		ListenSocket[i] = PGINVALID_SOCKET;
+	{
+		ListenSocket[i].socket = PGINVALID_SOCKET;
+		ListenSocket[i].isProxy = false;
+	}
 
 	on_proc_exit(CloseServerPorts, 0);
 
@@ -1214,17 +1226,17 @@ PostmasterMain(int argc, char *argv[])
 			char	   *curhost = (char *) lfirst(l);
 
 			if (strcmp(curhost, "*") == 0)
-				status = StreamServerPort(AF_UNSPEC, NULL,
+				socket = StreamServerPort(AF_UNSPEC, NULL,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 			else
-				status = StreamServerPort(AF_UNSPEC, curhost,
+				socket = StreamServerPort(AF_UNSPEC, curhost,
 										  (unsigned short) PostPortNumber,
 										  NULL,
 										  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful host addr in lockfile */
@@ -1238,6 +1250,30 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create listen socket for \"%s\"",
 								curhost)));
+
+			/* Also listen to the PROXY port on this address, if configured */
+			if (ProxyPortNumber)
+			{
+				if (strcmp(curhost, "*") == 0)
+					socket = StreamServerPort(AF_UNSPEC, NULL,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				else
+					socket = StreamServerPort(AF_UNSPEC, curhost,
+											  (unsigned short) ProxyPortNumber,
+											  NULL,
+											  ListenSocket, MAXLISTEN);
+				if (socket)
+				{
+					success++;
+					socket->isProxy = true;
+				}
+				else
+					ereport(WARNING,
+							(errmsg("could not create PROXY listen socket for \"%s\"",
+									curhost)));
+			}
 		}
 
 		if (!success && elemlist != NIL)
@@ -1250,7 +1286,7 @@ PostmasterMain(int argc, char *argv[])
 
 #ifdef USE_BONJOUR
 	/* Register for Bonjour only if we opened TCP socket(s) */
-	if (enable_bonjour && ListenSocket[0] != PGINVALID_SOCKET)
+	if (enable_bonjour && ListenSocket[0].socket != PGINVALID_SOCKET)
 	{
 		DNSServiceErrorType err;
 
@@ -1312,12 +1348,12 @@ PostmasterMain(int argc, char *argv[])
 		{
 			char	   *socketdir = (char *) lfirst(l);
 
-			status = StreamServerPort(AF_UNIX, NULL,
+			socket = StreamServerPort(AF_UNIX, NULL,
 									  (unsigned short) PostPortNumber,
 									  socketdir,
 									  ListenSocket, MAXLISTEN);
 
-			if (status == STATUS_OK)
+			if (socket)
 			{
 				success++;
 				/* record the first successful Unix socket in lockfile */
@@ -1328,9 +1364,23 @@ PostmasterMain(int argc, char *argv[])
 				ereport(WARNING,
 						(errmsg("could not create Unix-domain socket in directory \"%s\"",
 								socketdir)));
+
+			if (ProxyPortNumber)
+			{
+				socket = StreamServerPort(AF_UNIX, NULL,
+										  (unsigned short) ProxyPortNumber,
+										  socketdir,
+										  ListenSocket, MAXLISTEN);
+				if (socket)
+					socket->isProxy = true;
+				else
+					ereport(WARNING,
+							(errmsg("could not create Unix-domain PROXY socket for \"%s\"",
+									socketdir)));
+			}
 		}
 
-		if (!success && elemlist != NIL)
+		if (socket == NULL && elemlist != NIL)
 			ereport(FATAL,
 					(errmsg("could not create any Unix-domain sockets")));
 
@@ -1342,7 +1392,7 @@ PostmasterMain(int argc, char *argv[])
 	/*
 	 * check that we have some socket to listen on
 	 */
-	if (ListenSocket[0] == PGINVALID_SOCKET)
+	if (ListenSocket[0].socket == PGINVALID_SOCKET)
 		ereport(FATAL,
 				(errmsg("no socket created for listening")));
 
@@ -1497,10 +1547,10 @@ CloseServerPorts(int status, Datum arg)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -1789,15 +1839,17 @@ ServerLoop(void)
 
 			for (i = 0; i < MAXLISTEN; i++)
 			{
-				if (ListenSocket[i] == PGINVALID_SOCKET)
+				if (ListenSocket[i].socket == PGINVALID_SOCKET)
 					break;
-				if (FD_ISSET(ListenSocket[i], &rmask))
+				if (FD_ISSET(ListenSocket[i].socket, &rmask))
 				{
 					Port	   *port;
 
-					port = ConnCreate(ListenSocket[i]);
+					port = ConnCreate(ListenSocket[i].socket);
 					if (port)
 					{
+						port->isProxy = ListenSocket[i].isProxy;
+
 						BackendStartup(port);
 
 						/*
@@ -1965,7 +2017,7 @@ initMasks(fd_set *rmask)
 
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		int			fd = ListenSocket[i];
+		int			fd = ListenSocket[i].socket;
 
 		if (fd == PGINVALID_SOCKET)
 			break;
@@ -1978,6 +2030,284 @@ initMasks(fd_set *rmask)
 	return maxsock + 1;
 }
 
+static int
+UnwrapProxyConnection(Port *port)
+{
+	char		proxyver;
+	uint16		proxyaddrlen;
+	SockAddr	raddr_save;
+	SockAddr	laddr_save;
+	int			i;
+	bool		useproxy = false;
+
+	/*
+	 * These structs are from the PROXY protocol docs at
+	 * http://www.haproxy.org/download/1.8/doc/proxy-protocol.txt
+	 */
+	union
+	{
+		struct
+		{						/* for TCP/UDP over IPv4, len = 12 */
+			uint32		src_addr;
+			uint32		dst_addr;
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip4;
+		struct
+		{						/* for TCP/UDP over IPv6, len = 36 */
+			uint8		src_addr[16];
+			uint8		dst_addr[16];
+			uint16		src_port;
+			uint16		dst_port;
+		}			ip6;
+	}			proxyaddr;
+	struct
+	{
+		uint8		sig[12];	/* hex 0D 0A 0D 0A 00 0D 0A 51 55 49 54 0A */
+		uint8		ver_cmd;	/* protocol version and command */
+		uint8		fam;		/* protocol family and address */
+		uint16		len;		/* number of following bytes part of the
+								 * header */
+	}			proxyheader;
+
+	/*
+	 * Assert the size of the structs that are part of the protocol,
+	 * to defend against strange compilers.
+	 */
+	StaticAssertStmt(sizeof(proxyheader) == 16, "proxy header struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip4) == 12, "proxy address ipv4 struct has invalid size");
+	StaticAssertStmt(sizeof(proxyaddr.ip6) == 36, "proxy address ipv6 struct has invalid size");
+
+
+	/* Else if it's on our list of trusted proxies */
+	if (TrustedProxyServers)
+	{
+		for (i = 0; i < *((int *) TrustedProxyServers) * 2; i += 2)
+		{
+			if (port->raddr.addr.ss_family == TrustedProxyServers[i + 1].ss_family)
+			{
+				/*
+				 * Connection over unix sockets don't give us the source, so
+				 * just check if they're allowed at all.  For IP connections,
+				 * verify that it's an allowed address.
+				 */
+				if (port->raddr.addr.ss_family == AF_UNIX ||
+					pg_range_sockaddr(&port->raddr.addr,
+									  &TrustedProxyServers[i + 1],
+									  &TrustedProxyServers[i + 2]))
+				{
+					useproxy = true;
+					break;
+				}
+			}
+		}
+	}
+	if (!useproxy)
+	{
+		/*
+		 * Connection is not from one of our trusted proxies, so reject it.
+		 */
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("connection from unauthorized proxy server")));
+		return STATUS_ERROR;
+	}
+
+	/* Store a copy of the original address, for logging */
+	memcpy(&raddr_save, &port->raddr, sizeof(SockAddr));
+	memcpy(&laddr_save, &port->laddr, sizeof(SockAddr));
+
+	pq_startmsgread();
+
+	/*
+	 * PROXY requests always start with:
+	 * \x0D \x0A \x0D \x0A \x00 \x0D \x0A \x51 \x55 \x49 \x54 \x0A
+	 */
+
+	if (pq_getbytes((char *) &proxyheader, sizeof(proxyheader)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	if (memcmp(proxyheader.sig, "\x0d\x0a\x0d\x0a\x00\x0d\x0a\x51\x55\x49\x54\x0a", sizeof(proxyheader.sig)) != 0)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Proxy version is in the high 4 bits of the first byte */
+	proxyver = (proxyheader.ver_cmd & 0xF0) >> 4;
+	if (proxyver != 2)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol version: %x", proxyver)));
+		return STATUS_ERROR;
+	}
+
+	/*
+	 * Proxy command is in the low 4 bits of the first byte.
+	 * 0x00 = local, 0x01 = proxy, all others should be rejected
+	 */
+	if ((proxyheader.ver_cmd & 0x0F) == 0x00)
+	{
+		if (proxyheader.fam != 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family %x for local connection", proxyheader.fam)));
+			return STATUS_ERROR;
+		}
+	}
+	else if ((proxyheader.ver_cmd & 0x0F) == 0x01)
+	{
+		if (proxyheader.fam == 0)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("invalid proxy protocol family 0 for non-local connection")));
+			return STATUS_ERROR;
+		}
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol command: %x", (proxyheader.ver_cmd & 0x0f))));
+		return STATUS_ERROR;
+	}
+
+	proxyaddrlen = pg_ntoh16(proxyheader.len);
+
+	if (pq_getbytes((char *) &proxyaddr, proxyaddrlen > sizeof(proxyaddr) ? sizeof(proxyaddr) : proxyaddrlen) == EOF)
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("incomplete proxy packet")));
+		return STATUS_ERROR;
+	}
+
+	/* Connection family */
+	if (proxyheader.fam == 0)
+	{
+		/*
+		 * UNSPEC connection over LOCAL (verified above).
+		 * in this case we just ignore the address included.
+		 */
+	}
+	else if (proxyheader.fam == 0x11)
+	{
+		/* TCPv4 */
+		if (proxyaddrlen < 12)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET;
+		port->raddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->raddr.addr)->sin_addr.s_addr = proxyaddr.ip4.src_addr;
+		((struct sockaddr_in *) &port->raddr.addr)->sin_port = proxyaddr.ip4.src_port;
+
+		port->daddr.addr.ss_family = AF_INET;
+		port->daddr.salen = sizeof(struct sockaddr_in);
+		((struct sockaddr_in *) &port->daddr.addr)->sin_addr.s_addr = proxyaddr.ip4.dst_addr;
+		((struct sockaddr_in *) &port->daddr.addr)->sin_port = proxyaddr.ip4.dst_port;
+	}
+	else if (proxyheader.fam == 0x21)
+	{
+		/* TCPv6 */
+		if (proxyaddrlen < 36)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+		port->raddr.addr.ss_family = AF_INET6;
+		port->raddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->raddr.addr)->sin6_addr, proxyaddr.ip6.src_addr, 16);
+		((struct sockaddr_in6 *) &port->raddr.addr)->sin6_port = proxyaddr.ip6.src_port;
+
+
+		port->daddr.addr.ss_family = AF_INET6;
+		port->daddr.salen = sizeof(struct sockaddr_in6);
+		memcpy(&((struct sockaddr_in6 *) &port->daddr.addr)->sin6_addr, proxyaddr.ip6.dst_addr, 16);
+		((struct sockaddr_in6 *) &port->daddr.addr)->sin6_port = proxyaddr.ip6.dst_port;
+	}
+	else
+	{
+		ereport(COMMERROR,
+				(errcode(ERRCODE_PROTOCOL_VIOLATION),
+				 errmsg("invalid proxy protocol connection type: %x", proxyheader.fam)));
+		return STATUS_ERROR;
+	}
+
+	/* If there is any more header data present, skip past it */
+	if (proxyaddrlen > sizeof(proxyaddr))
+	{
+		if (pq_discardbytes(proxyaddrlen - sizeof(proxyaddr)) == EOF)
+		{
+			ereport(COMMERROR,
+					(errcode(ERRCODE_PROTOCOL_VIOLATION),
+					 errmsg("incomplete proxy packet")));
+			return STATUS_ERROR;
+		}
+	}
+
+	pq_endmsgread();
+
+	/*
+	 * Log what we've done if connection logging is enabled. We log the proxy
+	 * connection here, and let the normal connection logging mechanism log
+	 * the unwrapped connection.
+	 */
+	if (Log_connections)
+	{
+		char		remote_host[NI_MAXHOST];
+		char		remote_port[NI_MAXSERV];
+		char		proxy_host[NI_MAXHOST];
+		char		proxy_port[NI_MAXSERV];
+		int			ret;
+
+		remote_host[0] = '\0';
+		remote_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&raddr_save.addr, raddr_save.salen,
+									  remote_host, sizeof(remote_host),
+									  remote_port, sizeof(remote_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+		proxy_host[0] = '\0';
+		proxy_port[0] = '\0';
+		if ((ret = pg_getnameinfo_all(&laddr_save.addr, laddr_save.salen,
+									  proxy_host, sizeof(proxy_host),
+									  proxy_port, sizeof(proxy_port),
+									  (log_hostname ? 0 : NI_NUMERICHOST) | NI_NUMERICSERV)) != 0)
+			ereport(WARNING,
+					(errmsg_internal("pg_getnameinfo_all() failed: %s",
+									 gai_strerror(ret))));
+
+
+		ereport(LOG,
+				(errmsg("proxy connection from: host=%s port=%s (proxy host=%s port=%s)",
+						remote_host,
+						remote_port,
+						proxy_host,
+						proxy_port)));
+
+	}
+
+	return STATUS_OK;
+}
 
 /*
  * Read a client's startup packet and do something according to it.
@@ -2659,10 +2989,10 @@ ClosePostmasterPorts(bool am_syslogger)
 	 */
 	for (i = 0; i < MAXLISTEN; i++)
 	{
-		if (ListenSocket[i] != PGINVALID_SOCKET)
+		if (ListenSocket[i].socket != PGINVALID_SOCKET)
 		{
-			StreamClose(ListenSocket[i]);
-			ListenSocket[i] = PGINVALID_SOCKET;
+			StreamClose(ListenSocket[i].socket);
+			ListenSocket[i].socket = PGINVALID_SOCKET;
 		}
 	}
 
@@ -4455,6 +4785,31 @@ BackendInitialize(Port *port)
 	InitializeTimeouts();		/* establishes SIGALRM handler */
 	PG_SETMASK(&StartupBlockSig);
 
+	/*
+	 * Ready to begin client interaction.  We will give up and _exit(1) after
+	 * a time delay, so that a broken client can't hog a connection
+	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
+	 * against the time limit.
+	 *
+	 * If this is a proxy connection, we apply the timeout once while waiting
+	 * for the proxy header. It is then reapplied further down when we process
+	 * the startup packet, which means it can apply multiple times.
+	 *
+	 * For the time being we re-use AuthenticationTimeout for this, but it may
+	 * be considered for a separate tunable in the future.
+	 */
+	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
+
+	/* Check if this is a proxy connection and if so unwrap the proxying */
+	if (port->isProxy)
+	{
+		enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
+		if (UnwrapProxyConnection(port) != STATUS_OK)
+			proc_exit(0);
+		disable_timeout(STARTUP_PACKET_TIMEOUT, false);
+	}
+
+
 	/*
 	 * Get the remote host name and port for logging and status display.
 	 */
@@ -4507,27 +4862,20 @@ BackendInitialize(Port *port)
 		port->remote_hostname = strdup(remote_host);
 
 	/*
-	 * Ready to begin client interaction.  We will give up and _exit(1) after
-	 * a time delay, so that a broken client can't hog a connection
-	 * indefinitely.  PreAuthDelay and any DNS interactions above don't count
-	 * against the time limit.
+	 * Receive the startup packet (which might turn out to be a cancel request
+	 * packet).
 	 *
 	 * Note: AuthenticationTimeout is applied here while waiting for the
 	 * startup packet, and then again in InitPostgres for the duration of any
 	 * authentication operations.  So a hostile client could tie up the
-	 * process for nearly twice AuthenticationTimeout before we kick him off.
+	 * process for nearly twice (or three times in the case of a proxy connection)
+	 * AuthenticationTimeout before we kick him off.
 	 *
 	 * Note: because PostgresMain will call InitializeTimeouts again, the
 	 * registration of STARTUP_PACKET_TIMEOUT will be lost.  This is okay
 	 * since we never use it again after this function.
 	 */
-	RegisterTimeout(STARTUP_PACKET_TIMEOUT, StartupPacketTimeoutHandler);
 	enable_timeout_after(STARTUP_PACKET_TIMEOUT, AuthenticationTimeout * 1000);
-
-	/*
-	 * Receive the startup packet (which might turn out to be a cancel request
-	 * packet).
-	 */
 	status = ProcessStartupPacket(port, false, false);
 
 	/*
diff --git a/src/backend/utils/adt/network.c b/src/backend/utils/adt/network.c
index 0ab54316f8..9198a29f51 100644
--- a/src/backend/utils/adt/network.c
+++ b/src/backend/utils/adt/network.c
@@ -1802,6 +1802,8 @@ inet_client_port(PG_FUNCTION_ARGS)
 
 /*
  * IP address that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_addr(PG_FUNCTION_ARGS)
@@ -1813,7 +1815,7 @@ inet_server_addr(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1826,14 +1828,14 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 	local_host[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 local_host, sizeof(local_host),
 							 NULL, 0,
 							 NI_NUMERICHOST | NI_NUMERICSERV);
 	if (ret != 0)
 		PG_RETURN_NULL();
 
-	clean_ipv6_addr(port->laddr.addr.ss_family, local_host);
+	clean_ipv6_addr(port->daddr.addr.ss_family, local_host);
 
 	PG_RETURN_INET_P(network_in(local_host, false));
 }
@@ -1841,6 +1843,8 @@ inet_server_addr(PG_FUNCTION_ARGS)
 
 /*
  * port that the server accepted the connection on (NULL if Unix socket)
+ * If the connection is a PROXY connection, then this returns the IP address/port of
+ * the proxy server, and not the local connection!
  */
 Datum
 inet_server_port(PG_FUNCTION_ARGS)
@@ -1852,7 +1856,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 	if (port == NULL)
 		PG_RETURN_NULL();
 
-	switch (port->laddr.addr.ss_family)
+	switch (port->daddr.addr.ss_family)
 	{
 		case AF_INET:
 #ifdef HAVE_IPV6
@@ -1865,7 +1869,7 @@ inet_server_port(PG_FUNCTION_ARGS)
 
 	local_port[0] = '\0';
 
-	ret = pg_getnameinfo_all(&port->laddr.addr, port->laddr.salen,
+	ret = pg_getnameinfo_all(&port->daddr.addr, port->daddr.salen,
 							 NULL, 0,
 							 local_port, sizeof(local_port),
 							 NI_NUMERICHOST | NI_NUMERICSERV);
diff --git a/src/backend/utils/misc/guc.c b/src/backend/utils/misc/guc.c
index 1e3650184b..c21c39db24 100644
--- a/src/backend/utils/misc/guc.c
+++ b/src/backend/utils/misc/guc.c
@@ -52,10 +52,12 @@
 #include "commands/user.h"
 #include "commands/vacuum.h"
 #include "commands/variable.h"
+#include "common/ip.h"
 #include "common/string.h"
 #include "funcapi.h"
 #include "jit/jit.h"
 #include "libpq/auth.h"
+#include "libpq/ifaddr.h"
 #include "libpq/libpq.h"
 #include "libpq/pqformat.h"
 #include "miscadmin.h"
@@ -239,6 +241,8 @@ static bool check_recovery_target_lsn(char **newval, void **extra, GucSource sou
 static void assign_recovery_target_lsn(const char *newval, void *extra);
 static bool check_primary_slot_name(char **newval, void **extra, GucSource source);
 static bool check_default_with_oids(bool *newval, void **extra, GucSource source);
+static bool check_proxy_servers(char **newval, void **extra, GucSource source);
+static void assign_proxy_servers(const char *newval, void *extra);
 
 /* Private functions in guc-file.l that need to be called from guc.c */
 static ConfigVariable *ProcessConfigFileInternal(GucContext context,
@@ -2401,6 +2405,16 @@ static struct config_int ConfigureNamesInt[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_port", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the TCP port the server listens for PROXY connections on."),
+			NULL
+		},
+		&ProxyPortNumber,
+		0, 0, 65535,
+		NULL, NULL, NULL
+	},
+
 	{
 		{"unix_socket_permissions", PGC_POSTMASTER, CONN_AUTH_SETTINGS,
 			gettext_noop("Sets the access permissions of the Unix-domain socket."),
@@ -4399,6 +4413,17 @@ static struct config_string ConfigureNamesString[] =
 		NULL, NULL, NULL
 	},
 
+	{
+		{"proxy_servers", PGC_SIGHUP, CONN_AUTH_SETTINGS,
+			gettext_noop("Sets the addresses for trusted proxy servers."),
+			NULL,
+			GUC_LIST_INPUT
+		},
+		&TrustedProxyServersString,
+		"",
+		check_proxy_servers, assign_proxy_servers, NULL
+	},
+
 	{
 		/*
 		 * Can't be set by ALTER SYSTEM as it can lead to recursive definition
@@ -12680,4 +12705,118 @@ check_default_with_oids(bool *newval, void **extra, GucSource source)
 	return true;
 }
 
+static bool
+check_proxy_servers(char **newval, void **extra, GucSource source)
+{
+	char	   *rawstring;
+	List	   *elemlist;
+	ListCell   *l;
+	struct sockaddr_storage *myextra;
+
+	/* Special case when it's empty */
+	if (**newval == '\0')
+	{
+		*extra = NULL;
+		return true;
+	}
+
+	/* Need a modifiable copy of string */
+	rawstring = pstrdup(*newval);
+
+	/* Parse string into list of identifiers */
+	if (!SplitIdentifierString(rawstring, ',', &elemlist))
+	{
+		/* syntax error in list */
+		GUC_check_errdetail("List syntax is invalid.");
+		pfree(rawstring);
+		list_free(elemlist);
+		return false;
+	}
+
+	if (list_length(elemlist) == 0)
+	{
+		/* If it had only whitespace */
+		pfree(rawstring);
+		list_free(elemlist);
+
+		*extra = NULL;
+		return true;
+	}
+
+	/*
+	 * We store the result in an array of sockaddr_storage. The first entry is
+	 * just an overloaded int which holds the size of the array.
+	 */
+	myextra = (struct sockaddr_storage *) guc_malloc(ERROR, sizeof(struct sockaddr_storage) * (list_length(elemlist) * 2 + 1));
+	*((int *) &myextra[0]) = list_length(elemlist);
+
+	foreach(l, elemlist)
+	{
+		char	   *tok = (char *) lfirst(l);
+		char	   *netmasktok = NULL;
+		int			ret;
+		struct addrinfo *gai_result;
+		struct addrinfo hints;
+
+		/*
+		 * Unix sockets don't have endpoint addresses, so just flag them as
+		 * AF_UNIX
+		 */
+		if (pg_strcasecmp(tok, "unix") == 0)
+		{
+			myextra[foreach_current_index(l) * 2 + 1].ss_family = AF_UNIX;
+			continue;
+		}
+
+		netmasktok = strchr(tok, '/');
+		if (netmasktok)
+		{
+			*netmasktok = '\0';
+			netmasktok++;
+		}
+
+		memset((char *) &hints, 0, sizeof(hints));
+		hints.ai_flags = AI_NUMERICHOST;
+		hints.ai_family = AF_UNSPEC;
+
+		ret = pg_getaddrinfo_all(tok, NULL, &hints, &gai_result);
+		if (ret != 0 || gai_result == NULL)
+		{
+			GUC_check_errdetail("Invalid IP address %s", tok);
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+
+		memcpy((char *) &myextra[foreach_current_index(l) * 2 + 1], gai_result->ai_addr, gai_result->ai_addrlen);
+		pg_freeaddrinfo_all(hints.ai_family, gai_result);
+
+		/* A NULL netmasktok means the fully set hostmask */
+		if (pg_sockaddr_cidr_mask(&myextra[foreach_current_index(l) * 2 + 2], netmasktok, myextra[foreach_current_index(l) * 2 + 1].ss_family) != 0)
+		{
+			if (netmasktok)
+				GUC_check_errdetail("Invalid netmask %s", netmasktok);
+			else
+				GUC_check_errdetail("Could not create netmask");
+			pfree(rawstring);
+			list_free(elemlist);
+			free(myextra);
+			return false;
+		}
+	}
+
+	pfree(rawstring);
+	list_free(elemlist);
+	*extra = (void *) myextra;
+
+	return true;
+}
+
+static void
+assign_proxy_servers(const char *newval, void *extra)
+{
+	TrustedProxyServers = (struct sockaddr_storage *) extra;
+}
+
 #include "guc-file.c"
diff --git a/src/backend/utils/misc/postgresql.conf.sample b/src/backend/utils/misc/postgresql.conf.sample
index 4a094bb38b..079ab91e9b 100644
--- a/src/backend/utils/misc/postgresql.conf.sample
+++ b/src/backend/utils/misc/postgresql.conf.sample
@@ -62,6 +62,9 @@
 					# defaults to 'localhost'; use '*' for all
 					# (change requires restart)
 #port = 5432				# (change requires restart)
+#proxy_port = 0				# port to listen to for proxy connections
+					# (change requires restart)
+#proxy_servers = ''			# what proxy servers to trust
 #max_connections = 100			# (change requires restart)
 #superuser_reserved_connections = 3	# (change requires restart)
 #unix_socket_directories = '/tmp'	# comma-separated list of directories
diff --git a/src/include/libpq/libpq-be.h b/src/include/libpq/libpq-be.h
index dd3e5efba3..9188e78a88 100644
--- a/src/include/libpq/libpq-be.h
+++ b/src/include/libpq/libpq-be.h
@@ -126,9 +126,11 @@ typedef struct Port
 {
 	pgsocket	sock;			/* File descriptor */
 	bool		noblock;		/* is the socket in non-blocking mode? */
+	bool		isProxy;		/* is the connection using PROXY protocol */
 	ProtocolVersion proto;		/* FE/BE protocol version */
 	SockAddr	laddr;			/* local addr (postmaster) */
 	SockAddr	raddr;			/* remote addr (client) */
+	SockAddr    daddr;          /* destination addr (postmaster, or proxy server if proxy protocol used) */
 	char	   *remote_host;	/* name (or ip addr) of remote host */
 	char	   *remote_hostname;	/* name (not ip addr) of remote host, if
 									 * available */
diff --git a/src/include/libpq/libpq.h b/src/include/libpq/libpq.h
index d348a55812..16dfb47cd6 100644
--- a/src/include/libpq/libpq.h
+++ b/src/include/libpq/libpq.h
@@ -42,6 +42,12 @@ typedef struct
 
 extern const PGDLLIMPORT PQcommMethods *PqCommMethods;
 
+typedef struct
+{
+	pgsocket	socket;
+	bool		isProxy;
+} PQlistenSocket;
+
 #define pq_comm_reset() (PqCommMethods->comm_reset())
 #define pq_flush() (PqCommMethods->flush())
 #define pq_flush_if_writable() (PqCommMethods->flush_if_writable())
@@ -64,9 +70,9 @@ extern WaitEventSet *FeBeWaitSet;
 #define FeBeWaitSetLatchPos 1
 #define FeBeWaitSetNEvents 3
 
-extern int	StreamServerPort(int family, const char *hostName,
-							 unsigned short portNumber, const char *unixSocketDir,
-							 pgsocket ListenSocket[], int MaxListen);
+extern PQlistenSocket *StreamServerPort(int family, const char *hostName,
+										unsigned short portNumber, const char *unixSocketDir,
+										PQlistenSocket PQlistenSocket[], int MaxListen);
 extern int	StreamConnection(pgsocket server_fd, Port *port);
 extern void StreamClose(pgsocket sock);
 extern void TouchSocketFiles(void);
@@ -79,6 +85,7 @@ extern bool pq_is_reading_msg(void);
 extern int	pq_getmessage(StringInfo s, int maxlen);
 extern int	pq_getbyte(void);
 extern int	pq_peekbyte(void);
+extern int	pq_discardbytes(size_t len);
 extern int	pq_getbyte_if_available(unsigned char *c);
 extern bool pq_buffer_has_data(void);
 extern int	pq_putmessage_v2(char msgtype, const char *s, size_t len);
diff --git a/src/include/postmaster/postmaster.h b/src/include/postmaster/postmaster.h
index 324a30ec1a..bcde67dd00 100644
--- a/src/include/postmaster/postmaster.h
+++ b/src/include/postmaster/postmaster.h
@@ -17,10 +17,13 @@
 extern bool EnableSSL;
 extern int	ReservedBackends;
 extern PGDLLIMPORT int PostPortNumber;
+extern PGDLLIMPORT int ProxyPortNumber;
 extern int	Unix_socket_permissions;
 extern char *Unix_socket_group;
 extern char *Unix_socket_directories;
 extern char *ListenAddresses;
+extern char *TrustedProxyServersString;
+extern struct sockaddr_storage *TrustedProxyServers;
 extern bool ClientAuthInProgress;
 extern int	PreAuthDelay;
 extern int	AuthenticationTimeout;
diff --git a/src/test/Makefile b/src/test/Makefile
index 46275915ff..4ad030034c 100644
--- a/src/test/Makefile
+++ b/src/test/Makefile
@@ -12,7 +12,8 @@ subdir = src/test
 top_builddir = ../..
 include $(top_builddir)/src/Makefile.global
 
-SUBDIRS = perl regress isolation modules authentication recovery subscription
+SUBDIRS = perl regress isolation modules authentication recovery subscription \
+	  protocol
 
 # Test suites that are not safe by default but can be run if selected
 # by the user via the whitespace-separated list in variable
diff --git a/src/test/protocol/Makefile b/src/test/protocol/Makefile
new file mode 100644
index 0000000000..bda49d6ecb
--- /dev/null
+++ b/src/test/protocol/Makefile
@@ -0,0 +1,23 @@
+#-------------------------------------------------------------------------
+#
+# Makefile for src/test/protocol
+#
+# Portions Copyright (c) 1996-2021, PostgreSQL Global Development Group
+# Portions Copyright (c) 1994, Regents of the University of California
+#
+# src/test/protocol/Makefile
+#
+#-------------------------------------------------------------------------
+
+subdir = src/test/protocol
+top_builddir = ../../..
+include $(top_builddir)/src/Makefile.global
+
+check:
+	$(prove_check)
+
+installcheck:
+	$(prove_installcheck)
+
+clean distclean maintainer-clean:
+	rm -rf tmp_check
diff --git a/src/test/protocol/t/001_proxy.pl b/src/test/protocol/t/001_proxy.pl
new file mode 100644
index 0000000000..64619058c8
--- /dev/null
+++ b/src/test/protocol/t/001_proxy.pl
@@ -0,0 +1,151 @@
+use strict;
+use warnings;
+use PostgreSQL::Test::Cluster;
+use PostgreSQL::Test::Utils;
+use Test::More;
+use Socket qw(AF_INET AF_INET6 inet_pton);
+use IO::Socket;
+
+plan tests => 25;
+
+my $node = PostgreSQL::Test::Cluster->new('node');
+$node->init;
+$node->append_conf(
+	'postgresql.conf', qq{
+log_connections = on
+});
+$node->append_conf(
+	'pg_hba.conf', qq{
+host all all 11.22.33.44/32 trust
+host all all 1:2:3:4:5:6:0:9/128 trust
+});
+$node->append_conf('postgresql.conf', "proxy_port = " . ($node->port() + 1));
+
+$node->start;
+
+$node->safe_psql('postgres', 'CREATE USER proxytest;');
+
+sub make_message
+{
+	my ($msg) = @_;
+	return pack("Na*", length($msg) + 4, $msg);
+}
+
+sub read_packet
+{
+	my ($socket) = @_;
+	my $buf = "";
+	$socket->recv($buf, 1024);
+	return $buf;
+}
+
+
+# Test normal connection through localhost
+sub test_connection
+{
+	my ($socket, $proxy, $what, $shouldbe, $shouldfail, $extra) = @_;
+	ok($socket, $what);
+
+	my $startup = make_message(
+		pack("N(Z*Z*)*x", 196608, (user => "proxytest", database => "postgres")));
+
+	$extra = "" if !defined($extra);
+
+	if (defined($proxy))
+	{
+		my $p = "\x0D\x0A\x0D\x0A\x00\x0D\x0A\x51\x55\x49\x54\x0A\x21";
+		if ($proxy =~ ":")
+		{
+			# ipv6
+			$p .= "\x21";                        # TCP v6
+			$p .= pack "n", 36 + length($extra); # size
+			$p .= inet_pton(AF_INET6, $proxy);
+			$p .= "\0" x 16;                     # destination address
+		}
+		else
+		{
+			# ipv4
+			$p .= "\x11";                        # TCP v4
+			$p .= pack "n", 12 + length($extra); # size
+			$p .= inet_pton(AF_INET, $proxy);
+			$p .= "\0\0\0\0";                    # destination address
+		}
+		$p .= pack "n", 1919;                    # source port
+		$p .= pack "n", 0;
+		$p .= $extra;
+		print $socket $p;
+	}
+	print $socket $startup;
+
+	my $in = read_packet($socket);
+	if (defined($shouldfail))
+	{
+		isnt(substr($in, 0, 1), 'R', $what);
+	}
+	else
+	{
+		is(substr($in, 0, 1), 'R', $what);
+	}
+
+  SKIP:
+	{
+		skip "The rest of this test should fail", 3 if (defined($shouldfail));
+
+		is(substr($in, 8, 1), "\0", $what);
+
+		my ($resip, $resport) = split /\|/,
+		  $node->safe_psql('postgres',
+			"SELECT client_addr, client_port FROM pg_stat_activity WHERE pid != pg_backend_pid() AND backend_type='client backend'"
+		  );
+		is($resip, $shouldbe, $what);
+		if ($proxy)
+		{
+			is($resport, "1919", $what);
+		}
+		else
+		{
+			ok($resport, $what);
+		}
+	}
+
+	$socket->close();
+
+	return;
+}
+
+sub make_socket
+{
+	my ($port) = @_;
+	if ($PostgreSQL::Test::Cluster::use_tcp) {
+		return IO::Socket::INET->new(
+			PeerAddr => "127.0.0.1",
+			PeerPort => $port,
+			Proto    => "tcp",
+			Type     => SOCK_STREAM);
+	}
+	else {
+		return IO::Socket::UNIX->new(
+			Peer => $node->host() . "/.s.PGSQL." . $port,
+			Type => SOCK_STREAM);
+	}
+}
+
+# Test a regular connection first to make sure connecting etc works fine.
+test_connection(make_socket($node->port()),
+	undef, "normal connection", $PostgreSQL::Test::Cluster::use_tcp ? "127.0.0.1": "");
+
+# Make sure we can't make a proxy connection until it's allowed
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44", 1);
+
+# Allow proxy connections and test them
+$node->append_conf('postgresql.conf', "proxy_servers = 'unix, 127.0.0.1/32'");
+$node->restart();
+
+test_connection(make_socket($node->port() + 1),
+	"11.22.33.44", "proxy ipv4", "11.22.33.44");
+test_connection(make_socket($node->port() + 1),
+	"1:2:3:4:5:6::9", "proxy ipv6", "1:2:3:4:5:6:0:9");
+
+test_connection(make_socket($node->port() + 1),
+    "11.22.33.44", "proxy with extra", "11.22.33.44", undef, "abcdef"x100);
#43Peter Eisentraut
peter.eisentraut@enterprisedb.com
In reply to: Magnus Hagander (#42)
Re: PROXY protocol support

A general question on this feature: AFAICT, you can only send the proxy
header once at the beginning of the connection. So this wouldn't be of
use for PostgreSQL-protocol connection poolers (pgbouncer, pgpool),
where the same server connection can be used for clients from different
source addresses. Do I understand that correctly?

#44Magnus Hagander
magnus@hagander.net
In reply to: Peter Eisentraut (#43)
Re: PROXY protocol support

On Wed, Mar 9, 2022 at 5:23 PM Peter Eisentraut
<peter.eisentraut@enterprisedb.com> wrote:

A general question on this feature: AFAICT, you can only send the proxy
header once at the beginning of the connection. So this wouldn't be of
use for PostgreSQL-protocol connection poolers (pgbouncer, pgpool),
where the same server connection can be used for clients from different
source addresses. Do I understand that correctly?

Correct. It's only sent at connection startup, so if you're re-using
the connection it would also re-use the IP address of the first one.

For reusing the connection for multiple clients, you'd want something
different, like a "priviliged mode in the tunnel" that the pooler can
handle.

--
Magnus Hagander
Me: https://www.hagander.net/
Work: https://www.redpill-linpro.com/

#45wilfried roset
wilfried.roset@gmail.com
In reply to: Magnus Hagander (#44)
Re: PROXY protocol support

Hi,

I've been able to test the patch. Here is a recap of the experimentation.

# Setup

All tests have been done witch 3 VMs (PostgreSQL, HAproxy, psql client) on
Debian 11 communicating over private network.
* PostgreSQL have been built with proxy_protocol_11.patch applied on master branch (465ab24296).
* psql client is from postgresql-client-13 from Debian 11 repository.
* HAproxy version used is 2.5.5-1~bpo11+1 installed from https://haproxy.debian.net

# Configuration

PostgresSQL has been configured to listen only on its private IP. To enable
proxy protocol support `proxy_port` has been configured to `5431` and
`proxy_servers` to `10.0.0.0/24`. `log_connections` has been turned on to make
sure the correct IP address is logged. `log_min_duration_statement` has been
configured to 0 to log all queries. Finally `log_destination` has been
configured to `csvlog`.

pg_hba.conf is like this:

local all all trust
host all all 127.0.0.1/32 trust
host all all ::1/128 trust
local replication all trust
host replication all 127.0.0.1/32 trust
host replication all ::1/128 trust
host all all 10.0.0.208/32 md5

Where 10.0.0.208 is the IP host the psql client's VM.

HAproxy has two frontends, one for proxy protocol (port 5431) and one for
regular TCP traffic. The configuration looks like this:

listen postgresql
bind 10.0.0.222:5432
server pg 10.0.0.253:5432 check

listen postgresql_proxy
bind 10.0.0.222:5431
server pg 10.0.0.253:5431 send-proxy-v2

Where 10.0.0.222 is the IP of HAproxy's VM and 10.0.0.253 is the IP of
PostgreSQL's VM.

# Tests

* from psql's vm to haproxy on port 5432 (no proxy protocol)
--> connection denied by pg_hba.conf, as expected

* from psql's vm to postgresql's VM on port 5432 (no proxy protocol)
--> connection success with psql's vm ip in logfile and pg_stat_activity

* from psql's vm to postgresql's VM on port 5431 (proxy protocol)
--> unable to open a connection, as expected

* from psql's vm to haproxy on port 5431 (proxy protocol)
--> connection success with psql's vm ip in logfile and pg_stat_activity

I've also tested without proxy protocol enable (and pg_hba.conf updated
accordingly), PostgreSQL behave as expected.

# Conclusion

From my point of view the documentation is clear enough and the feature works
as expected.

#46Magnus Hagander
magnus@hagander.net
In reply to: wilfried roset (#45)
Re: PROXY protocol support

On Sat, Apr 2, 2022 at 12:17 AM wilfried roset <wilfried.roset@gmail.com>
wrote:

Hi,

I've been able to test the patch. Here is a recap of the experimentation.

# Setup

All tests have been done witch 3 VMs (PostgreSQL, HAproxy, psql client) on
Debian 11 communicating over private network.
* PostgreSQL have been built with proxy_protocol_11.patch applied on
master branch (465ab24296).
* psql client is from postgresql-client-13 from Debian 11 repository.
* HAproxy version used is 2.5.5-1~bpo11+1 installed from
https://haproxy.debian.net

# Configuration

PostgresSQL has been configured to listen only on its private IP. To enable
proxy protocol support `proxy_port` has been configured to `5431` and
`proxy_servers` to `10.0.0.0/24` <http://10.0.0.0/24&gt;. `log_connections`
has been turned on to make
sure the correct IP address is logged. `log_min_duration_statement` has
been
configured to 0 to log all queries. Finally `log_destination` has been
configured to `csvlog`.

pg_hba.conf is like this:

local all all trust
host all all 127.0.0.1/32 trust
host all all ::1/128 trust
local replication all trust
host replication all 127.0.0.1/32 trust
host replication all ::1/128 trust
host all all 10.0.0.208/32 md5

Where 10.0.0.208 is the IP host the psql client's VM.

HAproxy has two frontends, one for proxy protocol (port 5431) and one for
regular TCP traffic. The configuration looks like this:

listen postgresql
bind 10.0.0.222:5432
server pg 10.0.0.253:5432 check

listen postgresql_proxy
bind 10.0.0.222:5431
server pg 10.0.0.253:5431 send-proxy-v2

Where 10.0.0.222 is the IP of HAproxy's VM and 10.0.0.253 is the IP of
PostgreSQL's VM.

# Tests

* from psql's vm to haproxy on port 5432 (no proxy protocol)
--> connection denied by pg_hba.conf, as expected

* from psql's vm to postgresql's VM on port 5432 (no proxy protocol)
--> connection success with psql's vm ip in logfile and pg_stat_activity

* from psql's vm to postgresql's VM on port 5431 (proxy protocol)
--> unable to open a connection, as expected

* from psql's vm to haproxy on port 5431 (proxy protocol)
--> connection success with psql's vm ip in logfile and pg_stat_activity

I've also tested without proxy protocol enable (and pg_hba.conf updated
accordingly), PostgreSQL behave as expected.

# Conclusion

From my point of view the documentation is clear enough and the feature
works
as expected.

Hi!

Thanks for this review and testing!

I think it could do with at least noe more look-over at the source code
level as well at this point though since it's been sitting around for a
while, so it won't make it in for this deadline. But hopefully I can get it
in early in the next cycle!

--
Magnus Hagander
Me: https://www.hagander.net/ <http://www.hagander.net/&gt;
Work: https://www.redpill-linpro.com/ <http://www.redpill-linpro.com/&gt;

#47Jacob Champion
jchampion@timescale.com
In reply to: Magnus Hagander (#46)
Re: PROXY protocol support

This needs a rebase, but after that I expect it to be RfC.

--Jacob

The new status of this patch is: Waiting on Author

#48Julien Riou
julien@riou.xyz
In reply to: Jacob Champion (#47)
Re: PROXY protocol support

On 7/28/22 22:05, Jacob Champion wrote:

This needs a rebase, but after that I expect it to be RfC.

--Jacob

The new status of this patch is: Waiting on Author

Hello folks,

Thank you all for this awesome work!

I'm looking for this feature for years now. Last year, I've tried to
rebase the patch without success. Unfortunately, this is out of my league.

Magnus, please let me know if I can help.

Have a nice day,
Julien