src: Add tproxy support

This patch adds support for transparent proxy functionality which is
supported in ip, ip6 and inet tables.

The syntax is the following:
	tproxy [{|ip|ip6}] to {<ip address>|:<port>|<ip address>:<port>}

It looks for a socket listening on the specified address or port and
assigns it to the matching packet.

In an inet table, a packet matches for both families until address is
specified.
Network protocol family has to be specified **only** in inet tables if
address is specified.

As transparent proxy support is implemented for sockets with layer 4
information, a transport protocol header criterion has to be set in the
same rule. eg. 'meta l4proto tcp' or 'udp dport 4444'

Example ruleset:
	table ip x {
		chain y {
			type filter hook prerouting priority -150; policy accept;
			tcp dport ntp tproxy to 1.1.1.1
			udp dport ssh tproxy to :2222
		}
	}
	table ip6 x {
		chain y {
			type filter hook prerouting priority -150; policy accept;
			tcp dport ntp tproxy to [dead::beef]
			udp dport ssh tproxy to :2222
		}
	}
	table inet x {
		chain y {
			type filter hook prerouting priority -150; policy accept;
			tcp dport 321 tproxy to :ssh
			tcp dport 99 tproxy ip to 1.1.1.1:999
			udp dport 155 tproxy ip6 to [dead::beef]:smux
		}
	}

Signed-off-by: Máté Eckl <ecklm94@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

6 files changed, 267 insertions, 0 deletions

diff --git a/src/evaluate.c b/src/evaluate.c
index ae881ccd..da95cdf9 100644
--- a/src/evaluate.c
+++ b/src/evaluate.c
@@ -2478,6 +2478,86 @@ static int stmt_evaluate_nat(struct eval_ctx *ctx, struct stmt *stmt)
 	return 0;
 }
 
+static int stmt_evaluate_tproxy(struct eval_ctx *ctx, struct stmt *stmt)
+{
+	const struct datatype *dtype;
+	int err, len;
+
+	switch (ctx->pctx.family) {
+	case NFPROTO_IPV4:
+	case NFPROTO_IPV6:
+	case NFPROTO_INET:
+		break;
+	default:
+		return stmt_error(ctx, stmt,
+				  "tproxy is only supported for IPv4/IPv6/INET");
+	}
+
+	if (ctx->pctx.protocol[PROTO_BASE_TRANSPORT_HDR].desc == NULL)
+		return stmt_error(ctx, stmt, "Transparent proxy support requires"
+					     " transport protocol match");
+
+	if (!stmt->tproxy.addr && !stmt->tproxy.port)
+		return stmt_error(ctx, stmt, "Either address or port must be specified!");
+
+	if (ctx->pctx.family != NFPROTO_INET) {
+		if (stmt->tproxy.family != NFPROTO_UNSPEC)
+			return stmt_error(ctx, stmt, "Family can only be specified in inet tables.");
+		stmt->tproxy.family = ctx->pctx.family;
+	}
+	else {
+		const struct proto_desc *nproto =
+			ctx->pctx.protocol[PROTO_BASE_NETWORK_HDR].desc;
+		if ((nproto == &proto_ip && stmt->tproxy.family == NFPROTO_IPV6) ||
+		    (nproto == &proto_ip6 && stmt->tproxy.family == NFPROTO_IPV4))
+			/* this prevents us from rules like
+			 * ip protocol tcp tproxy ip6 to [dead::beef]
+			 */
+			return stmt_error(ctx, stmt,
+					  "Conflicting network layer protocols.");
+	}
+
+	if (stmt->tproxy.addr != NULL) {
+		if (stmt->tproxy.addr->ops->type == EXPR_RANGE)
+			return stmt_error(ctx, stmt, "Address ranges are not supported for tproxy.");
+		if (ctx->pctx.family == NFPROTO_INET) {
+			switch (stmt->tproxy.family) {
+			case NFPROTO_IPV4:
+				dtype = &ipaddr_type;
+				len   = 4 * BITS_PER_BYTE;
+				break;
+			case NFPROTO_IPV6:
+				dtype = &ip6addr_type;
+				len   = 16 * BITS_PER_BYTE;
+				break;
+			default:
+				return stmt_error(ctx, stmt,
+						  "Family must be specified in tproxy statement with address for inet tables.");
+			}
+			err = stmt_evaluate_arg(ctx, stmt, dtype, len,
+						BYTEORDER_BIG_ENDIAN,
+						&stmt->tproxy.addr);
+			if (err < 0)
+				return err;
+		}
+		else {
+			err = evaluate_addr(ctx, stmt, &stmt->tproxy.addr);
+			if (err < 0)
+				return err;
+		}
+	}
+
+	if (stmt->tproxy.port != NULL) {
+		if (stmt->tproxy.port->ops->type == EXPR_RANGE)
+			return stmt_error(ctx, stmt, "Port ranges are not supported for tproxy.");
+		err = nat_evaluate_transport(ctx, stmt, &stmt->tproxy.port);
+		if (err < 0)
+			return err;
+	}
+
+	return 0;
+}
+
 static int stmt_evaluate_dup(struct eval_ctx *ctx, struct stmt *stmt)
 {
 	int err;
@@ -2761,6 +2841,8 @@ int stmt_evaluate(struct eval_ctx *ctx, struct stmt *stmt)
 		return stmt_evaluate_reject(ctx, stmt);
 	case STMT_NAT:
 		return stmt_evaluate_nat(ctx, stmt);
+	case STMT_TPROXY:
+		return stmt_evaluate_tproxy(ctx, stmt);
 	case STMT_QUEUE:
 		return stmt_evaluate_queue(ctx, stmt);
 	case STMT_DUP:
diff --git a/src/netlink_delinearize.c b/src/netlink_delinearize.c
index 7e9765cf..c886ff98 100644
--- a/src/netlink_delinearize.c
+++ b/src/netlink_delinearize.c
@@ -969,6 +969,50 @@ out_err:
 	xfree(stmt);
 }
 
+static void netlink_parse_tproxy(struct netlink_parse_ctx *ctx,
+			      const struct location *loc,
+			      const struct nftnl_expr *nle)
+{
+	struct stmt *stmt;
+	struct expr *addr, *port;
+	enum nft_registers reg;
+
+	stmt = tproxy_stmt_alloc(loc);
+	stmt->tproxy.family = nftnl_expr_get_u32(nle, NFTNL_EXPR_TPROXY_FAMILY);
+	stmt->tproxy.table_family = ctx->table->handle.family;
+
+	reg = netlink_parse_register(nle, NFTNL_EXPR_TPROXY_REG_ADDR);
+	if (reg) {
+		addr = netlink_get_register(ctx, loc, reg);
+
+		switch (stmt->tproxy.family) {
+		case NFPROTO_IPV4:
+			expr_set_type(addr, &ipaddr_type, BYTEORDER_BIG_ENDIAN);
+			break;
+		case NFPROTO_IPV6:
+			expr_set_type(addr, &ip6addr_type, BYTEORDER_BIG_ENDIAN);
+			break;
+		default:
+			netlink_error(ctx, loc,
+				      "tproxy address must be IPv4 or IPv6");
+			goto err;
+		}
+		stmt->tproxy.addr = addr;
+	}
+
+	reg = netlink_parse_register(nle, NFTNL_EXPR_TPROXY_REG_PORT);
+	if (reg) {
+		port = netlink_get_register(ctx, loc, reg);
+		expr_set_type(port, &inet_service_type, BYTEORDER_BIG_ENDIAN);
+		stmt->tproxy.port = port;
+	}
+
+	ctx->stmt = stmt;
+	return;
+err:
+	xfree(stmt);
+}
+
 static void netlink_parse_masq(struct netlink_parse_ctx *ctx,
 			       const struct location *loc,
 			       const struct nftnl_expr *nle)
@@ -1362,6 +1406,7 @@ static const struct {
 	{ .name = "range",	.parse = netlink_parse_range },
 	{ .name = "reject",	.parse = netlink_parse_reject },
 	{ .name = "nat",	.parse = netlink_parse_nat },
+	{ .name = "tproxy",	.parse = netlink_parse_tproxy },
 	{ .name = "notrack",	.parse = netlink_parse_notrack },
 	{ .name = "masq",	.parse = netlink_parse_masq },
 	{ .name = "redir",	.parse = netlink_parse_redir },
@@ -2435,6 +2480,14 @@ static void rule_parse_postprocess(struct netlink_parse_ctx *ctx, struct rule *r
 				expr_postprocess(&rctx, &stmt->nat.proto);
 			}
 			break;
+		case STMT_TPROXY:
+			if (stmt->tproxy.addr)
+				expr_postprocess(&rctx, &stmt->tproxy.addr);
+			if (stmt->tproxy.port) {
+				payload_dependency_reset(&rctx.pdctx);
+				expr_postprocess(&rctx, &stmt->tproxy.port);
+			}
+			break;
 		case STMT_REJECT:
 			stmt_reject_postprocess(&rctx);
 			break;
diff --git a/src/netlink_linearize.c b/src/netlink_linearize.c
index 8471e837..aa00564a 100644
--- a/src/netlink_linearize.c
+++ b/src/netlink_linearize.c
@@ -1071,6 +1071,45 @@ static void netlink_gen_nat_stmt(struct netlink_linearize_ctx *ctx,
 	nftnl_rule_add_expr(ctx->nlr, nle);
 }
 
+static void netlink_gen_tproxy_stmt(struct netlink_linearize_ctx *ctx,
+				 const struct stmt *stmt)
+{
+	struct nftnl_expr *nle;
+	enum nft_registers addr_reg;
+	enum nft_registers port_reg;
+	int registers = 0;
+	const int family = stmt->tproxy.family;
+	int nftnl_reg_port;
+
+	nle = alloc_nft_expr("tproxy");
+
+	nftnl_expr_set_u32(nle, NFTNL_EXPR_TPROXY_FAMILY, family);
+
+	nftnl_reg_port = NFTNL_EXPR_TPROXY_REG_PORT;
+
+	if (stmt->tproxy.addr) {
+		addr_reg = get_register(ctx, NULL);
+		registers++;
+		netlink_gen_expr(ctx, stmt->tproxy.addr, addr_reg);
+		netlink_put_register(nle, NFTNL_EXPR_TPROXY_REG_ADDR,
+				     addr_reg);
+	}
+
+	if (stmt->tproxy.port) {
+		port_reg = get_register(ctx, NULL);
+		registers++;
+		netlink_gen_expr(ctx, stmt->tproxy.port, port_reg);
+		netlink_put_register(nle, nftnl_reg_port, port_reg);
+	}
+
+	while (registers > 0) {
+		release_register(ctx, NULL);
+		registers--;
+	}
+
+	nftnl_rule_add_expr(ctx->nlr, nle);
+}
+
 static void netlink_gen_dup_stmt(struct netlink_linearize_ctx *ctx,
 				 const struct stmt *stmt)
 {
@@ -1301,6 +1340,8 @@ static void netlink_gen_stmt(struct netlink_linearize_ctx *ctx,
 		return netlink_gen_reject_stmt(ctx, stmt);
 	case STMT_NAT:
 		return netlink_gen_nat_stmt(ctx, stmt);
+	case STMT_TPROXY:
+		return netlink_gen_tproxy_stmt(ctx, stmt);
 	case STMT_DUP:
 		return netlink_gen_dup_stmt(ctx, stmt);
 	case STMT_QUEUE:
diff --git a/src/parser_bison.y b/src/parser_bison.y
index 98bfebad..fe3c10ba 100644
--- a/src/parser_bison.y
+++ b/src/parser_bison.y
@@ -192,6 +192,8 @@ int nft_lex(void *, void *, void *);
 %token SOCKET			"socket"
 %token TRANSPARENT		"transparent"
 
+%token TPROXY			"tproxy"
+
 %token HOOK			"hook"
 %token DEVICE			"device"
 %token DEVICES			"devices"
@@ -572,6 +574,9 @@ int nft_lex(void *, void *, void *);
 %type <stmt>			nat_stmt nat_stmt_alloc masq_stmt masq_stmt_alloc redir_stmt redir_stmt_alloc
 %destructor { stmt_free($$); }	nat_stmt nat_stmt_alloc masq_stmt masq_stmt_alloc redir_stmt redir_stmt_alloc
 %type <val>			nf_nat_flags nf_nat_flag offset_opt
+%type <stmt>			tproxy_stmt
+%destructor { stmt_free($$); }	tproxy_stmt
+%type <val>				tproxy_family_spec
 %type <stmt>			queue_stmt queue_stmt_alloc
 %destructor { stmt_free($$); }	queue_stmt queue_stmt_alloc
 %type <val>			queue_stmt_flags queue_stmt_flag
@@ -2082,6 +2087,7 @@ stmt			:	verdict_stmt
 			|	quota_stmt
 			|	reject_stmt
 			|	nat_stmt
+			|	tproxy_stmt
 			|	queue_stmt
 			|	ct_stmt
 			|	masq_stmt
@@ -2477,6 +2483,44 @@ nat_stmt_alloc		:	SNAT	{ $$ = nat_stmt_alloc(&@$, NFT_NAT_SNAT); }
 			|	DNAT	{ $$ = nat_stmt_alloc(&@$, NFT_NAT_DNAT); }
 			;
 
+tproxy_family_spec	:	IP	{ $$ = NFPROTO_IPV4; }
+			|	IP6	{ $$ = NFPROTO_IPV6; }
+			;
+
+tproxy_stmt		:	TPROXY TO stmt_expr
+			{
+				$$ = tproxy_stmt_alloc(&@$);
+				$$->tproxy.family = NFPROTO_UNSPEC;
+				$$->tproxy.addr = $3;
+			}
+			|	TPROXY tproxy_family_spec TO stmt_expr
+			{
+				$$ = tproxy_stmt_alloc(&@$);
+				$$->tproxy.family = $2;
+				$$->tproxy.addr = $4;
+			}
+			|	TPROXY TO COLON stmt_expr
+			{
+				$$ = tproxy_stmt_alloc(&@$);
+				$$->tproxy.family = NFPROTO_UNSPEC;
+				$$->tproxy.port = $4;
+			}
+			|	TPROXY TO stmt_expr COLON stmt_expr
+			{
+				$$ = tproxy_stmt_alloc(&@$);
+				$$->tproxy.family = NFPROTO_UNSPEC;
+				$$->tproxy.addr = $3;
+				$$->tproxy.port = $5;
+			}
+			|	TPROXY tproxy_family_spec TO stmt_expr COLON stmt_expr
+			{
+				$$ = tproxy_stmt_alloc(&@$);
+				$$->tproxy.family = $2;
+				$$->tproxy.addr = $4;
+				$$->tproxy.port = $6;
+			}
+			;
+
 primary_stmt_expr	:	symbol_expr		{ $$ = $1; }
 			|	integer_expr		{ $$ = $1; }
 			|	boolean_expr		{ $$ = $1; }
diff --git a/src/scanner.l b/src/scanner.l
index ed01b5e7..703700fe 100644
--- a/src/scanner.l
+++ b/src/scanner.l
@@ -261,6 +261,8 @@ addrstring	({macaddr}|{ip4addr}|{ip6addr})
 "socket"		{ return SOCKET; }
 "transparent"		{ return TRANSPARENT;}
 
+"tproxy"		{ return TPROXY; }
+
 "accept"		{ return ACCEPT; }
 "drop"			{ return DROP; }
 "continue"		{ return CONTINUE; }
diff --git a/src/statement.c b/src/statement.c
index 6f5e6660..3040476f 100644
--- a/src/statement.c
+++ b/src/statement.c
@@ -760,6 +760,51 @@ struct stmt *fwd_stmt_alloc(const struct location *loc)
 	return stmt_alloc(loc, &fwd_stmt_ops);
 }
 
+static void tproxy_stmt_print(const struct stmt *stmt, struct output_ctx *octx)
+{
+	nft_print(octx, "tproxy");
+
+	if (stmt->tproxy.table_family == NFPROTO_INET &&
+	    stmt->tproxy.family != NFPROTO_UNSPEC)
+		nft_print(octx, " %s", nfproto_family_name(stmt->tproxy.family));
+	nft_print(octx, " to");
+	if (stmt->tproxy.addr) {
+		nft_print(octx, " ");
+		if (stmt->tproxy.addr->ops->type == EXPR_VALUE &&
+		    stmt->tproxy.addr->dtype->type == TYPE_IP6ADDR) {
+			nft_print(octx, "[");
+			expr_print(stmt->tproxy.addr, octx);
+			nft_print(octx, "]");
+		} else {
+			expr_print(stmt->tproxy.addr, octx);
+		}
+	}
+	if (stmt->tproxy.port && stmt->tproxy.port->ops->type == EXPR_VALUE) {
+		if (!stmt->tproxy.addr)
+			nft_print(octx, " ");
+		nft_print(octx, ":");
+		expr_print(stmt->tproxy.port, octx);
+	}
+}
+
+static void tproxy_stmt_destroy(struct stmt *stmt)
+{
+	expr_free(stmt->tproxy.addr);
+	expr_free(stmt->tproxy.port);
+}
+
+static const struct stmt_ops tproxy_stmt_ops = {
+	.type		= STMT_TPROXY,
+	.name		= "tproxy",
+	.print		= tproxy_stmt_print,
+	.destroy	= tproxy_stmt_destroy,
+};
+
+struct stmt *tproxy_stmt_alloc(const struct location *loc)
+{
+	return stmt_alloc(loc, &tproxy_stmt_ops);
+}
+
 static void xt_stmt_print(const struct stmt *stmt, struct output_ctx *octx)
 {
 	xt_stmt_xlate(stmt);