netfilter: nf_tables: add xfrm expression

supports fetching saddr/daddr of tunnel mode states, request id and spi.
If direction is 'in', use inbound skb secpath, else dst->xfrm.

Joint work with Máté Eckl.
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

include/uapi/linux/netfilter/nf_tables.h

@@ -1514,6 +1514,35 @@ enum nft_devices_attributes {
 };
 #define NFTA_DEVICE_MAX		(__NFTA_DEVICE_MAX - 1)
 
+/*
+ * enum nft_xfrm_attributes - nf_tables xfrm expr netlink attributes
+ *
+ * @NFTA_XFRM_DREG: destination register (NLA_U32)
+ * @NFTA_XFRM_KEY: enum nft_xfrm_keys (NLA_U32)
+ * @NFTA_XFRM_DIR: direction (NLA_U8)
+ * @NFTA_XFRM_SPNUM: index in secpath array (NLA_U32)
+ */
+enum nft_xfrm_attributes {
+	NFTA_XFRM_UNSPEC,
+	NFTA_XFRM_DREG,
+	NFTA_XFRM_KEY,
+	NFTA_XFRM_DIR,
+	NFTA_XFRM_SPNUM,
+	__NFTA_XFRM_MAX
+};
+#define NFTA_XFRM_MAX (__NFTA_XFRM_MAX - 1)
+
+enum nft_xfrm_keys {
+	NFT_XFRM_KEY_UNSPEC,
+	NFT_XFRM_KEY_DADDR_IP4,
+	NFT_XFRM_KEY_DADDR_IP6,
+	NFT_XFRM_KEY_SADDR_IP4,
+	NFT_XFRM_KEY_SADDR_IP6,
+	NFT_XFRM_KEY_REQID,
+	NFT_XFRM_KEY_SPI,
+	__NFT_XFRM_KEY_MAX,
+};
+#define NFT_XFRM_KEY_MAX (__NFT_XFRM_KEY_MAX - 1)
 
 /**
  * enum nft_trace_attributes - nf_tables trace netlink attributes

net/netfilter/Kconfig

@@ -625,6 +625,13 @@ config NFT_FIB_INET
 	  The lookup will be delegated to the IPv4 or IPv6 FIB depending
 	  on the protocol of the packet.
 
+config NFT_XFRM
+	tristate "Netfilter nf_tables xfrm/IPSec security association matching"
+	depends on XFRM
+	help
+	  This option adds an expression that you can use to extract properties
+	  of a packets security association.
+
 config NFT_SOCKET
 	tristate "Netfilter nf_tables socket match support"
 	depends on IPV6 || IPV6=n

net/netfilter/Makefile

@@ -113,6 +113,7 @@ obj-$(CONFIG_NFT_FIB_NETDEV)	+= nft_fib_netdev.o
 obj-$(CONFIG_NFT_SOCKET)	+= nft_socket.o
 obj-$(CONFIG_NFT_OSF)		+= nft_osf.o
 obj-$(CONFIG_NFT_TPROXY)	+= nft_tproxy.o
+obj-$(CONFIG_NFT_XFRM)		+= nft_xfrm.o
 
 # nf_tables netdev
 obj-$(CONFIG_NFT_DUP_NETDEV)	+= nft_dup_netdev.o

net/netfilter/nft_xfrm.c

+/*
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * Generic part shared by ipv4 and ipv6 backends.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/netlink.h>
+#include <linux/netfilter.h>
+#include <linux/netfilter/nf_tables.h>
+#include <net/netfilter/nf_tables_core.h>
+#include <net/netfilter/nf_tables.h>
+#include <linux/in.h>
+#include <net/xfrm.h>
+
+static const struct nla_policy nft_xfrm_policy[NFTA_XFRM_MAX + 1] = {
+	[NFTA_XFRM_KEY]		= { .type = NLA_U32 },
+	[NFTA_XFRM_DIR]		= { .type = NLA_U8 },
+	[NFTA_XFRM_SPNUM]	= { .type = NLA_U32 },
+	[NFTA_XFRM_DREG]	= { .type = NLA_U32 },
+};
+
+struct nft_xfrm {
+	enum nft_xfrm_keys	key:8;
+	enum nft_registers	dreg:8;
+	u8			dir;
+	u8			spnum;
+};
+
+static int nft_xfrm_get_init(const struct nft_ctx *ctx,
+			     const struct nft_expr *expr,
+			     const struct nlattr * const tb[])
+{
+	struct nft_xfrm *priv = nft_expr_priv(expr);
+	unsigned int len = 0;
+	u32 spnum = 0;
+	u8 dir;
+
+	if (!tb[NFTA_XFRM_KEY] || !tb[NFTA_XFRM_DIR] || !tb[NFTA_XFRM_DREG])
+		return -EINVAL;
+
+	switch (ctx->family) {
+	case NFPROTO_IPV4:
+	case NFPROTO_IPV6:
+	case NFPROTO_INET:
+		break;
+	default:
+		return -EOPNOTSUPP;
+	}
+
+	priv->key = ntohl(nla_get_u32(tb[NFTA_XFRM_KEY]));
+	switch (priv->key) {
+	case NFT_XFRM_KEY_REQID:
+	case NFT_XFRM_KEY_SPI:
+		len = sizeof(u32);
+		break;
+	case NFT_XFRM_KEY_DADDR_IP4:
+	case NFT_XFRM_KEY_SADDR_IP4:
+		len = sizeof(struct in_addr);
+		break;
+	case NFT_XFRM_KEY_DADDR_IP6:
+	case NFT_XFRM_KEY_SADDR_IP6:
+		len = sizeof(struct in6_addr);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	dir = nla_get_u8(tb[NFTA_XFRM_DIR]);
+	switch (dir) {
+	case XFRM_POLICY_IN:
+	case XFRM_POLICY_OUT:
+		priv->dir = dir;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (tb[NFTA_XFRM_SPNUM])
+		spnum = ntohl(nla_get_be32(tb[NFTA_XFRM_SPNUM]));
+
+	if (spnum >= XFRM_MAX_DEPTH)
+		return -ERANGE;
+
+	priv->spnum = spnum;
+
+	priv->dreg = nft_parse_register(tb[NFTA_XFRM_DREG]);
+	return nft_validate_register_store(ctx, priv->dreg, NULL,
+					   NFT_DATA_VALUE, len);
+}
+
+/* Return true if key asks for daddr/saddr and current
+ * state does have a valid address (BEET, TUNNEL).
+ */
+static bool xfrm_state_addr_ok(enum nft_xfrm_keys k, u8 family, u8 mode)
+{
+	switch (k) {
+	case NFT_XFRM_KEY_DADDR_IP4:
+	case NFT_XFRM_KEY_SADDR_IP4:
+		if (family == NFPROTO_IPV4)
+			break;
+		return false;
+	case NFT_XFRM_KEY_DADDR_IP6:
+	case NFT_XFRM_KEY_SADDR_IP6:
+		if (family == NFPROTO_IPV6)
+			break;
+		return false;
+	default:
+		return true;
+	}
+
+	return mode == XFRM_MODE_BEET || mode == XFRM_MODE_TUNNEL;
+}
+
+static void nft_xfrm_state_get_key(const struct nft_xfrm *priv,
+				   struct nft_regs *regs,
+				   const struct xfrm_state *state,
+				   u8 family)
+{
+	u32 *dest = &regs->data[priv->dreg];
+
+	if (!xfrm_state_addr_ok(priv->key, family, state->props.mode)) {
+		regs->verdict.code = NFT_BREAK;
+		return;
+	}
+
+	switch (priv->key) {
+	case NFT_XFRM_KEY_UNSPEC:
+	case __NFT_XFRM_KEY_MAX:
+		WARN_ON_ONCE(1);
+		break;
+	case NFT_XFRM_KEY_DADDR_IP4:
+		*dest = state->id.daddr.a4;
+		return;
+	case NFT_XFRM_KEY_DADDR_IP6:
+		memcpy(dest, &state->id.daddr.in6, sizeof(struct in6_addr));
+		return;
+	case NFT_XFRM_KEY_SADDR_IP4:
+		*dest = state->props.saddr.a4;
+		return;
+	case NFT_XFRM_KEY_SADDR_IP6:
+		memcpy(dest, &state->props.saddr.in6, sizeof(struct in6_addr));
+		return;
+	case NFT_XFRM_KEY_REQID:
+		*dest = state->props.reqid;
+		return;
+	case NFT_XFRM_KEY_SPI:
+		*dest = state->id.spi;
+		return;
+	}
+
+	regs->verdict.code = NFT_BREAK;
+}
+
+static void nft_xfrm_get_eval_in(const struct nft_xfrm *priv,
+				    struct nft_regs *regs,
+				    const struct nft_pktinfo *pkt)
+{
+	const struct sec_path *sp = pkt->skb->sp;
+	const struct xfrm_state *state;
+
+	if (sp == NULL || sp->len <= priv->spnum) {
+		regs->verdict.code = NFT_BREAK;
+		return;
+	}
+
+	state = sp->xvec[priv->spnum];
+	nft_xfrm_state_get_key(priv, regs, state, nft_pf(pkt));
+}
+
+static void nft_xfrm_get_eval_out(const struct nft_xfrm *priv,
+				  struct nft_regs *regs,
+				  const struct nft_pktinfo *pkt)
+{
+	const struct dst_entry *dst = skb_dst(pkt->skb);
+	int i;
+
+	for (i = 0; dst && dst->xfrm;
+	     dst = ((const struct xfrm_dst *)dst)->child, i++) {
+		if (i < priv->spnum)
+			continue;
+
+		nft_xfrm_state_get_key(priv, regs, dst->xfrm, nft_pf(pkt));
+		return;
+	}
+
+	regs->verdict.code = NFT_BREAK;
+}
+
+static void nft_xfrm_get_eval(const struct nft_expr *expr,
+			      struct nft_regs *regs,
+			      const struct nft_pktinfo *pkt)
+{
+	const struct nft_xfrm *priv = nft_expr_priv(expr);
+
+	switch (priv->dir) {
+	case XFRM_POLICY_IN:
+		nft_xfrm_get_eval_in(priv, regs, pkt);
+		break;
+	case XFRM_POLICY_OUT:
+		nft_xfrm_get_eval_out(priv, regs, pkt);
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		regs->verdict.code = NFT_BREAK;
+		break;
+	}
+}
+
+static int nft_xfrm_get_dump(struct sk_buff *skb,
+			     const struct nft_expr *expr)
+{
+	const struct nft_xfrm *priv = nft_expr_priv(expr);
+
+	if (nft_dump_register(skb, NFTA_XFRM_DREG, priv->dreg))
+		return -1;
+
+	if (nla_put_be32(skb, NFTA_XFRM_KEY, htonl(priv->key)))
+		return -1;
+	if (nla_put_u8(skb, NFTA_XFRM_DIR, priv->dir))
+		return -1;
+	if (nla_put_be32(skb, NFTA_XFRM_SPNUM, htonl(priv->spnum)))
+		return -1;
+
+	return 0;
+}
+
+static int nft_xfrm_validate(const struct nft_ctx *ctx, const struct nft_expr *expr,
+			     const struct nft_data **data)
+{
+	const struct nft_xfrm *priv = nft_expr_priv(expr);
+	unsigned int hooks;
+
+	switch (priv->dir) {
+	case XFRM_POLICY_IN:
+		hooks = (1 << NF_INET_FORWARD) |
+			(1 << NF_INET_LOCAL_IN) |
+			(1 << NF_INET_PRE_ROUTING);
+		break;
+	case XFRM_POLICY_OUT:
+		hooks = (1 << NF_INET_FORWARD) |
+			(1 << NF_INET_LOCAL_OUT) |
+			(1 << NF_INET_POST_ROUTING);
+		break;
+	default:
+		WARN_ON_ONCE(1);
+		return -EINVAL;
+	}
+
+	return nft_chain_validate_hooks(ctx->chain, hooks);
+}
+
+
+static struct nft_expr_type nft_xfrm_type;
+static const struct nft_expr_ops nft_xfrm_get_ops = {
+	.type		= &nft_xfrm_type,
+	.size		= NFT_EXPR_SIZE(sizeof(struct nft_xfrm)),
+	.eval		= nft_xfrm_get_eval,
+	.init		= nft_xfrm_get_init,
+	.dump		= nft_xfrm_get_dump,
+	.validate	= nft_xfrm_validate,
+};
+
+static struct nft_expr_type nft_xfrm_type __read_mostly = {
+	.name		= "xfrm",
+	.ops		= &nft_xfrm_get_ops,
+	.policy		= nft_xfrm_policy,
+	.maxattr	= NFTA_XFRM_MAX,
+	.owner		= THIS_MODULE,
+};
+
+static int __init nft_xfrm_module_init(void)
+{
+	return nft_register_expr(&nft_xfrm_type);
+}
+
+static void __exit nft_xfrm_module_exit(void)
+{
+	nft_unregister_expr(&nft_xfrm_type);
+}
+
+module_init(nft_xfrm_module_init);
+module_exit(nft_xfrm_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("nf_tables: xfrm/IPSec matching");
+MODULE_AUTHOR("Florian Westphal <fw@strlen.de>");
+MODULE_AUTHOR("Máté Eckl <ecklm94@gmail.com>");
+MODULE_ALIAS_NFT_EXPR("xfrm");