[PATCH nf-next v6 0/4] Netfilter egress hook

[PATCH nf-next v6 0/4] Netfilter egress hook

[Lukas Wunner]

Netfilter egress hook, 6th iteration

Changes:

* Perform netfilter egress classifying before tc egress classifying
  to achieve reverse order vis-a-vis ingress datapath.

* Avoid layering violations by way of new skb->nf_skip_egress flag.

* Add egress support to new nfnetlink_hook.c.


Link to previous version v5 (posted by Pablo):
https://lore.kernel.org/netdev/20210928095538.114207-1-pablo@netfilter.org/

Link to previous version v4:
https://lore.kernel.org/netdev/cover.1611304190.git.lukas@wunner.de/


Lukas Wunner (3):
  netfilter: Rename ingress hook include file
  netfilter: Generalize ingress hook include file
  netfilter: Introduce egress hook

Pablo Neira Ayuso (1):
  af_packet: Introduce egress hook

 drivers/net/ifb.c                 |   3 +
 include/linux/netdevice.h         |   4 +
 include/linux/netfilter_ingress.h |  58 ------------
 include/linux/netfilter_netdev.h  | 146 ++++++++++++++++++++++++++++++
 include/linux/skbuff.h            |   4 +
 include/uapi/linux/netfilter.h    |   1 +
 net/core/dev.c                    |  19 +++-
 net/netfilter/Kconfig             |  11 +++
 net/netfilter/core.c              |  34 ++++++-
 net/netfilter/nfnetlink_hook.c    |  16 +++-
 net/netfilter/nft_chain_filter.c  |   4 +-
 net/packet/af_packet.c            |  35 +++++++
 12 files changed, 265 insertions(+), 70 deletions(-)
 delete mode 100644 include/linux/netfilter_ingress.h
 create mode 100644 include/linux/netfilter_netdev.h

-- 
2.31.1

[PATCH nf-next v6 1/4] netfilter: Rename ingress hook include file

[Lukas Wunner]

Prepare for addition of a netfilter egress hook by renaming
<linux/netfilter_ingress.h> to <linux/netfilter_netdev.h>.

The egress hook also necessitates a refactoring of the include file,
but that is done in a separate commit to ease reviewing.

No functional change intended.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
---
 include/linux/{netfilter_ingress.h => netfilter_netdev.h} | 0
 net/core/dev.c                                            | 2 +-
 2 files changed, 1 insertion(+), 1 deletion(-)
 rename include/linux/{netfilter_ingress.h => netfilter_netdev.h} (100%)

diff --git a/include/linux/netfilter_ingress.h b/include/linux/netfilter_netdev.h
similarity index 100%
rename from include/linux/netfilter_ingress.h
rename to include/linux/netfilter_netdev.h
diff --git a/net/core/dev.c b/net/core/dev.c
index 62ddd7d6e00d..3b79121a6a07 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -140,7 +140,7 @@
 #include <linux/if_macvlan.h>
 #include <linux/errqueue.h>
 #include <linux/hrtimer.h>
-#include <linux/netfilter_ingress.h>
+#include <linux/netfilter_netdev.h>
 #include <linux/crash_dump.h>
 #include <linux/sctp.h>
 #include <net/udp_tunnel.h>
-- 
2.31.1

[PATCH nf-next v6 2/4] netfilter: Generalize ingress hook include file

[Lukas Wunner]

Prepare for addition of a netfilter egress hook by generalizing the
ingress hook include file.

No functional change intended.

Signed-off-by: Lukas Wunner <lukas@wunner.de>
---
 include/linux/netfilter_netdev.h | 20 +++++++++++---------
 net/core/dev.c                   |  2 +-
 2 files changed, 12 insertions(+), 10 deletions(-)

diff --git a/include/linux/netfilter_netdev.h b/include/linux/netfilter_netdev.h
index a13774be2eb5..5812b0fb0278 100644
--- a/include/linux/netfilter_netdev.h
+++ b/include/linux/netfilter_netdev.h
@@ -1,6 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
-#ifndef _NETFILTER_INGRESS_H_
-#define _NETFILTER_INGRESS_H_
+#ifndef _NETFILTER_NETDEV_H_
+#define _NETFILTER_NETDEV_H_
 
 #include <linux/netfilter.h>
 #include <linux/netdevice.h>
@@ -38,10 +38,6 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 	return ret;
 }
 
-static inline void nf_hook_ingress_init(struct net_device *dev)
-{
-	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
-}
 #else /* CONFIG_NETFILTER_INGRESS */
 static inline int nf_hook_ingress_active(struct sk_buff *skb)
 {
@@ -52,7 +48,13 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 {
 	return 0;
 }
-
-static inline void nf_hook_ingress_init(struct net_device *dev) {}
 #endif /* CONFIG_NETFILTER_INGRESS */
-#endif /* _NETFILTER_INGRESS_H_ */
+
+static inline void nf_hook_netdev_init(struct net_device *dev)
+{
+#ifdef CONFIG_NETFILTER_INGRESS
+	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
+#endif
+}
+
+#endif /* _NETFILTER_NETDEV_H_ */
diff --git a/net/core/dev.c b/net/core/dev.c
index 3b79121a6a07..cca23d024afe 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -10862,7 +10862,7 @@ struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
 	if (!dev->ethtool_ops)
 		dev->ethtool_ops = &default_ethtool_ops;
 
-	nf_hook_ingress_init(dev);
+	nf_hook_netdev_init(dev);
 
 	return dev;
 
-- 
2.31.1

[PATCH nf-next v6 3/4] netfilter: Introduce egress hook

[Lukas Wunner]

Support classifying packets with netfilter on egress to satisfy user
requirements such as:
* outbound security policies for containers (Laura)
* filtering and mangling intra-node Direct Server Return (DSR) traffic
  on a load balancer (Laura)
* filtering locally generated traffic coming in through AF_PACKET,
  such as local ARP traffic generated for clustering purposes or DHCP
  (Laura; the AF_PACKET plumbing is contained in a follow-up commit)
* L2 filtering from ingress and egress for AVB (Audio Video Bridging)
  and gPTP with nftables (Pablo)
* in the future: in-kernel NAT64/NAT46 (Pablo)

The egress hook introduced herein complements the ingress hook added by
commit e687ad60af09 ("netfilter: add netfilter ingress hook after
handle_ing() under unique static key").  A patch for nftables to hook up
egress rules from user space has been submitted separately, so users may
immediately take advantage of the feature.

Alternatively or in addition to netfilter, packets can be classified
with traffic control (tc).  On ingress, packets are classified first by
tc, then by netfilter.  On egress, the order is reversed for symmetry.
Conceptually, tc and netfilter can be thought of as layers, with
netfilter layered above tc.

Traffic control is capable of redirecting packets to another interface
(man 8 tc-mirred).  E.g., an ingress packet may be redirected from the
host namespace to a container via a veth connection:
tc ingress (host) -> tc egress (veth host) -> tc ingress (veth container)

In this case, netfilter egress classifying is not performed when leaving
the host namespace!  That's because the packet is still on the tc layer.
If tc redirects the packet to a physical interface in the host namespace
such that it leaves the system, the packet is never subjected to
netfilter egress classifying.  That is only logical since it hasn't
passed through netfilter ingress classifying either.

Packets can alternatively be redirected at the netfilter layer using
nft fwd.  Such a packet *is* subjected to netfilter egress classifying
since it has reached the netfilter layer.

Internally, the skb->nf_skip_egress flag controls whether netfilter is
invoked on egress by __dev_queue_xmit().  Because __dev_queue_xmit() may
be called recursively by tunnel drivers such as vxlan, the flag is
reverted to false after sch_handle_egress().  This ensures that
netfilter is applied both on the overlay and underlying network.

Interaction between tc and netfilter is possible by setting and querying
skb->mark.

If netfilter egress classifying is not enabled on any interface, it is
patched out of the data path by way of a static_key and doesn't make a
performance difference that is discernible from noise:

Before:             1537 1538 1538 1537 1538 1537 Mb/sec
After:              1536 1534 1539 1539 1539 1540 Mb/sec
Before + tc accept: 1418 1418 1418 1419 1419 1418 Mb/sec
After  + tc accept: 1419 1424 1418 1419 1422 1420 Mb/sec
Before + tc drop:   1620 1619 1619 1619 1620 1620 Mb/sec
After  + tc drop:   1616 1624 1625 1624 1622 1619 Mb/sec

When netfilter egress classifying is enabled on at least one interface,
a minimal performance penalty is incurred for every egress packet, even
if the interface it's transmitted over doesn't have any netfilter egress
rules configured.  That is caused by checking dev->nf_hooks_egress
against NULL.

Measurements were performed on a Core i7-3615QM.  Commands to reproduce:
ip link add dev foo type dummy
ip link set dev foo up
modprobe pktgen
echo "add_device foo" > /proc/net/pktgen/kpktgend_3
samples/pktgen/pktgen_bench_xmit_mode_queue_xmit.sh -i foo -n 400000000 -m "11:11:11:11:11:11" -d 1.1.1.1

Accept all traffic with tc:
tc qdisc add dev foo clsact
tc filter add dev foo egress bpf da bytecode '1,6 0 0 0,'

Drop all traffic with tc:
tc qdisc add dev foo clsact
tc filter add dev foo egress bpf da bytecode '1,6 0 0 2,'

Apply this patch when measuring packet drops to avoid errors in dmesg:
https://lore.kernel.org/netdev/a73dda33-57f4-95d8-ea51-ed483abd6a7a@iogearbox.net/

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: Laura García Liébana <nevola@gmail.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Thomas Graf <tgraf@suug.ch>
---
 drivers/net/ifb.c                |  3 ++
 include/linux/netdevice.h        |  4 ++
 include/linux/netfilter_netdev.h | 86 ++++++++++++++++++++++++++++++++
 include/linux/skbuff.h           |  4 ++
 include/uapi/linux/netfilter.h   |  1 +
 net/core/dev.c                   | 15 +++++-
 net/netfilter/Kconfig            | 11 ++++
 net/netfilter/core.c             | 34 +++++++++++--
 net/netfilter/nfnetlink_hook.c   | 16 ++++--
 net/netfilter/nft_chain_filter.c |  4 +-
 10 files changed, 168 insertions(+), 10 deletions(-)

diff --git a/drivers/net/ifb.c b/drivers/net/ifb.c
index e9258a9f3702..2c319dd27f29 100644
--- a/drivers/net/ifb.c
+++ b/drivers/net/ifb.c
@@ -31,6 +31,7 @@
 #include <linux/init.h>
 #include <linux/interrupt.h>
 #include <linux/moduleparam.h>
+#include <linux/netfilter_netdev.h>
 #include <net/pkt_sched.h>
 #include <net/net_namespace.h>
 
@@ -75,8 +76,10 @@ static void ifb_ri_tasklet(struct tasklet_struct *t)
 	}
 
 	while ((skb = __skb_dequeue(&txp->tq)) != NULL) {
+		/* Skip tc and netfilter to prevent redirection loop. */
 		skb->redirected = 0;
 		skb->tc_skip_classify = 1;
+		nf_skip_egress(skb, true);
 
 		u64_stats_update_begin(&txp->tsync);
 		txp->tx_packets++;
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index d79163208dfd..e9a48068f306 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -1861,6 +1861,7 @@ enum netdev_ml_priv_type {
  *	@xps_maps:	XXX: need comments on this one
  *	@miniq_egress:		clsact qdisc specific data for
  *				egress processing
+ *	@nf_hooks_egress:	netfilter hooks executed for egress packets
  *	@qdisc_hash:		qdisc hash table
  *	@watchdog_timeo:	Represents the timeout that is used by
  *				the watchdog (see dev_watchdog())
@@ -2161,6 +2162,9 @@ struct net_device {
 #ifdef CONFIG_NET_CLS_ACT
 	struct mini_Qdisc __rcu	*miniq_egress;
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	struct nf_hook_entries __rcu *nf_hooks_egress;
+#endif
 
 #ifdef CONFIG_NET_SCHED
 	DECLARE_HASHTABLE	(qdisc_hash, 4);
diff --git a/include/linux/netfilter_netdev.h b/include/linux/netfilter_netdev.h
index 5812b0fb0278..b71b57a83bb4 100644
--- a/include/linux/netfilter_netdev.h
+++ b/include/linux/netfilter_netdev.h
@@ -50,11 +50,97 @@ static inline int nf_hook_ingress(struct sk_buff *skb)
 }
 #endif /* CONFIG_NETFILTER_INGRESS */
 
+#ifdef CONFIG_NETFILTER_EGRESS
+static inline bool nf_hook_egress_active(void)
+{
+#ifdef CONFIG_JUMP_LABEL
+	if (!static_key_false(&nf_hooks_needed[NFPROTO_NETDEV][NF_NETDEV_EGRESS]))
+		return false;
+#endif
+	return true;
+}
+
+/**
+ * nf_hook_egress - classify packets before transmission
+ * @skb: packet to be classified
+ * @rc: result code which shall be returned by __dev_queue_xmit() on failure
+ * @dev: netdev whose egress hooks shall be applied to @skb
+ *
+ * Returns @skb on success or %NULL if the packet was consumed or filtered.
+ * Caller must hold rcu_read_lock.
+ *
+ * On ingress, packets are classified first by tc, then by netfilter.
+ * On egress, the order is reversed for symmetry.  Conceptually, tc and
+ * netfilter can be thought of as layers, with netfilter layered above tc:
+ * When tc redirects a packet to another interface, netfilter is not applied
+ * because the packet is on the tc layer.
+ *
+ * The nf_skip_egress flag controls whether netfilter is applied on egress.
+ * It is updated by __netif_receive_skb_core() and __dev_queue_xmit() when the
+ * packet passes through tc and netfilter.  Because __dev_queue_xmit() may be
+ * called recursively by tunnel drivers such as vxlan, the flag is reverted to
+ * false after sch_handle_egress().  This ensures that netfilter is applied
+ * both on the overlay and underlying network.
+ */
+static inline struct sk_buff *nf_hook_egress(struct sk_buff *skb, int *rc,
+					     struct net_device *dev)
+{
+	struct nf_hook_entries *e;
+	struct nf_hook_state state;
+	int ret;
+
+#ifdef CONFIG_NETFILTER_SKIP_EGRESS
+	if (skb->nf_skip_egress)
+		return skb;
+#endif
+
+	e = rcu_dereference(dev->nf_hooks_egress);
+	if (!e)
+		return skb;
+
+	nf_hook_state_init(&state, NF_NETDEV_EGRESS,
+			   NFPROTO_NETDEV, dev, NULL, NULL,
+			   dev_net(dev), NULL);
+	ret = nf_hook_slow(skb, &state, e, 0);
+
+	if (ret == 1) {
+		return skb;
+	} else if (ret < 0) {
+		*rc = NET_XMIT_DROP;
+		return NULL;
+	} else { /* ret == 0 */
+		*rc = NET_XMIT_SUCCESS;
+		return NULL;
+	}
+}
+#else /* CONFIG_NETFILTER_EGRESS */
+static inline bool nf_hook_egress_active(void)
+{
+	return false;
+}
+
+static inline struct sk_buff *nf_hook_egress(struct sk_buff *skb, int *rc,
+					     struct net_device *dev)
+{
+	return skb;
+}
+#endif /* CONFIG_NETFILTER_EGRESS */
+
+static inline void nf_skip_egress(struct sk_buff *skb, bool skip)
+{
+#ifdef CONFIG_NETFILTER_SKIP_EGRESS
+	skb->nf_skip_egress = skip;
+#endif
+}
+
 static inline void nf_hook_netdev_init(struct net_device *dev)
 {
 #ifdef CONFIG_NETFILTER_INGRESS
 	RCU_INIT_POINTER(dev->nf_hooks_ingress, NULL);
 #endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	RCU_INIT_POINTER(dev->nf_hooks_egress, NULL);
+#endif
 }
 
 #endif /* _NETFILTER_NETDEV_H_ */
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 841e2f0f5240..cb96f1e6460c 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -652,6 +652,7 @@ typedef unsigned char *sk_buff_data_t;
  *	@tc_at_ingress: used within tc_classify to distinguish in/egress
  *	@redirected: packet was redirected by packet classifier
  *	@from_ingress: packet was redirected from the ingress path
+ *	@nf_skip_egress: packet shall skip nf egress - see netfilter_netdev.h
  *	@peeked: this packet has been seen already, so stats have been
  *		done for it, don't do them again
  *	@nf_trace: netfilter packet trace flag
@@ -868,6 +869,9 @@ struct sk_buff {
 #ifdef CONFIG_NET_REDIRECT
 	__u8			from_ingress:1;
 #endif
+#ifdef CONFIG_NETFILTER_SKIP_EGRESS
+	__u8			nf_skip_egress:1;
+#endif
 #ifdef CONFIG_TLS_DEVICE
 	__u8			decrypted:1;
 #endif
diff --git a/include/uapi/linux/netfilter.h b/include/uapi/linux/netfilter.h
index ef9a44286e23..53411ccc69db 100644
--- a/include/uapi/linux/netfilter.h
+++ b/include/uapi/linux/netfilter.h
@@ -51,6 +51,7 @@ enum nf_inet_hooks {
 
 enum nf_dev_hooks {
 	NF_NETDEV_INGRESS,
+	NF_NETDEV_EGRESS,
 	NF_NETDEV_NUMHOOKS
 };
 
diff --git a/net/core/dev.c b/net/core/dev.c
index cca23d024afe..db1d3a46d98a 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3920,6 +3920,7 @@ EXPORT_SYMBOL(dev_loopback_xmit);
 static struct sk_buff *
 sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
 {
+#ifdef CONFIG_NET_CLS_ACT
 	struct mini_Qdisc *miniq = rcu_dereference_bh(dev->miniq_egress);
 	struct tcf_result cl_res;
 
@@ -3955,6 +3956,7 @@ sch_handle_egress(struct sk_buff *skb, int *ret, struct net_device *dev)
 	default:
 		break;
 	}
+#endif /* CONFIG_NET_CLS_ACT */
 
 	return skb;
 }
@@ -4148,13 +4150,20 @@ static int __dev_queue_xmit(struct sk_buff *skb, struct net_device *sb_dev)
 	qdisc_pkt_len_init(skb);
 #ifdef CONFIG_NET_CLS_ACT
 	skb->tc_at_ingress = 0;
-# ifdef CONFIG_NET_EGRESS
+#endif
+#ifdef CONFIG_NET_EGRESS
 	if (static_branch_unlikely(&egress_needed_key)) {
+		if (nf_hook_egress_active()) {
+			skb = nf_hook_egress(skb, &rc, dev);
+			if (!skb)
+				goto out;
+		}
+		nf_skip_egress(skb, true);
 		skb = sch_handle_egress(skb, &rc, dev);
 		if (!skb)
 			goto out;
+		nf_skip_egress(skb, false);
 	}
-# endif
 #endif
 	/* If device/qdisc don't need skb->dst, release it right now while
 	 * its hot in this cpu cache.
@@ -5296,6 +5305,7 @@ static int __netif_receive_skb_core(struct sk_buff **pskb, bool pfmemalloc,
 	if (static_branch_unlikely(&ingress_needed_key)) {
 		bool another = false;
 
+		nf_skip_egress(skb, true);
 		skb = sch_handle_ingress(skb, &pt_prev, &ret, orig_dev,
 					 &another);
 		if (another)
@@ -5303,6 +5313,7 @@ static int __netif_receive_skb_core(struct sk_buff **pskb, bool pfmemalloc,
 		if (!skb)
 			goto out;
 
+		nf_skip_egress(skb, false);
 		if (nf_ingress(skb, &pt_prev, &ret, orig_dev) < 0)
 			goto out;
 	}
diff --git a/net/netfilter/Kconfig b/net/netfilter/Kconfig
index 54395266339d..49c9fae9c62c 100644
--- a/net/netfilter/Kconfig
+++ b/net/netfilter/Kconfig
@@ -10,6 +10,17 @@ config NETFILTER_INGRESS
 	  This allows you to classify packets from ingress using the Netfilter
 	  infrastructure.
 
+config NETFILTER_EGRESS
+	bool "Netfilter egress support"
+	default y
+	select NET_EGRESS
+	help
+	  This allows you to classify packets before transmission using the
+	  Netfilter infrastructure.
+
+config NETFILTER_SKIP_EGRESS
+	def_bool NETFILTER_EGRESS && (NET_CLS_ACT || IFB)
+
 config NETFILTER_NETLINK
 	tristate
 
diff --git a/net/netfilter/core.c b/net/netfilter/core.c
index 63d032191e62..3a32a813fcde 100644
--- a/net/netfilter/core.c
+++ b/net/netfilter/core.c
@@ -316,6 +316,12 @@ nf_hook_entry_head(struct net *net, int pf, unsigned int hooknum,
 		if (dev && dev_net(dev) == net)
 			return &dev->nf_hooks_ingress;
 	}
+#endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	if (hooknum == NF_NETDEV_EGRESS) {
+		if (dev && dev_net(dev) == net)
+			return &dev->nf_hooks_egress;
+	}
 #endif
 	WARN_ON_ONCE(1);
 	return NULL;
@@ -344,6 +350,11 @@ static inline bool nf_ingress_hook(const struct nf_hook_ops *reg, int pf)
 	return false;
 }
 
+static inline bool nf_egress_hook(const struct nf_hook_ops *reg, int pf)
+{
+	return pf == NFPROTO_NETDEV && reg->hooknum == NF_NETDEV_EGRESS;
+}
+
 static void nf_static_key_inc(const struct nf_hook_ops *reg, int pf)
 {
 #ifdef CONFIG_JUMP_LABEL
@@ -383,9 +394,18 @@ static int __nf_register_net_hook(struct net *net, int pf,
 
 	switch (pf) {
 	case NFPROTO_NETDEV:
-		err = nf_ingress_check(net, reg, NF_NETDEV_INGRESS);
-		if (err < 0)
-			return err;
+#ifndef CONFIG_NETFILTER_INGRESS
+		if (reg->hooknum == NF_NETDEV_INGRESS)
+			return -EOPNOTSUPP;
+#endif
+#ifndef CONFIG_NETFILTER_EGRESS
+		if (reg->hooknum == NF_NETDEV_EGRESS)
+			return -EOPNOTSUPP;
+#endif
+		if ((reg->hooknum != NF_NETDEV_INGRESS &&
+		     reg->hooknum != NF_NETDEV_EGRESS) ||
+		    !reg->dev || dev_net(reg->dev) != net)
+			return -EINVAL;
 		break;
 	case NFPROTO_INET:
 		if (reg->hooknum != NF_INET_INGRESS)
@@ -417,6 +437,10 @@ static int __nf_register_net_hook(struct net *net, int pf,
 #ifdef CONFIG_NETFILTER_INGRESS
 	if (nf_ingress_hook(reg, pf))
 		net_inc_ingress_queue();
+#endif
+#ifdef CONFIG_NETFILTER_EGRESS
+	if (nf_egress_hook(reg, pf))
+		net_inc_egress_queue();
 #endif
 	nf_static_key_inc(reg, pf);
 
@@ -474,6 +498,10 @@ static void __nf_unregister_net_hook(struct net *net, int pf,
 #ifdef CONFIG_NETFILTER_INGRESS
 		if (nf_ingress_hook(reg, pf))
 			net_dec_ingress_queue();
+#endif
+#ifdef CONFIG_NETFILTER_EGRESS
+		if (nf_egress_hook(reg, pf))
+			net_dec_egress_queue();
 #endif
 		nf_static_key_dec(reg, pf);
 	} else {
diff --git a/net/netfilter/nfnetlink_hook.c b/net/netfilter/nfnetlink_hook.c
index f554e2ea32ee..d5c719c9e36c 100644
--- a/net/netfilter/nfnetlink_hook.c
+++ b/net/netfilter/nfnetlink_hook.c
@@ -185,7 +185,7 @@ static const struct nf_hook_entries *
 nfnl_hook_entries_head(u8 pf, unsigned int hook, struct net *net, const char *dev)
 {
 	const struct nf_hook_entries *hook_head = NULL;
-#ifdef CONFIG_NETFILTER_INGRESS
+#if defined(CONFIG_NETFILTER_INGRESS) || defined(CONFIG_NETFILTER_EGRESS)
 	struct net_device *netdev;
 #endif
 
@@ -221,9 +221,9 @@ nfnl_hook_entries_head(u8 pf, unsigned int hook, struct net *net, const char *de
 		hook_head = rcu_dereference(net->nf.hooks_decnet[hook]);
 		break;
 #endif
-#ifdef CONFIG_NETFILTER_INGRESS
+#if defined(CONFIG_NETFILTER_INGRESS) || defined(CONFIG_NETFILTER_EGRESS)
 	case NFPROTO_NETDEV:
-		if (hook != NF_NETDEV_INGRESS)
+		if (hook >= NF_NETDEV_NUMHOOKS)
 			return ERR_PTR(-EOPNOTSUPP);
 
 		if (!dev)
@@ -233,7 +233,15 @@ nfnl_hook_entries_head(u8 pf, unsigned int hook, struct net *net, const char *de
 		if (!netdev)
 			return ERR_PTR(-ENODEV);
 
-		return rcu_dereference(netdev->nf_hooks_ingress);
+#ifdef CONFIG_NETFILTER_INGRESS
+		if (hook == NF_NETDEV_INGRESS)
+			return rcu_dereference(netdev->nf_hooks_ingress);
+#endif
+#ifdef CONFIG_NETFILTER_EGRESS
+		if (hook == NF_NETDEV_EGRESS)
+			return rcu_dereference(netdev->nf_hooks_egress);
+#endif
+		fallthrough;
 #endif
 	default:
 		return ERR_PTR(-EPROTONOSUPPORT);
diff --git a/net/netfilter/nft_chain_filter.c b/net/netfilter/nft_chain_filter.c
index 5b02408a920b..680fe557686e 100644
--- a/net/netfilter/nft_chain_filter.c
+++ b/net/netfilter/nft_chain_filter.c
@@ -310,9 +310,11 @@ static const struct nft_chain_type nft_chain_filter_netdev = {
 	.name		= "filter",
 	.type		= NFT_CHAIN_T_DEFAULT,
 	.family		= NFPROTO_NETDEV,
-	.hook_mask	= (1 << NF_NETDEV_INGRESS),
+	.hook_mask	= (1 << NF_NETDEV_INGRESS) |
+			  (1 << NF_NETDEV_EGRESS),
 	.hooks		= {
 		[NF_NETDEV_INGRESS]	= nft_do_chain_netdev,
+		[NF_NETDEV_EGRESS]	= nft_do_chain_netdev,
 	},
 };
 
-- 
2.31.1

[PATCH nf-next v6 4/4] af_packet: Introduce egress hook

[Lukas Wunner]

From: Pablo Neira Ayuso <pablo@netfilter.org>

Add egress hook for AF_PACKET sockets that have the PACKET_QDISC_BYPASS
socket option set to on, which allows packets to escape without being
filtered in the egress path.

This patch only updates the AF_PACKET path, it does not update
dev_direct_xmit() so the XDP infrastructure has a chance to bypass
Netfilter.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
[lukas: acquire rcu_read_lock, fix typos, rebase]
Signed-off-by: Lukas Wunner <lukas@wunner.de>
---
 net/packet/af_packet.c | 35 +++++++++++++++++++++++++++++++++++
 1 file changed, 35 insertions(+)

diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index 2a2bc64f75cf..46943a18a10d 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -91,6 +91,7 @@
 #endif
 #include <linux/bpf.h>
 #include <net/compat.h>
+#include <linux/netfilter_netdev.h>
 
 #include "internal.h"
 
@@ -241,8 +242,42 @@ struct packet_skb_cb {
 static void __fanout_unlink(struct sock *sk, struct packet_sock *po);
 static void __fanout_link(struct sock *sk, struct packet_sock *po);
 
+#ifdef CONFIG_NETFILTER_EGRESS
+static noinline struct sk_buff *nf_hook_direct_egress(struct sk_buff *skb)
+{
+	struct sk_buff *next, *head = NULL, *tail;
+	int rc;
+
+	rcu_read_lock();
+	for (; skb != NULL; skb = next) {
+		next = skb->next;
+		skb_mark_not_on_list(skb);
+
+		if (!nf_hook_egress(skb, &rc, skb->dev))
+			continue;
+
+		if (!head)
+			head = skb;
+		else
+			tail->next = skb;
+
+		tail = skb;
+	}
+	rcu_read_unlock();
+
+	return head;
+}
+#endif
+
 static int packet_direct_xmit(struct sk_buff *skb)
 {
+#ifdef CONFIG_NETFILTER_EGRESS
+	if (nf_hook_egress_active()) {
+		skb = nf_hook_direct_egress(skb);
+		if (!skb)
+			return NET_XMIT_DROP;
+	}
+#endif
 	return dev_direct_xmit(skb, packet_pick_tx_queue(skb));
 }
 
-- 
2.31.1

Re: [PATCH nf-next v6 0/4] Netfilter egress hook

[Pablo Neira Ayuso]

On Fri, Oct 08, 2021 at 10:06:00PM +0200, Lukas Wunner wrote:
> Netfilter egress hook, 6th iteration
> 
> Changes:
> 
> * Perform netfilter egress classifying before tc egress classifying
>   to achieve reverse order vis-a-vis ingress datapath.
> 
> * Avoid layering violations by way of new skb->nf_skip_egress flag.
> 
> * Add egress support to new nfnetlink_hook.c.
> 
> 
> Link to previous version v5 (posted by Pablo):
> https://lore.kernel.org/netdev/20210928095538.114207-1-pablo@netfilter.org/
> 
> Link to previous version v4:
> https://lore.kernel.org/netdev/cover.1611304190.git.lukas@wunner.de/

Applied to nf-next, thanks.