[PATCH net-next] packet: add sockopt to ignore outgoing packets

[PATCH net-next] packet: add sockopt to ignore outgoing packets

[Vincent Whitchurch]

Currently, the only way to ignore outgoing packets on a packet socket is
via the BPF filter.  With MSG_ZEROCOPY, packets that are looped into
AF_PACKET are copied in dev_queue_xmit_nit(), and this copy happens even
if the filter run from packet_rcv() would reject them.  So the presence
of a packet socket on the interface takes away the benefits of
MSG_ZEROCOPY, even if the packet socket is not interested in outgoing
packets.  (Even when MSG_ZEROCOPY is not used, the skb is unnecessarily
cloned, but the cost for that is much lower.)

Add a socket option to allow AF_PACKET sockets to ignore outgoing
packets to solve this.  Note that the *BSDs already have something
similar: BIOCSSEESENT/BIOCSDIRECTION and BIOCSDIRFILT.

The first intended user is lldpd.

Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
---
 include/linux/netdevice.h      |  1 +
 include/uapi/linux/if_packet.h |  1 +
 net/core/dev.c                 |  3 +++
 net/packet/af_packet.c         | 15 +++++++++++++++
 4 files changed, 20 insertions(+)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index ca5ab98053c8..8ef14d9edc58 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -2317,6 +2317,7 @@ static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
 
 struct packet_type {
 	__be16			type;	/* This is really htons(ether_type). */
+	bool			ignore_outgoing;
 	struct net_device	*dev;	/* NULL is wildcarded here	     */
 	int			(*func) (struct sk_buff *,
 					 struct net_device *,
diff --git a/include/uapi/linux/if_packet.h b/include/uapi/linux/if_packet.h
index 67b61d91d89b..467b654bd4c7 100644
--- a/include/uapi/linux/if_packet.h
+++ b/include/uapi/linux/if_packet.h
@@ -57,6 +57,7 @@ struct sockaddr_ll {
 #define PACKET_QDISC_BYPASS		20
 #define PACKET_ROLLOVER_STATS		21
 #define PACKET_FANOUT_DATA		22
+#define PACKET_IGNORE_OUTGOING		23
 
 #define PACKET_FANOUT_HASH		0
 #define PACKET_FANOUT_LB		1
diff --git a/net/core/dev.c b/net/core/dev.c
index 325fc5088370..0addb4f0abfe 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -1947,6 +1947,9 @@ static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
 	if (!ptype->af_packet_priv || !skb->sk)
 		return false;
 
+	if (ptype->ignore_outgoing)
+		return true;
+
 	if (ptype->id_match)
 		return ptype->id_match(ptype, skb->sk);
 	else if ((struct sock *)ptype->af_packet_priv == skb->sk)
diff --git a/net/packet/af_packet.c b/net/packet/af_packet.c
index 5610061e7f2e..37bbafad724a 100644
--- a/net/packet/af_packet.c
+++ b/net/packet/af_packet.c
@@ -3805,6 +3805,18 @@ packet_setsockopt(struct socket *sock, int level, int optname, char __user *optv
 
 		return fanout_set_data(po, optval, optlen);
 	}
+	case PACKET_IGNORE_OUTGOING:
+	{
+		int val;
+
+		if (optlen != sizeof(val))
+			return -EINVAL;
+		if (copy_from_user(&val, optval, sizeof(val)))
+			return -EFAULT;
+
+		po->prot_hook.ignore_outgoing = !!val;
+		return 0;
+	}
 	case PACKET_TX_HAS_OFF:
 	{
 		unsigned int val;
@@ -3928,6 +3940,9 @@ static int packet_getsockopt(struct socket *sock, int level, int optname,
 			((u32)po->fanout->flags << 24)) :
 		       0);
 		break;
+	case PACKET_IGNORE_OUTGOING:
+		val = po->prot_hook.ignore_outgoing;
+		break;
 	case PACKET_ROLLOVER_STATS:
 		if (!po->rollover)
 			return -EINVAL;
-- 
2.11.0

Re: [PATCH net-next] packet: add sockopt to ignore outgoing packets

[Willem de Bruijn]

On Thu, Aug 30, 2018 at 6:12 AM Vincent Whitchurch
<vincent.whitchurch@axis.com> wrote:
>
> Currently, the only way to ignore outgoing packets on a packet socket is
> via the BPF filter.  With MSG_ZEROCOPY, packets that are looped into
> AF_PACKET are copied in dev_queue_xmit_nit(), and this copy happens even
> if the filter run from packet_rcv() would reject them.  So the presence
> of a packet socket on the interface takes away the benefits of
> MSG_ZEROCOPY, even if the packet socket is not interested in outgoing
> packets.  (Even when MSG_ZEROCOPY is not used, the skb is unnecessarily
> cloned, but the cost for that is much lower.)
>
> Add a socket option to allow AF_PACKET sockets to ignore outgoing
> packets to solve this.  Note that the *BSDs already have something
> similar: BIOCSSEESENT/BIOCSDIRECTION and BIOCSDIRFILT.
>
> The first intended user is lldpd.

Clear description of the use case, thanks. I don't see a simple alternative
to introducing a new socket option, either (a new ETH_P_xx protocol
wildcard different from ETH_P_ALL, perhaps).

> Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
> ---
>  include/linux/netdevice.h      |  1 +
>  include/uapi/linux/if_packet.h |  1 +
>  net/core/dev.c                 |  3 +++
>  net/packet/af_packet.c         | 15 +++++++++++++++
>  4 files changed, 20 insertions(+)
>
> diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
> index ca5ab98053c8..8ef14d9edc58 100644
> --- a/include/linux/netdevice.h
> +++ b/include/linux/netdevice.h
> @@ -2317,6 +2317,7 @@ static inline struct sk_buff *call_gro_receive_sk(gro_receive_sk_t cb,
>
>  struct packet_type {
>         __be16                  type;   /* This is really htons(ether_type). */
> +       bool                    ignore_outgoing;
>         struct net_device       *dev;   /* NULL is wildcarded here           */
>         int                     (*func) (struct sk_buff *,
>                                          struct net_device *,
> diff --git a/include/uapi/linux/if_packet.h b/include/uapi/linux/if_packet.h
> index 67b61d91d89b..467b654bd4c7 100644
> --- a/include/uapi/linux/if_packet.h
> +++ b/include/uapi/linux/if_packet.h
> @@ -57,6 +57,7 @@ struct sockaddr_ll {
>  #define PACKET_QDISC_BYPASS            20
>  #define PACKET_ROLLOVER_STATS          21
>  #define PACKET_FANOUT_DATA             22
> +#define PACKET_IGNORE_OUTGOING         23


>
>  #define PACKET_FANOUT_HASH             0
>  #define PACKET_FANOUT_LB               1
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 325fc5088370..0addb4f0abfe 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -1947,6 +1947,9 @@ static inline bool skb_loop_sk(struct packet_type *ptype, struct sk_buff *skb)
>         if (!ptype->af_packet_priv || !skb->sk)
>                 return false;
>
> +       if (ptype->ignore_outgoing)
> +               return true;
> +

This probably does not belong in skb_loop_sk, but in
dev_queue_xmit_nit directly.

>         if (ptype->id_match)
>                 return ptype->id_match(ptype, skb->sk);
>         else if ((struct sock *)ptype->af_packet_priv == skb->sk)