[patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[Jiri Pirko]

Per Davem's request, I prepared this patchset which introduces programmable
flow dissector. For current users of flow_keys, there is a wrapper
skb_flow_dissect_flow_keys which maintains the previous behaviour.
For purposes of cls_flower, couple of new dissection keys were introduced.

Note that this dissector can be also eventually used by openvswitch code.

Also, as a next step, I plan to get rid of *skb_flow_get_ports(export)
and *__skb_get_poff as their functionality can be now implemented by
skb_flow_dissect as well.

v2->v3:
- remove TCA_FLOWER_POLICE attr suggested by Jamal

v1->v2:
- move __skb_tx_hash rather to dev.c as suggested by Alex

Jiri Pirko (15):
  net: change name of flow_dissector header to match the .c file name
  flow_dissector: remove unused function flow_get_hlen declaration
  net: move *skb_get_poff declarations into correct header
  flow_dissector: fix doc for __skb_get_hash and remove couple of empty
    lines
  net: move __skb_get_hash function declaration to flow_dissector.h
  net: move __skb_tx_hash to dev.c
  net: move netdev_pick_tx and dependencies to net/core/dev.c
  flow_dissector: fix doc for skb_get_poff
  flow_dissector: introduce programable flow_dissector
  flow_dissect: use programable dissector in skb_flow_dissect and
    friends
  flow_dissector: add missing header includes
  flow_dissector: introduce support for ipv6 addressses
  flow_dissector: introduce support for Ethernet addresses
  flow_dissector: change port array into src,dst tuple
  tc: introduce Flower classifier

 drivers/net/bonding/bond_main.c                |  20 +-
 drivers/net/ethernet/cisco/enic/enic_clsf.c    |  29 +-
 drivers/net/ethernet/cisco/enic/enic_ethtool.c |  10 +-
 drivers/net/hyperv/netvsc_drv.c                |   8 +-
 include/linux/netdevice.h                      |   3 +
 include/linux/skbuff.h                         |  14 +-
 include/net/flow_dissector.h                   | 167 ++++++
 include/net/flow_keys.h                        |  61 ---
 include/net/ip.h                               |  10 +-
 include/net/ipv6.h                             |  10 +-
 include/uapi/linux/pkt_cls.h                   |  30 ++
 net/core/dev.c                                 | 106 ++++
 net/core/filter.c                              |   1 +
 net/core/flow_dissector.c                      | 380 ++++++++------
 net/ethernet/eth.c                             |   7 +-
 net/sched/Kconfig                              |  10 +
 net/sched/Makefile                             |   1 +
 net/sched/cls_flow.c                           |  22 +-
 net/sched/cls_flower.c                         | 688 +++++++++++++++++++++++++
 net/sched/sch_choke.c                          |   6 +-
 20 files changed, 1285 insertions(+), 298 deletions(-)
 create mode 100644 include/net/flow_dissector.h
 delete mode 100644 include/net/flow_keys.h
 create mode 100644 net/sched/cls_flower.c

-- 
2.4.0

[patch net-next v3 01/15] net: change name of flow_dissector header to match the .c file name

[Jiri Pirko]

add couple of empty lines on the way.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 drivers/net/bonding/bond_main.c             |  2 +-
 drivers/net/ethernet/cisco/enic/enic_clsf.c |  2 +-
 include/linux/skbuff.h                      |  2 +-
 include/net/flow_dissector.h                | 68 +++++++++++++++++++++++++++++
 include/net/flow_keys.h                     | 61 --------------------------
 include/net/ip.h                            |  2 +-
 include/net/ipv6.h                          |  2 +-
 net/core/flow_dissector.c                   |  2 +-
 net/sched/cls_flow.c                        |  2 +-
 net/sched/sch_choke.c                       |  2 +-
 10 files changed, 76 insertions(+), 69 deletions(-)
 create mode 100644 include/net/flow_dissector.h
 delete mode 100644 include/net/flow_keys.h

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index 2ee13be..eed2f86 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -76,7 +76,7 @@
 #include <net/netns/generic.h>
 #include <net/pkt_sched.h>
 #include <linux/rculist.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 #include <net/switchdev.h>
 #include <net/bonding.h>
 #include <net/bond_3ad.h>
diff --git a/drivers/net/ethernet/cisco/enic/enic_clsf.c b/drivers/net/ethernet/cisco/enic/enic_clsf.c
index 0be6850..380f049 100644
--- a/drivers/net/ethernet/cisco/enic/enic_clsf.c
+++ b/drivers/net/ethernet/cisco/enic/enic_clsf.c
@@ -5,7 +5,7 @@
 #include <linux/in.h>
 #include <linux/types.h>
 #include <linux/skbuff.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 #include "enic_res.h"
 #include "enic_clsf.h"
 
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 9c2f793..6cc1096 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -34,7 +34,7 @@
 #include <linux/dma-mapping.h>
 #include <linux/netdev_features.h>
 #include <linux/sched.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 
 /* A. Checksumming of received packets by device.
  *
diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
new file mode 100644
index 0000000..5e99a7b
--- /dev/null
+++ b/include/net/flow_dissector.h
@@ -0,0 +1,68 @@
+#ifndef _NET_FLOW_DISSECTOR_H
+#define _NET_FLOW_DISSECTOR_H
+
+/* struct flow_keys:
+ *	@src: source ip address in case of IPv4
+ *	      For IPv6 it contains 32bit hash of src address
+ *	@dst: destination ip address in case of IPv4
+ *	      For IPv6 it contains 32bit hash of dst address
+ *	@ports: port numbers of Transport header
+ *		port16[0]: src port number
+ *		port16[1]: dst port number
+ *	@thoff: Transport header offset
+ *	@n_proto: Network header protocol (eg. IPv4/IPv6)
+ *	@ip_proto: Transport header protocol (eg. TCP/UDP)
+ * All the members, except thoff, are in network byte order.
+ */
+struct flow_keys {
+	/* (src,dst) must be grouped, in the same way than in IP header */
+	__be32 src;
+	__be32 dst;
+	union {
+		__be32 ports;
+		__be16 port16[2];
+	};
+	u16	thoff;
+	__be16	n_proto;
+	u8	ip_proto;
+};
+
+bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
+			void *data, __be16 proto, int nhoff, int hlen);
+
+static inline bool skb_flow_dissect(const struct sk_buff *skb,
+				    struct flow_keys *flow)
+{
+	return __skb_flow_dissect(skb, flow, NULL, 0, 0, 0);
+}
+
+__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
+			    void *data, int hlen_proto);
+
+static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
+					int thoff, u8 ip_proto)
+{
+	return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
+}
+
+u32 flow_hash_from_keys(struct flow_keys *keys);
+
+unsigned int flow_get_hlen(const unsigned char *data, unsigned int max_len,
+			   __be16 protocol);
+
+/* struct flow_keys_digest:
+ *
+ * This structure is used to hold a digest of the full flow keys. This is a
+ * larger "hash" of a flow to allow definitively matching specific flows where
+ * the 32 bit skb->hash is not large enough. The size is limited to 16 bytes so
+ * that it can by used in CB of skb (see sch_choke for an example).
+ */
+#define FLOW_KEYS_DIGEST_LEN	16
+struct flow_keys_digest {
+	u8	data[FLOW_KEYS_DIGEST_LEN];
+};
+
+void make_flow_keys_digest(struct flow_keys_digest *digest,
+			   const struct flow_keys *flow);
+
+#endif
diff --git a/include/net/flow_keys.h b/include/net/flow_keys.h
deleted file mode 100644
index 6d6ef62..0000000
--- a/include/net/flow_keys.h
+++ /dev/null
@@ -1,61 +0,0 @@
-#ifndef _NET_FLOW_KEYS_H
-#define _NET_FLOW_KEYS_H
-
-/* struct flow_keys:
- *	@src: source ip address in case of IPv4
- *	      For IPv6 it contains 32bit hash of src address
- *	@dst: destination ip address in case of IPv4
- *	      For IPv6 it contains 32bit hash of dst address
- *	@ports: port numbers of Transport header
- *		port16[0]: src port number
- *		port16[1]: dst port number
- *	@thoff: Transport header offset
- *	@n_proto: Network header protocol (eg. IPv4/IPv6)
- *	@ip_proto: Transport header protocol (eg. TCP/UDP)
- * All the members, except thoff, are in network byte order.
- */
-struct flow_keys {
-	/* (src,dst) must be grouped, in the same way than in IP header */
-	__be32 src;
-	__be32 dst;
-	union {
-		__be32 ports;
-		__be16 port16[2];
-	};
-	u16	thoff;
-	__be16	n_proto;
-	u8	ip_proto;
-};
-
-bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
-			void *data, __be16 proto, int nhoff, int hlen);
-static inline bool skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow)
-{
-	return __skb_flow_dissect(skb, flow, NULL, 0, 0, 0);
-}
-__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
-			    void *data, int hlen_proto);
-static inline __be32 skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto)
-{
-	return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
-}
-u32 flow_hash_from_keys(struct flow_keys *keys);
-unsigned int flow_get_hlen(const unsigned char *data, unsigned int max_len,
-			   __be16 protocol);
-
-/* struct flow_keys_digest:
- *
- * This structure is used to hold a digest of the full flow keys. This is a
- * larger "hash" of a flow to allow definitively matching specific flows where
- * the 32 bit skb->hash is not large enough. The size is limited to 16 bytes so
- * that it can by used in CB of skb (see sch_choke for an example).
- */
-#define FLOW_KEYS_DIGEST_LEN	16
-struct flow_keys_digest {
-	u8	data[FLOW_KEYS_DIGEST_LEN];
-};
-
-void make_flow_keys_digest(struct flow_keys_digest *digest,
-			   const struct flow_keys *flow);
-
-#endif
diff --git a/include/net/ip.h b/include/net/ip.h
index d14af7e..562eb65 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -31,7 +31,7 @@
 #include <net/route.h>
 #include <net/snmp.h>
 #include <net/flow.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 
 struct sock;
 
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 53d25ef..9932b86 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -19,7 +19,7 @@
 #include <net/if_inet6.h>
 #include <net/ndisc.h>
 #include <net/flow.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 #include <net/snmp.h>
 
 #define SIN6_LEN_RFC2133	24
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index d3acc4d..eaa86b5 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -12,7 +12,7 @@
 #include <linux/if_tunnel.h>
 #include <linux/if_pppox.h>
 #include <linux/ppp_defs.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 #include <scsi/fc/fc_fcoe.h>
 
 /* copy saddr & daddr, possibly using 64bit load/store
diff --git a/net/sched/cls_flow.c b/net/sched/cls_flow.c
index a620c4e..4c5a922 100644
--- a/net/sched/cls_flow.c
+++ b/net/sched/cls_flow.c
@@ -26,7 +26,7 @@
 #include <net/pkt_cls.h>
 #include <net/ip.h>
 #include <net/route.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 
 #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
 #include <net/netfilter/nf_conntrack.h>
diff --git a/net/sched/sch_choke.c b/net/sched/sch_choke.c
index dfe3da7..2624e99 100644
--- a/net/sched/sch_choke.c
+++ b/net/sched/sch_choke.c
@@ -18,7 +18,7 @@
 #include <net/pkt_sched.h>
 #include <net/inet_ecn.h>
 #include <net/red.h>
-#include <net/flow_keys.h>
+#include <net/flow_dissector.h>
 
 /*
    CHOKe stateless AQM for fair bandwidth allocation
-- 
2.4.0

[patch net-next v3 02/15] flow_dissector: remove unused function flow_get_hlen declaration

[Jiri Pirko]

commit 56193d1bce ("net: Add function for parsing the header length out
of linear ethernet frames") added this function declaration but it is
defined nowhere.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/net/flow_dissector.h | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 5e99a7b..118ae69 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -47,9 +47,6 @@ static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
 
 u32 flow_hash_from_keys(struct flow_keys *keys);
 
-unsigned int flow_get_hlen(const unsigned char *data, unsigned int max_len,
-			   __be16 protocol);
-
 /* struct flow_keys_digest:
  *
  * This structure is used to hold a digest of the full flow keys. This is a
-- 
2.4.0

[patch net-next v3 03/15] net: move *skb_get_poff declarations into correct header

[Jiri Pirko]

Since these functions are defined in flow_dissector.c, move header
declarations from skbuff.h into flow_dissector.h

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/linux/skbuff.h       | 4 ----
 include/net/flow_dissector.h | 3 +++
 net/core/filter.c            | 1 +
 net/ethernet/eth.c           | 1 +
 4 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 6cc1096..8616bfe 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -3423,10 +3423,6 @@ struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
 				     unsigned int transport_len,
 				     __sum16(*skb_chkf)(struct sk_buff *skb));
 
-u32 skb_get_poff(const struct sk_buff *skb);
-u32 __skb_get_poff(const struct sk_buff *skb, void *data,
-		   const struct flow_keys *keys, int hlen);
-
 /**
  * skb_head_is_locked - Determine if the skb->head is locked down
  * @skb: skb to check
diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 118ae69..4570cca 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -46,6 +46,9 @@ static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
 }
 
 u32 flow_hash_from_keys(struct flow_keys *keys);
+u32 skb_get_poff(const struct sk_buff *skb);
+u32 __skb_get_poff(const struct sk_buff *skb, void *data,
+		   const struct flow_keys *keys, int hlen);
 
 /* struct flow_keys_digest:
  *
diff --git a/net/core/filter.c b/net/core/filter.c
index a831f19..6805717 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -36,6 +36,7 @@
 #include <net/netlink.h>
 #include <linux/skbuff.h>
 #include <net/sock.h>
+#include <net/flow_dissector.h>
 #include <linux/errno.h>
 #include <linux/timer.h>
 #include <asm/uaccess.h>
diff --git a/net/ethernet/eth.c b/net/ethernet/eth.c
index 9045e2a..9332a0a 100644
--- a/net/ethernet/eth.c
+++ b/net/ethernet/eth.c
@@ -58,6 +58,7 @@
 #include <net/ipv6.h>
 #include <net/ip.h>
 #include <net/dsa.h>
+#include <net/flow_dissector.h>
 #include <linux/uaccess.h>
 
 __setup("ether=", netdev_boot_setup);
-- 
2.4.0

[patch net-next v3 04/15] flow_dissector: fix doc for __skb_get_hash and remove couple of empty lines

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 net/core/flow_dissector.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index eaa86b5..0d9bc3a 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -337,8 +337,11 @@ void make_flow_keys_digest(struct flow_keys_digest *digest,
 }
 EXPORT_SYMBOL(make_flow_keys_digest);
 
-/*
- * __skb_get_hash: calculate a flow hash based on src/dst addresses
+/**
+ * __skb_get_hash: calculate a flow hash
+ * @skb: sk_buff to calculate flow hash from
+ *
+ * This function calculates a flow hash based on src/dst addresses
  * and src/dst port numbers.  Sets hash in skb to non-zero hash value
  * on success, zero indicates no valid hash.  Also, sets l4_hash in skb
  * if hash is a canonical 4-tuple hash over transport ports.
@@ -353,12 +356,9 @@ void __skb_get_hash(struct sk_buff *skb)
 	hash = ___skb_get_hash(skb, &keys, hashrnd);
 	if (!hash)
 		return;
-
 	if (keys.ports)
 		skb->l4_hash = 1;
-
 	skb->sw_hash = 1;
-
 	skb->hash = hash;
 }
 EXPORT_SYMBOL(__skb_get_hash);
-- 
2.4.0

[patch net-next v3 05/15] net: move __skb_get_hash function declaration to flow_dissector.h

[Jiri Pirko]

Since the definition of the function is in flow_dissector.c, it makes
sense to have the declaration in flow_dissector.h

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/linux/skbuff.h       | 1 -
 include/net/flow_dissector.h | 1 +
 2 files changed, 1 insertion(+), 1 deletion(-)

diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 8616bfe..55677e9 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -918,7 +918,6 @@ skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
 	skb->hash = hash;
 }
 
-void __skb_get_hash(struct sk_buff *skb);
 static inline __u32 skb_get_hash(struct sk_buff *skb)
 {
 	if (!skb->l4_hash && !skb->sw_hash)
diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 4570cca..e4ee761 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -46,6 +46,7 @@ static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
 }
 
 u32 flow_hash_from_keys(struct flow_keys *keys);
+void __skb_get_hash(struct sk_buff *skb);
 u32 skb_get_poff(const struct sk_buff *skb);
 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
 		   const struct flow_keys *keys, int hlen);
-- 
2.4.0

[patch net-next v3 06/15] net: move __skb_tx_hash to dev.c

[Jiri Pirko]

__skb_tx_hash function has no relation to flow_dissect so just move it
to dev.c

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/linux/netdevice.h |  3 +++
 include/linux/skbuff.h    |  3 ---
 net/core/dev.c            | 28 ++++++++++++++++++++++++++++
 net/core/flow_dissector.c | 28 ----------------------------
 4 files changed, 31 insertions(+), 31 deletions(-)

diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index a6d706b..662b55f 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -2847,6 +2847,9 @@ static inline int netif_set_xps_queue(struct net_device *dev,
 }
 #endif
 
+u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
+		  unsigned int num_tx_queues);
+
 /*
  * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
  * as a distribution range limit for the returned value.
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index 55677e9..bc207d4 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -3298,9 +3298,6 @@ static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
 	return skb->queue_mapping != 0;
 }
 
-u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
-		  unsigned int num_tx_queues);
-
 static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
 {
 #ifdef CONFIG_XFRM
diff --git a/net/core/dev.c b/net/core/dev.c
index e5f77c4..3e7838f 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2350,6 +2350,34 @@ void netif_device_attach(struct net_device *dev)
 }
 EXPORT_SYMBOL(netif_device_attach);
 
+/*
+ * Returns a Tx hash based on the given packet descriptor a Tx queues' number
+ * to be used as a distribution range.
+ */
+u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
+		  unsigned int num_tx_queues)
+{
+	u32 hash;
+	u16 qoffset = 0;
+	u16 qcount = num_tx_queues;
+
+	if (skb_rx_queue_recorded(skb)) {
+		hash = skb_get_rx_queue(skb);
+		while (unlikely(hash >= num_tx_queues))
+			hash -= num_tx_queues;
+		return hash;
+	}
+
+	if (dev->num_tc) {
+		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
+		qoffset = dev->tc_to_txq[tc].offset;
+		qcount = dev->tc_to_txq[tc].count;
+	}
+
+	return (u16) reciprocal_scale(skb_get_hash(skb), qcount) + qoffset;
+}
+EXPORT_SYMBOL(__skb_tx_hash);
+
 static void skb_warn_bad_offload(const struct sk_buff *skb)
 {
 	static const netdev_features_t null_features = 0;
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 0d9bc3a..07ca11d 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -371,34 +371,6 @@ __u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb)
 }
 EXPORT_SYMBOL(skb_get_hash_perturb);
 
-/*
- * Returns a Tx hash based on the given packet descriptor a Tx queues' number
- * to be used as a distribution range.
- */
-u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
-		  unsigned int num_tx_queues)
-{
-	u32 hash;
-	u16 qoffset = 0;
-	u16 qcount = num_tx_queues;
-
-	if (skb_rx_queue_recorded(skb)) {
-		hash = skb_get_rx_queue(skb);
-		while (unlikely(hash >= num_tx_queues))
-			hash -= num_tx_queues;
-		return hash;
-	}
-
-	if (dev->num_tc) {
-		u8 tc = netdev_get_prio_tc_map(dev, skb->priority);
-		qoffset = dev->tc_to_txq[tc].offset;
-		qcount = dev->tc_to_txq[tc].count;
-	}
-
-	return (u16) reciprocal_scale(skb_get_hash(skb), qcount) + qoffset;
-}
-EXPORT_SYMBOL(__skb_tx_hash);
-
 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
 		   const struct flow_keys *keys, int hlen)
 {
-- 
2.4.0

[patch net-next v3 07/15] net: move netdev_pick_tx and dependencies to net/core/dev.c

[Jiri Pirko]

next to its user. No relation to flow_dissector so it makes no sense to
have it in flow_dissector.c

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 net/core/dev.c            | 78 +++++++++++++++++++++++++++++++++++++++++++++++
 net/core/flow_dissector.c | 78 -----------------------------------------------
 2 files changed, 78 insertions(+), 78 deletions(-)

diff --git a/net/core/dev.c b/net/core/dev.c
index 3e7838f..8ef30be 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -2936,6 +2936,84 @@ int dev_loopback_xmit(struct sock *sk, struct sk_buff *skb)
 }
 EXPORT_SYMBOL(dev_loopback_xmit);
 
+static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
+{
+#ifdef CONFIG_XPS
+	struct xps_dev_maps *dev_maps;
+	struct xps_map *map;
+	int queue_index = -1;
+
+	rcu_read_lock();
+	dev_maps = rcu_dereference(dev->xps_maps);
+	if (dev_maps) {
+		map = rcu_dereference(
+		    dev_maps->cpu_map[skb->sender_cpu - 1]);
+		if (map) {
+			if (map->len == 1)
+				queue_index = map->queues[0];
+			else
+				queue_index = map->queues[reciprocal_scale(skb_get_hash(skb),
+									   map->len)];
+			if (unlikely(queue_index >= dev->real_num_tx_queues))
+				queue_index = -1;
+		}
+	}
+	rcu_read_unlock();
+
+	return queue_index;
+#else
+	return -1;
+#endif
+}
+
+static u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb)
+{
+	struct sock *sk = skb->sk;
+	int queue_index = sk_tx_queue_get(sk);
+
+	if (queue_index < 0 || skb->ooo_okay ||
+	    queue_index >= dev->real_num_tx_queues) {
+		int new_index = get_xps_queue(dev, skb);
+		if (new_index < 0)
+			new_index = skb_tx_hash(dev, skb);
+
+		if (queue_index != new_index && sk &&
+		    rcu_access_pointer(sk->sk_dst_cache))
+			sk_tx_queue_set(sk, new_index);
+
+		queue_index = new_index;
+	}
+
+	return queue_index;
+}
+
+struct netdev_queue *netdev_pick_tx(struct net_device *dev,
+				    struct sk_buff *skb,
+				    void *accel_priv)
+{
+	int queue_index = 0;
+
+#ifdef CONFIG_XPS
+	if (skb->sender_cpu == 0)
+		skb->sender_cpu = raw_smp_processor_id() + 1;
+#endif
+
+	if (dev->real_num_tx_queues != 1) {
+		const struct net_device_ops *ops = dev->netdev_ops;
+		if (ops->ndo_select_queue)
+			queue_index = ops->ndo_select_queue(dev, skb, accel_priv,
+							    __netdev_pick_tx);
+		else
+			queue_index = __netdev_pick_tx(dev, skb);
+
+		if (!accel_priv)
+			queue_index = netdev_cap_txqueue(dev, queue_index);
+	}
+
+	skb_set_queue_mapping(skb, queue_index);
+	return netdev_get_tx_queue(dev, queue_index);
+}
+
 /**
  *	__dev_queue_xmit - transmit a buffer
  *	@skb: buffer to transmit
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 07ca11d..04f8723 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -431,81 +431,3 @@ u32 skb_get_poff(const struct sk_buff *skb)
 
 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
 }
-
-static inline int get_xps_queue(struct net_device *dev, struct sk_buff *skb)
-{
-#ifdef CONFIG_XPS
-	struct xps_dev_maps *dev_maps;
-	struct xps_map *map;
-	int queue_index = -1;
-
-	rcu_read_lock();
-	dev_maps = rcu_dereference(dev->xps_maps);
-	if (dev_maps) {
-		map = rcu_dereference(
-		    dev_maps->cpu_map[skb->sender_cpu - 1]);
-		if (map) {
-			if (map->len == 1)
-				queue_index = map->queues[0];
-			else
-				queue_index = map->queues[reciprocal_scale(skb_get_hash(skb),
-									   map->len)];
-			if (unlikely(queue_index >= dev->real_num_tx_queues))
-				queue_index = -1;
-		}
-	}
-	rcu_read_unlock();
-
-	return queue_index;
-#else
-	return -1;
-#endif
-}
-
-static u16 __netdev_pick_tx(struct net_device *dev, struct sk_buff *skb)
-{
-	struct sock *sk = skb->sk;
-	int queue_index = sk_tx_queue_get(sk);
-
-	if (queue_index < 0 || skb->ooo_okay ||
-	    queue_index >= dev->real_num_tx_queues) {
-		int new_index = get_xps_queue(dev, skb);
-		if (new_index < 0)
-			new_index = skb_tx_hash(dev, skb);
-
-		if (queue_index != new_index && sk &&
-		    rcu_access_pointer(sk->sk_dst_cache))
-			sk_tx_queue_set(sk, new_index);
-
-		queue_index = new_index;
-	}
-
-	return queue_index;
-}
-
-struct netdev_queue *netdev_pick_tx(struct net_device *dev,
-				    struct sk_buff *skb,
-				    void *accel_priv)
-{
-	int queue_index = 0;
-
-#ifdef CONFIG_XPS
-	if (skb->sender_cpu == 0)
-		skb->sender_cpu = raw_smp_processor_id() + 1;
-#endif
-
-	if (dev->real_num_tx_queues != 1) {
-		const struct net_device_ops *ops = dev->netdev_ops;
-		if (ops->ndo_select_queue)
-			queue_index = ops->ndo_select_queue(dev, skb, accel_priv,
-							    __netdev_pick_tx);
-		else
-			queue_index = __netdev_pick_tx(dev, skb);
-
-		if (!accel_priv)
-			queue_index = netdev_cap_txqueue(dev, queue_index);
-	}
-
-	skb_set_queue_mapping(skb, queue_index);
-	return netdev_get_tx_queue(dev, queue_index);
-}
-- 
2.4.0

[patch net-next v3 08/15] flow_dissector: fix doc for skb_get_poff

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 net/core/flow_dissector.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 04f8723..d885e28 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -417,8 +417,12 @@ u32 __skb_get_poff(const struct sk_buff *skb, void *data,
 	return poff;
 }
 
-/* skb_get_poff() returns the offset to the payload as far as it could
- * be dissected. The main user is currently BPF, so that we can dynamically
+/**
+ * skb_get_poff - get the offset to the payload
+ * @skb: sk_buff to get the payload offset from
+ *
+ * The function will get the offset to the payload as far as it could
+ * be dissected.  The main user is currently BPF, so that we can dynamically
  * truncate packets without needing to push actual payload to the user
  * space and can analyze headers only, instead.
  */
-- 
2.4.0

[patch net-next v3 09/15] flow_dissector: introduce programable flow_dissector

[Jiri Pirko]

Introduce dissector infrastructure which allows user to specify which
parts of skb he wants to dissect.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/net/flow_dissector.h | 61 ++++++++++++++++++++++++++++++++++++++++++++
 net/core/flow_dissector.c    | 48 ++++++++++++++++++++++++++++++++++
 2 files changed, 109 insertions(+)

diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index e4ee761..20239e8 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -1,6 +1,64 @@
 #ifndef _NET_FLOW_DISSECTOR_H
 #define _NET_FLOW_DISSECTOR_H
 
+/**
+ * struct flow_dissector_key_basic:
+ * @thoff: Transport header offset
+ * @n_proto: Network header protocol (eg. IPv4/IPv6)
+ * @ip_proto: Transport header protocol (eg. TCP/UDP)
+ */
+struct flow_dissector_key_basic {
+	u16	thoff;
+	__be16	n_proto;
+	u8	ip_proto;
+};
+
+/**
+ * struct flow_dissector_key_addrs:
+ * @src: source ip address in case of IPv4
+ *	 For IPv6 it contains 32bit hash of src address
+ * @dst: destination ip address in case of IPv4
+ *	 For IPv6 it contains 32bit hash of dst address
+ */
+struct flow_dissector_key_addrs {
+	/* (src,dst) must be grouped, in the same way than in IP header */
+	__be32 src;
+	__be32 dst;
+};
+
+/**
+ * flow_dissector_key_tp_ports:
+ *	@ports: port numbers of Transport header
+ *		port16[0]: src port number
+ *		port16[1]: dst port number
+ */
+struct flow_dissector_key_ports {
+	union {
+		__be32 ports;
+		__be16 port16[2];
+	};
+};
+
+enum flow_dissector_key_id {
+	FLOW_DISSECTOR_KEY_BASIC, /* struct flow_dissector_key_basic */
+	FLOW_DISSECTOR_KEY_IPV4_ADDRS, /* struct flow_dissector_key_addrs */
+	FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, /* struct flow_dissector_key_addrs */
+	FLOW_DISSECTOR_KEY_PORTS, /* struct flow_dissector_key_ports */
+
+	FLOW_DISSECTOR_KEY_MAX,
+};
+
+struct flow_dissector_key {
+	enum flow_dissector_key_id key_id;
+	size_t offset; /* offset of struct flow_dissector_key_*
+			  in target the struct */
+};
+
+struct flow_dissector {
+	unsigned int used_keys; /* each bit repesents presence of one key id */
+	unsigned short int offset[FLOW_DISSECTOR_KEY_MAX];
+};
+
 /* struct flow_keys:
  *	@src: source ip address in case of IPv4
  *	      For IPv6 it contains 32bit hash of src address
@@ -27,6 +85,9 @@ struct flow_keys {
 	u8	ip_proto;
 };
 
+void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
+			     const struct flow_dissector_key *key,
+			     unsigned int key_count);
 bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
 			void *data, __be16 proto, int nhoff, int hlen);
 
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index d885e28..6883e22 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -1,3 +1,4 @@
+#include <linux/kernel.h>
 #include <linux/skbuff.h>
 #include <linux/export.h>
 #include <linux/ip.h>
@@ -15,6 +16,53 @@
 #include <net/flow_dissector.h>
 #include <scsi/fc/fc_fcoe.h>
 
+static bool skb_flow_dissector_uses_key(struct flow_dissector *flow_dissector,
+					enum flow_dissector_key_id key_id)
+{
+	return flow_dissector->used_keys & (1 << key_id);
+}
+
+static void skb_flow_dissector_set_key(struct flow_dissector *flow_dissector,
+				       enum flow_dissector_key_id key_id)
+{
+	flow_dissector->used_keys |= (1 << key_id);
+}
+
+static void *skb_flow_dissector_target(struct flow_dissector *flow_dissector,
+				       enum flow_dissector_key_id key_id,
+				       void *target_container)
+{
+	return ((char *) target_container) + flow_dissector->offset[key_id];
+}
+
+void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
+			     const struct flow_dissector_key *key,
+			     unsigned int key_count)
+{
+	unsigned int i;
+
+	memset(flow_dissector, 0, sizeof(*flow_dissector));
+
+	for (i = 0; i < key_count; i++, key++) {
+		/* User should make sure that every key target offset is withing
+		 * boundaries of unsigned short.
+		 */
+		BUG_ON(key->offset > USHRT_MAX);
+		BUG_ON(skb_flow_dissector_uses_key(flow_dissector,
+						   key->key_id));
+
+		skb_flow_dissector_set_key(flow_dissector, key->key_id);
+		flow_dissector->offset[key->key_id] = key->offset;
+	}
+
+	/* Ensure that the dissector always includes basic key. That way
+	 * we are able to avoid handling lack of it in fast path.
+	 */
+	BUG_ON(!skb_flow_dissector_uses_key(flow_dissector,
+					    FLOW_DISSECTOR_KEY_BASIC));
+}
+EXPORT_SYMBOL(skb_flow_dissector_init);
+
 /* copy saddr & daddr, possibly using 64bit load/store
  * Equivalent to :	flow->src = iph->saddr;
  *			flow->dst = iph->daddr;
-- 
2.4.0

[patch net-next v3 10/15] flow_dissect: use programable dissector in skb_flow_dissect and friends

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 drivers/net/bonding/bond_main.c                |  18 +--
 drivers/net/ethernet/cisco/enic/enic_clsf.c    |  27 ++--
 drivers/net/ethernet/cisco/enic/enic_ethtool.c |  10 +-
 drivers/net/hyperv/netvsc_drv.c                |   8 +-
 include/linux/skbuff.h                         |   4 +-
 include/net/flow_dissector.h                   |  63 ++++----
 include/net/ip.h                               |   8 +-
 include/net/ipv6.h                             |   8 +-
 net/core/flow_dissector.c                      | 199 +++++++++++++++++--------
 net/ethernet/eth.c                             |   6 +-
 net/sched/cls_flow.c                           |  20 +--
 net/sched/sch_choke.c                          |   4 +-
 12 files changed, 229 insertions(+), 146 deletions(-)

diff --git a/drivers/net/bonding/bond_main.c b/drivers/net/bonding/bond_main.c
index eed2f86..f7d2e39 100644
--- a/drivers/net/bonding/bond_main.c
+++ b/drivers/net/bonding/bond_main.c
@@ -3054,16 +3054,16 @@ static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
 	int noff, proto = -1;
 
 	if (bond->params.xmit_policy > BOND_XMIT_POLICY_LAYER23)
-		return skb_flow_dissect(skb, fk);
+		return skb_flow_dissect_flow_keys(skb, fk);
 
-	fk->ports = 0;
+	fk->ports.ports = 0;
 	noff = skb_network_offset(skb);
 	if (skb->protocol == htons(ETH_P_IP)) {
 		if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph))))
 			return false;
 		iph = ip_hdr(skb);
-		fk->src = iph->saddr;
-		fk->dst = iph->daddr;
+		fk->addrs.src = iph->saddr;
+		fk->addrs.dst = iph->daddr;
 		noff += iph->ihl << 2;
 		if (!ip_is_fragment(iph))
 			proto = iph->protocol;
@@ -3071,15 +3071,15 @@ static bool bond_flow_dissect(struct bonding *bond, struct sk_buff *skb,
 		if (unlikely(!pskb_may_pull(skb, noff + sizeof(*iph6))))
 			return false;
 		iph6 = ipv6_hdr(skb);
-		fk->src = (__force __be32)ipv6_addr_hash(&iph6->saddr);
-		fk->dst = (__force __be32)ipv6_addr_hash(&iph6->daddr);
+		fk->addrs.src = (__force __be32)ipv6_addr_hash(&iph6->saddr);
+		fk->addrs.dst = (__force __be32)ipv6_addr_hash(&iph6->daddr);
 		noff += sizeof(*iph6);
 		proto = iph6->nexthdr;
 	} else {
 		return false;
 	}
 	if (bond->params.xmit_policy == BOND_XMIT_POLICY_LAYER34 && proto >= 0)
-		fk->ports = skb_flow_get_ports(skb, noff, proto);
+		fk->ports.ports = skb_flow_get_ports(skb, noff, proto);
 
 	return true;
 }
@@ -3105,8 +3105,8 @@ u32 bond_xmit_hash(struct bonding *bond, struct sk_buff *skb)
 	    bond->params.xmit_policy == BOND_XMIT_POLICY_ENCAP23)
 		hash = bond_eth_hash(skb);
 	else
-		hash = (__force u32)flow.ports;
-	hash ^= (__force u32)flow.dst ^ (__force u32)flow.src;
+		hash = (__force u32)flow.ports.ports;
+	hash ^= (__force u32)flow.addrs.dst ^ (__force u32)flow.addrs.src;
 	hash ^= (hash >> 16);
 	hash ^= (hash >> 8);
 
diff --git a/drivers/net/ethernet/cisco/enic/enic_clsf.c b/drivers/net/ethernet/cisco/enic/enic_clsf.c
index 380f049..d3d25c7 100644
--- a/drivers/net/ethernet/cisco/enic/enic_clsf.c
+++ b/drivers/net/ethernet/cisco/enic/enic_clsf.c
@@ -22,7 +22,7 @@ int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq)
 	int res;
 	struct filter data;
 
-	switch (keys->ip_proto) {
+	switch (keys->basic.ip_proto) {
 	case IPPROTO_TCP:
 		data.u.ipv4.protocol = PROTO_TCP;
 		break;
@@ -33,10 +33,10 @@ int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq)
 		return -EPROTONOSUPPORT;
 	};
 	data.type = FILTER_IPV4_5TUPLE;
-	data.u.ipv4.src_addr = ntohl(keys->src);
-	data.u.ipv4.dst_addr = ntohl(keys->dst);
-	data.u.ipv4.src_port = ntohs(keys->port16[0]);
-	data.u.ipv4.dst_port = ntohs(keys->port16[1]);
+	data.u.ipv4.src_addr = ntohl(keys->addrs.src);
+	data.u.ipv4.dst_addr = ntohl(keys->addrs.dst);
+	data.u.ipv4.src_port = ntohs(keys->ports.port16[0]);
+	data.u.ipv4.dst_port = ntohs(keys->ports.port16[1]);
 	data.u.ipv4.flags = FILTER_FIELDS_IPV4_5TUPLE;
 
 	spin_lock_bh(&enic->devcmd_lock);
@@ -158,11 +158,11 @@ static struct enic_rfs_fltr_node *htbl_key_search(struct hlist_head *h,
 	struct enic_rfs_fltr_node *tpos;
 
 	hlist_for_each_entry(tpos, h, node)
-		if (tpos->keys.src == k->src &&
-		    tpos->keys.dst == k->dst &&
-		    tpos->keys.ports == k->ports &&
-		    tpos->keys.ip_proto == k->ip_proto &&
-		    tpos->keys.n_proto == k->n_proto)
+		if (tpos->keys.addrs.src == k->addrs.src &&
+		    tpos->keys.addrs.dst == k->addrs.dst &&
+		    tpos->keys.ports.ports == k->ports.ports &&
+		    tpos->keys.basic.ip_proto == k->basic.ip_proto &&
+		    tpos->keys.basic.n_proto == k->basic.n_proto)
 			return tpos;
 	return NULL;
 }
@@ -177,9 +177,10 @@ int enic_rx_flow_steer(struct net_device *dev, const struct sk_buff *skb,
 	int res, i;
 
 	enic = netdev_priv(dev);
-	res = skb_flow_dissect(skb, &keys);
-	if (!res || keys.n_proto != htons(ETH_P_IP) ||
-	    (keys.ip_proto != IPPROTO_TCP && keys.ip_proto != IPPROTO_UDP))
+	res = skb_flow_dissect_flow_keys(skb, &keys);
+	if (!res || keys.basic.n_proto != htons(ETH_P_IP) ||
+	    (keys.basic.ip_proto != IPPROTO_TCP &&
+	     keys.basic.ip_proto != IPPROTO_UDP))
 		return -EPROTONOSUPPORT;
 
 	tbl_idx = skb_get_hash_raw(skb) & ENIC_RFS_FLW_MASK;
diff --git a/drivers/net/ethernet/cisco/enic/enic_ethtool.c b/drivers/net/ethernet/cisco/enic/enic_ethtool.c
index 28d9ca6..7588f8d 100644
--- a/drivers/net/ethernet/cisco/enic/enic_ethtool.c
+++ b/drivers/net/ethernet/cisco/enic/enic_ethtool.c
@@ -334,7 +334,7 @@ static int enic_grxclsrule(struct enic *enic, struct ethtool_rxnfc *cmd)
 	n = htbl_fltr_search(enic, (u16)fsp->location);
 	if (!n)
 		return -EINVAL;
-	switch (n->keys.ip_proto) {
+	switch (n->keys.basic.ip_proto) {
 	case IPPROTO_TCP:
 		fsp->flow_type = TCP_V4_FLOW;
 		break;
@@ -346,16 +346,16 @@ static int enic_grxclsrule(struct enic *enic, struct ethtool_rxnfc *cmd)
 		break;
 	}
 
-	fsp->h_u.tcp_ip4_spec.ip4src = n->keys.src;
+	fsp->h_u.tcp_ip4_spec.ip4src = n->keys.addrs.src;
 	fsp->m_u.tcp_ip4_spec.ip4src = (__u32)~0;
 
-	fsp->h_u.tcp_ip4_spec.ip4dst = n->keys.dst;
+	fsp->h_u.tcp_ip4_spec.ip4dst = n->keys.addrs.dst;
 	fsp->m_u.tcp_ip4_spec.ip4dst = (__u32)~0;
 
-	fsp->h_u.tcp_ip4_spec.psrc = n->keys.port16[0];
+	fsp->h_u.tcp_ip4_spec.psrc = n->keys.ports.port16[0];
 	fsp->m_u.tcp_ip4_spec.psrc = (__u16)~0;
 
-	fsp->h_u.tcp_ip4_spec.pdst = n->keys.port16[1];
+	fsp->h_u.tcp_ip4_spec.pdst = n->keys.ports.port16[1];
 	fsp->m_u.tcp_ip4_spec.pdst = (__u16)~0;
 
 	fsp->ring_cookie = n->rq_id;
diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 5993c7e..8e5fe88 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -196,12 +196,12 @@ static bool netvsc_set_hash(u32 *hash, struct sk_buff *skb)
 	struct flow_keys flow;
 	int data_len;
 
-	if (!skb_flow_dissect(skb, &flow) ||
-	    !(flow.n_proto == htons(ETH_P_IP) ||
-	      flow.n_proto == htons(ETH_P_IPV6)))
+	if (!skb_flow_dissect_flow_keys(skb, &flow) ||
+	    !(flow.basic.n_proto == htons(ETH_P_IP) ||
+	      flow.basic.n_proto == htons(ETH_P_IPV6)))
 		return false;
 
-	if (flow.ip_proto == IPPROTO_TCP)
+	if (flow.basic.ip_proto == IPPROTO_TCP)
 		data_len = 12;
 	else
 		data_len = 8;
diff --git a/include/linux/skbuff.h b/include/linux/skbuff.h
index bc207d4..0661db7 100644
--- a/include/linux/skbuff.h
+++ b/include/linux/skbuff.h
@@ -1935,8 +1935,8 @@ static inline void skb_probe_transport_header(struct sk_buff *skb,
 
 	if (skb_transport_header_was_set(skb))
 		return;
-	else if (skb_flow_dissect(skb, &keys))
-		skb_set_transport_header(skb, keys.thoff);
+	else if (skb_flow_dissect_flow_keys(skb, &keys))
+		skb_set_transport_header(skb, keys.basic.thoff);
 	else
 		skb_set_transport_header(skb, offset_hint);
 }
diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 20239e8..0c8d406 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -59,42 +59,47 @@ struct flow_dissector {
 	unsigned short int offset[FLOW_DISSECTOR_KEY_MAX];
 };
 
-/* struct flow_keys:
- *	@src: source ip address in case of IPv4
- *	      For IPv6 it contains 32bit hash of src address
- *	@dst: destination ip address in case of IPv4
- *	      For IPv6 it contains 32bit hash of dst address
- *	@ports: port numbers of Transport header
- *		port16[0]: src port number
- *		port16[1]: dst port number
- *	@thoff: Transport header offset
- *	@n_proto: Network header protocol (eg. IPv4/IPv6)
- *	@ip_proto: Transport header protocol (eg. TCP/UDP)
- * All the members, except thoff, are in network byte order.
- */
-struct flow_keys {
-	/* (src,dst) must be grouped, in the same way than in IP header */
-	__be32 src;
-	__be32 dst;
-	union {
-		__be32 ports;
-		__be16 port16[2];
-	};
-	u16	thoff;
-	__be16	n_proto;
-	u8	ip_proto;
-};
-
 void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
 			     const struct flow_dissector_key *key,
 			     unsigned int key_count);
-bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
+
+bool __skb_flow_dissect(const struct sk_buff *skb,
+			struct flow_dissector *flow_dissector,
+			void *target_container,
 			void *data, __be16 proto, int nhoff, int hlen);
 
 static inline bool skb_flow_dissect(const struct sk_buff *skb,
-				    struct flow_keys *flow)
+				    struct flow_dissector *flow_dissector,
+				    void *target_container)
+{
+	return __skb_flow_dissect(skb, flow_dissector, target_container,
+				  NULL, 0, 0, 0);
+}
+
+struct flow_keys {
+	struct flow_dissector_key_addrs addrs;
+	struct flow_dissector_key_ports ports;
+	struct flow_dissector_key_basic basic;
+};
+
+extern struct flow_dissector flow_keys_dissector;
+extern struct flow_dissector flow_keys_buf_dissector;
+
+static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
+					      struct flow_keys *flow)
+{
+	memset(flow, 0, sizeof(*flow));
+	return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
+				  NULL, 0, 0, 0);
+}
+
+static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
+						  void *data, __be16 proto,
+						  int nhoff, int hlen)
 {
-	return __skb_flow_dissect(skb, flow, NULL, 0, 0, 0);
+	memset(flow, 0, sizeof(*flow));
+	return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
+				  data, proto, nhoff, hlen);
 }
 
 __be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
diff --git a/include/net/ip.h b/include/net/ip.h
index 562eb65..b0443d4 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -360,10 +360,10 @@ static inline void inet_set_txhash(struct sock *sk)
 	struct inet_sock *inet = inet_sk(sk);
 	struct flow_keys keys;
 
-	keys.src = inet->inet_saddr;
-	keys.dst = inet->inet_daddr;
-	keys.port16[0] = inet->inet_sport;
-	keys.port16[1] = inet->inet_dport;
+	keys.addrs.src = inet->inet_saddr;
+	keys.addrs.dst = inet->inet_daddr;
+	keys.ports.port16[0] = inet->inet_sport;
+	keys.ports.port16[1] = inet->inet_dport;
 
 	sk->sk_txhash = flow_hash_from_keys(&keys);
 }
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9932b86..9eed976 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -698,10 +698,10 @@ static inline void ip6_set_txhash(struct sock *sk)
 	struct ipv6_pinfo *np = inet6_sk(sk);
 	struct flow_keys keys;
 
-	keys.src = (__force __be32)ipv6_addr_hash(&np->saddr);
-	keys.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr);
-	keys.port16[0] = inet->inet_sport;
-	keys.port16[1] = inet->inet_dport;
+	keys.addrs.src = (__force __be32)ipv6_addr_hash(&np->saddr);
+	keys.addrs.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr);
+	keys.ports.port16[0] = inet->inet_sport;
+	keys.ports.port16[1] = inet->inet_dport;
 
 	sk->sk_txhash = flow_hash_from_keys(&keys);
 }
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 6883e22..6a49aca 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -13,6 +13,7 @@
 #include <linux/if_tunnel.h>
 #include <linux/if_pppox.h>
 #include <linux/ppp_defs.h>
+#include <linux/stddef.h>
 #include <net/flow_dissector.h>
 #include <scsi/fc/fc_fcoe.h>
 
@@ -63,17 +64,6 @@ void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
 }
 EXPORT_SYMBOL(skb_flow_dissector_init);
 
-/* copy saddr & daddr, possibly using 64bit load/store
- * Equivalent to :	flow->src = iph->saddr;
- *			flow->dst = iph->daddr;
- */
-static void iph_to_flow_copy_addrs(struct flow_keys *flow, const struct iphdr *iph)
-{
-	BUILD_BUG_ON(offsetof(typeof(*flow), dst) !=
-		     offsetof(typeof(*flow), src) + sizeof(flow->src));
-	memcpy(&flow->src, &iph->saddr, sizeof(flow->src) + sizeof(flow->dst));
-}
-
 /**
  * __skb_flow_get_ports - extract the upper layer ports and return them
  * @skb: sk_buff to extract the ports from
@@ -111,17 +101,27 @@ EXPORT_SYMBOL(__skb_flow_get_ports);
 /**
  * __skb_flow_dissect - extract the flow_keys struct and return it
  * @skb: sk_buff to extract the flow from, can be NULL if the rest are specified
+ * @flow_dissector: list of keys to dissect
+ * @target_container: target structure to put dissected values into
  * @data: raw buffer pointer to the packet, if NULL use skb->data
  * @proto: protocol for which to get the flow, if @data is NULL use skb->protocol
  * @nhoff: network header offset, if @data is NULL use skb_network_offset(skb)
  * @hlen: packet header length, if @data is NULL use skb_headlen(skb)
  *
- * The function will try to retrieve the struct flow_keys from either the skbuff
- * or a raw buffer specified by the rest parameters
+ * The function will try to retrieve individual keys into target specified
+ * by flow_dissector from either the skbuff or a raw buffer specified by the
+ * rest parameters.
+ *
+ * Caller must take care of zeroing target container memory.
  */
-bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
+bool __skb_flow_dissect(const struct sk_buff *skb,
+			struct flow_dissector *flow_dissector,
+			void *target_container,
 			void *data, __be16 proto, int nhoff, int hlen)
 {
+	struct flow_dissector_key_basic *key_basic;
+	struct flow_dissector_key_addrs *key_addrs;
+	struct flow_dissector_key_ports *key_ports;
 	u8 ip_proto;
 
 	if (!data) {
@@ -131,7 +131,12 @@ bool __skb_flow_dissect(const struct sk_buff *skb, struct flow_keys *flow,
 		hlen = skb_headlen(skb);
 	}
 
-	memset(flow, 0, sizeof(*flow));
+	/* It is ensured by skb_flow_dissector_init() that basic key will
+	 * be always present.
+	 */
+	key_basic = skb_flow_dissector_target(flow_dissector,
+					      FLOW_DISSECTOR_KEY_BASIC,
+					      target_container);
 
 again:
 	switch (proto) {
@@ -148,14 +153,13 @@ ip:
 		if (ip_is_fragment(iph))
 			ip_proto = 0;
 
-		/* skip the address processing if skb is NULL.  The assumption
-		 * here is that if there is no skb we are not looking for flow
-		 * info but lengths and protocols.
-		 */
-		if (!skb)
+		if (!skb_flow_dissector_uses_key(flow_dissector,
+						 FLOW_DISSECTOR_KEY_IPV4_ADDRS))
 			break;
-
-		iph_to_flow_copy_addrs(flow, iph);
+		key_addrs = skb_flow_dissector_target(flow_dissector,
+						      FLOW_DISSECTOR_KEY_IPV4_ADDRS,
+						      target_container);
+		memcpy(key_addrs, &iph->saddr, sizeof(*key_addrs));
 		break;
 	}
 	case htons(ETH_P_IPV6): {
@@ -171,12 +175,15 @@ ipv6:
 		ip_proto = iph->nexthdr;
 		nhoff += sizeof(struct ipv6hdr);
 
-		/* see comment above in IPv4 section */
-		if (!skb)
+		if (!skb_flow_dissector_uses_key(flow_dissector,
+						 FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS))
 			break;
+		key_addrs = skb_flow_dissector_target(flow_dissector,
+						      FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS,
+						      target_container);
 
-		flow->src = (__force __be32)ipv6_addr_hash(&iph->saddr);
-		flow->dst = (__force __be32)ipv6_addr_hash(&iph->daddr);
+		key_addrs->src = (__force __be32)ipv6_addr_hash(&iph->saddr);
+		key_addrs->dst = (__force __be32)ipv6_addr_hash(&iph->daddr);
 
 		flow_label = ip6_flowlabel(iph);
 		if (flow_label) {
@@ -184,10 +191,18 @@ ipv6:
 			 * use that to represent the ports without any
 			 * further dissection.
 			 */
-			flow->n_proto = proto;
-			flow->ip_proto = ip_proto;
-			flow->ports = flow_label;
-			flow->thoff = (u16)nhoff;
+
+			key_basic->n_proto = proto;
+			key_basic->ip_proto = ip_proto;
+			key_basic->thoff = (u16)nhoff;
+
+			if (!skb_flow_dissector_uses_key(flow_dissector,
+							 FLOW_DISSECTOR_KEY_PORTS))
+				break;
+			key_ports = skb_flow_dissector_target(flow_dissector,
+							      FLOW_DISSECTOR_KEY_PORTS,
+							      target_container);
+			key_ports->ports = flow_label;
 
 			return true;
 		}
@@ -234,14 +249,22 @@ ipv6:
 		hdr = __skb_header_pointer(skb, nhoff, sizeof(_hdr), data, hlen, &_hdr);
 		if (!hdr)
 			return false;
-		flow->src = hdr->srcnode;
-		flow->dst = 0;
-		flow->n_proto = proto;
-		flow->thoff = (u16)nhoff;
+		key_basic->n_proto = proto;
+		key_basic->thoff = (u16)nhoff;
+
+		if (skb_flow_dissector_uses_key(flow_dissector,
+						FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS)) {
+			return true;
+			key_addrs = skb_flow_dissector_target(flow_dissector,
+							      FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS,
+							      target_container);
+			key_addrs->src = hdr->srcnode;
+			key_addrs->dst = 0;
+		}
 		return true;
 	}
 	case htons(ETH_P_FCOE):
-		flow->thoff = (u16)(nhoff + FCOE_HEADER_LEN);
+		key_basic->thoff = (u16)(nhoff + FCOE_HEADER_LEN);
 		/* fall through */
 	default:
 		return false;
@@ -296,14 +319,24 @@ ipv6:
 		break;
 	}
 
-	flow->n_proto = proto;
-	flow->ip_proto = ip_proto;
-	flow->thoff = (u16) nhoff;
-
-	/* unless skb is set we don't need to record port info */
-	if (skb)
-		flow->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
-						   data, hlen);
+	/* It is ensured by skb_flow_dissector_init() that basic key will
+	 * be always present.
+	 */
+	key_basic = skb_flow_dissector_target(flow_dissector,
+					      FLOW_DISSECTOR_KEY_BASIC,
+					      target_container);
+	key_basic->n_proto = proto;
+	key_basic->ip_proto = ip_proto;
+	key_basic->thoff = (u16) nhoff;
+
+	if (skb_flow_dissector_uses_key(flow_dissector,
+					FLOW_DISSECTOR_KEY_PORTS)) {
+		key_ports = skb_flow_dissector_target(flow_dissector,
+						      FLOW_DISSECTOR_KEY_PORTS,
+						      target_container);
+		key_ports->ports = __skb_flow_get_ports(skb, nhoff, ip_proto,
+							data, hlen);
+	}
 
 	return true;
 }
@@ -325,16 +358,16 @@ static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
 	u32 hash;
 
 	/* get a consistent hash (same value on both flow directions) */
-	if (((__force u32)keys->dst < (__force u32)keys->src) ||
-	    (((__force u32)keys->dst == (__force u32)keys->src) &&
-	     ((__force u16)keys->port16[1] < (__force u16)keys->port16[0]))) {
-		swap(keys->dst, keys->src);
-		swap(keys->port16[0], keys->port16[1]);
+	if (((__force u32)keys->addrs.dst < (__force u32)keys->addrs.src) ||
+	    (((__force u32)keys->addrs.dst == (__force u32)keys->addrs.src) &&
+	     ((__force u16)keys->ports.port16[1] < (__force u16)keys->ports.port16[0]))) {
+		swap(keys->addrs.dst, keys->addrs.src);
+		swap(keys->ports.port16[0], keys->ports.port16[1]);
 	}
 
-	hash = __flow_hash_3words((__force u32)keys->dst,
-				  (__force u32)keys->src,
-				  (__force u32)keys->ports,
+	hash = __flow_hash_3words((__force u32)keys->addrs.dst,
+				  (__force u32)keys->addrs.src,
+				  (__force u32)keys->ports.ports,
 				  keyval);
 	if (!hash)
 		hash = 1;
@@ -352,7 +385,7 @@ EXPORT_SYMBOL(flow_hash_from_keys);
 static inline u32 ___skb_get_hash(const struct sk_buff *skb,
 				  struct flow_keys *keys, u32 keyval)
 {
-	if (!skb_flow_dissect(skb, keys))
+	if (!skb_flow_dissect_flow_keys(skb, keys))
 		return 0;
 
 	return __flow_hash_from_keys(keys, keyval);
@@ -377,11 +410,11 @@ void make_flow_keys_digest(struct flow_keys_digest *digest,
 
 	memset(digest, 0, sizeof(*digest));
 
-	data->n_proto = flow->n_proto;
-	data->ip_proto = flow->ip_proto;
-	data->ports = flow->ports;
-	data->src = flow->src;
-	data->dst = flow->dst;
+	data->n_proto = flow->basic.n_proto;
+	data->ip_proto = flow->basic.ip_proto;
+	data->ports = flow->ports.ports;
+	data->src = flow->addrs.src;
+	data->dst = flow->addrs.dst;
 }
 EXPORT_SYMBOL(make_flow_keys_digest);
 
@@ -404,7 +437,7 @@ void __skb_get_hash(struct sk_buff *skb)
 	hash = ___skb_get_hash(skb, &keys, hashrnd);
 	if (!hash)
 		return;
-	if (keys.ports)
+	if (keys.ports.ports)
 		skb->l4_hash = 1;
 	skb->sw_hash = 1;
 	skb->hash = hash;
@@ -422,9 +455,9 @@ EXPORT_SYMBOL(skb_get_hash_perturb);
 u32 __skb_get_poff(const struct sk_buff *skb, void *data,
 		   const struct flow_keys *keys, int hlen)
 {
-	u32 poff = keys->thoff;
+	u32 poff = keys->basic.thoff;
 
-	switch (keys->ip_proto) {
+	switch (keys->basic.ip_proto) {
 	case IPPROTO_TCP: {
 		/* access doff as u8 to avoid unaligned access */
 		const u8 *doff;
@@ -478,8 +511,52 @@ u32 skb_get_poff(const struct sk_buff *skb)
 {
 	struct flow_keys keys;
 
-	if (!skb_flow_dissect(skb, &keys))
+	if (!skb_flow_dissect_flow_keys(skb, &keys))
 		return 0;
 
 	return __skb_get_poff(skb, skb->data, &keys, skb_headlen(skb));
 }
+
+static const struct flow_dissector_key flow_keys_dissector_keys[] = {
+	{
+		.key_id = FLOW_DISSECTOR_KEY_BASIC,
+		.offset = offsetof(struct flow_keys, basic),
+	},
+	{
+		.key_id = FLOW_DISSECTOR_KEY_IPV4_ADDRS,
+		.offset = offsetof(struct flow_keys, addrs),
+	},
+	{
+		.key_id = FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS,
+		.offset = offsetof(struct flow_keys, addrs),
+	},
+	{
+		.key_id = FLOW_DISSECTOR_KEY_PORTS,
+		.offset = offsetof(struct flow_keys, ports),
+	},
+};
+
+static const struct flow_dissector_key flow_keys_buf_dissector_keys[] = {
+	{
+		.key_id = FLOW_DISSECTOR_KEY_BASIC,
+		.offset = offsetof(struct flow_keys, basic),
+	},
+};
+
+struct flow_dissector flow_keys_dissector __read_mostly;
+EXPORT_SYMBOL(flow_keys_dissector);
+
+struct flow_dissector flow_keys_buf_dissector __read_mostly;
+
+static int __init init_default_flow_dissectors(void)
+{
+	skb_flow_dissector_init(&flow_keys_dissector,
+				flow_keys_dissector_keys,
+				ARRAY_SIZE(flow_keys_dissector_keys));
+	skb_flow_dissector_init(&flow_keys_buf_dissector,
+				flow_keys_buf_dissector_keys,
+				ARRAY_SIZE(flow_keys_buf_dissector_keys));
+	return 0;
+}
+
+late_initcall_sync(init_default_flow_dissectors);
diff --git a/net/ethernet/eth.c b/net/ethernet/eth.c
index 9332a0a..c3325bd 100644
--- a/net/ethernet/eth.c
+++ b/net/ethernet/eth.c
@@ -131,9 +131,9 @@ u32 eth_get_headlen(void *data, unsigned int len)
 		return len;
 
 	/* parse any remaining L2/L3 headers, check for L4 */
-	if (!__skb_flow_dissect(NULL, &keys, data,
-				eth->h_proto, sizeof(*eth), len))
-		return max_t(u32, keys.thoff, sizeof(*eth));
+	if (!skb_flow_dissect_flow_keys_buf(&keys, data, eth->h_proto,
+					    sizeof(*eth), len))
+		return max_t(u32, keys.basic.thoff, sizeof(*eth));
 
 	/* parse for any L4 headers */
 	return min_t(u32, __skb_get_poff(NULL, data, &keys, len), len);
diff --git a/net/sched/cls_flow.c b/net/sched/cls_flow.c
index 4c5a922..4b3e3e3 100644
--- a/net/sched/cls_flow.c
+++ b/net/sched/cls_flow.c
@@ -68,35 +68,35 @@ static inline u32 addr_fold(void *addr)
 
 static u32 flow_get_src(const struct sk_buff *skb, const struct flow_keys *flow)
 {
-	if (flow->src)
-		return ntohl(flow->src);
+	if (flow->addrs.src)
+		return ntohl(flow->addrs.src);
 	return addr_fold(skb->sk);
 }
 
 static u32 flow_get_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 {
-	if (flow->dst)
-		return ntohl(flow->dst);
+	if (flow->addrs.dst)
+		return ntohl(flow->addrs.dst);
 	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 }
 
 static u32 flow_get_proto(const struct sk_buff *skb, const struct flow_keys *flow)
 {
-	return flow->ip_proto;
+	return flow->basic.ip_proto;
 }
 
 static u32 flow_get_proto_src(const struct sk_buff *skb, const struct flow_keys *flow)
 {
-	if (flow->ports)
-		return ntohs(flow->port16[0]);
+	if (flow->ports.ports)
+		return ntohs(flow->ports.port16[0]);
 
 	return addr_fold(skb->sk);
 }
 
 static u32 flow_get_proto_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 {
-	if (flow->ports)
-		return ntohs(flow->port16[1]);
+	if (flow->ports.ports)
+		return ntohs(flow->ports.port16[1]);
 
 	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 }
@@ -295,7 +295,7 @@ static int flow_classify(struct sk_buff *skb, const struct tcf_proto *tp,
 
 		keymask = f->keymask;
 		if (keymask & FLOW_KEYS_NEEDED)
-			skb_flow_dissect(skb, &flow_keys);
+			skb_flow_dissect_flow_keys(skb, &flow_keys);
 
 		for (n = 0; n < f->nkeys; n++) {
 			key = ffs(keymask) - 1;
diff --git a/net/sched/sch_choke.c b/net/sched/sch_choke.c
index 2624e99..93d5742 100644
--- a/net/sched/sch_choke.c
+++ b/net/sched/sch_choke.c
@@ -170,13 +170,13 @@ static bool choke_match_flow(struct sk_buff *skb1,
 
 	if (!choke_skb_cb(skb1)->keys_valid) {
 		choke_skb_cb(skb1)->keys_valid = 1;
-		skb_flow_dissect(skb1, &temp);
+		skb_flow_dissect_flow_keys(skb1, &temp);
 		make_flow_keys_digest(&choke_skb_cb(skb1)->keys, &temp);
 	}
 
 	if (!choke_skb_cb(skb2)->keys_valid) {
 		choke_skb_cb(skb2)->keys_valid = 1;
-		skb_flow_dissect(skb2, &temp);
+		skb_flow_dissect_flow_keys(skb2, &temp);
 		make_flow_keys_digest(&choke_skb_cb(skb2)->keys, &temp);
 	}
 
-- 
2.4.0

[patch net-next v3 11/15] flow_dissector: add missing header includes

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/net/flow_dissector.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 0c8d406..1add891 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -1,6 +1,9 @@
 #ifndef _NET_FLOW_DISSECTOR_H
 #define _NET_FLOW_DISSECTOR_H
 
+#include <linux/types.h>
+#include <linux/skbuff.h>
+
 /**
  * struct flow_dissector_key_basic:
  * @thoff: Transport header offset
-- 
2.4.0

[patch net-next v3 12/15] flow_dissector: introduce support for ipv6 addressses

[Jiri Pirko]

So far, only hashes made out of ipv6 addresses could be dissected. This
patch introduces support for dissection of full ipv6 addresses.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/net/flow_dissector.h | 13 +++++++++++++
 net/core/flow_dissector.c    | 29 +++++++++++++++++++++--------
 2 files changed, 34 insertions(+), 8 deletions(-)

diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 1add891..586b123 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -3,6 +3,7 @@
 
 #include <linux/types.h>
 #include <linux/skbuff.h>
+#include <linux/in6.h>
 
 /**
  * struct flow_dissector_key_basic:
@@ -42,11 +43,23 @@ struct flow_dissector_key_ports {
 	};
 };
 
+/**
+ * struct flow_dissector_key_ipv6_addrs:
+ * @src: source ip address
+ * @dst: destination ip address
+ */
+struct flow_dissector_key_ipv6_addrs {
+	/* (src,dst) must be grouped, in the same way than in IP header */
+	struct in6_addr src;
+	struct in6_addr dst;
+};
+
 enum flow_dissector_key_id {
 	FLOW_DISSECTOR_KEY_BASIC, /* struct flow_dissector_key_basic */
 	FLOW_DISSECTOR_KEY_IPV4_ADDRS, /* struct flow_dissector_key_addrs */
 	FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, /* struct flow_dissector_key_addrs */
 	FLOW_DISSECTOR_KEY_PORTS, /* struct flow_dissector_key_ports */
+	FLOW_DISSECTOR_KEY_IPV6_ADDRS, /* struct flow_dissector_key_ipv6_addrs */
 
 	FLOW_DISSECTOR_KEY_MAX,
 };
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 6a49aca..1b95d5c 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -175,16 +175,29 @@ ipv6:
 		ip_proto = iph->nexthdr;
 		nhoff += sizeof(struct ipv6hdr);
 
-		if (!skb_flow_dissector_uses_key(flow_dissector,
-						 FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS))
-			break;
-		key_addrs = skb_flow_dissector_target(flow_dissector,
-						      FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS,
-						      target_container);
+		if (skb_flow_dissector_uses_key(flow_dissector,
+						FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS)) {
+			key_addrs = skb_flow_dissector_target(flow_dissector,
+							      FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS,
+							      target_container);
 
-		key_addrs->src = (__force __be32)ipv6_addr_hash(&iph->saddr);
-		key_addrs->dst = (__force __be32)ipv6_addr_hash(&iph->daddr);
+			key_addrs->src = (__force __be32)ipv6_addr_hash(&iph->saddr);
+			key_addrs->dst = (__force __be32)ipv6_addr_hash(&iph->daddr);
+			goto flow_label;
+		}
+		if (skb_flow_dissector_uses_key(flow_dissector,
+						FLOW_DISSECTOR_KEY_IPV6_ADDRS)) {
+			struct flow_dissector_key_ipv6_addrs *key_ipv6_addrs;
 
+			key_ipv6_addrs = skb_flow_dissector_target(flow_dissector,
+								   FLOW_DISSECTOR_KEY_IPV6_ADDRS,
+								   target_container);
+
+			memcpy(key_ipv6_addrs, &iph->saddr, sizeof(*key_ipv6_addrs));
+			goto flow_label;
+		}
+		break;
+flow_label:
 		flow_label = ip6_flowlabel(iph);
 		if (flow_label) {
 			/* Awesome, IPv6 packet has a flow label so we can
-- 
2.4.0

[patch net-next v3 13/15] flow_dissector: introduce support for Ethernet addresses

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 include/net/flow_dissector.h | 13 +++++++++++++
 net/core/flow_dissector.c    | 12 ++++++++++++
 2 files changed, 25 insertions(+)

diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 586b123..5eac987 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -4,6 +4,7 @@
 #include <linux/types.h>
 #include <linux/skbuff.h>
 #include <linux/in6.h>
+#include <uapi/linux/if_ether.h>
 
 /**
  * struct flow_dissector_key_basic:
@@ -54,12 +55,24 @@ struct flow_dissector_key_ipv6_addrs {
 	struct in6_addr dst;
 };
 
+/**
+ * struct flow_dissector_key_eth_addrs:
+ * @src: source Ethernet address
+ * @dst: destination Ethernet address
+ */
+struct flow_dissector_key_eth_addrs {
+	/* (dst,src) must be grouped, in the same way than in ETH header */
+	unsigned char dst[ETH_ALEN];
+	unsigned char src[ETH_ALEN];
+};
+
 enum flow_dissector_key_id {
 	FLOW_DISSECTOR_KEY_BASIC, /* struct flow_dissector_key_basic */
 	FLOW_DISSECTOR_KEY_IPV4_ADDRS, /* struct flow_dissector_key_addrs */
 	FLOW_DISSECTOR_KEY_IPV6_HASH_ADDRS, /* struct flow_dissector_key_addrs */
 	FLOW_DISSECTOR_KEY_PORTS, /* struct flow_dissector_key_ports */
 	FLOW_DISSECTOR_KEY_IPV6_ADDRS, /* struct flow_dissector_key_ipv6_addrs */
+	FLOW_DISSECTOR_KEY_ETH_ADDRS, /* struct flow_dissector_key_eth_addrs */
 
 	FLOW_DISSECTOR_KEY_MAX,
 };
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 1b95d5c..7a0b391 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -14,6 +14,7 @@
 #include <linux/if_pppox.h>
 #include <linux/ppp_defs.h>
 #include <linux/stddef.h>
+#include <linux/if_ether.h>
 #include <net/flow_dissector.h>
 #include <scsi/fc/fc_fcoe.h>
 
@@ -138,6 +139,17 @@ bool __skb_flow_dissect(const struct sk_buff *skb,
 					      FLOW_DISSECTOR_KEY_BASIC,
 					      target_container);
 
+	if (skb_flow_dissector_uses_key(flow_dissector,
+					FLOW_DISSECTOR_KEY_ETH_ADDRS)) {
+		struct ethhdr *eth = eth_hdr(skb);
+		struct flow_dissector_key_eth_addrs *key_eth_addrs;
+
+		key_eth_addrs = skb_flow_dissector_target(flow_dissector,
+							  FLOW_DISSECTOR_KEY_ETH_ADDRS,
+							  target_container);
+		memcpy(key_eth_addrs, &eth->h_dest, sizeof(*key_eth_addrs));
+	}
+
 again:
 	switch (proto) {
 	case htons(ETH_P_IP): {
-- 
2.4.0

[patch net-next v3 14/15] flow_dissector: change port array into src,dst tuple

[Jiri Pirko]

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
---
 drivers/net/ethernet/cisco/enic/enic_clsf.c    | 4 ++--
 drivers/net/ethernet/cisco/enic/enic_ethtool.c | 4 ++--
 include/net/flow_dissector.h                   | 9 ++++++---
 include/net/ip.h                               | 4 ++--
 include/net/ipv6.h                             | 4 ++--
 net/core/flow_dissector.c                      | 4 ++--
 net/sched/cls_flow.c                           | 4 ++--
 7 files changed, 18 insertions(+), 15 deletions(-)

diff --git a/drivers/net/ethernet/cisco/enic/enic_clsf.c b/drivers/net/ethernet/cisco/enic/enic_clsf.c
index d3d25c7..6739ebc 100644
--- a/drivers/net/ethernet/cisco/enic/enic_clsf.c
+++ b/drivers/net/ethernet/cisco/enic/enic_clsf.c
@@ -35,8 +35,8 @@ int enic_addfltr_5t(struct enic *enic, struct flow_keys *keys, u16 rq)
 	data.type = FILTER_IPV4_5TUPLE;
 	data.u.ipv4.src_addr = ntohl(keys->addrs.src);
 	data.u.ipv4.dst_addr = ntohl(keys->addrs.dst);
-	data.u.ipv4.src_port = ntohs(keys->ports.port16[0]);
-	data.u.ipv4.dst_port = ntohs(keys->ports.port16[1]);
+	data.u.ipv4.src_port = ntohs(keys->ports.src);
+	data.u.ipv4.dst_port = ntohs(keys->ports.dst);
 	data.u.ipv4.flags = FILTER_FIELDS_IPV4_5TUPLE;
 
 	spin_lock_bh(&enic->devcmd_lock);
diff --git a/drivers/net/ethernet/cisco/enic/enic_ethtool.c b/drivers/net/ethernet/cisco/enic/enic_ethtool.c
index 7588f8d..117c096 100644
--- a/drivers/net/ethernet/cisco/enic/enic_ethtool.c
+++ b/drivers/net/ethernet/cisco/enic/enic_ethtool.c
@@ -352,10 +352,10 @@ static int enic_grxclsrule(struct enic *enic, struct ethtool_rxnfc *cmd)
 	fsp->h_u.tcp_ip4_spec.ip4dst = n->keys.addrs.dst;
 	fsp->m_u.tcp_ip4_spec.ip4dst = (__u32)~0;
 
-	fsp->h_u.tcp_ip4_spec.psrc = n->keys.ports.port16[0];
+	fsp->h_u.tcp_ip4_spec.psrc = n->keys.ports.src;
 	fsp->m_u.tcp_ip4_spec.psrc = (__u16)~0;
 
-	fsp->h_u.tcp_ip4_spec.pdst = n->keys.ports.port16[1];
+	fsp->h_u.tcp_ip4_spec.pdst = n->keys.ports.dst;
 	fsp->m_u.tcp_ip4_spec.pdst = (__u16)~0;
 
 	fsp->ring_cookie = n->rq_id;
diff --git a/include/net/flow_dissector.h b/include/net/flow_dissector.h
index 5eac987..bac9c14 100644
--- a/include/net/flow_dissector.h
+++ b/include/net/flow_dissector.h
@@ -34,13 +34,16 @@ struct flow_dissector_key_addrs {
 /**
  * flow_dissector_key_tp_ports:
  *	@ports: port numbers of Transport header
- *		port16[0]: src port number
- *		port16[1]: dst port number
+ *		src: source port number
+ *		dst: destination port number
  */
 struct flow_dissector_key_ports {
 	union {
 		__be32 ports;
-		__be16 port16[2];
+		struct {
+			__be16 src;
+			__be16 dst;
+		};
 	};
 };
 
diff --git a/include/net/ip.h b/include/net/ip.h
index b0443d4..0ed6d76 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -362,8 +362,8 @@ static inline void inet_set_txhash(struct sock *sk)
 
 	keys.addrs.src = inet->inet_saddr;
 	keys.addrs.dst = inet->inet_daddr;
-	keys.ports.port16[0] = inet->inet_sport;
-	keys.ports.port16[1] = inet->inet_dport;
+	keys.ports.src = inet->inet_sport;
+	keys.ports.dst = inet->inet_dport;
 
 	sk->sk_txhash = flow_hash_from_keys(&keys);
 }
diff --git a/include/net/ipv6.h b/include/net/ipv6.h
index 9eed976..aab8190 100644
--- a/include/net/ipv6.h
+++ b/include/net/ipv6.h
@@ -700,8 +700,8 @@ static inline void ip6_set_txhash(struct sock *sk)
 
 	keys.addrs.src = (__force __be32)ipv6_addr_hash(&np->saddr);
 	keys.addrs.dst = (__force __be32)ipv6_addr_hash(&sk->sk_v6_daddr);
-	keys.ports.port16[0] = inet->inet_sport;
-	keys.ports.port16[1] = inet->inet_dport;
+	keys.ports.src = inet->inet_sport;
+	keys.ports.dst = inet->inet_dport;
 
 	sk->sk_txhash = flow_hash_from_keys(&keys);
 }
diff --git a/net/core/flow_dissector.c b/net/core/flow_dissector.c
index 7a0b391..204d09c 100644
--- a/net/core/flow_dissector.c
+++ b/net/core/flow_dissector.c
@@ -385,9 +385,9 @@ static inline u32 __flow_hash_from_keys(struct flow_keys *keys, u32 keyval)
 	/* get a consistent hash (same value on both flow directions) */
 	if (((__force u32)keys->addrs.dst < (__force u32)keys->addrs.src) ||
 	    (((__force u32)keys->addrs.dst == (__force u32)keys->addrs.src) &&
-	     ((__force u16)keys->ports.port16[1] < (__force u16)keys->ports.port16[0]))) {
+	     ((__force u16)keys->ports.dst < (__force u16)keys->ports.src))) {
 		swap(keys->addrs.dst, keys->addrs.src);
-		swap(keys->ports.port16[0], keys->ports.port16[1]);
+		swap(keys->ports.src, keys->ports.dst);
 	}
 
 	hash = __flow_hash_3words((__force u32)keys->addrs.dst,
diff --git a/net/sched/cls_flow.c b/net/sched/cls_flow.c
index 4b3e3e3..b435992 100644
--- a/net/sched/cls_flow.c
+++ b/net/sched/cls_flow.c
@@ -88,7 +88,7 @@ static u32 flow_get_proto(const struct sk_buff *skb, const struct flow_keys *flo
 static u32 flow_get_proto_src(const struct sk_buff *skb, const struct flow_keys *flow)
 {
 	if (flow->ports.ports)
-		return ntohs(flow->ports.port16[0]);
+		return ntohs(flow->ports.src);
 
 	return addr_fold(skb->sk);
 }
@@ -96,7 +96,7 @@ static u32 flow_get_proto_src(const struct sk_buff *skb, const struct flow_keys
 static u32 flow_get_proto_dst(const struct sk_buff *skb, const struct flow_keys *flow)
 {
 	if (flow->ports.ports)
-		return ntohs(flow->ports.port16[1]);
+		return ntohs(flow->ports.dst);
 
 	return addr_fold(skb_dst(skb)) ^ (__force u16) tc_skb_protocol(skb);
 }
-- 
2.4.0

[patch net-next v3 15/15] tc: introduce Flower classifier

[Jiri Pirko]

This patch introduces a flow-based filter. So far, the very essential
packet fields are supported.

This patch is only the first step. There is a lot of potential performance
improvements possible to implement. Also a lot of features are missing
now. They will be addressed in follow-up patches.

Signed-off-by: Jiri Pirko <jiri@resnulli.us>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
---
v2->v1:
- removed TCA_FLOWER_POLICE suggested by Jamal
---
 include/uapi/linux/pkt_cls.h |  30 ++
 net/sched/Kconfig            |  10 +
 net/sched/Makefile           |   1 +
 net/sched/cls_flower.c       | 688 +++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 729 insertions(+)
 create mode 100644 net/sched/cls_flower.c

diff --git a/include/uapi/linux/pkt_cls.h b/include/uapi/linux/pkt_cls.h
index 596ffa0..c3d0c24 100644
--- a/include/uapi/linux/pkt_cls.h
+++ b/include/uapi/linux/pkt_cls.h
@@ -409,6 +409,36 @@ enum {
 
 #define TCA_BPF_MAX (__TCA_BPF_MAX - 1)
 
+/* Flower classifier */
+
+enum {
+	TCA_FLOWER_UNSPEC,
+	TCA_FLOWER_CLASSID,
+	TCA_FLOWER_INDEV,
+	TCA_FLOWER_ACT,
+	TCA_FLOWER_KEY_ETH_DST,		/* ETH_ALEN */
+	TCA_FLOWER_KEY_ETH_DST_MASK,	/* ETH_ALEN */
+	TCA_FLOWER_KEY_ETH_SRC,		/* ETH_ALEN */
+	TCA_FLOWER_KEY_ETH_SRC_MASK,	/* ETH_ALEN */
+	TCA_FLOWER_KEY_ETH_TYPE,	/* be16 */
+	TCA_FLOWER_KEY_IP_PROTO,	/* u8 */
+	TCA_FLOWER_KEY_IPV4_SRC,	/* be32 */
+	TCA_FLOWER_KEY_IPV4_SRC_MASK,	/* be32 */
+	TCA_FLOWER_KEY_IPV4_DST,	/* be32 */
+	TCA_FLOWER_KEY_IPV4_DST_MASK,	/* be32 */
+	TCA_FLOWER_KEY_IPV6_SRC,	/* struct in6_addr */
+	TCA_FLOWER_KEY_IPV6_SRC_MASK,	/* struct in6_addr */
+	TCA_FLOWER_KEY_IPV6_DST,	/* struct in6_addr */
+	TCA_FLOWER_KEY_IPV6_DST_MASK,	/* struct in6_addr */
+	TCA_FLOWER_KEY_TCP_SRC,		/* be16 */
+	TCA_FLOWER_KEY_TCP_DST,		/* be16 */
+	TCA_FLOWER_KEY_UDP_SRC,		/* be16 */
+	TCA_FLOWER_KEY_UDP_DST,		/* be16 */
+	__TCA_FLOWER_MAX,
+};
+
+#define TCA_FLOWER_MAX (__TCA_FLOWER_MAX - 1)
+
 /* Extended Matches */
 
 struct tcf_ematch_tree_hdr {
diff --git a/net/sched/Kconfig b/net/sched/Kconfig
index 2274e72..5fd1c2f 100644
--- a/net/sched/Kconfig
+++ b/net/sched/Kconfig
@@ -477,6 +477,16 @@ config NET_CLS_BPF
 	  To compile this code as a module, choose M here: the module will
 	  be called cls_bpf.
 
+config NET_CLS_FLOWER
+	tristate "Flower classifier"
+	select NET_CLS
+	---help---
+	  If you say Y here, you will be able to classify packets based on
+	  a configurable combination of packet keys and masks.
+
+	  To compile this code as a module, choose M here: the module will
+	  be called cls_flower.
+
 config NET_EMATCH
 	bool "Extended Matches"
 	select NET_CLS
diff --git a/net/sched/Makefile b/net/sched/Makefile
index 7ca7f4c..690c1689 100644
--- a/net/sched/Makefile
+++ b/net/sched/Makefile
@@ -56,6 +56,7 @@ obj-$(CONFIG_NET_CLS_BASIC)	+= cls_basic.o
 obj-$(CONFIG_NET_CLS_FLOW)	+= cls_flow.o
 obj-$(CONFIG_NET_CLS_CGROUP)	+= cls_cgroup.o
 obj-$(CONFIG_NET_CLS_BPF)	+= cls_bpf.o
+obj-$(CONFIG_NET_CLS_FLOWER)	+= cls_flower.o
 obj-$(CONFIG_NET_EMATCH)	+= ematch.o
 obj-$(CONFIG_NET_EMATCH_CMP)	+= em_cmp.o
 obj-$(CONFIG_NET_EMATCH_NBYTE)	+= em_nbyte.o
diff --git a/net/sched/cls_flower.c b/net/sched/cls_flower.c
new file mode 100644
index 0000000..9bc654c
--- /dev/null
+++ b/net/sched/cls_flower.c
@@ -0,0 +1,688 @@
+/*
+ * net/sched/cls_flower.c		Flower classifier
+ *
+ * Copyright (c) 2015 Jiri Pirko <jiri@resnulli.us>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License as published by
+ * the Free Software Foundation; either version 2 of the License, or
+ * (at your option) any later version.
+ */
+
+#include <linux/kernel.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/rhashtable.h>
+
+#include <linux/if_ether.h>
+#include <linux/in6.h>
+#include <linux/ip.h>
+
+#include <net/sch_generic.h>
+#include <net/pkt_cls.h>
+#include <net/ip.h>
+#include <net/flow_dissector.h>
+
+struct fl_flow_key {
+	int	indev_ifindex;
+	struct flow_dissector_key_basic basic;
+	struct flow_dissector_key_eth_addrs eth;
+	union {
+		struct flow_dissector_key_addrs ipv4;
+		struct flow_dissector_key_ipv6_addrs ipv6;
+	};
+	struct flow_dissector_key_ports tp;
+} __aligned(BITS_PER_LONG / 8); /* Ensure that we can do comparisons as longs. */
+
+struct fl_flow_mask_range {
+	unsigned short int start;
+	unsigned short int end;
+};
+
+struct fl_flow_mask {
+	struct fl_flow_key key;
+	struct fl_flow_mask_range range;
+	struct rcu_head	rcu;
+};
+
+struct cls_fl_head {
+	struct rhashtable ht;
+	struct fl_flow_mask mask;
+	struct flow_dissector dissector;
+	u32 hgen;
+	bool mask_assigned;
+	struct list_head filters;
+	struct rhashtable_params ht_params;
+	struct rcu_head rcu;
+};
+
+struct cls_fl_filter {
+	struct rhash_head ht_node;
+	struct fl_flow_key mkey;
+	struct tcf_exts exts;
+	struct tcf_result res;
+	struct fl_flow_key key;
+	struct list_head list;
+	u32 handle;
+	struct rcu_head	rcu;
+};
+
+static unsigned short int fl_mask_range(const struct fl_flow_mask *mask)
+{
+	return mask->range.end - mask->range.start;
+}
+
+static void fl_mask_update_range(struct fl_flow_mask *mask)
+{
+	const u8 *bytes = (const u8 *) &mask->key;
+	size_t size = sizeof(mask->key);
+	size_t i, first = 0, last = size - 1;
+
+	for (i = 0; i < sizeof(mask->key); i++) {
+		if (bytes[i]) {
+			if (!first && i)
+				first = i;
+			last = i;
+		}
+	}
+	mask->range.start = rounddown(first, sizeof(long));
+	mask->range.end = roundup(last + 1, sizeof(long));
+}
+
+static void *fl_key_get_start(struct fl_flow_key *key,
+			      const struct fl_flow_mask *mask)
+{
+	return (u8 *) key + mask->range.start;
+}
+
+static void fl_set_masked_key(struct fl_flow_key *mkey, struct fl_flow_key *key,
+			      struct fl_flow_mask *mask)
+{
+	const long *lkey = fl_key_get_start(key, mask);
+	const long *lmask = fl_key_get_start(&mask->key, mask);
+	long *lmkey = fl_key_get_start(mkey, mask);
+	int i;
+
+	for (i = 0; i < fl_mask_range(mask); i += sizeof(long))
+		*lmkey++ = *lkey++ & *lmask++;
+}
+
+static void fl_clear_masked_range(struct fl_flow_key *key,
+				  struct fl_flow_mask *mask)
+{
+	memset(fl_key_get_start(key, mask), 0, fl_mask_range(mask));
+}
+
+static int fl_classify(struct sk_buff *skb, const struct tcf_proto *tp,
+		       struct tcf_result *res)
+{
+	struct cls_fl_head *head = rcu_dereference_bh(tp->root);
+	struct cls_fl_filter *f;
+	struct fl_flow_key skb_key;
+	struct fl_flow_key skb_mkey;
+
+	fl_clear_masked_range(&skb_key, &head->mask);
+	skb_key.indev_ifindex = skb->skb_iif;
+	/* skb_flow_dissect() does not set n_proto in case an unknown protocol,
+	 * so do it rather here.
+	 */
+	skb_key.basic.n_proto = skb->protocol;
+	skb_flow_dissect(skb, &head->dissector, &skb_key);
+
+	fl_set_masked_key(&skb_mkey, &skb_key, &head->mask);
+
+	f = rhashtable_lookup_fast(&head->ht,
+				   fl_key_get_start(&skb_mkey, &head->mask),
+				   head->ht_params);
+	if (f) {
+		*res = f->res;
+		return tcf_exts_exec(skb, &f->exts, res);
+	}
+	return -1;
+}
+
+static int fl_init(struct tcf_proto *tp)
+{
+	struct cls_fl_head *head;
+
+	head = kzalloc(sizeof(*head), GFP_KERNEL);
+	if (!head)
+		return -ENOBUFS;
+
+	INIT_LIST_HEAD_RCU(&head->filters);
+	rcu_assign_pointer(tp->root, head);
+
+	return 0;
+}
+
+static void fl_destroy_filter(struct rcu_head *head)
+{
+	struct cls_fl_filter *f = container_of(head, struct cls_fl_filter, rcu);
+
+	tcf_exts_destroy(&f->exts);
+	kfree(f);
+}
+
+static bool fl_destroy(struct tcf_proto *tp, bool force)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *f, *next;
+
+	if (!force && !list_empty(&head->filters))
+		return false;
+
+	list_for_each_entry_safe(f, next, &head->filters, list) {
+		list_del_rcu(&f->list);
+		call_rcu(&f->rcu, fl_destroy_filter);
+	}
+	RCU_INIT_POINTER(tp->root, NULL);
+	if (head->mask_assigned)
+		rhashtable_destroy(&head->ht);
+	kfree_rcu(head, rcu);
+	return true;
+}
+
+static unsigned long fl_get(struct tcf_proto *tp, u32 handle)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *f;
+
+	list_for_each_entry(f, &head->filters, list)
+		if (f->handle == handle)
+			return (unsigned long) f;
+	return 0;
+}
+
+static const struct nla_policy fl_policy[TCA_FLOWER_MAX + 1] = {
+	[TCA_FLOWER_UNSPEC]		= { .type = NLA_UNSPEC },
+	[TCA_FLOWER_CLASSID]		= { .type = NLA_U32 },
+	[TCA_FLOWER_INDEV]		= { .type = NLA_STRING,
+					    .len = IFNAMSIZ },
+	[TCA_FLOWER_KEY_ETH_DST]	= { .len = ETH_ALEN },
+	[TCA_FLOWER_KEY_ETH_DST_MASK]	= { .len = ETH_ALEN },
+	[TCA_FLOWER_KEY_ETH_SRC]	= { .len = ETH_ALEN },
+	[TCA_FLOWER_KEY_ETH_SRC_MASK]	= { .len = ETH_ALEN },
+	[TCA_FLOWER_KEY_ETH_TYPE]	= { .type = NLA_U16 },
+	[TCA_FLOWER_KEY_IP_PROTO]	= { .type = NLA_U8 },
+	[TCA_FLOWER_KEY_IPV4_SRC]	= { .type = NLA_U32 },
+	[TCA_FLOWER_KEY_IPV4_SRC_MASK]	= { .type = NLA_U32 },
+	[TCA_FLOWER_KEY_IPV4_DST]	= { .type = NLA_U32 },
+	[TCA_FLOWER_KEY_IPV4_DST_MASK]	= { .type = NLA_U32 },
+	[TCA_FLOWER_KEY_IPV6_SRC]	= { .len = sizeof(struct in6_addr) },
+	[TCA_FLOWER_KEY_IPV6_SRC_MASK]	= { .len = sizeof(struct in6_addr) },
+	[TCA_FLOWER_KEY_IPV6_DST]	= { .len = sizeof(struct in6_addr) },
+	[TCA_FLOWER_KEY_IPV6_DST_MASK]	= { .len = sizeof(struct in6_addr) },
+	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
+	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
+	[TCA_FLOWER_KEY_TCP_SRC]	= { .type = NLA_U16 },
+	[TCA_FLOWER_KEY_TCP_DST]	= { .type = NLA_U16 },
+};
+
+static void fl_set_key_val(struct nlattr **tb,
+			   void *val, int val_type,
+			   void *mask, int mask_type, int len)
+{
+	if (!tb[val_type])
+		return;
+	memcpy(val, nla_data(tb[val_type]), len);
+	if (mask_type == TCA_FLOWER_UNSPEC || !tb[mask_type])
+		memset(mask, 0xff, len);
+	else
+		memcpy(mask, nla_data(tb[mask_type]), len);
+}
+
+static int fl_set_key(struct net *net, struct nlattr **tb,
+		      struct fl_flow_key *key, struct fl_flow_key *mask)
+{
+	int err;
+
+	if (tb[TCA_FLOWER_INDEV]) {
+		err = tcf_change_indev(net, tb[TCA_FLOWER_INDEV]);
+		if (err < 0)
+			return err;
+		key->indev_ifindex = err;
+		mask->indev_ifindex = 0xffffffff;
+	}
+
+	fl_set_key_val(tb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
+		       mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
+		       sizeof(key->eth.dst));
+	fl_set_key_val(tb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
+		       mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
+		       sizeof(key->eth.src));
+	fl_set_key_val(tb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
+		       &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
+		       sizeof(key->basic.n_proto));
+	if (key->basic.n_proto == htons(ETH_P_IP) ||
+	    key->basic.n_proto == htons(ETH_P_IPV6)) {
+		fl_set_key_val(tb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
+			       &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
+			       sizeof(key->basic.ip_proto));
+	}
+	if (key->basic.n_proto == htons(ETH_P_IP)) {
+		fl_set_key_val(tb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
+			       &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
+			       sizeof(key->ipv4.src));
+		fl_set_key_val(tb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
+			       &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
+			       sizeof(key->ipv4.dst));
+	} else if (key->basic.n_proto == htons(ETH_P_IPV6)) {
+		fl_set_key_val(tb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
+			       &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
+			       sizeof(key->ipv6.src));
+		fl_set_key_val(tb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
+			       &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
+			       sizeof(key->ipv6.dst));
+	}
+	if (key->basic.ip_proto == IPPROTO_TCP) {
+		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
+			       &mask->tp.src, TCA_FLOWER_UNSPEC,
+			       sizeof(key->tp.src));
+		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
+			       &mask->tp.dst, TCA_FLOWER_UNSPEC,
+			       sizeof(key->tp.dst));
+	} else if (key->basic.ip_proto == IPPROTO_UDP) {
+		fl_set_key_val(tb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
+			       &mask->tp.src, TCA_FLOWER_UNSPEC,
+			       sizeof(key->tp.src));
+		fl_set_key_val(tb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
+			       &mask->tp.dst, TCA_FLOWER_UNSPEC,
+			       sizeof(key->tp.dst));
+	}
+
+	return 0;
+}
+
+static bool fl_mask_eq(struct fl_flow_mask *mask1,
+		       struct fl_flow_mask *mask2)
+{
+	const long *lmask1 = fl_key_get_start(&mask1->key, mask1);
+	const long *lmask2 = fl_key_get_start(&mask2->key, mask2);
+
+	return !memcmp(&mask1->range, &mask2->range, sizeof(mask1->range)) &&
+	       !memcmp(lmask1, lmask2, fl_mask_range(mask1));
+}
+
+static const struct rhashtable_params fl_ht_params = {
+	.key_offset = offsetof(struct cls_fl_filter, mkey), /* base offset */
+	.head_offset = offsetof(struct cls_fl_filter, ht_node),
+	.automatic_shrinking = true,
+};
+
+static int fl_init_hashtable(struct cls_fl_head *head,
+			     struct fl_flow_mask *mask)
+{
+	head->ht_params = fl_ht_params;
+	head->ht_params.key_len = fl_mask_range(mask);
+	head->ht_params.key_offset += mask->range.start;
+
+	return rhashtable_init(&head->ht, &head->ht_params);
+}
+
+#define FL_KEY_MEMBER_OFFSET(member) offsetof(struct fl_flow_key, member)
+#define FL_KEY_MEMBER_SIZE(member) (sizeof(((struct fl_flow_key *) 0)->member))
+#define FL_KEY_MEMBER_END_OFFSET(member)					\
+	(FL_KEY_MEMBER_OFFSET(member) + FL_KEY_MEMBER_SIZE(member))
+
+#define FL_KEY_IN_RANGE(mask, member)						\
+        (FL_KEY_MEMBER_OFFSET(member) <= (mask)->range.end &&			\
+         FL_KEY_MEMBER_END_OFFSET(member) >= (mask)->range.start)
+
+#define FL_KEY_SET(keys, cnt, id, member)					\
+	do {									\
+		keys[cnt].key_id = id;						\
+		keys[cnt].offset = FL_KEY_MEMBER_OFFSET(member);		\
+		cnt++;								\
+	} while(0);
+
+#define FL_KEY_SET_IF_IN_RANGE(mask, keys, cnt, id, member)			\
+	do {									\
+		if (FL_KEY_IN_RANGE(mask, member))				\
+			FL_KEY_SET(keys, cnt, id, member);			\
+	} while(0);
+
+static void fl_init_dissector(struct cls_fl_head *head,
+			      struct fl_flow_mask *mask)
+{
+	struct flow_dissector_key keys[FLOW_DISSECTOR_KEY_MAX];
+	size_t cnt = 0;
+
+	FL_KEY_SET(keys, cnt, FLOW_DISSECTOR_KEY_BASIC, basic);
+	FL_KEY_SET_IF_IN_RANGE(mask, keys, cnt,
+			       FLOW_DISSECTOR_KEY_ETH_ADDRS, eth);
+	FL_KEY_SET_IF_IN_RANGE(mask, keys, cnt,
+			       FLOW_DISSECTOR_KEY_IPV4_ADDRS, ipv4);
+	FL_KEY_SET_IF_IN_RANGE(mask, keys, cnt,
+			       FLOW_DISSECTOR_KEY_IPV6_ADDRS, ipv6);
+	FL_KEY_SET_IF_IN_RANGE(mask, keys, cnt,
+			       FLOW_DISSECTOR_KEY_PORTS, tp);
+
+	skb_flow_dissector_init(&head->dissector, keys, cnt);
+}
+
+static int fl_check_assign_mask(struct cls_fl_head *head,
+				struct fl_flow_mask *mask)
+{
+	int err;
+
+	if (head->mask_assigned) {
+		if (!fl_mask_eq(&head->mask, mask))
+			return -EINVAL;
+		else
+			return 0;
+	}
+
+	/* Mask is not assigned yet. So assign it and init hashtable
+	 * according to that.
+	 */
+	err = fl_init_hashtable(head, mask);
+	if (err)
+		return err;
+	memcpy(&head->mask, mask, sizeof(head->mask));
+	head->mask_assigned = true;
+
+	fl_init_dissector(head, mask);
+
+	return 0;
+}
+
+static int fl_set_parms(struct net *net, struct tcf_proto *tp,
+			struct cls_fl_filter *f, struct fl_flow_mask *mask,
+			unsigned long base, struct nlattr **tb,
+			struct nlattr *est, bool ovr)
+{
+	struct tcf_exts e;
+	int err;
+
+	tcf_exts_init(&e, TCA_FLOWER_ACT, 0);
+	err = tcf_exts_validate(net, tp, tb, est, &e, ovr);
+	if (err < 0)
+		return err;
+
+	if (tb[TCA_FLOWER_CLASSID]) {
+		f->res.classid = nla_get_u32(tb[TCA_FLOWER_CLASSID]);
+		tcf_bind_filter(tp, &f->res, base);
+	}
+
+	err = fl_set_key(net, tb, &f->key, &mask->key);
+	if (err)
+		goto errout;
+
+	fl_mask_update_range(mask);
+	fl_set_masked_key(&f->mkey, &f->key, mask);
+
+	tcf_exts_change(tp, &f->exts, &e);
+
+	return 0;
+errout:
+	tcf_exts_destroy(&e);
+	return err;
+}
+
+static u32 fl_grab_new_handle(struct tcf_proto *tp,
+			      struct cls_fl_head *head)
+{
+	unsigned int i = 0x80000000;
+	u32 handle;
+
+	do {
+		if (++head->hgen == 0x7FFFFFFF)
+			head->hgen = 1;
+	} while (--i > 0 && fl_get(tp, head->hgen));
+
+	if (unlikely(i == 0)) {
+		pr_err("Insufficient number of handles\n");
+		handle = 0;
+	} else {
+		handle = head->hgen;
+	}
+
+	return handle;
+}
+
+static int fl_change(struct net *net, struct sk_buff *in_skb,
+		     struct tcf_proto *tp, unsigned long base,
+		     u32 handle, struct nlattr **tca,
+		     unsigned long *arg, bool ovr)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *fold = (struct cls_fl_filter *) *arg;
+	struct cls_fl_filter *fnew;
+	struct nlattr *tb[TCA_FLOWER_MAX + 1];
+	struct fl_flow_mask mask = {};
+	int err;
+
+	if (!tca[TCA_OPTIONS])
+		return -EINVAL;
+
+	err = nla_parse_nested(tb, TCA_FLOWER_MAX, tca[TCA_OPTIONS], fl_policy);
+	if (err < 0)
+		return err;
+
+	if (fold && handle && fold->handle != handle)
+		return -EINVAL;
+
+	fnew = kzalloc(sizeof(*fnew), GFP_KERNEL);
+	if (!fnew)
+		return -ENOBUFS;
+
+	tcf_exts_init(&fnew->exts, TCA_FLOWER_ACT, 0);
+
+	if (!handle) {
+		handle = fl_grab_new_handle(tp, head);
+		if (!handle) {
+			err = -EINVAL;
+			goto errout;
+		}
+	}
+	fnew->handle = handle;
+
+	err = fl_set_parms(net, tp, fnew, &mask, base, tb, tca[TCA_RATE], ovr);
+	if (err)
+		goto errout;
+
+	err = fl_check_assign_mask(head, &mask);
+	if (err)
+		goto errout;
+
+	err = rhashtable_insert_fast(&head->ht, &fnew->ht_node,
+				     head->ht_params);
+	if (err)
+		goto errout;
+	if (fold)
+		rhashtable_remove_fast(&head->ht, &fold->ht_node,
+				       head->ht_params);
+
+	*arg = (unsigned long) fnew;
+
+	if (fold) {
+		list_replace_rcu(&fnew->list, &fold->list);
+		tcf_unbind_filter(tp, &fold->res);
+		call_rcu(&fold->rcu, fl_destroy_filter);
+	} else {
+		list_add_tail_rcu(&fnew->list, &head->filters);
+	}
+
+	return 0;
+
+errout:
+	kfree(fnew);
+	return err;
+}
+
+static int fl_delete(struct tcf_proto *tp, unsigned long arg)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *f = (struct cls_fl_filter *) arg;
+
+	rhashtable_remove_fast(&head->ht, &f->ht_node,
+			       head->ht_params);
+	list_del_rcu(&f->list);
+	tcf_unbind_filter(tp, &f->res);
+	call_rcu(&f->rcu, fl_destroy_filter);
+	return 0;
+}
+
+static void fl_walk(struct tcf_proto *tp, struct tcf_walker *arg)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *f;
+
+	list_for_each_entry_rcu(f, &head->filters, list) {
+		if (arg->count < arg->skip)
+			goto skip;
+		if (arg->fn(tp, (unsigned long) f, arg) < 0) {
+			arg->stop = 1;
+			break;
+		}
+skip:
+		arg->count++;
+	}
+}
+
+static int fl_dump_key_val(struct sk_buff *skb,
+			   void *val, int val_type,
+			   void *mask, int mask_type, int len)
+{
+	int err;
+
+	if (!memchr_inv(mask, 0, len))
+		return 0;
+	err = nla_put(skb, val_type, len, val);
+	if (err)
+		return err;
+	if (mask_type != TCA_FLOWER_UNSPEC) {
+		err = nla_put(skb, mask_type, len, mask);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int fl_dump(struct net *net, struct tcf_proto *tp, unsigned long fh,
+		   struct sk_buff *skb, struct tcmsg *t)
+{
+	struct cls_fl_head *head = rtnl_dereference(tp->root);
+	struct cls_fl_filter *f = (struct cls_fl_filter *) fh;
+	struct nlattr *nest;
+	struct fl_flow_key *key, *mask;
+
+	if (!f)
+		return skb->len;
+
+	t->tcm_handle = f->handle;
+
+	nest = nla_nest_start(skb, TCA_OPTIONS);
+	if (!nest)
+		goto nla_put_failure;
+
+	if (f->res.classid &&
+	    nla_put_u32(skb, TCA_FLOWER_CLASSID, f->res.classid))
+		goto nla_put_failure;
+
+	key = &f->key;
+	mask = &head->mask.key;
+
+	if (mask->indev_ifindex) {
+		struct net_device *dev;
+
+		dev = __dev_get_by_index(net, key->indev_ifindex);
+		if (dev && nla_put_string(skb, TCA_FLOWER_INDEV, dev->name))
+			goto nla_put_failure;
+	}
+
+	if (fl_dump_key_val(skb, key->eth.dst, TCA_FLOWER_KEY_ETH_DST,
+			    mask->eth.dst, TCA_FLOWER_KEY_ETH_DST_MASK,
+			    sizeof(key->eth.dst)) ||
+	    fl_dump_key_val(skb, key->eth.src, TCA_FLOWER_KEY_ETH_SRC,
+			    mask->eth.src, TCA_FLOWER_KEY_ETH_SRC_MASK,
+			    sizeof(key->eth.src)) ||
+	    fl_dump_key_val(skb, &key->basic.n_proto, TCA_FLOWER_KEY_ETH_TYPE,
+			    &mask->basic.n_proto, TCA_FLOWER_UNSPEC,
+			    sizeof(key->basic.n_proto)))
+		goto nla_put_failure;
+	if ((key->basic.n_proto == htons(ETH_P_IP) ||
+	     key->basic.n_proto == htons(ETH_P_IPV6)) &&
+	    fl_dump_key_val(skb, &key->basic.ip_proto, TCA_FLOWER_KEY_IP_PROTO,
+			    &mask->basic.ip_proto, TCA_FLOWER_UNSPEC,
+			    sizeof(key->basic.ip_proto)))
+		goto nla_put_failure;
+
+	if (key->basic.n_proto == htons(ETH_P_IP) &&
+	    (fl_dump_key_val(skb, &key->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC,
+			     &mask->ipv4.src, TCA_FLOWER_KEY_IPV4_SRC_MASK,
+			     sizeof(key->ipv4.src)) ||
+	     fl_dump_key_val(skb, &key->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST,
+			     &mask->ipv4.dst, TCA_FLOWER_KEY_IPV4_DST_MASK,
+			     sizeof(key->ipv4.dst))))
+		goto nla_put_failure;
+	else if (key->basic.n_proto == htons(ETH_P_IPV6) &&
+		 (fl_dump_key_val(skb, &key->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC,
+				  &mask->ipv6.src, TCA_FLOWER_KEY_IPV6_SRC_MASK,
+				  sizeof(key->ipv6.src)) ||
+		  fl_dump_key_val(skb, &key->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST,
+				  &mask->ipv6.dst, TCA_FLOWER_KEY_IPV6_DST_MASK,
+				  sizeof(key->ipv6.dst))))
+		goto nla_put_failure;
+
+	if (key->basic.ip_proto == IPPROTO_TCP &&
+	    (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_TCP_SRC,
+			     &mask->tp.src, TCA_FLOWER_UNSPEC,
+			     sizeof(key->tp.src)) ||
+	     fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_TCP_DST,
+			     &mask->tp.dst, TCA_FLOWER_UNSPEC,
+			     sizeof(key->tp.dst))))
+		goto nla_put_failure;
+	else if (key->basic.ip_proto == IPPROTO_UDP &&
+		 (fl_dump_key_val(skb, &key->tp.src, TCA_FLOWER_KEY_UDP_SRC,
+				  &mask->tp.src, TCA_FLOWER_UNSPEC,
+				  sizeof(key->tp.src)) ||
+		  fl_dump_key_val(skb, &key->tp.dst, TCA_FLOWER_KEY_UDP_DST,
+				  &mask->tp.dst, TCA_FLOWER_UNSPEC,
+				  sizeof(key->tp.dst))))
+		goto nla_put_failure;
+
+	if (tcf_exts_dump(skb, &f->exts))
+		goto nla_put_failure;
+
+	nla_nest_end(skb, nest);
+
+	if (tcf_exts_dump_stats(skb, &f->exts) < 0)
+		goto nla_put_failure;
+
+	return skb->len;
+
+nla_put_failure:
+	nla_nest_cancel(skb, nest);
+	return -1;
+}
+
+static struct tcf_proto_ops cls_fl_ops __read_mostly = {
+	.kind		= "flower",
+	.classify	= fl_classify,
+	.init		= fl_init,
+	.destroy	= fl_destroy,
+	.get		= fl_get,
+	.change		= fl_change,
+	.delete		= fl_delete,
+	.walk		= fl_walk,
+	.dump		= fl_dump,
+	.owner		= THIS_MODULE,
+};
+
+static int __init cls_fl_init(void)
+{
+	return register_tcf_proto_ops(&cls_fl_ops);
+}
+
+static void __exit cls_fl_exit(void)
+{
+	unregister_tcf_proto_ops(&cls_fl_ops);
+}
+
+module_init(cls_fl_init);
+module_exit(cls_fl_exit);
+
+MODULE_AUTHOR("Jiri Pirko <jiri@resnulli.us>");
+MODULE_DESCRIPTION("Flower classifier");
+MODULE_LICENSE("GPL v2");
-- 
2.4.0

Re: [patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[David Miller]

From: Jiri Pirko <jiri@resnulli.us>
Date: Tue, 12 May 2015 14:56:06 +0200

> Per Davem's request, I prepared this patchset which introduces programmable
> flow dissector. For current users of flow_keys, there is a wrapper
> skb_flow_dissect_flow_keys which maintains the previous behaviour.
> For purposes of cls_flower, couple of new dissection keys were introduced.
> 
> Note that this dissector can be also eventually used by openvswitch code.
> 
> Also, as a next step, I plan to get rid of *skb_flow_get_ports(export)
> and *__skb_get_poff as their functionality can be now implemented by
> skb_flow_dissect as well.
> 
> v2->v3:
> - remove TCA_FLOWER_POLICE attr suggested by Jamal
> 
> v1->v2:
> - move __skb_tx_hash rather to dev.c as suggested by Alex

Ok, assuming this passes all of my build tests, I'll push this into
net-next.

I'm sure there will be some performance improvements possible, and I
hope you will look into making sure this new programmable classifier
is as light weight as possible.

Anyways, thanks a lot.

Re: [patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[Tom Herbert]

On Wed, May 13, 2015 at 3:21 PM, David Miller <davem@davemloft.net> wrote:
> From: Jiri Pirko <jiri@resnulli.us>
> Date: Tue, 12 May 2015 14:56:06 +0200
>
>> Per Davem's request, I prepared this patchset which introduces programmable
>> flow dissector. For current users of flow_keys, there is a wrapper
>> skb_flow_dissect_flow_keys which maintains the previous behaviour.
>> For purposes of cls_flower, couple of new dissection keys were introduced.
>>
>> Note that this dissector can be also eventually used by openvswitch code.
>>
>> Also, as a next step, I plan to get rid of *skb_flow_get_ports(export)
>> and *__skb_get_poff as their functionality can be now implemented by
>> skb_flow_dissect as well.
>>
>> v2->v3:
>> - remove TCA_FLOWER_POLICE attr suggested by Jamal
>>
>> v1->v2:
>> - move __skb_tx_hash rather to dev.c as suggested by Alex
>
> Ok, assuming this passes all of my build tests, I'll push this into
> net-next.
>
> I'm sure there will be some performance improvements possible, and I
> hope you will look into making sure this new programmable classifier
> is as light weight as possible.
>
> Anyways, thanks a lot.

David,

I still have concerns about making flow_dissector more complex like
this. This still seems like it should this programmable logic be done
in a separate function. We call flow_dissector at least once per
packet via skb_get_hash, it is in the critical path, and adding
several conditionals can only slow it down and provides no new value
to skb_get_hash. At the very least can we at least get some
performance numbers to show impact of this?

Thanks,
Tom

Re: [patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[David Miller]

From: Tom Herbert <tom@herbertland.com>
Date: Wed, 13 May 2015 15:27:59 -0400

>> I'm sure there will be some performance improvements possible, and I
>> hope you will look into making sure this new programmable classifier
>> is as light weight as possible.
 ...

> I still have concerns about making flow_dissector more complex like
> this. This still seems like it should this programmable logic be done
> in a separate function. We call flow_dissector at least once per
> packet via skb_get_hash, it is in the critical path, and adding
> several conditionals can only slow it down and provides no new value
> to skb_get_hash. At the very least can we at least get some
> performance numbers to show impact of this?

The part of what I said to Jiri above is meant exactly to ensure that
he handles this.

If we need a specialized fast path for the skb_get_hash() code paths,
so be it.

But I'm not going to denouce his entire efforts for something that
hasn't even been shown to be an issue yet.  And if it is, I'm sure
Jiri will work to address it.

Re: [patch net-next v3 00/15] introduce programable flow dissector and cls_flower

[Jiri Pirko]

Wed, May 13, 2015 at 10:01:48PM CEST, davem@davemloft.net wrote:
>From: Tom Herbert <tom@herbertland.com>
>Date: Wed, 13 May 2015 15:27:59 -0400
>
>>> I'm sure there will be some performance improvements possible, and I
>>> hope you will look into making sure this new programmable classifier
>>> is as light weight as possible.
> ...
>
>> I still have concerns about making flow_dissector more complex like
>> this. This still seems like it should this programmable logic be done
>> in a separate function. We call flow_dissector at least once per
>> packet via skb_get_hash, it is in the critical path, and adding
>> several conditionals can only slow it down and provides no new value
>> to skb_get_hash. At the very least can we at least get some
>> performance numbers to show impact of this?
>
>The part of what I said to Jiri above is meant exactly to ensure that
>he handles this.
>
>If we need a specialized fast path for the skb_get_hash() code paths,
>so be it.
>
>But I'm not going to denouce his entire efforts for something that
>hasn't even been shown to be an issue yet.  And if it is, I'm sure
>Jiri will work to address it.

Yes, sure thing.