Merge branch 'ovs_flowids'

Joe Stringer says:

====================
openvswitch: Introduce 128-bit unique flow identifiers.

This series extends the openvswitch datapath interface for flow commands to use
128-bit unique identifiers as an alternative to the netlink-formatted flow key.
This significantly reduces the cost of assembling messages between the kernel
and userspace, in particular improving Open vSwitch revalidation performance by
40% or more.

v14:
- Perform lookup using unmasked key in legacy case.
- Fix minor checkpatch.pl style violations.

v13:
- Embed sw_flow_id in sw_flow to save memory allocation in UFID case.
- Malloc unmasked key for id in non-UFID case.
- Fix bug where non-UFID case could double-serialize keys.

v12:
- Userspace patches fully merged into Open vSwitch master
- New minor refactor patches (2,3,4)
- Merge unmasked_key, ufid representation of flow identifier in sw_flow
- Improve memory allocation sizes when serializing ufid
- Handle corner case where a flow_new is requested with a flow that has an
  identical ufid as an existing flow, but a different flow key
- Limit UFID to between 1-16 octets inclusive.
- Add various helper functions to improve readibility

v11:
- Pushed most of the prerequisite patches for this series to OVS master.
- Split out openvswitch.h interface changes from datapath implementation
- Datapath implementation to be reviewed on net-next, separately

v10:
- New patch allowing datapath to serialize masked keys
- Simplify datapath interface by accepting UFID or flow_key, but not both
- Flows set up with UFID must be queried/deleted using UFID
- Reduce sw_flow memory usage for UFID
- Don't periodically rehash UFID table in linux datapath
- Remove kernel_only UFID in linux datapath

v9:
- No kernel changes

v8:
- Rename UID -> UFID
- Fix null dereference in datapath when paired with older userspace
- All patches are reviewed/acked except datapath changes.

v7:
- Remove OVS_DP_F_INDEX_BY_UID
- Rework datapath UID serialization for variable length UIDs

v6:
- Reduce netlink conversions for all datapaths
- Various bugfixes

v5:
- Various bugfixes
- Improve logging

v4:
- Datapath memory leak fixes
- Enable UID-based terse dumping and deleting by default
- Various fixes

RFCv3:
- Add datapath implementation
====================
Signed-off-by: David S. Miller <davem@davemloft.net>

Documentation/networking/openvswitch.txt

@@ -131,6 +131,19 @@ performs best-effort detection of overlapping wildcarded flows and may reject
 some but not all of them. However, this behavior may change in future versions.
 
 
+Unique flow identifiers
+-----------------------
+
+An alternative to using the original match portion of a key as the handle for
+flow identification is a unique flow identifier, or "UFID". UFIDs are optional
+for both the kernel and user space program.
+
+User space programs that support UFID are expected to provide it during flow
+setup in addition to the flow, then refer to the flow using the UFID for all
+future operations. The kernel is not required to index flows by the original
+flow key if a UFID is specified.
+
+
 Basic rule for evolving flow keys
 ---------------------------------
 

include/net/genetlink.h

@@ -205,6 +205,23 @@ static inline struct nlmsghdr *genlmsg_nlhdr(void *user_hdr,
 				   NLMSG_HDRLEN);
 }
 
+/**
+ * genlmsg_parse - parse attributes of a genetlink message
+ * @nlh: netlink message header
+ * @family: genetlink message family
+ * @tb: destination array with maxtype+1 elements
+ * @maxtype: maximum attribute type to be expected
+ * @policy: validation policy
+ * */
+static inline int genlmsg_parse(const struct nlmsghdr *nlh,
+				const struct genl_family *family,
+				struct nlattr *tb[], int maxtype,
+				const struct nla_policy *policy)
+{
+	return nlmsg_parse(nlh, family->hdrsize + GENL_HDRLEN, tb, maxtype,
+			   policy);
+}
+
 /**
  * genl_dump_check_consistent - check if sequence is consistent and advertise if not
  * @cb: netlink callback structure that stores the sequence number

include/uapi/linux/openvswitch.h

@@ -459,6 +459,14 @@ struct ovs_key_nd {
  * a wildcarded match. Omitting attribute is treated as wildcarding all
  * corresponding fields. Optional for all requests. If not present,
  * all flow key bits are exact match bits.
+ * @OVS_FLOW_ATTR_UFID: A value between 1-16 octets specifying a unique
+ * identifier for the flow. Causes the flow to be indexed by this value rather
+ * than the value of the %OVS_FLOW_ATTR_KEY attribute. Optional for all
+ * requests. Present in notifications if the flow was created with this
+ * attribute.
+ * @OVS_FLOW_ATTR_UFID_FLAGS: A 32-bit value of OR'd %OVS_UFID_F_*
+ * flags that provide alternative semantics for flow installation and
+ * retrieval. Optional for all requests.
  *
  * These attributes follow the &struct ovs_header within the Generic Netlink
  * payload for %OVS_FLOW_* commands.
@@ -474,11 +482,23 @@ enum ovs_flow_attr {
 	OVS_FLOW_ATTR_MASK,      /* Sequence of OVS_KEY_ATTR_* attributes. */
 	OVS_FLOW_ATTR_PROBE,     /* Flow operation is a feature probe, error
 				  * logging should be suppressed. */
+	OVS_FLOW_ATTR_UFID,      /* Variable length unique flow identifier. */
+	OVS_FLOW_ATTR_UFID_FLAGS,/* u32 of OVS_UFID_F_*. */
 	__OVS_FLOW_ATTR_MAX
 };
 
 #define OVS_FLOW_ATTR_MAX (__OVS_FLOW_ATTR_MAX - 1)
 
+/**
+ * Omit attributes for notifications.
+ *
+ * If a datapath request contains an %OVS_UFID_F_OMIT_* flag, then the datapath
+ * may omit the corresponding %OVS_FLOW_ATTR_* from the response.
+ */
+#define OVS_UFID_F_OMIT_KEY      (1 << 0)
+#define OVS_UFID_F_OMIT_MASK     (1 << 1)
+#define OVS_UFID_F_OMIT_ACTIONS  (1 << 2)
+
 /**
  * enum ovs_sample_attr - Attributes for %OVS_ACTION_ATTR_SAMPLE action.
  * @OVS_SAMPLE_ATTR_PROBABILITY: 32-bit fraction of packets to sample with

net/openvswitch/flow.h

@@ -197,6 +197,16 @@ struct sw_flow_match {
 	struct sw_flow_mask *mask;
 };
 
+#define MAX_UFID_LENGTH 16 /* 128 bits */
+
+struct sw_flow_id {
+	u32 ufid_len;
+	union {
+		u32 ufid[MAX_UFID_LENGTH / 4];
+		struct sw_flow_key *unmasked_key;
+	};
+};
+
 struct sw_flow_actions {
 	struct rcu_head rcu;
 	u32 actions_len;
@@ -213,13 +223,15 @@ struct flow_stats {
 
 struct sw_flow {
 	struct rcu_head rcu;
-	struct hlist_node hash_node[2];
+	struct {
+		struct hlist_node node[2];
 		u32 hash;
+	} flow_table, ufid_table;
 	int stats_last_writer;		/* NUMA-node id of the last writer on
 					 * 'stats[0]'.
 					 */
 	struct sw_flow_key key;
-	struct sw_flow_key unmasked_key;
+	struct sw_flow_id id;
 	struct sw_flow_mask *mask;
 	struct sw_flow_actions __rcu *sf_acts;
 	struct flow_stats __rcu *stats[]; /* One for each NUMA node.  First one
@@ -243,6 +255,16 @@ struct arp_eth_header {
 	unsigned char       ar_tip[4];		/* target IP address        */
 } __packed;
 
+static inline bool ovs_identifier_is_ufid(const struct sw_flow_id *sfid)
+{
+	return sfid->ufid_len;
+}
+
+static inline bool ovs_identifier_is_key(const struct sw_flow_id *sfid)
+{
+	return !ovs_identifier_is_ufid(sfid);
+}
+
 void ovs_flow_stats_update(struct sw_flow *, __be16 tcp_flags,
 			   const struct sk_buff *);
 void ovs_flow_stats_get(const struct sw_flow *, struct ovs_flow_stats *,

net/openvswitch/flow_netlink.c

@@ -1180,6 +1180,59 @@ int ovs_nla_get_match(struct sw_flow_match *match,
 	return err;
 }
 
+static size_t get_ufid_len(const struct nlattr *attr, bool log)
+{
+	size_t len;
+
+	if (!attr)
+		return 0;
+
+	len = nla_len(attr);
+	if (len < 1 || len > MAX_UFID_LENGTH) {
+		OVS_NLERR(log, "ufid size %u bytes exceeds the range (1, %d)",
+			  nla_len(attr), MAX_UFID_LENGTH);
+		return 0;
+	}
+
+	return len;
+}
+
+/* Initializes 'flow->ufid', returning true if 'attr' contains a valid UFID,
+ * or false otherwise.
+ */
+bool ovs_nla_get_ufid(struct sw_flow_id *sfid, const struct nlattr *attr,
+		      bool log)
+{
+	sfid->ufid_len = get_ufid_len(attr, log);
+	if (sfid->ufid_len)
+		memcpy(sfid->ufid, nla_data(attr), sfid->ufid_len);
+
+	return sfid->ufid_len;
+}
+
+int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
+			   const struct sw_flow_key *key, bool log)
+{
+	struct sw_flow_key *new_key;
+
+	if (ovs_nla_get_ufid(sfid, ufid, log))
+		return 0;
+
+	/* If UFID was not provided, use unmasked key. */
+	new_key = kmalloc(sizeof(*new_key), GFP_KERNEL);
+	if (!new_key)
+		return -ENOMEM;
+	memcpy(new_key, key, sizeof(*key));
+	sfid->unmasked_key = new_key;
+
+	return 0;
+}
+
+u32 ovs_nla_get_ufid_flags(const struct nlattr *attr)
+{
+	return attr ? nla_get_u32(attr) : 0;
+}
+
 /**
  * ovs_nla_get_flow_metadata - parses Netlink attributes into a flow key.
  * @key: Receives extracted in_port, priority, tun_key and skb_mark.
@@ -1216,12 +1269,12 @@ int ovs_nla_get_flow_metadata(const struct nlattr *attr,
 	return metadata_from_nlattrs(&match, &attrs, a, false, log);
 }
 
-int ovs_nla_put_flow(const struct sw_flow_key *swkey,
-		     const struct sw_flow_key *output, struct sk_buff *skb)
+static int __ovs_nla_put_key(const struct sw_flow_key *swkey,
+			     const struct sw_flow_key *output, bool is_mask,
+			     struct sk_buff *skb)
 {
 	struct ovs_key_ethernet *eth_key;
 	struct nlattr *nla, *encap;
-	bool is_mask = (swkey != output);
 
 	if (nla_put_u32(skb, OVS_KEY_ATTR_RECIRC_ID, output->recirc_id))
 		goto nla_put_failure;
@@ -1431,6 +1484,49 @@ int ovs_nla_put_flow(const struct sw_flow_key *swkey,
 	return -EMSGSIZE;
 }
 
+int ovs_nla_put_key(const struct sw_flow_key *swkey,
+		    const struct sw_flow_key *output, int attr, bool is_mask,
+		    struct sk_buff *skb)
+{
+	int err;
+	struct nlattr *nla;
+
+	nla = nla_nest_start(skb, attr);
+	if (!nla)
+		return -EMSGSIZE;
+	err = __ovs_nla_put_key(swkey, output, is_mask, skb);
+	if (err)
+		return err;
+	nla_nest_end(skb, nla);
+
+	return 0;
+}
+
+/* Called with ovs_mutex or RCU read lock. */
+int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb)
+{
+	if (ovs_identifier_is_ufid(&flow->id))
+		return nla_put(skb, OVS_FLOW_ATTR_UFID, flow->id.ufid_len,
+			       flow->id.ufid);
+
+	return ovs_nla_put_key(flow->id.unmasked_key, flow->id.unmasked_key,
+			       OVS_FLOW_ATTR_KEY, false, skb);
+}
+
+/* Called with ovs_mutex or RCU read lock. */
+int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb)
+{
+	return ovs_nla_put_key(&flow->mask->key, &flow->key,
+				OVS_FLOW_ATTR_KEY, false, skb);
+}
+
+/* Called with ovs_mutex or RCU read lock. */
+int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb)
+{
+	return ovs_nla_put_key(&flow->key, &flow->mask->key,
+				OVS_FLOW_ATTR_MASK, true, skb);
+}
+
 #define MAX_ACTIONS_BUFSIZE	(32 * 1024)
 
 static struct sw_flow_actions *nla_alloc_flow_actions(int size, bool log)

net/openvswitch/flow_netlink.h

@@ -43,16 +43,25 @@ size_t ovs_key_attr_size(void);
 void ovs_match_init(struct sw_flow_match *match,
 		    struct sw_flow_key *key, struct sw_flow_mask *mask);
 
-int ovs_nla_put_flow(const struct sw_flow_key *,
-		     const struct sw_flow_key *, struct sk_buff *);
+int ovs_nla_put_key(const struct sw_flow_key *, const struct sw_flow_key *,
+		    int attr, bool is_mask, struct sk_buff *);
 int ovs_nla_get_flow_metadata(const struct nlattr *, struct sw_flow_key *,
 			      bool log);
 
+int ovs_nla_put_identifier(const struct sw_flow *flow, struct sk_buff *skb);
+int ovs_nla_put_masked_key(const struct sw_flow *flow, struct sk_buff *skb);
+int ovs_nla_put_mask(const struct sw_flow *flow, struct sk_buff *skb);
+
 int ovs_nla_get_match(struct sw_flow_match *, const struct nlattr *key,
 		      const struct nlattr *mask, bool log);
 int ovs_nla_put_egress_tunnel_key(struct sk_buff *,
 				  const struct ovs_tunnel_info *);
 
+bool ovs_nla_get_ufid(struct sw_flow_id *, const struct nlattr *, bool log);
+int ovs_nla_get_identifier(struct sw_flow_id *sfid, const struct nlattr *ufid,
+			   const struct sw_flow_key *key, bool log);
+u32 ovs_nla_get_ufid_flags(const struct nlattr *attr);
+
 int ovs_nla_copy_actions(const struct nlattr *attr,
 			 const struct sw_flow_key *key,
 			 struct sw_flow_actions **sfa, bool log);

net/openvswitch/flow_table.h

@@ -47,9 +47,11 @@ struct table_instance {
 
 struct flow_table {
 	struct table_instance __rcu *ti;
+	struct table_instance __rcu *ufid_ti;
 	struct list_head mask_list;
 	unsigned long last_rehash;
 	unsigned int count;
+	unsigned int ufid_count;
 };
 
 extern struct kmem_cache *flow_stats_cache;
@@ -78,8 +80,10 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
 				    const struct sw_flow_key *);
 struct sw_flow *ovs_flow_tbl_lookup_exact(struct flow_table *tbl,
 					  const struct sw_flow_match *match);
-bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
-			       const struct sw_flow_match *match);
+struct sw_flow *ovs_flow_tbl_lookup_ufid(struct flow_table *,
+					 const struct sw_flow_id *);
+
+bool ovs_flow_cmp(const struct sw_flow *, const struct sw_flow_match *);
 
 void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
 		       const struct sw_flow_mask *mask);