Skip to contentSkip to conversation

Open vSwitch

GIT net-next New Jesse Gross 3 participants 13 messages Nov 2, 2013 – Nov 4, 2013 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

A set of updates for net-next/3.13. Major changes are:
* Restructure flow handling code to be more logically organized and easier to read.
* Rehashing of the flow table is moved from a workqueue to flow installation time. Before, heavy load could block the workqueue for excessive periods of time.
* Additional debugging information is provided to help diagnose megaflows.
* It's now possible to match on TCP flags.

The following changes since commit 272b98c6455f00884f0350f775c5342358ebb73f:

Linux 3.12-rc1 (2013-09-16 16:17:51 -0400)

are available in the git repository at:

git://git.kernel.org/pub/scm/linux/kernel/git/jesse/openvswitch.git master

for you to fetch changes up to 8ddd094675cfd453fc9838caa46ea108a4107183:

openvswitch: Use flow hash during flow lookup operation. (2013-11-01 18:43:46 -0700)

----------------------------------------------------------------
Andy Zhou (1):
openvswitch: collect mega flow mask stats

Jarno Rajahalme (2):
openvswitch: Widen TCP flags handling.
openvswitch: TCP flags matching support.

Pravin B Shelar (6):
openvswitch: Move flow table rehashing to flow install.
openvswitch: Restructure datapath.c and flow.c
openvswitch: Move mega-flow list out of rehashing struct.
openvswitch: Simplify mega-flow APIs.
openvswitch: Enable all GSO features on internal port.
openvswitch: Use flow hash during flow lookup operation.

Wei Yongjun (2):
openvswitch: remove duplicated include from vport-vxlan.c
openvswitch: remove duplicated include from vport-gre.c

 include/uapi/linux/openvswitch.h     |   18 +-
 net/openvswitch/Makefile             |    2 +
 net/openvswitch/datapath.c           |  668 ++------------
 net/openvswitch/datapath.h           |    9 +-
 net/openvswitch/flow.c               | 1605 +--------------------------------
 net/openvswitch/flow.h               |  132 +--
 net/openvswitch/flow_netlink.c       | 1630 ++++++++++++++++++++++++++++++++++
 net/openvswitch/flow_netlink.h       |   60 ++
 net/openvswitch/flow_table.c         |  592 ++++++++++++
 net/openvswitch/flow_table.h         |   81 ++
 net/openvswitch/vport-gre.c          |    2 -
 net/openvswitch/vport-internal_dev.c |    2 +-
 net/openvswitch/vport-vxlan.c        |    1 -
 13 files changed, 2511 insertions(+), 2291 deletions(-)
 create mode 100644 net/openvswitch/flow_netlink.c
 create mode 100644 net/openvswitch/flow_netlink.h
 create mode 100644 net/openvswitch/flow_table.c
 create mode 100644 net/openvswitch/flow_table.h

A message that is not in the archive (sent off-list or not yet received): 1383378230-59624-1-git-send-email-jesse-l0M0P4e3n4LQT0dZR+AlfA@public.gmane.org

#2 lore raw

PATCH 1/11 openvswitch: Move flow table rehashing to flow install.

From: Pravin B Shelar <[email protected]>

Rehashing in ovs-workqueue can cause ovs-mutex lock contentions in case of heavy flow setups where both needs ovs-mutex. So by moving rehashing to flow-setup we can eliminate contention. This also simplify ovs locking and reduces dependence on workqueue.

---
 net/openvswitch/datapath.c | 50 ++++++++++------------------------------------
 net/openvswitch/datapath.h |  2 ++
 2 files changed, 13 insertions(+), 39 deletions(-)
net/openvswitch/datapath.c +11 -39
index 2aa13bd..2e1a9c2 100644
@@ -60,8 +60,6 @@
 
 
 #define REHASH_FLOW_INTERVAL (10 * 60 * HZ)
-static void rehash_flow_table(struct work_struct *work);
-static DECLARE_DELAYED_WORK(rehash_flow_wq, rehash_flow_table);
 
 int ovs_net_id __read_mostly;
 
@@ -1289,22 +1287,25 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 	/* Check if this is a duplicate flow */
 	flow = ovs_flow_lookup(table, &key);
 	if (!flow) {
+		struct flow_table *new_table = NULL;
 		struct sw_flow_mask *mask_p;
+
 		/* Bail out if we're not allowed to create a new flow. */
 		error = -ENOENT;
 		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
 			goto err_unlock_ovs;
 
 		/* Expand table, if necessary, to make room. */
-		if (ovs_flow_tbl_need_to_expand(table)) {
-			struct flow_table *new_table;
-
+		if (ovs_flow_tbl_need_to_expand(table))
 			new_table = ovs_flow_tbl_expand(table);
-			if (!IS_ERR(new_table)) {
-				rcu_assign_pointer(dp->table, new_table);
-				ovs_flow_tbl_destroy(table, true);
-				table = ovsl_dereference(dp->table);
-			}
+		else if (time_after(jiffies, dp->last_rehash + REHASH_FLOW_INTERVAL))
+			new_table = ovs_flow_tbl_rehash(table);
+
+		if (new_table && !IS_ERR(new_table)) {
+			rcu_assign_pointer(dp->table, new_table);
+			ovs_flow_tbl_destroy(table, true);
+			table = ovsl_dereference(dp->table);
+			dp->last_rehash = jiffies;
 		}
 
 		/* Allocate flow. */
@@ -2336,32 +2337,6 @@ error:
 	return err;
 }
 
-static void rehash_flow_table(struct work_struct *work)
-{
-	struct datapath *dp;
-	struct net *net;
-
-	ovs_lock();
-	rtnl_lock();
-	for_each_net(net) {
-		struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
-
-		list_for_each_entry(dp, &ovs_net->dps, list_node) {
-			struct flow_table *old_table = ovsl_dereference(dp->table);
-			struct flow_table *new_table;
-
-			new_table = ovs_flow_tbl_rehash(old_table);
-			if (!IS_ERR(new_table)) {
-				rcu_assign_pointer(dp->table, new_table);
-				ovs_flow_tbl_destroy(old_table, true);
-			}
-		}
-	}
-	rtnl_unlock();
-	ovs_unlock();
-	schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
-}
-
 static int __net_init ovs_init_net(struct net *net)
 {
 	struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
@@ -2419,8 +2394,6 @@ static int __init dp_init(void)
 	if (err < 0)
 		goto error_unreg_notifier;
 
-	schedule_delayed_work(&rehash_flow_wq, REHASH_FLOW_INTERVAL);
-
 	return 0;
 
 error_unreg_notifier:
@@ -2437,7 +2410,6 @@ error:
 
 static void dp_cleanup(void)
 {
-	cancel_delayed_work_sync(&rehash_flow_wq);
 	dp_unregister_genl(ARRAY_SIZE(dp_genl_families));
 	unregister_netdevice_notifier(&ovs_dp_device_notifier);
 	unregister_pernet_device(&ovs_net_ops);
net/openvswitch/datapath.h +2 -0
index 4d109c1..2c15541 100644
@@ -62,6 +62,7 @@ struct dp_stats_percpu {
  * ovs_mutex and RCU.
  * @stats_percpu: Per-CPU datapath statistics.
  * @net: Reference to net namespace.
+ * @last_rehash: Timestamp of last rehash.
  *
  * Context: See the comment on locking at the top of datapath.c for additional
  * locking information.
@@ -83,6 +84,7 @@ struct datapath {
 	/* Network namespace ref. */
 	struct net *net;
 #endif
+	unsigned long last_rehash;
 };
 
 /**
Signature · 1 line
1.8.3.2
#3 lore raw

PATCH 2/11 openvswitch: remove duplicated include from vport-vxlan.c

In reply to Jesse Gross

From: Wei Yongjun <[email protected]>

Remove duplicated include.

---
 net/openvswitch/vport-vxlan.c | 1 -
 1 file changed, 1 deletion(-)
net/openvswitch/vport-vxlan.c +0 -1
index a481c03..b0da394 100644
@@ -29,7 +29,6 @@
 #include <net/ip.h>
 #include <net/udp.h>
 #include <net/ip_tunnels.h>
-#include <net/udp.h>
 #include <net/rtnetlink.h>
 #include <net/route.h>
 #include <net/dsfield.h>
Signature · 1 line
1.8.3.2
#4 lore raw

PATCH 3/11 openvswitch: remove duplicated include from vport-gre.c

In reply to Jesse Gross

From: Wei Yongjun <[email protected]>

Remove duplicated include.

---
 net/openvswitch/vport-gre.c | 2 --
 1 file changed, 2 deletions(-)
net/openvswitch/vport-gre.c +0 -2
index c99dea5..a3d6951 100644
@@ -24,8 +24,6 @@
 #include <linux/if_tunnel.h>
 #include <linux/if_vlan.h>
 #include <linux/in.h>
-#include <linux/if_vlan.h>
-#include <linux/in.h>
 #include <linux/in_route.h>
 #include <linux/inetdevice.h>
 #include <linux/jhash.h>
Signature · 1 line
1.8.3.2
#5 lore raw

PATCH 4/11 openvswitch: Restructure datapath.c and flow.c

In reply to Jesse Gross

From: Pravin B Shelar <[email protected]>

Over the time datapath.c and flow.c has became pretty large files. Following patch restructures functionality of component into three different components:

flow.c: contains flow extract.
flow_netlink.c: netlink flow api.
flow_table.c: flow table api.

This patch restructures code without changing logic.

---
 net/openvswitch/Makefile       |    2 +
 net/openvswitch/datapath.c     |  528 +------------
 net/openvswitch/datapath.h     |    1 +
 net/openvswitch/flow.c         | 1605 +---------------------------------------
 net/openvswitch/flow.h         |  128 +---
 net/openvswitch/flow_netlink.c | 1603 +++++++++++++++++++++++++++++++++++++++
 net/openvswitch/flow_netlink.h |   60 ++
 net/openvswitch/flow_table.c   |  517 +++++++++++++
 net/openvswitch/flow_table.h   |   91 +++
 9 files changed, 2354 insertions(+), 2181 deletions(-)
 create mode 100644 net/openvswitch/flow_netlink.c
 create mode 100644 net/openvswitch/flow_netlink.h
 create mode 100644 net/openvswitch/flow_table.c
 create mode 100644 net/openvswitch/flow_table.h
net/openvswitch/Makefile +2 -0
index ea36e99..3591cb5 100644
@@ -9,6 +9,8 @@ openvswitch-y := \
 	datapath.o \
 	dp_notify.o \
 	flow.o \
+	flow_netlink.o \
+	flow_table.o \
 	vport.o \
 	vport-internal_dev.o \
 	vport-netdev.o
net/openvswitch/datapath.c +30 -498

670 lines not on this page · raw message

net/openvswitch/datapath.h +1 -0
index 2c15541..a6982ef 100644
@@ -27,6 +27,7 @@
 #include <linux/u64_stats_sync.h>
 
 #include "flow.h"
+#include "flow_table.h"
 #include "vport.h"
 
 #define DP_MAX_PORTS           USHRT_MAX
net/openvswitch/flow.c +24 -1581

1,636 lines not on this page · raw message

net/openvswitch/flow.h +26 -102

154 lines not on this page · raw message

net/openvswitch/flow_netlink.c new file +1603 -0

525 lines not on this page · raw message

net/openvswitch/flow_netlink.h new file +60 -0

60 lines not on this page · raw message

net/openvswitch/flow_table.c new file +517 -0

517 lines not on this page · raw message

net/openvswitch/flow_table.h new file +91 -0

91 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.8.3.2

Part of the diff is not shown here. View the raw message.

#6 lore raw

PATCH 5/11 openvswitch: Move mega-flow list out of rehashing struct.

In reply to Jesse Gross

From: Pravin B Shelar <[email protected]>

ovs-flow rehash does not touch mega flow list. Following patch moves it dp struct datapath. Avoid one extra indirection for accessing mega-flow list head on every packet receive.

---
 net/openvswitch/datapath.c   |  77 ++++------------
 net/openvswitch/datapath.h   |   6 +-
 net/openvswitch/flow_table.c | 205 ++++++++++++++++++++++++++-----------------
 net/openvswitch/flow_table.h |  32 +++----
 4 files changed, 155 insertions(+), 165 deletions(-)
net/openvswitch/datapath.c +17 -60
index 72e6874..60b9be3 100644
@@ -59,8 +59,6 @@
 #include "vport-internal_dev.h"
 #include "vport-netdev.h"
 
-#define REHASH_FLOW_INTERVAL (10 * 60 * HZ)
-
 int ovs_net_id __read_mostly;
 
 static void ovs_notify(struct sk_buff *skb, struct genl_info *info,
@@ -163,7 +161,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
 {
 	struct datapath *dp = container_of(rcu, struct datapath, rcu);
 
-	ovs_flow_tbl_destroy((__force struct flow_table *)dp->table, false);
+	ovs_flow_tbl_destroy(&dp->table, false);
 	free_percpu(dp->stats_percpu);
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp->ports);
@@ -235,7 +233,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 	}
 
 	/* Look up flow. */
-	flow = ovs_flow_tbl_lookup(rcu_dereference(dp->table), &key);
+	flow = ovs_flow_tbl_lookup(&dp->table, &key);
 	if (unlikely(!flow)) {
 		struct dp_upcall_info upcall;
 
@@ -453,23 +451,6 @@ out:
 	return err;
 }
 
-/* Called with ovs_mutex. */
-static int flush_flows(struct datapath *dp)
-{
-	struct flow_table *old_table;
-	struct flow_table *new_table;
-
-	old_table = ovsl_dereference(dp->table);
-	new_table = ovs_flow_tbl_alloc(TBL_MIN_BUCKETS);
-	if (!new_table)
-		return -ENOMEM;
-
-	rcu_assign_pointer(dp->table, new_table);
-
-	ovs_flow_tbl_destroy(old_table, true);
-	return 0;
-}
-
 static void clear_stats(struct sw_flow *flow)
 {
 	flow->used = 0;
@@ -584,11 +565,9 @@ static struct genl_ops dp_packet_genl_ops[] = {
 
 static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
 {
-	struct flow_table *table;
 	int i;
 
-	table = rcu_dereference_check(dp->table, lockdep_ovsl_is_held());
-	stats->n_flows = ovs_flow_tbl_count(table);
+	stats->n_flows = ovs_flow_tbl_count(&dp->table);
 
 	stats->n_hit = stats->n_missed = stats->n_lost = 0;
 	for_each_possible_cpu(i) {
@@ -773,7 +752,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 	struct sw_flow_mask mask;
 	struct sk_buff *reply;
 	struct datapath *dp;
-	struct flow_table *table;
 	struct sw_flow_actions *acts = NULL;
 	struct sw_flow_match match;
 	int error;
@@ -814,12 +792,9 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 	if (!dp)
 		goto err_unlock_ovs;
 
-	table = ovsl_dereference(dp->table);
-
 	/* Check if this is a duplicate flow */
-	flow = ovs_flow_tbl_lookup(table, &key);
+	flow = ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow) {
-		struct flow_table *new_table = NULL;
 		struct sw_flow_mask *mask_p;
 
 		/* Bail out if we're not allowed to create a new flow. */
@@ -827,19 +802,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
 			goto err_unlock_ovs;
 
-		/* Expand table, if necessary, to make room. */
-		if (ovs_flow_tbl_need_to_expand(table))
-			new_table = ovs_flow_tbl_expand(table);
-		else if (time_after(jiffies, dp->last_rehash + REHASH_FLOW_INTERVAL))
-			new_table = ovs_flow_tbl_rehash(table);
-
-		if (new_table && !IS_ERR(new_table)) {
-			rcu_assign_pointer(dp->table, new_table);
-			ovs_flow_tbl_destroy(table, true);
-			table = ovsl_dereference(dp->table);
-			dp->last_rehash = jiffies;
-		}
-
 		/* Allocate flow. */
 		flow = ovs_flow_alloc();
 		if (IS_ERR(flow)) {
@@ -852,7 +814,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		flow->unmasked_key = key;
 
 		/* Make sure mask is unique in the system */
-		mask_p = ovs_sw_flow_mask_find(table, &mask);
+		mask_p = ovs_sw_flow_mask_find(&dp->table, &mask);
 		if (!mask_p) {
 			/* Allocate a new mask if none exsits. */
 			mask_p = ovs_sw_flow_mask_alloc();
@@ -860,7 +822,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 				goto err_flow_free;
 			mask_p->key = mask.key;
 			mask_p->range = mask.range;
-			ovs_sw_flow_mask_insert(table, mask_p);
+			ovs_sw_flow_mask_insert(&dp->table, mask_p);
 		}
 
 		ovs_sw_flow_mask_add_ref(mask_p);
@@ -868,7 +830,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		rcu_assign_pointer(flow->sf_acts, acts);
 
 		/* Put flow in bucket. */
-		ovs_flow_tbl_insert(table, flow);
+		ovs_flow_tbl_insert(&dp->table, flow);
 
 		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
 						info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -936,7 +898,6 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 	struct sk_buff *reply;
 	struct sw_flow *flow;
 	struct datapath *dp;
-	struct flow_table *table;
 	struct sw_flow_match match;
 	int err;
 
@@ -957,8 +918,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 		goto unlock;
 	}
 
-	table = ovsl_dereference(dp->table);
-	flow = ovs_flow_tbl_lookup(table, &key);
+	flow = ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
 		err = -ENOENT;
 		goto unlock;
@@ -986,7 +946,6 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	struct sk_buff *reply;
 	struct sw_flow *flow;
 	struct datapath *dp;
-	struct flow_table *table;
 	struct sw_flow_match match;
 	int err;
 
@@ -998,7 +957,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	}
 
 	if (!a[OVS_FLOW_ATTR_KEY]) {
-		err = flush_flows(dp);
+		err = ovs_flow_tbl_flush(&dp->table);
 		goto unlock;
 	}
 
@@ -1007,8 +966,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	if (err)
 		goto unlock;
 
-	table = ovsl_dereference(dp->table);
-	flow = ovs_flow_tbl_lookup(table, &key);
+	flow = ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
 		err = -ENOENT;
 		goto unlock;
@@ -1020,7 +978,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 		goto unlock;
 	}
 
-	ovs_flow_tbl_remove(table, flow);
+	ovs_flow_tbl_remove(&dp->table, flow);
 
 	err = ovs_flow_cmd_fill_info(flow, dp, reply, info->snd_portid,
 				     info->snd_seq, 0, OVS_FLOW_CMD_DEL);
@@ -1039,8 +997,8 @@ unlock:
 static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
 {
 	struct ovs_header *ovs_header = genlmsg_data(nlmsg_data(cb->nlh));
+	struct table_instance *ti;
 	struct datapath *dp;
-	struct flow_table *table;
 
 	rcu_read_lock();
 	dp = get_dp(sock_net(skb->sk), ovs_header->dp_ifindex);
@@ -1049,14 +1007,14 @@ static int ovs_flow_cmd_dump(struct sk_buff *skb, struct netlink_callback *cb)
 		return -ENODEV;
 	}
 
-	table = rcu_dereference(dp->table);
+	ti = rcu_dereference(dp->table.ti);
 	for (;;) {
 		struct sw_flow *flow;
 		u32 bucket, obj;
 
 		bucket = cb->args[0];
 		obj = cb->args[1];
-		flow = ovs_flow_tbl_dump_next(table, &bucket, &obj);
+		flow = ovs_flow_tbl_dump_next(ti, &bucket, &obj);
 		if (!flow)
 			break;
 
@@ -1220,9 +1178,8 @@ static int ovs_dp_cmd_new(struct sk_buff *skb, struct genl_info *info)
 	ovs_dp_set_net(dp, hold_net(sock_net(skb->sk)));
 
 	/* Allocate table. */
-	err = -ENOMEM;
-	rcu_assign_pointer(dp->table, ovs_flow_tbl_alloc(TBL_MIN_BUCKETS));
-	if (!dp->table)
+	err = ovs_flow_tbl_init(&dp->table);
+	if (err)
 		goto err_free_dp;
 
 	dp->stats_percpu = alloc_percpu(struct dp_stats_percpu);
@@ -1279,7 +1236,7 @@ err_destroy_ports_array:
 err_destroy_percpu:
 	free_percpu(dp->stats_percpu);
 err_destroy_table:
-	ovs_flow_tbl_destroy(ovsl_dereference(dp->table), false);
+	ovs_flow_tbl_destroy(&dp->table, false);
 err_free_dp:
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp);
net/openvswitch/datapath.h +2 -4
index a6982ef..acfd4af 100644
@@ -58,12 +58,11 @@ struct dp_stats_percpu {
  * struct datapath - datapath for flow-based packet switching
  * @rcu: RCU callback head for deferred destruction.
  * @list_node: Element in global 'dps' list.
- * @table: Current flow table.  Protected by ovs_mutex and RCU.
+ * @table: flow table.
  * @ports: Hash table for ports.  %OVSP_LOCAL port always exists.  Protected by
  * ovs_mutex and RCU.
  * @stats_percpu: Per-CPU datapath statistics.
  * @net: Reference to net namespace.
- * @last_rehash: Timestamp of last rehash.
  *
  * Context: See the comment on locking at the top of datapath.c for additional
  * locking information.
@@ -73,7 +72,7 @@ struct datapath {
 	struct list_head list_node;
 
 	/* Flow table. */
-	struct flow_table __rcu *table;
+	struct flow_table table;
 
 	/* Switch ports. */
 	struct hlist_head *ports;
@@ -85,7 +84,6 @@ struct datapath {
 	/* Network namespace ref. */
 	struct net *net;
 #endif
-	unsigned long last_rehash;
 };
 
 /**
net/openvswitch/flow_table.c +123 -82

355 lines not on this page · raw message

net/openvswitch/flow_table.h +13 -19
index d7a1144..5d1abe5 100644
@@ -36,42 +36,36 @@
 
 #include "flow.h"
 
-#define TBL_MIN_BUCKETS		1024
-
-struct flow_table {
+struct table_instance {
 	struct flex_array *buckets;
-	unsigned int count, n_buckets;
+	unsigned int n_buckets;
 	struct rcu_head rcu;
-	struct list_head *mask_list;
 	int node_ver;
 	u32 hash_seed;
 	bool keep_flows;
 };
 
+struct flow_table {
+	struct table_instance __rcu *ti;
+	struct list_head mask_list;
+	unsigned long last_rehash;
+	unsigned int count;
+};
+
 int ovs_flow_init(void);
 void ovs_flow_exit(void);
 
 struct sw_flow *ovs_flow_alloc(void);
 void ovs_flow_free(struct sw_flow *, bool deferred);
 
-static inline int ovs_flow_tbl_count(struct flow_table *table)
-{
-	return table->count;
-}
-
-static inline int ovs_flow_tbl_need_to_expand(struct flow_table *table)
-{
-	return (table->count > table->n_buckets);
-}
-
-struct flow_table *ovs_flow_tbl_alloc(int new_size);
-struct flow_table *ovs_flow_tbl_expand(struct flow_table *table);
-struct flow_table *ovs_flow_tbl_rehash(struct flow_table *table);
+int ovs_flow_tbl_init(struct flow_table *);
+int ovs_flow_tbl_count(struct flow_table *table);
 void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred);
+int ovs_flow_tbl_flush(struct flow_table *flow_table);
 
 void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow);
 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
-struct sw_flow *ovs_flow_tbl_dump_next(struct flow_table *table,
+struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *table,
 				       u32 *bucket, u32 *idx);
 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
 				    const struct sw_flow_key *);
Signature · 1 line
1.8.3.2
#7 lore raw

PATCH 6/11 openvswitch: Simplify mega-flow APIs.

In reply to Jesse Gross

From: Pravin B Shelar <[email protected]>

Hides mega-flow implementation in flow_table.c rather than datapath.c.

---
 net/openvswitch/datapath.c   |  27 +++------
 net/openvswitch/flow_table.c | 138 +++++++++++++++++++++++++------------------
 net/openvswitch/flow_table.h |  12 +---
 3 files changed, 89 insertions(+), 88 deletions(-)
net/openvswitch/datapath.c +7 -20
index 60b9be3..cf27097 100644
@@ -161,7 +161,7 @@ static void destroy_dp_rcu(struct rcu_head *rcu)
 {
 	struct datapath *dp = container_of(rcu, struct datapath, rcu);
 
-	ovs_flow_tbl_destroy(&dp->table, false);
+	ovs_flow_tbl_destroy(&dp->table);
 	free_percpu(dp->stats_percpu);
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp->ports);
@@ -795,8 +795,6 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 	/* Check if this is a duplicate flow */
 	flow = ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow) {
-		struct sw_flow_mask *mask_p;
-
 		/* Bail out if we're not allowed to create a new flow. */
 		error = -ENOENT;
 		if (info->genlhdr->cmd == OVS_FLOW_CMD_SET)
@@ -812,25 +810,14 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 
 		flow->key = masked_key;
 		flow->unmasked_key = key;
-
-		/* Make sure mask is unique in the system */
-		mask_p = ovs_sw_flow_mask_find(&dp->table, &mask);
-		if (!mask_p) {
-			/* Allocate a new mask if none exsits. */
-			mask_p = ovs_sw_flow_mask_alloc();
-			if (!mask_p)
-				goto err_flow_free;
-			mask_p->key = mask.key;
-			mask_p->range = mask.range;
-			ovs_sw_flow_mask_insert(&dp->table, mask_p);
-		}
-
-		ovs_sw_flow_mask_add_ref(mask_p);
-		flow->mask = mask_p;
 		rcu_assign_pointer(flow->sf_acts, acts);
 
 		/* Put flow in bucket. */
-		ovs_flow_tbl_insert(&dp->table, flow);
+		error = ovs_flow_tbl_insert(&dp->table, flow, &mask);
+		if (error) {
+			acts = NULL;
+			goto err_flow_free;
+		}
 
 		reply = ovs_flow_cmd_build_info(flow, dp, info->snd_portid,
 						info->snd_seq, OVS_FLOW_CMD_NEW);
@@ -1236,7 +1223,7 @@ err_destroy_ports_array:
 err_destroy_percpu:
 	free_percpu(dp->stats_percpu);
 err_destroy_table:
-	ovs_flow_tbl_destroy(&dp->table, false);
+	ovs_flow_tbl_destroy(&dp->table);
 err_free_dp:
 	release_net(ovs_dp_get_net(dp));
 	kfree(dp);
net/openvswitch/flow_table.c +79 -59

199 lines not on this page · raw message

net/openvswitch/flow_table.h +3 -9
index 5d1abe5..4db5f78 100644
@@ -60,10 +60,11 @@ void ovs_flow_free(struct sw_flow *, bool deferred);
 
 int ovs_flow_tbl_init(struct flow_table *);
 int ovs_flow_tbl_count(struct flow_table *table);
-void ovs_flow_tbl_destroy(struct flow_table *table, bool deferred);
+void ovs_flow_tbl_destroy(struct flow_table *table);
 int ovs_flow_tbl_flush(struct flow_table *flow_table);
 
-void ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow);
+int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
+			struct sw_flow_mask *mask);
 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
 struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *table,
 				       u32 *bucket, u32 *idx);
@@ -73,13 +74,6 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
 bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
 			       struct sw_flow_match *match);
 
-struct sw_flow_mask *ovs_sw_flow_mask_alloc(void);
-void ovs_sw_flow_mask_add_ref(struct sw_flow_mask *);
-void ovs_sw_flow_mask_del_ref(struct sw_flow_mask *, bool deferred);
-void ovs_sw_flow_mask_insert(struct flow_table *, struct sw_flow_mask *);
-struct sw_flow_mask *ovs_sw_flow_mask_find(const struct flow_table *,
-					   const struct sw_flow_mask *);
 void ovs_flow_mask_key(struct sw_flow_key *dst, const struct sw_flow_key *src,
 		       const struct sw_flow_mask *mask);
-
 #endif /* flow_table.h */
Signature · 1 line
1.8.3.2
#8 lore raw

PATCH 7/11 openvswitch: collect mega flow mask stats

In reply to Jesse Gross

From: Andy Zhou <[email protected]>

Collect mega flow mask stats. ovs-dpctl show command can be used to display them for debugging and performance tuning.

---
 include/uapi/linux/openvswitch.h | 17 ++++++++++++++---
 net/openvswitch/datapath.c       | 38 +++++++++++++++++++++++++++++++-------
 net/openvswitch/datapath.h       |  4 ++++
 net/openvswitch/flow_table.c     | 16 +++++++++++++++-
 net/openvswitch/flow_table.h     |  4 +++-
 5 files changed, 67 insertions(+), 12 deletions(-)
include/uapi/linux/openvswitch.h +14 -3
index a74d375..2cc4644 100644
@@ -63,15 +63,18 @@ enum ovs_datapath_cmd {
  * not be sent.
  * @OVS_DP_ATTR_STATS: Statistics about packets that have passed through the
  * datapath.  Always present in notifications.
+ * @OVS_DP_ATTR_MEGAFLOW_STATS: Statistics about mega flow masks usage for the
+ * datapath. Always present in notifications.
  *
  * These attributes follow the &struct ovs_header within the Generic Netlink
  * payload for %OVS_DP_* commands.
  */
 enum ovs_datapath_attr {
 	OVS_DP_ATTR_UNSPEC,
-	OVS_DP_ATTR_NAME,       /* name of dp_ifindex netdev */
-	OVS_DP_ATTR_UPCALL_PID, /* Netlink PID to receive upcalls */
-	OVS_DP_ATTR_STATS,      /* struct ovs_dp_stats */
+	OVS_DP_ATTR_NAME,		/* name of dp_ifindex netdev */
+	OVS_DP_ATTR_UPCALL_PID,		/* Netlink PID to receive upcalls */
+	OVS_DP_ATTR_STATS,		/* struct ovs_dp_stats */
+	OVS_DP_ATTR_MEGAFLOW_STATS,	/* struct ovs_dp_megaflow_stats */
 	__OVS_DP_ATTR_MAX
 };
 
@@ -84,6 +87,14 @@ struct ovs_dp_stats {
 	__u64 n_flows;           /* Number of flows present */
 };
 
+struct ovs_dp_megaflow_stats {
+	__u64 n_mask_hit;	 /* Number of masks used for flow lookups. */
+	__u32 n_masks;		 /* Number of masks for the datapath. */
+	__u32 pad0;		 /* Pad for future expension. */
+	__u64 pad1;		 /* Pad for future expension. */
+	__u64 pad2;		 /* Pad for future expension. */
+};
+
 struct ovs_vport_stats {
 	__u64   rx_packets;		/* total packets received       */
 	__u64   tx_packets;		/* total packets transmitted    */
net/openvswitch/datapath.c +31 -7
index cf27097..5bc5a4e 100644
@@ -221,6 +221,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 	struct dp_stats_percpu *stats;
 	struct sw_flow_key key;
 	u64 *stats_counter;
+	u32 n_mask_hit;
 	int error;
 
 	stats = this_cpu_ptr(dp->stats_percpu);
@@ -233,7 +234,7 @@ void ovs_dp_process_received_packet(struct vport *p, struct sk_buff *skb)
 	}
 
 	/* Look up flow. */
-	flow = ovs_flow_tbl_lookup(&dp->table, &key);
+	flow = ovs_flow_tbl_lookup(&dp->table, &key, &n_mask_hit);
 	if (unlikely(!flow)) {
 		struct dp_upcall_info upcall;
 
@@ -258,6 +259,7 @@ out:
 	/* Update datapath statistics. */
 	u64_stats_update_begin(&stats->sync);
 	(*stats_counter)++;
+	stats->n_mask_hit += n_mask_hit;
 	u64_stats_update_end(&stats->sync);
 }
 
@@ -563,13 +565,18 @@ static struct genl_ops dp_packet_genl_ops[] = {
 	}
 };
 
-static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
+static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats,
+			 struct ovs_dp_megaflow_stats *mega_stats)
 {
 	int i;
 
+	memset(mega_stats, 0, sizeof(*mega_stats));
+
 	stats->n_flows = ovs_flow_tbl_count(&dp->table);
+	mega_stats->n_masks = ovs_flow_tbl_num_masks(&dp->table);
 
 	stats->n_hit = stats->n_missed = stats->n_lost = 0;
+
 	for_each_possible_cpu(i) {
 		const struct dp_stats_percpu *percpu_stats;
 		struct dp_stats_percpu local_stats;
@@ -585,6 +592,7 @@ static void get_dp_stats(struct datapath *dp, struct ovs_dp_stats *stats)
 		stats->n_hit += local_stats.n_hit;
 		stats->n_missed += local_stats.n_missed;
 		stats->n_lost += local_stats.n_lost;
+		mega_stats->n_mask_hit += local_stats.n_mask_hit;
 	}
 }
 
@@ -743,6 +751,14 @@ static struct sk_buff *ovs_flow_cmd_build_info(struct sw_flow *flow,
 	return skb;
 }
 
+static struct sw_flow *__ovs_flow_tbl_lookup(struct flow_table *tbl,
+					      const struct sw_flow_key *key)
+{
+	u32 __always_unused n_mask_hit;
+
+	return ovs_flow_tbl_lookup(tbl, key, &n_mask_hit);
+}
+
 static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 {
 	struct nlattr **a = info->attrs;
@@ -793,7 +809,7 @@ static int ovs_flow_cmd_new_or_set(struct sk_buff *skb, struct genl_info *info)
 		goto err_unlock_ovs;
 
 	/* Check if this is a duplicate flow */
-	flow = ovs_flow_tbl_lookup(&dp->table, &key);
+	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow) {
 		/* Bail out if we're not allowed to create a new flow. */
 		error = -ENOENT;
@@ -905,7 +921,7 @@ static int ovs_flow_cmd_get(struct sk_buff *skb, struct genl_info *info)
 		goto unlock;
 	}
 
-	flow = ovs_flow_tbl_lookup(&dp->table, &key);
+	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
 		err = -ENOENT;
 		goto unlock;
@@ -953,7 +969,7 @@ static int ovs_flow_cmd_del(struct sk_buff *skb, struct genl_info *info)
 	if (err)
 		goto unlock;
 
-	flow = ovs_flow_tbl_lookup(&dp->table, &key);
+	flow = __ovs_flow_tbl_lookup(&dp->table, &key);
 	if (!flow || !ovs_flow_cmp_unmasked_key(flow, &match)) {
 		err = -ENOENT;
 		goto unlock;
@@ -1067,6 +1083,7 @@ static size_t ovs_dp_cmd_msg_size(void)
 
 	msgsize += nla_total_size(IFNAMSIZ);
 	msgsize += nla_total_size(sizeof(struct ovs_dp_stats));
+	msgsize += nla_total_size(sizeof(struct ovs_dp_megaflow_stats));
 
 	return msgsize;
 }
@@ -1076,6 +1093,7 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
 {
 	struct ovs_header *ovs_header;
 	struct ovs_dp_stats dp_stats;
+	struct ovs_dp_megaflow_stats dp_megaflow_stats;
 	int err;
 
 	ovs_header = genlmsg_put(skb, portid, seq, &dp_datapath_genl_family,
@@ -1091,8 +1109,14 @@ static int ovs_dp_cmd_fill_info(struct datapath *dp, struct sk_buff *skb,
 	if (err)
 		goto nla_put_failure;
 
-	get_dp_stats(dp, &dp_stats);
-	if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats), &dp_stats))
+	get_dp_stats(dp, &dp_stats, &dp_megaflow_stats);
+	if (nla_put(skb, OVS_DP_ATTR_STATS, sizeof(struct ovs_dp_stats),
+			&dp_stats))
+		goto nla_put_failure;
+
+	if (nla_put(skb, OVS_DP_ATTR_MEGAFLOW_STATS,
+			sizeof(struct ovs_dp_megaflow_stats),
+			&dp_megaflow_stats))
 		goto nla_put_failure;
 
 	return genlmsg_end(skb, ovs_header);
net/openvswitch/datapath.h +4 -0
index acfd4af..d3d14a58 100644
@@ -46,11 +46,15 @@
  * @n_lost: Number of received packets that had no matching flow in the flow
  * table that could not be sent to userspace (normally due to an overflow in
  * one of the datapath's queues).
+ * @n_mask_hit: Number of masks looked up for flow match.
+ *   @n_mask_hit / (@n_hit + @n_missed)  will be the average masks looked
+ *   up per packet.
  */
 struct dp_stats_percpu {
 	u64 n_hit;
 	u64 n_missed;
 	u64 n_lost;
+	u64 n_mask_hit;
 	struct u64_stats_sync sync;
 };
 
net/openvswitch/flow_table.c +15 -1
index 036e019..536b4d2 100644
@@ -430,13 +430,16 @@ static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
 }
 
 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
-				    const struct sw_flow_key *key)
+				    const struct sw_flow_key *key,
+				    u32 *n_mask_hit)
 {
 	struct table_instance *ti = rcu_dereference(tbl->ti);
 	struct sw_flow_mask *mask;
 	struct sw_flow *flow;
 
+	*n_mask_hit = 0;
 	list_for_each_entry_rcu(mask, &tbl->mask_list, list) {
+		(*n_mask_hit)++;
 		flow = masked_flow_lookup(ti, key, mask);
 		if (flow)  /* Found */
 			return flow;
@@ -444,6 +447,17 @@ struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *tbl,
 	return NULL;
 }
 
+int ovs_flow_tbl_num_masks(const struct flow_table *table)
+{
+	struct sw_flow_mask *mask;
+	int num = 0;
+
+	list_for_each_entry(mask, &table->mask_list, list)
+		num++;
+
+	return num;
+}
+
 static struct table_instance *table_instance_expand(struct table_instance *ti)
 {
 	return table_instance_rehash(ti, ti->n_buckets * 2);
net/openvswitch/flow_table.h +3 -1
index 4db5f78..fbe45d5 100644
@@ -66,10 +66,12 @@ int ovs_flow_tbl_flush(struct flow_table *flow_table);
 int ovs_flow_tbl_insert(struct flow_table *table, struct sw_flow *flow,
 			struct sw_flow_mask *mask);
 void ovs_flow_tbl_remove(struct flow_table *table, struct sw_flow *flow);
+int  ovs_flow_tbl_num_masks(const struct flow_table *table);
 struct sw_flow *ovs_flow_tbl_dump_next(struct table_instance *table,
 				       u32 *bucket, u32 *idx);
 struct sw_flow *ovs_flow_tbl_lookup(struct flow_table *,
-				    const struct sw_flow_key *);
+				    const struct sw_flow_key *,
+				    u32 *n_mask_hit);
 
 bool ovs_flow_cmp_unmasked_key(const struct sw_flow *flow,
 			       struct sw_flow_match *match);
Signature · 1 line
1.8.3.2
#9 lore raw

PATCH 8/11 openvswitch: Enable all GSO features on internal port.

In reply to Jesse Gross

From: Pravin B Shelar <[email protected]>

OVS already can handle all types of segmentation offloads that are supported by the kernel.
Following patch specifically enables UDP and IPV6 segmentation offloads.

---
 net/openvswitch/vport-internal_dev.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/openvswitch/vport-internal_dev.c +1 -1
index 98d3edb..729c687 100644
@@ -134,7 +134,7 @@ static void do_setup(struct net_device *netdev)
 	netdev->tx_queue_len = 0;
 
 	netdev->features = NETIF_F_LLTX | NETIF_F_SG | NETIF_F_FRAGLIST |
-			   NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_TSO;
+			   NETIF_F_HIGHDMA | NETIF_F_HW_CSUM | NETIF_F_GSO_SOFTWARE;
 
 	netdev->vlan_features = netdev->features;
 	netdev->features |= NETIF_F_HW_VLAN_CTAG_TX;
Signature · 1 line
1.8.3.2
#10 lore raw

PATCH 9/11 openvswitch: Widen TCP flags handling.

In reply to Jesse Gross

From: Jarno Rajahalme <[email protected]>

Widen TCP flags handling from 7 bits (uint8_t) to 12 bits (uint16_t). The kernel interface remains at 8 bits, which makes no functional difference now, as none of the higher bits is currently of interest to the userspace.

---
 net/openvswitch/datapath.c | 2 +-
 net/openvswitch/flow.c     | 8 +++-----
 net/openvswitch/flow.h     | 2 +-
 3 files changed, 5 insertions(+), 7 deletions(-)
net/openvswitch/datapath.c +1 -1
index 5bc5a4e..1408adc 100644
@@ -671,7 +671,7 @@ static int ovs_flow_cmd_fill_info(struct sw_flow *flow, struct datapath *dp,
 	used = flow->used;
 	stats.n_packets = flow->packet_count;
 	stats.n_bytes = flow->byte_count;
-	tcp_flags = flow->tcp_flags;
+	tcp_flags = (u8)ntohs(flow->tcp_flags);
 	spin_unlock_bh(&flow->lock);
 
 	if (used &&
net/openvswitch/flow.c +3 -5
index 617810f..b73c768 100644
@@ -58,19 +58,17 @@ u64 ovs_flow_used_time(unsigned long flow_jiffies)
 	return cur_ms - idle_ms;
 }
 
-#define TCP_FLAGS_OFFSET 13
-#define TCP_FLAG_MASK 0x3f
+#define TCP_FLAGS_BE16(tp) (*(__be16 *)&tcp_flag_word(tp) & htons(0x0FFF))
 
 void ovs_flow_used(struct sw_flow *flow, struct sk_buff *skb)
 {
-	u8 tcp_flags = 0;
+	__be16 tcp_flags = 0;
 
 	if ((flow->key.eth.type == htons(ETH_P_IP) ||
 	     flow->key.eth.type == htons(ETH_P_IPV6)) &&
 	    flow->key.ip.proto == IPPROTO_TCP &&
 	    likely(skb->len >= skb_transport_offset(skb) + sizeof(struct tcphdr))) {
-		u8 *tcp = (u8 *)tcp_hdr(skb);
-		tcp_flags = *(tcp + TCP_FLAGS_OFFSET) & TCP_FLAG_MASK;
+		tcp_flags = TCP_FLAGS_BE16(tcp_hdr(skb));
 	}
 
 	spin_lock(&flow->lock);
net/openvswitch/flow.h +1 -1
index 098fd1d..204e0cc 100644
@@ -158,7 +158,7 @@ struct sw_flow {
 	unsigned long used;	/* Last used time (in jiffies). */
 	u64 packet_count;	/* Number of packets matched. */
 	u64 byte_count;		/* Number of bytes matched. */
-	u8 tcp_flags;		/* Union of seen TCP flags. */
+	__be16 tcp_flags;	/* Union of seen TCP flags. */
 };
 
 struct arp_eth_header {
Signature · 1 line
1.8.3.2
#11 lore raw

PATCH 10/11 openvswitch: TCP flags matching support.

In reply to Jesse Gross

From: Jarno Rajahalme <[email protected]>

tcp_flags=flags/mask
Bitwise match on TCP flags. The flags and mask are 16-bit num‐
bers written in decimal or in hexadecimal prefixed by 0x. Each
1-bit in mask requires that the corresponding bit in port must match. Each 0-bit in mask causes the corresponding bit to be ignored.

TCP protocol currently defines 9 flag bits, and additional 3 bits are reserved (must be transmitted as zero), see RFCs 793, 3168, and 3540. The flag bits are, numbering from the least significant bit:

0: FIN No more data from sender.

1: SYN Synchronize sequence numbers.

2: RST Reset the connection.

3: PSH Push function.

4: ACK Acknowledgement field significant.

5: URG Urgent pointer field significant.

6: ECE ECN Echo.

7: CWR Congestion Windows Reduced.

8: NS Nonce Sum.

9-11: Reserved.

12-15: Not matchable, must be zero.

---
 include/uapi/linux/openvswitch.h |  1 +
 net/openvswitch/flow.c           |  2 ++
 net/openvswitch/flow.h           |  2 ++
 net/openvswitch/flow_netlink.c   | 31 +++++++++++++++++++++++++++++--
 4 files changed, 34 insertions(+), 2 deletions(-)
include/uapi/linux/openvswitch.h +1 -0
index 2cc4644..d120f9f 100644
@@ -271,6 +271,7 @@ enum ovs_key_attr {
 	OVS_KEY_ATTR_SKB_MARK,  /* u32 skb mark */
 	OVS_KEY_ATTR_TUNNEL,    /* Nested set of ovs_tunnel attributes */
 	OVS_KEY_ATTR_SCTP,      /* struct ovs_key_sctp */
+	OVS_KEY_ATTR_TCP_FLAGS,	/* be16 TCP flags. */
 
 #ifdef __KERNEL__
 	OVS_KEY_ATTR_IPV4_TUNNEL,  /* struct ovs_key_ipv4_tunnel */
net/openvswitch/flow.c +2 -0
index b73c768..b409f52 100644
@@ -428,6 +428,7 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 				struct tcphdr *tcp = tcp_hdr(skb);
 				key->ipv4.tp.src = tcp->source;
 				key->ipv4.tp.dst = tcp->dest;
+				key->ipv4.tp.flags = TCP_FLAGS_BE16(tcp);
 			}
 		} else if (key->ip.proto == IPPROTO_UDP) {
 			if (udphdr_ok(skb)) {
@@ -496,6 +497,7 @@ int ovs_flow_extract(struct sk_buff *skb, u16 in_port, struct sw_flow_key *key)
 				struct tcphdr *tcp = tcp_hdr(skb);
 				key->ipv6.tp.src = tcp->source;
 				key->ipv6.tp.dst = tcp->dest;
+				key->ipv6.tp.flags = TCP_FLAGS_BE16(tcp);
 			}
 		} else if (key->ip.proto == NEXTHDR_UDP) {
 			if (udphdr_ok(skb)) {
net/openvswitch/flow.h +2 -0
index 204e0cc..1510f51 100644
@@ -93,6 +93,7 @@ struct sw_flow_key {
 				struct {
 					__be16 src;		/* TCP/UDP/SCTP source port. */
 					__be16 dst;		/* TCP/UDP/SCTP destination port. */
+					__be16 flags;		/* TCP flags. */
 				} tp;
 				struct {
 					u8 sha[ETH_ALEN];	/* ARP source hardware address. */
@@ -109,6 +110,7 @@ struct sw_flow_key {
 			struct {
 				__be16 src;		/* TCP/UDP/SCTP source port. */
 				__be16 dst;		/* TCP/UDP/SCTP destination port. */
+				__be16 flags;		/* TCP flags. */
 			} tp;
 			struct {
 				struct in6_addr target;	/* ND target address. */
net/openvswitch/flow_netlink.c +29 -2
index e04649c..2bc1bc1 100644
@@ -114,6 +114,7 @@ static bool match_validate(const struct sw_flow_match *match,
 	mask_allowed &= ~((1 << OVS_KEY_ATTR_IPV4)
 			| (1 << OVS_KEY_ATTR_IPV6)
 			| (1 << OVS_KEY_ATTR_TCP)
+			| (1 << OVS_KEY_ATTR_TCP_FLAGS)
 			| (1 << OVS_KEY_ATTR_UDP)
 			| (1 << OVS_KEY_ATTR_SCTP)
 			| (1 << OVS_KEY_ATTR_ICMP)
@@ -154,8 +155,11 @@ static bool match_validate(const struct sw_flow_match *match,
 
 			if (match->key->ip.proto == IPPROTO_TCP) {
 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
-				if (match->mask && (match->mask->key.ip.proto == 0xff))
+				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
+				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
+					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
+				}
 			}
 
 			if (match->key->ip.proto == IPPROTO_ICMP) {
@@ -186,8 +190,11 @@ static bool match_validate(const struct sw_flow_match *match,
 
 			if (match->key->ip.proto == IPPROTO_TCP) {
 				key_expected |= 1 << OVS_KEY_ATTR_TCP;
-				if (match->mask && (match->mask->key.ip.proto == 0xff))
+				key_expected |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
+				if (match->mask && (match->mask->key.ip.proto == 0xff)) {
 					mask_allowed |= 1 << OVS_KEY_ATTR_TCP;
+					mask_allowed |= 1 << OVS_KEY_ATTR_TCP_FLAGS;
+				}
 			}
 
 			if (match->key->ip.proto == IPPROTO_ICMPV6) {
@@ -235,6 +242,7 @@ static const int ovs_key_lens[OVS_KEY_ATTR_MAX + 1] = {
 	[OVS_KEY_ATTR_IPV4] = sizeof(struct ovs_key_ipv4),
 	[OVS_KEY_ATTR_IPV6] = sizeof(struct ovs_key_ipv6),
 	[OVS_KEY_ATTR_TCP] = sizeof(struct ovs_key_tcp),
+	[OVS_KEY_ATTR_TCP_FLAGS] = sizeof(__be16),
 	[OVS_KEY_ATTR_UDP] = sizeof(struct ovs_key_udp),
 	[OVS_KEY_ATTR_SCTP] = sizeof(struct ovs_key_sctp),
 	[OVS_KEY_ATTR_ICMP] = sizeof(struct ovs_key_icmp),
@@ -634,6 +642,19 @@ static int ovs_key_from_nlattrs(struct sw_flow_match *match,  u64 attrs,
 		attrs &= ~(1 << OVS_KEY_ATTR_TCP);
 	}
 
+	if (attrs & (1 << OVS_KEY_ATTR_TCP_FLAGS)) {
+		if (orig_attrs & (1 << OVS_KEY_ATTR_IPV4)) {
+			SW_FLOW_KEY_PUT(match, ipv4.tp.flags,
+					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
+					is_mask);
+		} else {
+			SW_FLOW_KEY_PUT(match, ipv6.tp.flags,
+					nla_get_be16(a[OVS_KEY_ATTR_TCP_FLAGS]),
+					is_mask);
+		}
+		attrs &= ~(1 << OVS_KEY_ATTR_TCP_FLAGS);
+	}
+
 	if (attrs & (1 << OVS_KEY_ATTR_UDP)) {
 		const struct ovs_key_udp *udp_key;
 
@@ -1004,9 +1025,15 @@ int ovs_nla_put_flow(const struct sw_flow_key *swkey,
 			if (swkey->eth.type == htons(ETH_P_IP)) {
 				tcp_key->tcp_src = output->ipv4.tp.src;
 				tcp_key->tcp_dst = output->ipv4.tp.dst;
+				if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
+						 output->ipv4.tp.flags))
+					goto nla_put_failure;
 			} else if (swkey->eth.type == htons(ETH_P_IPV6)) {
 				tcp_key->tcp_src = output->ipv6.tp.src;
 				tcp_key->tcp_dst = output->ipv6.tp.dst;
+				if (nla_put_be16(skb, OVS_KEY_ATTR_TCP_FLAGS,
+						 output->ipv6.tp.flags))
+					goto nla_put_failure;
 			}
 		} else if (swkey->ip.proto == IPPROTO_UDP) {
 			struct ovs_key_udp *udp_key;
Signature · 1 line
1.8.3.2
#12 lore raw

PATCH 11/11 openvswitch: Use flow hash during flow lookup operation.

In reply to Jesse Gross

From: Pravin B Shelar <[email protected]>

Flow->hash can be used to detect hash collisions and avoid flow key compare in flow lookup.

---
 net/openvswitch/flow_table.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/openvswitch/flow_table.c +1 -1
index 536b4d2..e425427 100644
@@ -421,7 +421,7 @@ static struct sw_flow *masked_flow_lookup(struct table_instance *ti,
 	hash = flow_hash(&masked_key, key_start, key_end);
 	head = find_bucket(ti, hash);
 	hlist_for_each_entry_rcu(flow, head, hash_node[ti->node_ver]) {
-		if (flow->mask == mask &&
+		if (flow->mask == mask && flow->hash == hash &&
 		    flow_cmp_masked_key(flow, &masked_key,
 					  key_start, key_end))
 			return flow;
Signature · 1 line
1.8.3.2
maintainer
#13 lore raw
In reply to Jesse Gross

From: Jesse Gross <[email protected]>
Date: Sat, 2 Nov 2013 00:43:39 -0700

Quoted 8 lines

A set of updates for net-next/3.13. Major changes are:
* Restructure flow handling code to be more logically organized and easier to read.
* Rehashing of the flow table is moved from a workqueue to flow installation time. Before, heavy load could block the workqueue for excessive periods of time.
* Additional debugging information is provided to help diagnose megaflows.
* It's now possible to match on TCP flags.

Looks good, pulled, thanks Jesse.