vswitchd: Make the maximum size of MAC learning tables user-configurable.

We've had a couple of requests for this over the years.  It's easy to do,
so let's implement it.

Signed-off-by: Ben Pfaff <blp@nicira.com>
Acked-by: Kyle Mestery <kmestery@cisco.com>

diff --git a/NEWS b/NEWS
index e1d4070..2255805 100644 (file)
--- a/NEWS
+++ b/NEWS
@@ -1,5 +1,6 @@
 post-v1.9.0
 --------------------
+    - The maximum size of the MAC learning table is now configurable.
     - New support for the VXLAN tunnel protocol (see the IETF draft here:
       http://tools.ietf.org/html/draft-mahalingam-dutt-dcops-vxlan-02).
     - New "vlog/disable-rate-limit" and "vlog/enable-rate-limit" commands

diff --git a/lib/mac-learning.c b/lib/mac-learning.c
index 3c541af..f609d48 100644 (file)
--- a/lib/mac-learning.c
+++ b/lib/mac-learning.c
@@ -110,7 +110,8 @@ normalize_idle_time(unsigned int idle_time)
 }
 
 /* Creates and returns a new MAC learning table with an initial MAC aging
- * timeout of 'idle_time' seconds. */
+ * timeout of 'idle_time' seconds and an initial maximum of MAC_DEFAULT_MAX
+ * entries. */
 struct mac_learning *
 mac_learning_create(unsigned int idle_time)
 {
@@ -122,6 +123,7 @@ mac_learning_create(unsigned int idle_time)
     ml->secret = random_uint32();
     ml->flood_vlans = NULL;
     ml->idle_time = normalize_idle_time(idle_time);
+    ml->max_entries = MAC_DEFAULT_MAX;
     return ml;
 }
 
@@ -176,6 +178,16 @@ mac_learning_set_idle_time(struct mac_learning *ml, unsigned int idle_time)
     }
 }
 
+/* Sets the maximum number of entries in 'ml' to 'max_entries', adjusting it
+ * to be within a reasonable range. */
+void
+mac_learning_set_max_entries(struct mac_learning *ml, size_t max_entries)
+{
+    ml->max_entries = (max_entries < 10 ? 10
+                       : max_entries > 1000 * 1000 ? 1000 * 1000
+                       : max_entries);
+}
+
 static bool
 is_learning_vlan(const struct mac_learning *ml, uint16_t vlan)
 {
@@ -212,7 +224,7 @@ mac_learning_insert(struct mac_learning *ml,
     if (!e) {
         uint32_t hash = mac_table_hash(ml, src_mac, vlan);
 
-        if (hmap_count(&ml->table) >= MAC_MAX) {
+        if (hmap_count(&ml->table) >= ml->max_entries) {
             get_lru(ml, &e);
             mac_learning_expire(ml, e);
         }
@@ -311,7 +323,9 @@ void
 mac_learning_run(struct mac_learning *ml, struct tag_set *set)
 {
     struct mac_entry *e;
-    while (get_lru(ml, &e) && time_now() >= e->expires) {
+    while (get_lru(ml, &e)
+           && (hmap_count(&ml->table) > ml->max_entries
+               || time_now() >= e->expires)) {
         COVERAGE_INC(mac_learning_expired);
         if (set) {
             tag_set_add(set, e->tag);
@@ -323,7 +337,9 @@ mac_learning_run(struct mac_learning *ml, struct tag_set *set)
 void
 mac_learning_wait(struct mac_learning *ml)
 {
-    if (!list_is_empty(&ml->lrus)) {
+    if (hmap_count(&ml->table) > ml->max_entries) {
+        poll_immediate_wake();
+    } else if (!list_is_empty(&ml->lrus)) {
         struct mac_entry *e = mac_entry_from_lru_node(ml->lrus.next);
         poll_timer_wait_until(e->expires * 1000LL);
     }

diff --git a/lib/mac-learning.h b/lib/mac-learning.h
index 8f8fd45..284e7f6 100644 (file)
--- a/lib/mac-learning.h
+++ b/lib/mac-learning.h
@@ -26,7 +26,8 @@
 
 struct mac_learning;
 
-#define MAC_MAX 2048
+/* Default maximum size of a MAC learning table, in entries. */
+#define MAC_DEFAULT_MAX 2048
 
 /* Time, in seconds, before expiring a mac_entry due to inactivity. */
 #define MAC_ENTRY_DEFAULT_IDLE_TIME 300
@@ -83,6 +84,7 @@ struct mac_learning {
     uint32_t secret;            /* Secret for randomizing hash table. */
     unsigned long *flood_vlans; /* Bitmap of learning disabled VLANs. */
     unsigned int idle_time;     /* Max age before deleting an entry. */
+    size_t max_entries;         /* Max number of learned MACs. */
 };
 
 /* Basics. */
@@ -96,6 +98,7 @@ void mac_learning_wait(struct mac_learning *);
 bool mac_learning_set_flood_vlans(struct mac_learning *,
                                   const unsigned long *bitmap);
 void mac_learning_set_idle_time(struct mac_learning *, unsigned int idle_time);
+void mac_learning_set_max_entries(struct mac_learning *, size_t max_entries);
 
 /* Learning. */
 bool mac_learning_may_learn(const struct mac_learning *,

diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
index d32f6d1..ca0a065 100644 (file)
--- a/ofproto/ofproto-dpif.c
+++ b/ofproto/ofproto-dpif.c
@@ -2762,10 +2762,12 @@ forward_bpdu_changed(struct ofproto *ofproto_)
 }
 
 static void
-set_mac_idle_time(struct ofproto *ofproto_, unsigned int idle_time)
+set_mac_table_config(struct ofproto *ofproto_, unsigned int idle_time,
+                     size_t max_entries)
 {
     struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
     mac_learning_set_idle_time(ofproto->ml, idle_time);
+    mac_learning_set_max_entries(ofproto->ml, max_entries);
 }
 \f
 /* Ports. */
@@ -7940,6 +7942,6 @@ const struct ofproto_class ofproto_dpif_class = {
     set_flood_vlans,
     is_mirror_output_bundle,
     forward_bpdu_changed,
-    set_mac_idle_time,
+    set_mac_table_config,
     set_realdev,
 };

diff --git a/ofproto/ofproto-provider.h b/ofproto/ofproto-provider.h
index 8f6f6b1..bc0105f 100644 (file)
--- a/ofproto/ofproto-provider.h
+++ b/ofproto/ofproto-provider.h
@@ -1279,9 +1279,14 @@ struct ofproto_class {
      * will be invoked. */
     void (*forward_bpdu_changed)(struct ofproto *ofproto);
 
-    /* Sets the MAC aging timeout for the OFPP_NORMAL action to 'idle_time',
-     * in seconds. */
-    void (*set_mac_idle_time)(struct ofproto *ofproto, unsigned int idle_time);
+    /* Sets the MAC aging timeout for the OFPP_NORMAL action to 'idle_time', in
+     * seconds, and the maximum number of MAC table entries to
+     * 'max_entries'.
+     *
+     * An implementation that doesn't support configuring these features may
+     * set this function to NULL or implement it as a no-op. */
+    void (*set_mac_table_config)(struct ofproto *ofproto,
+                                 unsigned int idle_time, size_t max_entries);
 
 /* Linux VLAN device support (e.g. "eth0.10" for VLAN 10.)
  *

diff --git a/ofproto/ofproto.c b/ofproto/ofproto.c
index 1130eb7..504b422 100644 (file)
--- a/ofproto/ofproto.c
+++ b/ofproto/ofproto.c
@@ -585,12 +585,15 @@ ofproto_set_forward_bpdu(struct ofproto *ofproto, bool forward_bpdu)
 }
 
 /* Sets the MAC aging timeout for the OFPP_NORMAL action on 'ofproto' to
- * 'idle_time', in seconds. */
+ * 'idle_time', in seconds, and the maximum number of MAC table entries to
+ * 'max_entries'. */
 void
-ofproto_set_mac_idle_time(struct ofproto *ofproto, unsigned idle_time)
+ofproto_set_mac_table_config(struct ofproto *ofproto, unsigned idle_time,
+                             size_t max_entries)
 {
-    if (ofproto->ofproto_class->set_mac_idle_time) {
-        ofproto->ofproto_class->set_mac_idle_time(ofproto, idle_time);
+    if (ofproto->ofproto_class->set_mac_table_config) {
+        ofproto->ofproto_class->set_mac_table_config(ofproto, idle_time,
+                                                     max_entries);
     }
 }
 

diff --git a/ofproto/ofproto.h b/ofproto/ofproto.h
index e4abe52..dc5d9ce 100644 (file)
--- a/ofproto/ofproto.h
+++ b/ofproto/ofproto.h
@@ -229,7 +229,8 @@ void ofproto_set_extra_in_band_remotes(struct ofproto *,
 void ofproto_set_in_band_queue(struct ofproto *, int queue_id);
 void ofproto_set_flow_eviction_threshold(struct ofproto *, unsigned threshold);
 void ofproto_set_forward_bpdu(struct ofproto *, bool forward_bpdu);
-void ofproto_set_mac_idle_time(struct ofproto *, unsigned idle_time);
+void ofproto_set_mac_table_config(struct ofproto *, unsigned idle_time,
+                                  size_t max_entries);
 void ofproto_set_desc(struct ofproto *,
                       const char *mfr_desc, const char *hw_desc,
                       const char *sw_desc, const char *serial_desc,

diff --git a/tests/ofproto-dpif.at b/tests/ofproto-dpif.at
index 6693f1b..6a4dc23 100644 (file)
--- a/tests/ofproto-dpif.at
+++ b/tests/ofproto-dpif.at
@@ -930,6 +930,68 @@ AT_CHECK_UNQUOTED([ovs-appctl fdb/show br1 | sed 's/[[0-9]]\{1,\}$/?/'], [0], [d
 OVS_VSWITCHD_STOP
 AT_CLEANUP
 
+AT_SETUP([ofproto-dpif - MAC table overflow])
+OVS_VSWITCHD_START(
+  [set bridge br0 fail-mode=standalone other-config:mac-table-size=10 -- \
+   add-port br0 p1 -- set Interface p1 type=dummy -- \
+   add-port br0 p2 -- set Interface p2 type=dummy -- \
+   add-port br0 p3 -- set Interface p3 type=dummy])
+
+arp='eth_type(0x0806),arp(sip=192.168.0.1,tip=192.168.0.2,op=1,sha=50:54:00:00:00:05,tha=00:00:00:00:00:00)'
+
+AT_CHECK([ovs-appctl time/stop])
+
+# Trace 10 ARP packets arriving on p3, to create MAC learning entries.
+for i in 0 1 2 3 4 5 6 7 8 9; do
+    OFPROTO_TRACE(
+      [br0],
+      [in_port(3),eth(src=50:54:00:00:00:0$i,dst=ff:ff:ff:ff:ff:ff),$arp],
+      [-generate],
+      [1,2,100])
+    ovs-appctl time/warp 1000
+done
+
+# Check for the MAC learning entries.
+AT_CHECK_UNQUOTED([ovs-appctl fdb/show br0 | sed 's/ *[[0-9]]\{1,\}$//' | sort],
+  [0], [dnl
+    3     0  50:54:00:00:00:00
+    3     0  50:54:00:00:00:01
+    3     0  50:54:00:00:00:02
+    3     0  50:54:00:00:00:03
+    3     0  50:54:00:00:00:04
+    3     0  50:54:00:00:00:05
+    3     0  50:54:00:00:00:06
+    3     0  50:54:00:00:00:07
+    3     0  50:54:00:00:00:08
+    3     0  50:54:00:00:00:09
+ port  VLAN  MAC                Age
+])
+
+# Trace another ARP packet on another MAC.
+OFPROTO_TRACE(
+  [br0],
+  [in_port(3),eth(src=50:54:00:00:00:10,dst=ff:ff:ff:ff:ff:ff),$arp],
+  [-generate],
+  [1,2,100])
+
+# Check that the new one chased the oldest one out of the table.
+AT_CHECK_UNQUOTED([ovs-appctl fdb/show br0 | sed 's/[[0-9]]\{1,\}$/?/' | sort],
+  [0], [dnl
+    3     0  50:54:00:00:00:01    ?
+    3     0  50:54:00:00:00:02    ?
+    3     0  50:54:00:00:00:03    ?
+    3     0  50:54:00:00:00:04    ?
+    3     0  50:54:00:00:00:05    ?
+    3     0  50:54:00:00:00:06    ?
+    3     0  50:54:00:00:00:07    ?
+    3     0  50:54:00:00:00:08    ?
+    3     0  50:54:00:00:00:09    ?
+    3     0  50:54:00:00:00:10    ?
+ port  VLAN  MAC                Age
+])
+OVS_VSWITCHD_STOP
+AT_CLEANUP
+
 dnl Test that basic NetFlow reports flow statistics correctly:
 dnl - The initial packet of a flow are correctly accounted.
 dnl - Later packets within a flow are correctly accounted.

diff --git a/vswitchd/bridge.c b/vswitchd/bridge.c
index b1d2feb..53dd2f6 100644 (file)
--- a/vswitchd/bridge.c
+++ b/vswitchd/bridge.c
@@ -182,7 +182,7 @@ static void bridge_configure_datapath_id(struct bridge *);
 static void bridge_configure_flow_eviction_threshold(struct bridge *);
 static void bridge_configure_netflow(struct bridge *);
 static void bridge_configure_forward_bpdu(struct bridge *);
-static void bridge_configure_mac_idle_time(struct bridge *);
+static void bridge_configure_mac_table(struct bridge *);
 static void bridge_configure_sflow(struct bridge *, int *sflow_bridge_number);
 static void bridge_configure_stp(struct bridge *);
 static void bridge_configure_tables(struct bridge *);
@@ -591,7 +591,7 @@ bridge_reconfigure_continue(const struct ovsrec_open_vswitch *ovs_cfg)
         bridge_configure_mirrors(br);
         bridge_configure_flow_eviction_threshold(br);
         bridge_configure_forward_bpdu(br);
-        bridge_configure_mac_idle_time(br);
+        bridge_configure_mac_table(br);
         bridge_configure_remotes(br, managers, n_managers);
         bridge_configure_netflow(br);
         bridge_configure_sflow(br, &sflow_bridge_number);
@@ -1493,18 +1493,27 @@ bridge_configure_forward_bpdu(struct bridge *br)
                                            false));
 }
 
-/* Set MAC aging time for 'br'. */
+/* Set MAC learning table configuration for 'br'. */
 static void
-bridge_configure_mac_idle_time(struct bridge *br)
+bridge_configure_mac_table(struct bridge *br)
 {
     const char *idle_time_str;
     int idle_time;
 
+    const char *mac_table_size_str;
+    int mac_table_size;
+
     idle_time_str = smap_get(&br->cfg->other_config, "mac-aging-time");
     idle_time = (idle_time_str && atoi(idle_time_str)
                  ? atoi(idle_time_str)
                  : MAC_ENTRY_DEFAULT_IDLE_TIME);
-    ofproto_set_mac_idle_time(br->ofproto, idle_time);
+
+    mac_table_size_str = smap_get(&br->cfg->other_config, "mac-table-size");
+    mac_table_size = (mac_table_size_str && atoi(mac_table_size_str)
+                      ? atoi(mac_table_size_str)
+                      : MAC_DEFAULT_MAX);
+
+    ofproto_set_mac_table_config(br->ofproto, idle_time, mac_table_size);
 }
 
 static void

diff --git a/vswitchd/vswitch.xml b/vswitchd/vswitch.xml
index 31c15b4..c78899f 100644 (file)
--- a/vswitchd/vswitch.xml
+++ b/vswitchd/vswitch.xml
@@ -623,6 +623,15 @@
           transmit packets.
         </p>
       </column>
+
+      <column name="other_config" key="mac-table-size"
+              type='{"type": "integer", "minInteger": 1}'>
+        <p>
+          The maximum number of MAC addresses to learn.  The default is
+          currently 2048.  The value, if specified, is forced into a reasonable
+          range, currently 10 to 1,000,000.
+        </p>
+      </column>
     </group>
 
     <group title="Bridge Status">