ofproto-dpif: Don't poll ports when nothing changes

Previously, as part of ofproto-dpif run() processing, we would loop
through all ports and poll for changes to carrier, bfd, cfm and lacp
status. This information is used to determine whether bundles may be
enabled, and to perform revalidation when needed.

This patch makes the bfd, cfm, lacp and stp modules aware of the new
global connectivity_seq, notifying on changes in port status. We can
then use connectivity_seq to check if anything has changed before
looping through all ports in ofproto-dpif. In a test environment of 5000
internal ports and 50 tunnel ports with bfd, this reduces average CPU
usage of the main thread from about 35% to about 25%.

Signed-off-by: Joe Stringer <joestringer@nicira.com>
Signed-off-by: Ethan Jackson <ethan@nicira.com>
Acked-by: Ethan Jackson <ethan@nicira.com>

diff --git a/lib/bfd.c b/lib/bfd.c
index 1df5acd..5ac20d6 100644 (file)
--- a/lib/bfd.c
+++ b/lib/bfd.c
@@ -21,6 +21,7 @@
 #include <netinet/ip.h>
 
 #include "byte-order.h"
+#include "connectivity.h"
 #include "csum.h"
 #include "dpif.h"
 #include "dynamic-string.h"
@@ -37,6 +38,7 @@
 #include "packets.h"
 #include "poll-loop.h"
 #include "random.h"
+#include "seq.h"
 #include "smap.h"
 #include "timeval.h"
 #include "unaligned.h"
@@ -505,8 +507,8 @@ bfd_run(struct bfd *bfd) OVS_EXCLUDED(mutex)
 
     if (bfd->state > STATE_DOWN && now >= bfd->detect_time) {
         bfd_set_state(bfd, STATE_DOWN, DIAG_EXPIRED);
-        bfd_forwarding__(bfd);
     }
+    bfd_forwarding__(bfd);
 
     /* Decay may only happen when state is STATE_UP, bfd->decay_min_rx is
      * configured, and decay_detect_time is reached. */
@@ -851,6 +853,7 @@ bfd_forwarding__(struct bfd *bfd) OVS_REQUIRES(mutex)
                             && bfd->rmt_diag != DIAG_RCPATH_DOWN;
     if (bfd->last_forwarding != last_forwarding) {
         bfd->flap_count++;
+        seq_change(connectivity_seq_get());
     }
     return bfd->last_forwarding;
 }
@@ -1052,6 +1055,8 @@ bfd_set_state(struct bfd *bfd, enum state state, enum diag diag)
         if (bfd->state == STATE_UP && bfd->decay_min_rx) {
             bfd_decay_update(bfd);
         }
+
+        seq_change(connectivity_seq_get());
     }
 }
 

diff --git a/lib/cfm.c b/lib/cfm.c
index 9c65b34..8ba0c44 100644 (file)
--- a/lib/cfm.c
+++ b/lib/cfm.c
@@ -22,6 +22,7 @@
 #include <string.h>
 
 #include "byte-order.h"
+#include "connectivity.h"
 #include "dynamic-string.h"
 #include "flow.h"
 #include "hash.h"
@@ -31,6 +32,7 @@
 #include "packets.h"
 #include "poll-loop.h"
 #include "random.h"
+#include "seq.h"
 #include "timer.h"
 #include "timeval.h"
 #include "unixctl.h"
@@ -396,6 +398,7 @@ cfm_run(struct cfm *cfm) OVS_EXCLUDED(mutex)
         long long int interval = cfm_fault_interval(cfm);
         struct remote_mp *rmp, *rmp_next;
         bool old_cfm_fault = cfm->fault;
+        bool old_rmp_opup = cfm->remote_opup;
         bool demand_override;
         bool rmp_set_opup = false;
         bool rmp_set_opdown = false;
@@ -420,6 +423,7 @@ cfm_run(struct cfm *cfm) OVS_EXCLUDED(mutex)
                 cfm->health = 0;
             } else {
                 int exp_ccm_recvd;
+                int old_health = cfm->health;
 
                 rmp = CONTAINER_OF(hmap_first(&cfm->remote_mps),
                                    struct remote_mp, node);
@@ -434,6 +438,10 @@ cfm_run(struct cfm *cfm) OVS_EXCLUDED(mutex)
                 cfm->health = MIN(cfm->health, 100);
                 rmp->num_health_ccm = 0;
                 ovs_assert(cfm->health >= 0 && cfm->health <= 100);
+
+                if (cfm->health != old_health) {
+                    seq_change(connectivity_seq_get());
+                }
             }
             cfm->health_interval = 0;
         }
@@ -476,6 +484,10 @@ cfm_run(struct cfm *cfm) OVS_EXCLUDED(mutex)
             cfm->remote_opup = true;
         }
 
+        if (old_rmp_opup != cfm->remote_opup) {
+            seq_change(connectivity_seq_get());
+        }
+
         if (hmap_is_empty(&cfm->remote_mps)) {
             cfm->fault |= CFM_FAULT_RECV;
         }
@@ -497,6 +509,8 @@ cfm_run(struct cfm *cfm) OVS_EXCLUDED(mutex)
             if (old_cfm_fault == false || cfm->fault == false) {
                 cfm->flap_count++;
             }
+
+            seq_change(connectivity_seq_get());
         }
 
         cfm->booted = true;

diff --git a/lib/lacp.c b/lib/lacp.c
index fce65b3..07f1e10 100644 (file)
--- a/lib/lacp.c
+++ b/lib/lacp.c
@@ -18,12 +18,14 @@
 
 #include <stdlib.h>
 
+#include "connectivity.h"
 #include "dynamic-string.h"
 #include "hash.h"
 #include "hmap.h"
 #include "ofpbuf.h"
 #include "packets.h"
 #include "poll-loop.h"
+#include "seq.h"
 #include "shash.h"
 #include "timer.h"
 #include "timeval.h"
@@ -509,11 +511,16 @@ lacp_run(struct lacp *lacp, lacp_send_pdu *send_pdu) OVS_EXCLUDED(mutex)
     ovs_mutex_lock(&mutex);
     HMAP_FOR_EACH (slave, node, &lacp->slaves) {
         if (timer_expired(&slave->rx)) {
+            enum slave_status old_status = slave->status;
+
             if (slave->status == LACP_CURRENT) {
                 slave_set_expired(slave);
             } else if (slave->status == LACP_EXPIRED) {
                 slave_set_defaulted(slave);
             }
+            if (slave->status != old_status) {
+                seq_change(connectivity_seq_get());
+            }
         }
     }
 
@@ -544,6 +551,7 @@ lacp_run(struct lacp *lacp, lacp_send_pdu *send_pdu) OVS_EXCLUDED(mutex)
                         : LACP_SLOW_TIME_TX);
 
             timer_set_duration(&slave->tx, duration);
+            seq_change(connectivity_seq_get());
         }
     }
     ovs_mutex_unlock(&mutex);

diff --git a/lib/stp.c b/lib/stp.c
index 725bfff..8b9d26a 100644 (file)
--- a/lib/stp.c
+++ b/lib/stp.c
@@ -26,8 +26,10 @@
 #include <inttypes.h>
 #include <stdlib.h>
 #include "byte-order.h"
+#include "connectivity.h"
 #include "ofpbuf.h"
 #include "packets.h"
+#include "seq.h"
 #include "unixctl.h"
 #include "util.h"
 #include "vlog.h"
@@ -1127,6 +1129,7 @@ stp_configuration_update(struct stp *stp) OVS_REQUIRES(mutex)
 {
     stp_root_selection(stp);
     stp_designated_port_selection(stp);
+    seq_change(connectivity_seq_get());
 }
 
 static bool
@@ -1257,6 +1260,7 @@ stp_set_port_state(struct stp_port *p, enum stp_state state)
         if (p < p->stp->first_changed_port) {
             p->stp->first_changed_port = p;
         }
+        seq_change(connectivity_seq_get());
     }
     p->state = state;
 }
@@ -1275,6 +1279,7 @@ stp_topology_change_detection(struct stp *stp) OVS_REQUIRES(mutex)
     }
     stp->fdb_needs_flush = true;
     stp->topology_change_detected = true;
+    seq_change(connectivity_seq_get());
     VLOG_INFO_RL(&rl, "%s: detected topology change.", stp->name);
 }
 

diff --git a/ofproto/ofproto-dpif.c b/ofproto/ofproto-dpif.c
index 086f676..616aa3c 100644 (file)
--- a/ofproto/ofproto-dpif.c
+++ b/ofproto/ofproto-dpif.c
@@ -25,6 +25,7 @@
 #include "bond.h"
 #include "bundle.h"
 #include "byte-order.h"
+#include "connectivity.h"
 #include "connmgr.h"
 #include "coverage.h"
 #include "cfm.h"
@@ -57,6 +58,7 @@
 #include "ofproto-dpif-upcall.h"
 #include "ofproto-dpif-xlate.h"
 #include "poll-loop.h"
+#include "seq.h"
 #include "simap.h"
 #include "smap.h"
 #include "timer.h"
@@ -505,6 +507,7 @@ struct ofproto_dpif {
     struct sset ghost_ports;       /* Ports with no datapath port. */
     struct sset port_poll_set;     /* Queued names for port_poll() reply. */
     int port_poll_errno;           /* Last errno for port_poll() reply. */
+    uint64_t change_seq;           /* Connectivity status changes. */
 
     /* Per ofproto's dpif stats. */
     uint64_t n_hit;
@@ -1262,6 +1265,7 @@ construct(struct ofproto *ofproto_)
     sset_init(&ofproto->ghost_ports);
     sset_init(&ofproto->port_poll_set);
     ofproto->port_poll_errno = 0;
+    ofproto->change_seq = 0;
 
     SHASH_FOR_EACH_SAFE (node, next, &init_ofp_ports) {
         struct iface_hint *iface_hint = node->data;
@@ -1467,8 +1471,8 @@ static int
 run(struct ofproto *ofproto_)
 {
     struct ofproto_dpif *ofproto = ofproto_dpif_cast(ofproto_);
-    struct ofport_dpif *ofport;
     struct ofbundle *bundle;
+    uint64_t new_seq;
     int error;
 
     if (mbridge_need_revalidate(ofproto->mbridge)) {
@@ -1501,8 +1505,15 @@ run(struct ofproto *ofproto_)
         dpif_ipfix_run(ofproto->ipfix);
     }
 
-    HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
-        port_run(ofport);
+    new_seq = seq_read(connectivity_seq_get());
+    if (ofproto->change_seq != new_seq) {
+        struct ofport_dpif *ofport;
+
+        HMAP_FOR_EACH (ofport, up.hmap_node, &ofproto->up.ports) {
+            port_run(ofport);
+        }
+
+        ofproto->change_seq = new_seq;
     }
     HMAP_FOR_EACH (bundle, hmap_node, &ofproto->bundles) {
         bundle_run(bundle);