[ovs-dev] [PATCH ovn v4 5/5] controller: Improve ct zone handling.

Odintsov Vladislav VlOdintsov at croc.ru
Sat Aug 21 08:10:32 UTC 2021


Thanks Numan for the quick fix.

Though you’ve already sent a patch, just in case info:
I was using a normal distributed router, LRP with no gateway chassis.
Codebase - master branch.

Regards,
Vladislav Odintsov

On 21 Aug 2021, at 02:37, Numan Siddique <numans at ovn.org<mailto:numans at ovn.org>> wrote:

On Fri, Aug 20, 2021 at 7:28 PM Numan Siddique <numans at ovn.org<mailto:numans at ovn.org>> wrote:

On Fri, Aug 20, 2021 at 7:19 PM Odintsov Vladislav <VlOdintsov at croc.ru<mailto:VlOdintsov at croc.ru>> wrote:

Hi Numan,

It looks like this patch introduces a regression.
In my scenario with this commit my quite complicated virtual topology is non-working and I have to call recompute to make things working.
Debugging my problem I’ve found next problem:

If we create LR and LRP, we can see that for this LRP a ct_zone id is allocated (ovn-appctl -t ovn-controller ct-zone-list).
If we call ovn-appctl -t ovn-controller recompute, this zone id is deleted from the list.

Recompute is also solves a problem in my setup.
I’ve tried to revert commit [1] and zone id for LRP was no longer allocated.
I tried to look through the code but couldn’t understand the reason for this behaviour.

Can you please take a look if there can be a possible bug?

[1] https://github.com/ovn-org/ovn/commit/6fb87aad8c353bda13910ecfccc0c8cfa5ac3fdf

Hi Vladislav,

Thanks for reporting this.  If recompute solves your problem then
definitely there is an I-P bug.

Can you give a bit more details ?  Is this LR a normal router ? or
gateway router ?
Is there a gateway router port associated with the LR ?

Is there a script to reproduce the issue ?  I'll try to reproduce this locally.

Are you using OVN 21.06 ?  or master ?  Is it possible to try it out
with the master ?

I was able to reproduce the issue.  I'll submit the fix.

Thanks
Numan


Thanks
Numan


Regards,
Vladislav Odintsov

On 2 Aug 2021, at 23:49, numans at ovn.org<mailto:numans at ovn.org> wrote:

From: Numan Siddique <numans at ovn.org<mailto:numans at ovn.org>>

Prior to this patch, ovn-controller generates a zone id for each
OVS interface which has external_ids:iface-id set even if there is
no corresponding logical port for it.  This patch now changes the
way we allocate the zone id.  A zone id is allocated only if
there is an OVS interface with external_ids:iface-id and the
corresponding logical port is claimed.  We use the runtime data
(rt_data->lbinding_data.lports) for this.

This patch also improves the ct_zones_runtime_data_handler()
by using the tracked datapath data to allocate the zone ids
for newly claimed lports or to free up the zone id for
the released lports instead of triggering a full recompute.

And finally this patch also adds a ct zone handler for pflow_output
engine.  This handler falls back to recompute if the ct zone engine
was recomputed.  Otherwise it returns true.

Acked-by: Han Zhou <hzhou at ovn.org<mailto:hzhou at ovn.org>>
Reviewed-by: Mark Michelson <mmichels at redhat.com<mailto:mmichels at redhat.com>>
Signed-off-by: Numan Siddique <numans at ovn.org<mailto:numans at ovn.org>>
---
controller/ovn-controller.c | 135 +++++++++++++++++++++++++++---------
lib/inc-proc-eng.h          |   4 ++
tests/ovn.at<http://ovn.at>                |   2 +-
3 files changed, 108 insertions(+), 33 deletions(-)

diff --git a/controller/ovn-controller.c b/controller/ovn-controller.c
index d69147c04..ee28f2612 100644
--- a/controller/ovn-controller.c
+++ b/controller/ovn-controller.c
@@ -624,8 +624,32 @@ add_pending_ct_zone_entry(struct shash *pending_ct_zones,
   }
}

+static bool
+alloc_id_to_ct_zone(const char *zone_name, struct simap *ct_zones,
+                    unsigned long *ct_zone_bitmap, int *scan_start,
+                    struct shash *pending_ct_zones)
+{
+    /* We assume that there are 64K zones and that we own them all. */
+    int zone = bitmap_scan(ct_zone_bitmap, 0, *scan_start, MAX_CT_ZONES + 1);
+    if (zone == MAX_CT_ZONES + 1) {
+        static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
+        VLOG_WARN_RL(&rl, "exhausted all ct zones");
+        return false;
+    }
+
+    *scan_start = zone + 1;
+
+    add_pending_ct_zone_entry(pending_ct_zones, CT_ZONE_OF_QUEUED,
+                              zone, true, zone_name);
+
+    bitmap_set1(ct_zone_bitmap, zone);
+    simap_put(ct_zones, zone_name, zone);
+    return true;
+}
+
static void
-update_ct_zones(const struct sset *lports, const struct hmap *local_datapaths,
+update_ct_zones(const struct shash *binding_lports,
+                const struct hmap *local_datapaths,
               struct simap *ct_zones, unsigned long *ct_zone_bitmap,
               struct shash *pending_ct_zones)
{
@@ -636,8 +660,9 @@ update_ct_zones(const struct sset *lports, const struct hmap *local_datapaths,
   struct simap req_snat_zones = SIMAP_INITIALIZER(&req_snat_zones);
   unsigned long unreq_snat_zones[BITMAP_N_LONGS(MAX_CT_ZONES)];

-    SSET_FOR_EACH(user, lports) {
-        sset_add(&all_users, user);
+    struct shash_node *shash_node;
+    SHASH_FOR_EACH (shash_node, binding_lports) {
+        sset_add(&all_users, shash_node->name);
   }

   /* Local patched datapath (gateway routers) need zones assigned. */
@@ -725,26 +750,12 @@ update_ct_zones(const struct sset *lports, const struct hmap *local_datapaths,
   /* Assign a unique zone id for each logical port and two zones
    * to a gateway router. */
   SSET_FOR_EACH(user, &all_users) {
-        int zone;
-
       if (simap_contains(ct_zones, user)) {
           continue;
       }

-        /* We assume that there are 64K zones and that we own them all. */
-        zone = bitmap_scan(ct_zone_bitmap, 0, scan_start, MAX_CT_ZONES + 1);
-        if (zone == MAX_CT_ZONES + 1) {
-            static struct vlog_rate_limit rl = VLOG_RATE_LIMIT_INIT(1, 1);
-            VLOG_WARN_RL(&rl, "exhausted all ct zones");
-            break;
-        }
-        scan_start = zone + 1;
-
-        add_pending_ct_zone_entry(pending_ct_zones, CT_ZONE_OF_QUEUED,
-                                  zone, true, user);
-
-        bitmap_set1(ct_zone_bitmap, zone);
-        simap_put(ct_zones, user, zone);
+        alloc_id_to_ct_zone(user, ct_zones, ct_zone_bitmap, &scan_start,
+                            pending_ct_zones);
   }

   simap_destroy(&req_snat_zones);
@@ -1769,6 +1780,9 @@ struct ed_type_ct_zones {
   unsigned long bitmap[BITMAP_N_LONGS(MAX_CT_ZONES)];
   struct shash pending;
   struct simap current;
+
+    /* Tracked data. */
+    bool recomputed;
};

static void *
@@ -1791,6 +1805,13 @@ en_ct_zones_init(struct engine_node *node, struct engine_arg *arg OVS_UNUSED)
   return data;
}

+static void
+en_ct_zones_clear_tracked_data(void *data_)
+{
+    struct ed_type_ct_zones *data = data_;
+    data->recomputed = false;
+}
+
static void
en_ct_zones_cleanup(void *data)
{
@@ -1807,11 +1828,12 @@ en_ct_zones_run(struct engine_node *node, void *data)
   struct ed_type_runtime_data *rt_data =
       engine_get_input_data("runtime_data", node);

-    update_ct_zones(&rt_data->local_lports, &rt_data->local_datapaths,
+    update_ct_zones(&rt_data->lbinding_data.lports, &rt_data->local_datapaths,
                   &ct_zones_data->current, ct_zones_data->bitmap,
                   &ct_zones_data->pending);


+    ct_zones_data->recomputed = true;
   engine_set_node_state(node, EN_UPDATED);
}

@@ -1869,7 +1891,7 @@ ct_zones_datapath_binding_handler(struct engine_node *node, void *data)
}

static bool
-ct_zones_runtime_data_handler(struct engine_node *node, void *data OVS_UNUSED)
+ct_zones_runtime_data_handler(struct engine_node *node, void *data)
{
   struct ed_type_runtime_data *rt_data =
       engine_get_input_data("runtime_data", node);
@@ -1879,24 +1901,53 @@ ct_zones_runtime_data_handler(struct engine_node *node, void *data OVS_UNUSED)
       return false;
   }

-    /* If local_lports have changed then fall back to full recompute. */
-    if (rt_data->local_lports_changed) {
-        return false;
-    }
+    struct ed_type_ct_zones *ct_zones_data = data;

   struct hmap *tracked_dp_bindings = &rt_data->tracked_dp_bindings;
   struct tracked_datapath *tdp;
+    int scan_start = 0;
+
+    bool updated = false;
+
   HMAP_FOR_EACH (tdp, node, tracked_dp_bindings) {
       if (tdp->tracked_type == TRACKED_RESOURCE_NEW) {
           /* A new datapath has been added. Fall back to full recompute. */
           return false;
       }

-        /* When an lport is claimed or released because of port binding,
-         * changes we don't have to compute the ct zone entries for these.
-         * That is because we generate the ct zone entries for each local
-         * OVS interface which has external_ids:iface-id set.  For the local
-         * OVS interface changes, rt_data->local_ports_changed will be true. */
+        struct shash_node *shash_node;
+        SHASH_FOR_EACH (shash_node, &tdp->lports) {
+            struct tracked_lport *t_lport = shash_node->data;
+            if (t_lport->tracked_type == TRACKED_RESOURCE_NEW) {
+                if (!simap_contains(&ct_zones_data->current,
+                                    t_lport->pb->logical_port)) {
+                    alloc_id_to_ct_zone(t_lport->pb->logical_port,
+                                        &ct_zones_data->current,
+                                        ct_zones_data->bitmap, &scan_start,
+                                        &ct_zones_data->pending);
+                    updated = true;
+                }
+            } else if (t_lport->tracked_type == TRACKED_RESOURCE_REMOVED) {
+                struct simap_node *ct_zone =
+                    simap_find(&ct_zones_data->current,
+                               t_lport->pb->logical_port);
+                if (ct_zone) {
+                    add_pending_ct_zone_entry(
+                        &ct_zones_data->pending, CT_ZONE_DB_QUEUED,
+                        ct_zone->data, false, ct_zone->name);
+
+                    bitmap_set0(ct_zones_data->bitmap, ct_zone->data);
+                    simap_delete(&ct_zones_data->current, ct_zone);
+                    updated = true;
+                }
+            } else {
+                OVS_NOT_REACHED();
+            }
+        }
+    }
+
+    if (updated) {
+        engine_set_node_state(node, EN_UPDATED);
   }

   return true;
@@ -2868,6 +2919,25 @@ pflow_output_runtime_data_handler(struct engine_node *node, void *data)
   return true;
}

+static bool
+pflow_output_ct_zones_handler(struct engine_node *node OVS_UNUSED,
+                                    void *data OVS_UNUSED)
+{
+    struct ed_type_ct_zones *ct_zones_data =
+        engine_get_input_data("ct_zones", node);
+
+    /* If ct_zones engine node was recomputed, then fall back to full
+     * recompute of pflow_output.  Otherwise there is no need to do
+     * anything for the following reasons:
+     *   - When an lport is claimed, ct zone handler for the
+     *     runtime_data handler allocates the zone id for the lport (and it is
+     *     saved in the br-int external_ids).
+     *   - pflow_output handler for the runtime data  adds
+     *     the physical flows for the claimed lport.
+     * */
+    return !ct_zones_data->recomputed;
+}
+
static void *
en_flow_output_init(struct engine_node *node OVS_UNUSED,
                   struct engine_arg *arg OVS_UNUSED)
@@ -3123,7 +3193,7 @@ main(int argc, char *argv[])
   stopwatch_create(BFD_RUN_STOPWATCH_NAME, SW_MS);

   /* Define inc-proc-engine nodes. */
-    ENGINE_NODE_CUSTOM_DATA(ct_zones, "ct_zones");
+    ENGINE_NODE_CUSTOM_WITH_CLEAR_TRACK_DATA(ct_zones, "ct_zones");
   ENGINE_NODE_WITH_CLEAR_TRACK_DATA(runtime_data, "runtime_data");
   ENGINE_NODE(non_vif_data, "non_vif_data");
   ENGINE_NODE(mff_ovn_geneve, "mff_ovn_geneve");
@@ -3164,7 +3234,8 @@ main(int argc, char *argv[])
    */
   engine_add_input(&en_pflow_output, &en_non_vif_data,
                    NULL);
-    engine_add_input(&en_pflow_output, &en_ct_zones, NULL);
+    engine_add_input(&en_pflow_output, &en_ct_zones,
+                     pflow_output_ct_zones_handler);
   engine_add_input(&en_pflow_output, &en_sb_chassis,
                    pflow_lflow_output_sb_chassis_handler);

diff --git a/lib/inc-proc-eng.h b/lib/inc-proc-eng.h
index 7e9f5bb70..859b30a71 100644
--- a/lib/inc-proc-eng.h
+++ b/lib/inc-proc-eng.h
@@ -302,6 +302,10 @@ void engine_ovsdb_node_add_index(struct engine_node *, const char *name,
#define ENGINE_NODE_CUSTOM_DATA(NAME, NAME_STR) \
   ENGINE_NODE_DEF(NAME, NAME_STR)

+#define ENGINE_NODE_CUSTOM_WITH_CLEAR_TRACK_DATA(NAME, NAME_STR) \
+    ENGINE_NODE_CUSTOM_DATA(NAME, NAME_STR) \
+    en_##NAME.clear_tracked_data = en_##NAME##_clear_tracked_data;
+
#define ENGINE_NODE(NAME, NAME_STR) \
   static bool (*en_##NAME##_is_valid)(struct engine_node *node) = NULL; \
   ENGINE_NODE_DEF(NAME, NAME_STR)
diff --git a/tests/ovn.at<http://ovn.at> b/tests/ovn.at<http://ovn.at>
index a207f5e12..7ae136ad9 100644
--- a/tests/ovn.at<http://ovn.at>
+++ b/tests/ovn.at<http://ovn.at>
@@ -23527,7 +23527,7 @@ OVS_WAIT_UNTIL([test $(ovs-ofctl dump-flows br-int table=38,metadata=${sw0_dpkey
reg15=0x${p1_dpkey} | grep REG13 | wc -l) -eq 0])

p1_zoneid=$(as hv1 ovs-vsctl get bridge br-int external_ids:ct-zone-sw0-p1 | sed 's/"//g')
-AT_CHECK([test ! -z $p1_zoneid])
+AT_CHECK([test -z $p1_zoneid])

OVN_CLEANUP([hv1])
AT_CLEANUP
--
2.31.1

_______________________________________________
dev mailing list
dev at openvswitch.org<mailto:dev at openvswitch.org>
https://mail.openvswitch.org/mailman/listinfo/ovs-dev

_______________________________________________
dev mailing list
dev at openvswitch.org<mailto:dev at openvswitch.org>
https://mail.openvswitch.org/mailman/listinfo/ovs-dev



More information about the dev mailing list