[master]staticroutebfd process implementation (#13789)

* [BFD] staticroutebfd implementation
* To enable the BFD for static route

HLD: sonic-net/SONiC#1216

Author: qiwang4

dockers/docker-fpm-frr/frr/supervisord/critical_processes.j2

@@ -9,4 +9,5 @@ program:pimd
 program:frrcfgd
 {%- else %}
 program:bgpcfgd
+program:staticroutebfd
 {%- endif %}

dockers/docker-fpm-frr/frr/supervisord/supervisord.conf.j2

@@ -141,6 +141,20 @@ stderr_logfile=syslog
 dependent_startup=true
 dependent_startup_wait_for=bgpd:running
 
+{% if DEVICE_METADATA.localhost.frr_mgmt_framework_config is defined and DEVICE_METADATA.localhost.frr_mgmt_framework_config == "true" %}
+{% else %}
+[program:staticroutebfd]
+command=/usr/local/bin/staticroutebfd
+priority=6
+autostart=false
+autorestart=false
+startsecs=0
+stdout_logfile=syslog
+stderr_logfile=syslog
+dependent_startup=true
+dependent_startup_wait_for=bgpd:running
+{% endif %}
+
 [program:bgpmon]
 command=/usr/local/bin/bgpmon
 priority=6

src/sonic-bgpcfgd/bgpcfgd/managers_static_rt.py

@@ -25,6 +25,7 @@ def __init__(self, common_objs, db, table):
         self.directory.subscribe([("CONFIG_DB", swsscommon.CFG_DEVICE_METADATA_TABLE_NAME, "localhost/bgp_asn"),], self.on_bgp_asn_change)
         self.static_routes = {}
         self.vrf_pending_redistribution = set()
+        self.config_db = None
 
     OP_DELETE = 'DELETE'
     OP_ADD = 'ADD'
@@ -41,7 +42,17 @@ def set_handler(self, key, data):
         intf_list   = arg_list(data['ifname']) if 'ifname' in data else None
         dist_list   = arg_list(data['distance']) if 'distance' in data else None
         nh_vrf_list = arg_list(data['nexthop-vrf']) if 'nexthop-vrf' in data else None
-        route_tag   = self.ROUTE_ADVERTISE_DISABLE_TAG if 'advertise' in data and data['advertise'] == "false" else self.ROUTE_ADVERTISE_ENABLE_TAG
+        bfd_enable  = arg_list(data['bfd']) if 'bfd' in data else None
+        route_tag   = self.ROUTE_ADVERTISE_DISABLE_TAG if 'advertise' in data and data['advertise'] == "false" else self.ROUTE_ADVERTISE_ENABLE_TAG 
+
+        # bfd enabled route would be handled in staticroutebfd, skip here
+        if bfd_enable and bfd_enable[0].lower() == "true":
+            log_debug("{} static route {} bfd flag is true".format(self.db_name, key))
+            tmp_nh_set, tmp_route_tag = self.static_routes.get(vrf, {}).get(ip_prefix, (IpNextHopSet(is_ipv6), route_tag))
+            if tmp_nh_set: #clear nexthop set if it is not empty
+                log_debug("{} static route {} bfd flag is true, cur_nh is not empty, clear it".format(self.db_name, key))
+                self.static_routes.setdefault(vrf, {}).pop(ip_prefix, None)
+            return True
 
         try:
             ip_nh_set = IpNextHopSet(is_ipv6, bkh_list, nh_list, intf_list, dist_list, nh_vrf_list)
@@ -60,19 +71,59 @@ def set_handler(self, key, data):
 
         if cmd_list:
             self.cfg_mgr.push_list(cmd_list)
-            log_debug("Static route {} is scheduled for updates".format(key))
+            log_debug("{} Static route {} is scheduled for updates. {}".format(self.db_name, key, str(cmd_list)))
         else:
-            log_debug("Nothing to update for static route {}".format(key))
+            log_debug("{} Nothing to update for static route {}".format(self.db_name, key))
 
         self.static_routes.setdefault(vrf, {})[ip_prefix] = (ip_nh_set, route_tag)
 
         return True
 
 
+    def skip_appl_del(self, vrf, ip_prefix):
+        """
+        If a static route is bfd enabled, the processed static route is written into application DB by staticroutebfd.
+        When we disable bfd for that route at runtime, that static route entry will be removed from APPL_DB STATIC_ROUTE_TABLE.
+        In the case, the StaticRouteMgr(appl_db) cannot uninstall the static route from FRR if the static route is still in CONFIG_DB,
+        so need this checking (skip appl_db deletion) to avoid race condition between StaticRouteMgr(appl_db) and StaticRouteMgr(config_db)
+        For more detailed information:
+        https://github.com/sonic-net/SONiC/blob/master/doc/static-route/SONiC_static_route_bfd_hld.md#bfd-field-changes-from-true-to-false
+
+        :param vrf: vrf from the split_key(key) return
+        :param ip_prefix: ip_prefix from the split_key(key) return
+        :return: True if the deletion comes from APPL_DB and the vrf|ip_prefix exists in CONFIG_DB, otherwise return False
+        """
+        if self.db_name == "CONFIG_DB":
+            return False
+
+        if self.config_db is None:
+            self.config_db = swsscommon.SonicV2Connector()
+            self.config_db.connect(self.config_db.CONFIG_DB)
+
+        #just pop local cache if the route exist in config_db
+        cfg_key = "STATIC_ROUTE|" + vrf + "|" + ip_prefix
+        nexthop = self.config_db.get(self.config_db.CONFIG_DB, cfg_key, "nexthop")
+        if nexthop and len(nexthop)>0:
+            self.static_routes.setdefault(vrf, {}).pop(ip_prefix, None)
+            return True
+
+        if vrf == "default":
+            cfg_key = "STATIC_ROUTE|" + ip_prefix
+        nexthop = self.config_db.get(self.config_db.CONFIG_DB, cfg_key, "nexthop")
+        if nexthop and len(nexthop)>0:
+            self.static_routes.setdefault(vrf, {}).pop(ip_prefix, None)
+            return True
+
+        return False
+
     def del_handler(self, key):
         vrf, ip_prefix = self.split_key(key)
         is_ipv6 = TemplateFabric.is_ipv6(ip_prefix)
 
+        if self.skip_appl_del(vrf, ip_prefix):
+            log_debug("{} ignore appl_db static route deletion because of key {} exist in config_db".format(self.db_name, key))
+            return
+
         ip_nh_set = IpNextHopSet(is_ipv6)
         cur_nh_set, route_tag = self.static_routes.get(vrf, {}).get(ip_prefix, (IpNextHopSet(is_ipv6), self.ROUTE_ADVERTISE_DISABLE_TAG))
         cmd_list = self.static_route_commands(ip_nh_set, cur_nh_set, ip_prefix, vrf, route_tag, route_tag)
@@ -85,9 +136,9 @@ def del_handler(self, key):
 
         if cmd_list:
             self.cfg_mgr.push_list(cmd_list)
-            log_debug("Static route {} is scheduled for updates".format(key))
+            log_debug("{} Static route {} is scheduled for updates. {}".format(self.db_name, key, str(cmd_list)))
         else:
-            log_debug("Nothing to update for static route {}".format(key))
+            log_debug("{} Nothing to update for static route {}".format(self.db_name, key))
 
         self.static_routes.setdefault(vrf, {}).pop(ip_prefix, None)
 

src/sonic-bgpcfgd/setup.py

@@ -11,6 +11,7 @@
     entry_points = {
         'console_scripts': [
             'bgpcfgd = bgpcfgd.main:main',
+            'staticroutebfd = staticroutebfd.main:main',
             'bgpmon = bgpmon.bgpmon:main',
         ]
     },