]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/net/necp_client.c
xnu-4903.221.2.tar.gz
[apple/xnu.git] / bsd / net / necp_client.c
index 41e6efaa84524fbbe6154257d1a38f4c13065832..814f2f0be88b4348b6d5e94beeb1270ed70de168 100644 (file)
 
 extern u_int32_t necp_debug;
 
-// proc_best_name() is declared here in advance of it landing in a header file.
-// See comment in kern_proc.c
-extern char *proc_best_name(proc_t p);
-
 static int noop_read(struct fileproc *, struct uio *, int, vfs_context_t);
 static int noop_write(struct fileproc *, struct uio *, int, vfs_context_t);
 static int noop_ioctl(struct fileproc *, unsigned long, caddr_t,
@@ -192,15 +188,17 @@ extern unsigned int get_maxmtu(struct rtentry *);
 #define NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT                    0x00040
 #define NECP_PARSED_PARAMETERS_FIELD_PROHIBITED_AGENT          0x00080
 #define NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT           0x00100
-#define NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE       0x00200
-#define NECP_PARSED_PARAMETERS_FIELD_PROHIBITED_AGENT_TYPE     0x00400
-#define NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE      0x00800
-#define NECP_PARSED_PARAMETERS_FIELD_FLAGS                                     0x01000
-#define NECP_PARSED_PARAMETERS_FIELD_IP_PROTOCOL                       0x02000
-#define NECP_PARSED_PARAMETERS_FIELD_EFFECTIVE_PID                     0x04000
-#define NECP_PARSED_PARAMETERS_FIELD_EFFECTIVE_UUID                    0x08000
-#define NECP_PARSED_PARAMETERS_FIELD_TRAFFIC_CLASS                     0x10000
-#define NECP_PARSED_PARAMETERS_FIELD_LOCAL_PORT                                0x20000
+#define NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT                     0x00200
+#define NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE       0x00400
+#define NECP_PARSED_PARAMETERS_FIELD_PROHIBITED_AGENT_TYPE     0x00800
+#define NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE      0x01000
+#define NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE                0x02000
+#define NECP_PARSED_PARAMETERS_FIELD_FLAGS                                     0x04000
+#define NECP_PARSED_PARAMETERS_FIELD_IP_PROTOCOL                       0x08000
+#define NECP_PARSED_PARAMETERS_FIELD_EFFECTIVE_PID                     0x10000
+#define NECP_PARSED_PARAMETERS_FIELD_EFFECTIVE_UUID                    0x20000
+#define NECP_PARSED_PARAMETERS_FIELD_TRAFFIC_CLASS                     0x40000
+#define NECP_PARSED_PARAMETERS_FIELD_LOCAL_PORT                                0x80000
 
 #define NECP_MAX_PARSED_PARAMETERS 16
 struct necp_client_parsed_parameters {
@@ -215,9 +213,11 @@ struct necp_client_parsed_parameters {
        struct necp_client_parameter_netagent_type required_netagent_types[NECP_MAX_PARSED_PARAMETERS];
        struct necp_client_parameter_netagent_type prohibited_netagent_types[NECP_MAX_PARSED_PARAMETERS];
        struct necp_client_parameter_netagent_type preferred_netagent_types[NECP_MAX_PARSED_PARAMETERS];
+       struct necp_client_parameter_netagent_type avoided_netagent_types[NECP_MAX_PARSED_PARAMETERS];
        uuid_t required_netagents[NECP_MAX_PARSED_PARAMETERS];
        uuid_t prohibited_netagents[NECP_MAX_PARSED_PARAMETERS];
        uuid_t preferred_netagents[NECP_MAX_PARSED_PARAMETERS];
+       uuid_t avoided_netagents[NECP_MAX_PARSED_PARAMETERS];
        u_int16_t ip_protocol;
        pid_t effective_pid;
        uuid_t effective_uuid;
@@ -226,7 +226,7 @@ struct necp_client_parsed_parameters {
 
 static bool
 necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_parameters,
-                                                                  u_int *return_ifindex);
+                                                                  u_int *return_ifindex, bool *validate_agents);
 
 static bool
 necp_ifnet_matches_local_address(struct ifnet *ifp, struct sockaddr *sa);
@@ -234,7 +234,8 @@ necp_ifnet_matches_local_address(struct ifnet *ifp, struct sockaddr *sa);
 static bool
 necp_ifnet_matches_parameters(struct ifnet *ifp,
                                                          struct necp_client_parsed_parameters *parsed_parameters,
-                                                         u_int32_t *preferred_count, bool ignore_require_if);
+                                                         u_int32_t *preferred_count,
+                                                         bool secondary_interface);
 
 static const struct fileops necp_fd_ops = {
        .fo_type = DTYPE_NETPOLICY,
@@ -254,6 +255,8 @@ struct necp_client_assertion {
 
 struct necp_client_flow_header {
        struct necp_tlv_header outer_header;
+       struct necp_tlv_header flow_id_tlv_header;
+       uuid_t flow_id;
        struct necp_tlv_header flags_tlv_header;
        u_int32_t flags_value;
        struct necp_tlv_header interface_tlv_header;
@@ -274,10 +277,64 @@ struct necp_client_nexus_flow_header {
 } __attribute__((__packed__));
 
 
+struct necp_client_flow {
+       LIST_ENTRY(necp_client_flow) flow_chain;
+       unsigned invalid : 1;
+       unsigned nexus : 1; // If true, flow is a nexus; if false, flow is attached to socket
+       unsigned socket : 1;
+       unsigned viable : 1;
+       unsigned assigned : 1;
+       unsigned has_protoctl_event : 1;
+       unsigned check_tcp_heuristics : 1;
+       unsigned _reserved : 1;
+       union {
+               uuid_t nexus_agent;
+               struct {
+                       void *socket_handle;
+                       necp_client_flow_cb cb;
+               };
+       } u;
+       uint32_t interface_index;
+       uint16_t interface_flags;
+       uint32_t necp_flow_flags;
+       struct necp_client_flow_protoctl_event protoctl_event;
+       union necp_sockaddr_union local_addr;
+       union necp_sockaddr_union remote_addr;
+
+       size_t assigned_results_length;
+       u_int8_t *assigned_results;
+};
+
+struct necp_client_flow_registration {
+       RB_ENTRY(necp_client_flow_registration) fd_link;
+       RB_ENTRY(necp_client_flow_registration) global_link;
+       RB_ENTRY(necp_client_flow_registration) client_link;
+       LIST_ENTRY(necp_client_flow_registration) collect_stats_chain;
+       uuid_t registration_id;
+       u_int32_t flags;
+       unsigned flow_result_read : 1;
+       unsigned defunct : 1;
+       void *interface_handle;
+       necp_client_flow_cb interface_cb;
+       struct necp_client *client;
+       LIST_HEAD(_necp_registration_flow_list, necp_client_flow) flow_list;
+       u_int64_t last_interface_details __attribute__((aligned(sizeof(u_int64_t))));
+};
+
+static int necp_client_flow_id_cmp(struct necp_client_flow_registration *flow0, struct necp_client_flow_registration *flow1);
+
+RB_HEAD(_necp_client_flow_tree, necp_client_flow_registration);
+RB_PROTOTYPE_PREV(_necp_client_flow_tree, necp_client_flow_registration, client_link, necp_client_flow_id_cmp);
+RB_GENERATE_PREV(_necp_client_flow_tree, necp_client_flow_registration, client_link, necp_client_flow_id_cmp);
+
+#define NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT 4
+#define NECP_CLIENT_MAX_INTERFACE_OPTIONS 16
+
+#define NECP_CLIENT_INTERFACE_OPTION_EXTRA_COUNT (NECP_CLIENT_MAX_INTERFACE_OPTIONS - NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT)
+
 struct necp_client {
        RB_ENTRY(necp_client) link;
        RB_ENTRY(necp_client) global_link;
-       LIST_ENTRY(necp_client) collect_stats_chain;
 
        decl_lck_mtx_data(, lock);
        decl_lck_mtx_data(, route_lock);
@@ -285,10 +342,9 @@ struct necp_client {
 
        uuid_t client_id;
        unsigned result_read : 1;
-       unsigned flow_result_read : 1;
        unsigned allow_multiple_flows : 1;
+       unsigned legacy_client_is_flow : 1;
 
-       unsigned defunct : 1;
        unsigned background : 1;
        unsigned background_update : 1;
        unsigned platform_binary : 1;
@@ -301,13 +357,18 @@ struct necp_client {
        u_int16_t ip_protocol;
        int proc_pid;
 
-       LIST_HEAD(_necp_client_flow_list, necp_client_flow) flow_list;
+       struct _necp_client_flow_tree flow_registrations;
        LIST_HEAD(_necp_client_assertion_list, necp_client_assertion) assertion_list;
 
        struct rtentry *current_route;
 
-       void *interface_handle;
-       void (*interface_cb)(void *handle, int action, struct necp_client_flow *flow);
+       struct necp_client_interface_option interface_options[NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT];
+       struct necp_client_interface_option *extra_interface_options;
+       u_int8_t interface_option_count; // Number in interface_options + extra_interface_options
+
+       struct necp_client_result_netagent failed_trigger_agent;
+
+       void *agent_handle;
 
        size_t parameters_length;
        u_int8_t parameters[0];
@@ -331,18 +392,19 @@ necp_client_add_assertion(struct necp_client *client, uuid_t netagent_uuid);
 static bool
 necp_client_remove_assertion(struct necp_client *client, uuid_t netagent_uuid);
 
-LIST_HEAD(_necp_client_list, necp_client);
-static struct _necp_client_list necp_collect_stats_client_list;
+LIST_HEAD(_necp_flow_registration_list, necp_client_flow_registration);
+static struct _necp_flow_registration_list necp_collect_stats_flow_list;
 
-struct necp_client_defunct {
-       LIST_ENTRY(necp_client_defunct) chain;
+struct necp_flow_defunct {
+       LIST_ENTRY(necp_flow_defunct) chain;
 
-       uuid_t client_id;
+       uuid_t flow_id;
        uuid_t nexus_agent;
+       void *agent_handle;
        int proc_pid;
 };
 
-LIST_HEAD(_necp_client_defunct_list, necp_client_defunct);
+LIST_HEAD(_necp_flow_defunct_list, necp_flow_defunct);
 
 static int necp_client_id_cmp(struct necp_client *client0, struct necp_client *client1);
 
@@ -354,7 +416,16 @@ RB_HEAD(_necp_client_global_tree, necp_client);
 RB_PROTOTYPE_PREV(_necp_client_global_tree, necp_client, global_link, necp_client_id_cmp);
 RB_GENERATE_PREV(_necp_client_global_tree, necp_client, global_link, necp_client_id_cmp);
 
+RB_HEAD(_necp_fd_flow_tree, necp_client_flow_registration);
+RB_PROTOTYPE_PREV(_necp_fd_flow_tree, necp_client_flow_registration, fd_link, necp_client_flow_id_cmp);
+RB_GENERATE_PREV(_necp_fd_flow_tree, necp_client_flow_registration, fd_link, necp_client_flow_id_cmp);
+
+RB_HEAD(_necp_client_flow_global_tree, necp_client_flow_registration);
+RB_PROTOTYPE_PREV(_necp_client_flow_global_tree, necp_client_flow_registration, global_link, necp_client_flow_id_cmp);
+RB_GENERATE_PREV(_necp_client_flow_global_tree, necp_client_flow_registration, global_link, necp_client_flow_id_cmp);
+
 static struct _necp_client_global_tree necp_client_global_tree;
+static struct _necp_client_flow_global_tree necp_client_flow_global_tree;
 
 struct necp_client_update {
        TAILQ_ENTRY(necp_client_update) chain;
@@ -366,10 +437,15 @@ struct necp_client_update {
 };
 
 
+#define        NAIF_ATTACHED   0x1     // arena is attached to list
+#define        NAIF_REDIRECT   0x2     // arena mmap has been redirected
+#define        NAIF_DEFUNCT    0x4     // arena is now defunct
+
 struct necp_fd_data {
        u_int8_t necp_fd_type;
        LIST_ENTRY(necp_fd_data) chain;
        struct _necp_client_tree clients;
+       struct _necp_fd_flow_tree flows;
        TAILQ_HEAD(_necp_client_update_list, necp_client_update) update_list;
        int update_count;
        int flags;
@@ -392,11 +468,17 @@ static LIST_HEAD(_necp_fd_observer_list, necp_fd_data) necp_fd_observer_list;
 static unsigned int necp_client_fd_size;       /* size of zone element */
 static struct zone *necp_client_fd_zone;       /* zone for necp_fd_data */
 
-#define        NECP_FLOW_ZONE_MAX                      512
-#define        NECP_FLOW_ZONE_NAME                     "necp.flow"
+#define        NECP_FLOW_ZONE_NAME                                     "necp.flow"
+#define        NECP_FLOW_REGISTRATION_ZONE_NAME        "necp.flowregistration"
 
 static unsigned int necp_flow_size;            /* size of necp_client_flow */
-static struct mcache *necp_flow_cache;         /* cache for necp_client_flow */
+static struct mcache *necp_flow_cache; /* cache for necp_client_flow */
+
+static unsigned int necp_flow_registration_size;       /* size of necp_client_flow_registration */
+static struct mcache *necp_flow_registration_cache;    /* cache for necp_client_flow_registration */
+
+#define        NECP_ARENA_INFO_ZONE_MAX                128
+#define        NECP_ARENA_INFO_ZONE_NAME               "necp.arenainfo"
 
 
 static lck_grp_attr_t  *necp_fd_grp_attr       = NULL;
@@ -406,6 +488,7 @@ static      lck_grp_t               *necp_fd_mtx_grp        = NULL;
 decl_lck_rw_data(static, necp_fd_lock);
 decl_lck_rw_data(static, necp_observer_lock);
 decl_lck_rw_data(static, necp_client_tree_lock);
+decl_lck_rw_data(static, necp_flow_tree_lock);
 decl_lck_rw_data(static, necp_collect_stats_list_lock);
 
 #define NECP_STATS_LIST_LOCK_EXCLUSIVE() lck_rw_lock_exclusive(&necp_collect_stats_list_lock)
@@ -415,6 +498,12 @@ decl_lck_rw_data(static, necp_collect_stats_list_lock);
 #define NECP_CLIENT_TREE_LOCK_EXCLUSIVE() lck_rw_lock_exclusive(&necp_client_tree_lock)
 #define NECP_CLIENT_TREE_LOCK_SHARED() lck_rw_lock_shared(&necp_client_tree_lock)
 #define NECP_CLIENT_TREE_UNLOCK() lck_rw_done(&necp_client_tree_lock)
+#define NECP_CLIENT_TREE_ASSERT_LOCKED() LCK_RW_ASSERT(&necp_client_tree_lock, LCK_RW_ASSERT_HELD)
+
+#define NECP_FLOW_TREE_LOCK_EXCLUSIVE() lck_rw_lock_exclusive(&necp_flow_tree_lock)
+#define NECP_FLOW_TREE_LOCK_SHARED() lck_rw_lock_shared(&necp_flow_tree_lock)
+#define NECP_FLOW_TREE_UNLOCK() lck_rw_done(&necp_flow_tree_lock)
+#define NECP_FLOW_TREE_ASSERT_LOCKED() LCK_RW_ASSERT(&necp_flow_tree_lock, LCK_RW_ASSERT_HELD)
 
 #define NECP_FD_LIST_LOCK_EXCLUSIVE() lck_rw_lock_exclusive(&necp_fd_lock)
 #define NECP_FD_LIST_LOCK_SHARED() lck_rw_lock_shared(&necp_fd_lock)
@@ -428,7 +517,8 @@ decl_lck_rw_data(static, necp_collect_stats_list_lock);
 
 // Take NECP_FD_LIST_LOCK when accessing or modifying the necp_fd_list
 // Take NECP_CLIENT_TREE_LOCK when accessing or modifying the necp_client_global_tree
-// Take NECP_STATS_LIST_LOCK when accessing or modifying the necp_collect_stats_client_list
+// Take NECP_FLOW_TREE_LOCK when accessing or modifying the necp_client_flow_global_tree
+// Take NECP_STATS_LIST_LOCK when accessing or modifying the necp_collect_stats_flow_list
 // Take NECP_FD_LOCK when accessing or modifying an necp_fd_data entry
 // Take NECP_CLIENT_LOCK when accessing or modifying a single necp_client
 // Take NECP_CLIENT_ROUTE_LOCK when accessing or modifying a client's route
@@ -438,8 +528,9 @@ decl_lck_rw_data(static, necp_collect_stats_list_lock);
 // 2. NECP_FD_LOCK (any)
 // 3. NECP_CLIENT_TREE_LOCK
 // 4. NECP_CLIENT_LOCK (any)
-// 5. NECP_STATS_LIST_LOCK
-// 6. NECP_CLIENT_ROUTE_LOCK (any)
+// 5. NECP_FLOW_TREE_LOCK
+// 6. NECP_STATS_LIST_LOCK
+// 7. NECP_CLIENT_ROUTE_LOCK (any)
 
 static thread_call_t necp_client_update_tcall;
 
@@ -489,6 +580,19 @@ necp_fd_notify(struct necp_fd_data *fd_data, bool locked)
        }
 }
 
+static inline bool
+necp_client_has_unread_flows(struct necp_client *client)
+{
+       NECP_CLIENT_ASSERT_LOCKED(client);
+       struct necp_client_flow_registration *flow_registration = NULL;
+       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+               if (!flow_registration->flow_result_read) {
+                       return true;
+               }
+       }
+       return false;
+}
+
 static int
 necp_fd_poll(struct necp_fd_data *fd_data, int events, void *wql, struct proc *p, int is_kevent)
 {
@@ -508,7 +612,7 @@ necp_fd_poll(struct necp_fd_data *fd_data, int events, void *wql, struct proc *p
                        bool has_unread_clients = FALSE;
                        RB_FOREACH(client, _necp_client_tree, &fd_data->clients) {
                                NECP_CLIENT_LOCK(client);
-                               if (!client->result_read || !client->flow_result_read) {
+                               if (!client->result_read || necp_client_has_unread_flows(client)) {
                                        has_unread_clients = TRUE;
                                }
                                NECP_CLIENT_UNLOCK(client);
@@ -526,14 +630,96 @@ necp_fd_poll(struct necp_fd_data *fd_data, int events, void *wql, struct proc *p
        return (revents);
 }
 
+static inline void
+necp_generate_client_id(uuid_t client_id, bool is_flow)
+{
+       uuid_generate_random(client_id);
+
+       if (is_flow) {
+               client_id[9] |= 0x01;
+       } else {
+               client_id[9] &= ~0x01;
+       }
+}
+
+static inline bool
+necp_client_id_is_flow(uuid_t client_id)
+{
+       return (client_id[9] & 0x01);
+}
+
 static struct necp_client *
-necp_client_fd_find_client_and_lock(struct necp_fd_data *client_fd, uuid_t client_id)
+necp_find_client_and_lock(uuid_t client_id)
+{
+       NECP_CLIENT_TREE_ASSERT_LOCKED();
+
+       struct necp_client *client = NULL;
+
+       if (necp_client_id_is_flow(client_id)) {
+               NECP_FLOW_TREE_LOCK_SHARED();
+               struct necp_client_flow_registration find;
+               uuid_copy(find.registration_id, client_id);
+               struct necp_client_flow_registration *flow = RB_FIND(_necp_client_flow_global_tree, &necp_client_flow_global_tree, &find);
+               if (flow != NULL) {
+                       client = flow->client;
+               }
+               NECP_FLOW_TREE_UNLOCK();
+       } else {
+               struct necp_client find;
+               uuid_copy(find.client_id, client_id);
+               client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
+       }
+
+       if (client != NULL) {
+               NECP_CLIENT_LOCK(client);
+       }
+
+       return (client);
+}
+
+static struct necp_client_flow_registration *
+necp_client_find_flow(struct necp_client *client, uuid_t flow_id)
+{
+       NECP_CLIENT_ASSERT_LOCKED(client);
+       struct necp_client_flow_registration *flow = NULL;
+
+       if (necp_client_id_is_flow(flow_id)) {
+               struct necp_client_flow_registration find;
+               uuid_copy(find.registration_id, flow_id);
+               flow = RB_FIND(_necp_client_flow_tree, &client->flow_registrations, &find);
+       } else {
+               flow = RB_ROOT(&client->flow_registrations);
+       }
+
+       return (flow);
+}
+
+static struct necp_client *
+necp_client_fd_find_client_unlocked(struct necp_fd_data *client_fd, uuid_t client_id)
 {
-       struct necp_client find;
        NECP_FD_ASSERT_LOCKED(client_fd);
-       uuid_copy(find.client_id, client_id);
-       struct necp_client *client = RB_FIND(_necp_client_tree, &client_fd->clients, &find);
+       struct necp_client *client = NULL;
+
+       if (necp_client_id_is_flow(client_id)) {
+               struct necp_client_flow_registration find;
+               uuid_copy(find.registration_id, client_id);
+               struct necp_client_flow_registration *flow = RB_FIND(_necp_fd_flow_tree, &client_fd->flows, &find);
+               if (flow != NULL) {
+                       client = flow->client;
+               }
+       } else {
+               struct necp_client find;
+               uuid_copy(find.client_id, client_id);
+               client = RB_FIND(_necp_client_tree, &client_fd->clients, &find);
+       }
+
+       return (client);
+}
 
+static struct necp_client *
+necp_client_fd_find_client_and_lock(struct necp_fd_data *client_fd, uuid_t client_id)
+{
+       struct necp_client *client = necp_client_fd_find_client_unlocked(client_fd, client_id);
        if (client != NULL) {
                NECP_CLIENT_LOCK(client);
        }
@@ -547,6 +733,12 @@ necp_client_id_cmp(struct necp_client *client0, struct necp_client *client1)
        return (uuid_compare(client0->client_id, client1->client_id));
 }
 
+static inline int
+necp_client_flow_id_cmp(struct necp_client_flow_registration *flow0, struct necp_client_flow_registration *flow1)
+{
+       return (uuid_compare(flow0->registration_id, flow1->registration_id));
+}
+
 static int
 necpop_select(struct fileproc *fp, int which, void *wql, vfs_context_t ctx)
 {
@@ -630,8 +822,6 @@ necp_fd_knrtouch(struct knote *kn, struct kevent_internal_s *kev)
        fd_data = (struct necp_fd_data *)kn->kn_hook;
 
        NECP_FD_LOCK(fd_data);
-       if ((kn->kn_status & KN_UDATA_SPECIFIC) == 0)
-               kn->kn_udata = kev->udata;
        revents = necp_fd_poll(fd_data, POLLIN, NULL, current_proc(), 1);
        NECP_FD_UNLOCK(fd_data);
 
@@ -681,58 +871,47 @@ necpop_kqfilter(struct fileproc *fp, struct knote *kn,
        return ((revents & POLLIN) != 0);
 }
 
+#define INTERFACE_FLAGS_SHIFT   32
+#define INTERFACE_FLAGS_MASK    0xffff
+#define INTERFACE_INDEX_SHIFT   0
+#define INTERFACE_INDEX_MASK    0xffffffff
 
-static bool
-necp_set_client_defunct(struct necp_client *client)
+static uint64_t
+combine_interface_details(uint32_t interface_index, uint16_t interface_flags)
 {
-       bool updated = FALSE;
-       u_int32_t flags = 0;
-       u_int32_t value_size = 0;
-
-       client->defunct = TRUE;
-
-       u_int8_t *flags_pointer = necp_buffer_get_tlv_value(client->result, 0, &value_size);
-       if (flags_pointer && value_size == sizeof(flags)) {
-               memcpy(&flags, flags_pointer, value_size);
-
-               flags |= NECP_CLIENT_RESULT_FLAG_DEFUNCT;
-
-               (void)necp_buffer_write_tlv_if_different(client->result, NECP_CLIENT_RESULT_FLAGS,
-                                                                                                sizeof(flags), &flags, &updated, client->result, sizeof(client->result));
-
-        if (updated) {
-            client->result_read = FALSE;
-        }
-       }
-
-       return (updated);
+    return (((uint64_t)interface_flags & INTERFACE_FLAGS_MASK) << INTERFACE_FLAGS_SHIFT |
+            ((uint64_t)interface_index & INTERFACE_INDEX_MASK) << INTERFACE_INDEX_SHIFT);
 }
 
+
 static void
-necp_defunct_client_for_policy(struct necp_client *client,
-                                                          struct _necp_client_defunct_list *defunct_list)
+necp_defunct_flow_registration(struct necp_client *client,
+                                                          struct necp_client_flow_registration *flow_registration,
+                                                          struct _necp_flow_defunct_list *defunct_list)
 {
        NECP_CLIENT_ASSERT_LOCKED(client);
-       
-       if (!client->defunct) {
+
+       if (!flow_registration->defunct) {
                bool needs_defunct = false;
                struct necp_client_flow *search_flow = NULL;
-               LIST_FOREACH(search_flow, &client->flow_list, flow_chain) {
+               LIST_FOREACH(search_flow, &flow_registration->flow_list, flow_chain) {
                        if (search_flow->nexus &&
-                               !uuid_is_null(search_flow->u.nexus_agent) &&
-                               search_flow->requested_nexus) {
+                               !uuid_is_null(search_flow->u.nexus_agent)) {
 
                                // Save defunct values for the nexus
                                if (defunct_list != NULL) {
                                        // Sleeping alloc won't fail; copy only what's necessary
-                                       struct necp_client_defunct *client_defunct = _MALLOC(sizeof (struct necp_client_defunct),
-                                                                                                                                                M_NECP, M_WAITOK | M_ZERO);
-                                       uuid_copy(client_defunct->nexus_agent, search_flow->u.nexus_agent);
-                                       uuid_copy(client_defunct->client_id, client->client_id);
-                                       client_defunct->proc_pid = client->proc_pid;
+                                       struct necp_flow_defunct *flow_defunct = _MALLOC(sizeof (struct necp_flow_defunct),
+                                                                                                                                        M_NECP, M_WAITOK | M_ZERO);
+                                       uuid_copy(flow_defunct->nexus_agent, search_flow->u.nexus_agent);
+                                       uuid_copy(flow_defunct->flow_id, ((flow_registration->flags & NECP_CLIENT_FLOW_FLAGS_USE_CLIENT_ID) ?
+                                                                                                         client->client_id :
+                                                                                                         flow_registration->registration_id));
+                                       flow_defunct->proc_pid = client->proc_pid;
+                                       flow_defunct->agent_handle = client->agent_handle;
 
                                        // Add to the list provided by caller
-                                       LIST_INSERT_HEAD(defunct_list, client_defunct, chain);
+                                       LIST_INSERT_HEAD(defunct_list, flow_defunct, chain);
                                }
 
                                needs_defunct = true;
@@ -740,12 +919,25 @@ necp_defunct_client_for_policy(struct necp_client *client,
                }
 
                if (needs_defunct) {
+
                        // Only set defunct if there was some assigned flow
-                       client->defunct = true;
+                       flow_registration->defunct = true;
                }
        }
 }
 
+static void
+necp_defunct_client_for_policy(struct necp_client *client,
+                                                          struct _necp_flow_defunct_list *defunct_list)
+{
+       NECP_CLIENT_ASSERT_LOCKED(client);
+
+       struct necp_client_flow_registration *flow_registration = NULL;
+       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+               necp_defunct_flow_registration(client, flow_registration, defunct_list);
+       }
+}
+
 static void
 necp_client_free(struct necp_client *client)
 {
@@ -753,6 +945,9 @@ necp_client_free(struct necp_client *client)
 
        NECP_CLIENT_UNLOCK(client);
 
+       FREE(client->extra_interface_options, M_NECP);
+       client->extra_interface_options = NULL;
+
        lck_mtx_destroy(&client->route_lock, necp_fd_mtx_grp);
        lck_mtx_destroy(&client->lock, necp_fd_mtx_grp);
 
@@ -927,32 +1122,26 @@ necp_client_update_observer_remove(struct necp_client *client)
 }
 
 static void
-necp_destroy_client(struct necp_client *client, pid_t pid, bool abort)
+necp_destroy_client_flow_registration(struct necp_client *client,
+                                                                         struct necp_client_flow_registration *flow_registration,
+                                                                         pid_t pid, bool abort)
 {
-       NECP_CLIENT_ASSERT_UNLOCKED(client);
-
-       necp_client_update_observer_remove(client);
-
-       NECP_CLIENT_LOCK(client);
+       NECP_CLIENT_ASSERT_LOCKED(client);
 
-       // Free route
-       NECP_CLIENT_ROUTE_LOCK(client);
-       if (client->current_route != NULL) {
-               rtfree(client->current_route);
-               client->current_route = NULL;
-       }
-       NECP_CLIENT_ROUTE_UNLOCK(client);
 
-       // Remove flow assignments
        struct necp_client_flow *search_flow = NULL;
        struct necp_client_flow *temp_flow = NULL;
-       LIST_FOREACH_SAFE(search_flow, &client->flow_list, flow_chain, temp_flow) {
+       LIST_FOREACH_SAFE(search_flow, &flow_registration->flow_list, flow_chain, temp_flow) {
                if (search_flow->nexus &&
-                       !uuid_is_null(search_flow->u.nexus_agent) &&
-                       search_flow->requested_nexus) {
+                       !uuid_is_null(search_flow->u.nexus_agent)) {
                        // Note that if we had defuncted the client earlier, this would result in a harmless ENOENT
-                       int netagent_error = netagent_client_message(search_flow->u.nexus_agent, client->client_id, pid,
-                           abort ? NETAGENT_MESSAGE_TYPE_ABORT_NEXUS : NETAGENT_MESSAGE_TYPE_CLOSE_NEXUS);
+                       int netagent_error = netagent_client_message(search_flow->u.nexus_agent,
+                                                                                                                ((flow_registration->flags & NECP_CLIENT_FLOW_FLAGS_USE_CLIENT_ID) ?
+                                                                                                                 client->client_id :
+                                                                                                                 flow_registration->registration_id),
+                                                                                                                pid, client->agent_handle,
+                                                                                                                (abort ? NETAGENT_MESSAGE_TYPE_ABORT_NEXUS :
+                                                                                                                 NETAGENT_MESSAGE_TYPE_CLOSE_NEXUS));
                        if (netagent_error != 0 && netagent_error != ENOENT) {
                                NECPLOG(LOG_ERR, "necp_client_remove close nexus error (%d)", netagent_error);
                        }
@@ -971,11 +1160,42 @@ necp_destroy_client(struct necp_client *client, pid_t pid, bool abort)
                mcache_free(necp_flow_cache, search_flow);
        }
 
+       RB_REMOVE(_necp_client_flow_tree, &client->flow_registrations, flow_registration);
+       flow_registration->client = NULL;
+
+       mcache_free(necp_flow_registration_cache, flow_registration);
+}
+
+static void
+necp_destroy_client(struct necp_client *client, pid_t pid, bool abort)
+{
+       NECP_CLIENT_ASSERT_UNLOCKED(client);
+
+       necp_client_update_observer_remove(client);
+
+       NECP_CLIENT_LOCK(client);
+
+       // Free route
+       NECP_CLIENT_ROUTE_LOCK(client);
+       if (client->current_route != NULL) {
+               rtfree(client->current_route);
+               client->current_route = NULL;
+       }
+       NECP_CLIENT_ROUTE_UNLOCK(client);
+
+       // Remove flow assignments
+       struct necp_client_flow_registration *flow_registration = NULL;
+       struct necp_client_flow_registration *temp_flow_registration = NULL;
+       RB_FOREACH_SAFE(flow_registration, _necp_client_flow_tree, &client->flow_registrations, temp_flow_registration) {
+               necp_destroy_client_flow_registration(client, flow_registration, pid, abort);
+       }
+
        // Remove agent assertions
        struct necp_client_assertion *search_assertion = NULL;
        struct necp_client_assertion *temp_assertion = NULL;
        LIST_FOREACH_SAFE(search_assertion, &client->assertion_list, assertion_chain, temp_assertion) {
-               int netagent_error = netagent_client_message(search_assertion->asserted_netagent, client->client_id, pid, NETAGENT_MESSAGE_TYPE_CLIENT_UNASSERT);
+               int netagent_error = netagent_client_message(search_assertion->asserted_netagent, client->client_id, pid,
+                                                                                                        client->agent_handle, NETAGENT_MESSAGE_TYPE_CLIENT_UNASSERT);
                if (netagent_error != 0) {
                        NECPLOG((netagent_error == ENOENT ? LOG_DEBUG : LOG_ERR),
                                        "necp_client_remove unassert agent error (%d)", netagent_error);
@@ -1018,6 +1238,16 @@ necpop_close(struct fileglob *fg, vfs_context_t ctx)
 
                NECP_FD_LOCK(fd_data);
                pid_t pid = fd_data->proc_pid;
+
+               struct necp_client_flow_registration *flow_registration = NULL;
+               struct necp_client_flow_registration *temp_flow_registration = NULL;
+               RB_FOREACH_SAFE(flow_registration, _necp_fd_flow_tree, &fd_data->flows, temp_flow_registration) {
+                       NECP_FLOW_TREE_LOCK_EXCLUSIVE();
+                       RB_REMOVE(_necp_client_flow_global_tree, &necp_client_flow_global_tree, flow_registration);
+                       NECP_FLOW_TREE_UNLOCK();
+                       RB_REMOVE(_necp_fd_flow_tree, &fd_data->flows, flow_registration);
+               }
+
                struct necp_client *client = NULL;
                struct necp_client *temp_client = NULL;
                RB_FOREACH_SAFE(client, _necp_client_tree, &fd_data->clients, temp_client) {
@@ -1089,75 +1319,143 @@ necp_find_fd_data(int fd, struct necp_fd_data **fd_data)
        }
        *fd_data = (struct necp_fd_data *)fp->f_fglob->fg_data;
 
+       if ((*fd_data)->necp_fd_type != necp_fd_type_client) {
+               // Not a client fd, ignore
+               error = EINVAL;
+               goto done;
+       }
+
 done:
        proc_fdunlock(p);
        return (error);
 }
 
 
-static void
-necp_client_add_socket_flow(struct necp_client *client, struct inpcb *inp)
-{
-       struct necp_client_flow *new_flow = mcache_alloc(necp_flow_cache, MCR_SLEEP);
-       if (new_flow == NULL) {
-               NECPLOG0(LOG_ERR, "Failed to allocate socket flow");
-               return;
-       }
-
-       memset(new_flow, 0, sizeof(*new_flow));
-
-       new_flow->socket = TRUE;
-       new_flow->u.socket_handle = inp;
-       new_flow->u.cb = inp->necp_cb;
-
-       OSIncrementAtomic(&necp_socket_flow_count);
-
-       LIST_INSERT_HEAD(&client->flow_list, new_flow, flow_chain);
-}
-
-static void
-necp_client_add_interface_flow(struct necp_client *client, uint32_t interface_index)
+static struct necp_client_flow *
+necp_client_add_interface_flow(struct necp_client_flow_registration *flow_registration,
+                                                          uint32_t interface_index)
 {
        struct necp_client_flow *new_flow = mcache_alloc(necp_flow_cache, MCR_SLEEP);
        if (new_flow == NULL) {
                NECPLOG0(LOG_ERR, "Failed to allocate interface flow");
-               return;
+               return NULL;
        }
 
        memset(new_flow, 0, sizeof(*new_flow));
 
        // Neither nexus nor socket
        new_flow->interface_index = interface_index;
-       new_flow->u.socket_handle = client->interface_handle;
-       new_flow->u.cb = client->interface_cb;
+       new_flow->u.socket_handle = flow_registration->interface_handle;
+       new_flow->u.cb = flow_registration->interface_cb;
 
        OSIncrementAtomic(&necp_if_flow_count);
 
-       LIST_INSERT_HEAD(&client->flow_list, new_flow, flow_chain);
+       LIST_INSERT_HEAD(&flow_registration->flow_list, new_flow, flow_chain);
+
+       return new_flow;
 }
 
-static void
-necp_client_add_interface_flow_if_needed(struct necp_client *client, uint32_t interface_index)
+static struct necp_client_flow *
+necp_client_add_interface_flow_if_needed(struct necp_client *client,
+                                                                                struct necp_client_flow_registration *flow_registration,
+                                                                                uint32_t interface_index)
 {
        if (!client->allow_multiple_flows ||
                interface_index == IFSCOPE_NONE) {
                // Interface not set, or client not allowed to use this mode
-               return;
+               return NULL;
        }
 
        struct necp_client_flow *flow = NULL;
-       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
+       LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
                if (!flow->nexus && !flow->socket && flow->interface_index == interface_index) {
                        // Already have the flow
                        flow->invalid = FALSE;
+                       flow->u.socket_handle = flow_registration->interface_handle;
+                       flow->u.cb = flow_registration->interface_cb;
+                       return NULL;
+               }
+       }
+       return necp_client_add_interface_flow(flow_registration, interface_index);
+}
+
+static void
+necp_client_add_interface_option_if_needed(struct necp_client *client,
+                                                                                  uint32_t interface_index,
+                                                                                  uint32_t interface_generation,
+                                                                                  uuid_t *nexus_agent)
+{
+       if (interface_index == IFSCOPE_NONE ||
+               (client->interface_option_count != 0 && !client->allow_multiple_flows)) {
+               // Interface not set, or client not allowed to use this mode
+               return;
+       }
+
+       if (client->interface_option_count >= NECP_CLIENT_MAX_INTERFACE_OPTIONS) {
+               // Cannot take any more interface options
+               return;
+       }
 
-                       flow->u.socket_handle = client->interface_handle;
-                       flow->u.cb = client->interface_cb;
-                       return;
+       // Check if already present
+       for (u_int32_t option_i = 0; option_i < client->interface_option_count; option_i++) {
+               if (option_i < NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT) {
+                       struct necp_client_interface_option *option = &client->interface_options[option_i];
+                       if (option->interface_index == interface_index) {
+                               if (nexus_agent == NULL) {
+                                       return;
+                               }
+                               if (uuid_compare(option->nexus_agent, *nexus_agent) == 0) {
+                                       return;
+                               }
+                               if (uuid_is_null(option->nexus_agent)) {
+                                       uuid_copy(option->nexus_agent, *nexus_agent);
+                                       return;
+                               }
+                               // If we get to this point, this is a new nexus flow
+                       }
+               } else {
+                       struct necp_client_interface_option *option = &client->extra_interface_options[option_i - NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT];
+                       if (option->interface_index == interface_index) {
+                               if (nexus_agent == NULL) {
+                                       return;
+                               }
+                               if (uuid_compare(option->nexus_agent, *nexus_agent) == 0) {
+                                       return;
+                               }
+                               if (uuid_is_null(option->nexus_agent)) {
+                                       uuid_copy(option->nexus_agent, *nexus_agent);
+                                       return;
+                               }
+                               // If we get to this point, this is a new nexus flow
+                       }
                }
        }
 
-       necp_client_add_interface_flow(client, interface_index);
+       // Add a new entry
+       if (client->interface_option_count < NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT) {
+               // Add to static
+               struct necp_client_interface_option *option = &client->interface_options[client->interface_option_count];
+               option->interface_index = interface_index;
+               option->interface_generation = interface_generation;
+               if (nexus_agent != NULL) {
+                       uuid_copy(option->nexus_agent, *nexus_agent);
+               }
+               client->interface_option_count++;
+       } else {
+               // Add to extra
+               if (client->extra_interface_options == NULL) {
+                       client->extra_interface_options = _MALLOC(sizeof(struct necp_client_interface_option) * NECP_CLIENT_INTERFACE_OPTION_EXTRA_COUNT, M_NECP, M_WAITOK | M_ZERO);
+               }
+               if (client->extra_interface_options != NULL) {
+                       struct necp_client_interface_option *option = &client->extra_interface_options[client->interface_option_count - NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT];
+                       option->interface_index = interface_index;
+                       option->interface_generation = interface_generation;
+                       if (nexus_agent != NULL) {
+                               uuid_copy(option->nexus_agent, *nexus_agent);
+                       }
+                       client->interface_option_count++;
+               }
+       }
 }
 
 static bool
@@ -1179,66 +1477,110 @@ necp_client_flow_is_viable(proc_t proc, struct necp_client *client,
                        result.routing_result != NECP_KERNEL_POLICY_RESULT_DROP);
 }
 
+static void
+necp_flow_add_interface_flows(proc_t proc,
+                                                         struct necp_client *client,
+                                                         struct necp_client_flow_registration *flow_registration,
+                                                         bool send_initial)
+{
+       // Traverse all interfaces and add a tracking flow if needed
+       for (u_int32_t option_i = 0; option_i < client->interface_option_count; option_i++) {
+               if (option_i < NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT) {
+                       struct necp_client_interface_option *option = &client->interface_options[option_i];
+                       struct necp_client_flow *flow = necp_client_add_interface_flow_if_needed(client, flow_registration, option->interface_index);
+                       if (flow != NULL && send_initial) {
+                               flow->viable = necp_client_flow_is_viable(proc, client, flow);
+                               if (flow->viable && flow->u.cb) {
+                                       bool viable = flow->viable;
+                                       flow->u.cb(flow_registration->interface_handle, NECP_CLIENT_CBACTION_INITIAL, flow->interface_index, flow->necp_flow_flags, &viable);
+                                       flow->viable = viable;
+                               }
+                       }
+               } else {
+                       struct necp_client_interface_option *option = &client->extra_interface_options[option_i - NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT];
+                       struct necp_client_flow *flow = necp_client_add_interface_flow_if_needed(client, flow_registration, option->interface_index);
+                       if (flow != NULL && send_initial) {
+                               flow->viable = necp_client_flow_is_viable(proc, client, flow);
+                               if (flow->viable && flow->u.cb) {
+                                       bool viable = flow->viable;
+                                       flow->u.cb(flow_registration->interface_handle, NECP_CLIENT_CBACTION_INITIAL, flow->interface_index, flow->necp_flow_flags, &viable);
+                                       flow->viable = viable;
+                               }
+                       }
+               }
+       }
+}
+
 static bool
 necp_client_update_flows(proc_t proc,
                                                 struct necp_client *client,
-                                                struct _necp_client_defunct_list *defunct_list,
-                                                bool *defuncted_by_flow)
+                                                struct _necp_flow_defunct_list *defunct_list)
 {
        NECP_CLIENT_ASSERT_LOCKED(client);
 
        bool client_updated = FALSE;
        struct necp_client_flow *flow = NULL;
        struct necp_client_flow *temp_flow = NULL;
-       LIST_FOREACH_SAFE(flow, &client->flow_list, flow_chain, temp_flow) {
-               // Check policy result for flow
-               int old_flags = flow->necp_flow_flags;
-               bool viable = necp_client_flow_is_viable(proc, client, flow);
-
-               // TODO: Defunct nexus flows that are blocked by policy
-
-               if (flow->viable != viable) {
-                       flow->viable = viable;
-                       client_updated = TRUE;
+       struct necp_client_flow_registration *flow_registration = NULL;
+       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+               if (flow_registration->interface_cb != NULL) {
+                       // Add any interface flows that are not already tracked
+                       necp_flow_add_interface_flows(proc, client, flow_registration, false);
                }
 
-               if ((old_flags & (NECP_CLIENT_RESULT_FLAG_HAS_IPV4 | NECP_CLIENT_RESULT_FLAG_HAS_IPV6)) !=
-                   (flow->necp_flow_flags & (NECP_CLIENT_RESULT_FLAG_HAS_IPV4 | NECP_CLIENT_RESULT_FLAG_HAS_IPV6))) {
-                       client_updated = TRUE;
-               }
+               LIST_FOREACH_SAFE(flow, &flow_registration->flow_list, flow_chain, temp_flow) {
+                       // Check policy result for flow
+                       int old_flags = flow->necp_flow_flags;
+                       bool viable = necp_client_flow_is_viable(proc, client, flow);
 
-               if (flow->viable && client_updated && (flow->socket || (!flow->socket && !flow->nexus)) && flow->u.cb) {
-                       flow->u.cb(flow->u.socket_handle, NECP_CLIENT_CBACTION_VIABLE, flow);
-               }
+                       // TODO: Defunct nexus flows that are blocked by policy
 
-               if (!flow->viable || flow->invalid) {
-                       if (client_updated && (flow->socket || (!flow->socket && !flow->nexus)) && flow->u.cb) {
-                               flow->u.cb(flow->u.socket_handle, NECP_CLIENT_CBACTION_NONVIABLE, flow);
+                       if (flow->viable != viable) {
+                               flow->viable = viable;
+                               client_updated = TRUE;
+                       }
+
+                       if ((old_flags & NECP_CLIENT_RESULT_FLAG_FORCE_UPDATE) !=
+                               (flow->necp_flow_flags & NECP_CLIENT_RESULT_FLAG_FORCE_UPDATE)) {
+                               client_updated = TRUE;
                        }
-                       // The callback might change the viable-flag of the
-                       // flow depending on its policy. Thus, we need to
-                       // check again the flags after the callback.
-               }
 
-               (void)defunct_list;
-               (void)defuncted_by_flow;
+                       if (flow->viable && client_updated && (flow->socket || (!flow->socket && !flow->nexus)) && flow->u.cb) {
+                               bool flow_viable = flow->viable;
+                               flow->u.cb(flow->u.socket_handle, NECP_CLIENT_CBACTION_VIABLE, flow->interface_index, flow->necp_flow_flags, &viable);
+                               flow->viable = flow_viable;
+                       }
 
-               // Handle flows that no longer match
-               if (!flow->viable || flow->invalid) {
-                       // Drop them as long as they aren't assigned data
-                       if (!flow->requested_nexus && !flow->assigned) {
-                               if (flow->assigned_results != NULL) {
-                                       FREE(flow->assigned_results, M_NETAGENT);
-                                       flow->assigned_results = NULL;
-                                       client_updated = TRUE;
+                       if (!flow->viable || flow->invalid) {
+                               if (client_updated && (flow->socket || (!flow->socket && !flow->nexus)) && flow->u.cb) {
+                                       bool flow_viable = flow->viable;
+                                       flow->u.cb(flow->u.socket_handle, NECP_CLIENT_CBACTION_NONVIABLE, flow->interface_index, flow->necp_flow_flags, &viable);
+                                       flow->viable = flow_viable;
                                }
-                               LIST_REMOVE(flow, flow_chain);
-                               if (flow->socket) {
-                                       OSDecrementAtomic(&necp_socket_flow_count);
-                               } else {
-                                       OSDecrementAtomic(&necp_if_flow_count);
+                               // The callback might change the viable-flag of the
+                               // flow depending on its policy. Thus, we need to
+                               // check the flags again after the callback.
+                       }
+
+                       (void)defunct_list;
+
+                       // Handle flows that no longer match
+                       if (!flow->viable || flow->invalid) {
+                               // Drop them as long as they aren't assigned data
+                               if (!flow->nexus && !flow->assigned) {
+                                       if (flow->assigned_results != NULL) {
+                                               FREE(flow->assigned_results, M_NETAGENT);
+                                               flow->assigned_results = NULL;
+                                               client_updated = TRUE;
+                                       }
+                                       LIST_REMOVE(flow, flow_chain);
+                                       if (flow->socket) {
+                                               OSDecrementAtomic(&necp_socket_flow_count);
+                                       } else {
+                                               OSDecrementAtomic(&necp_if_flow_count);
+                                       }
+                                       mcache_free(necp_flow_cache, flow);
                                }
-                               mcache_free(necp_flow_cache, flow);
                        }
                }
        }
@@ -1249,23 +1591,29 @@ necp_client_update_flows(proc_t proc,
 static void
 necp_client_mark_all_nonsocket_flows_as_invalid(struct necp_client *client)
 {
+       struct necp_client_flow_registration *flow_registration = NULL;
        struct necp_client_flow *flow = NULL;
-       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-               if (!flow->socket) { // Socket flows are not marked as invalid
-                       flow->invalid = TRUE;
+       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+               LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
+                       if (!flow->socket) { // Socket flows are not marked as invalid
+                               flow->invalid = TRUE;
+                       }
                }
        }
+
+       // Reset option count every update
+       client->interface_option_count = 0;
 }
 
 static bool
-necp_netagent_applies_to_client(__unused struct necp_client *client,
+necp_netagent_applies_to_client(struct necp_client *client,
                                                                const struct necp_client_parsed_parameters *parameters,
-                                                               uuid_t netagent_uuid, bool allow_nexus,
-                                                               uint32_t interface_index, u_int16_t interface_flags)
+                                                               uuid_t *netagent_uuid, bool allow_nexus,
+                                                               uint32_t interface_index, uint32_t interface_generation)
 {
-#pragma unused(interface_index, interface_flags)
+#pragma unused(interface_index, interface_generation)
        bool applies = FALSE;
-       u_int32_t flags = netagent_get_flags(netagent_uuid);
+       u_int32_t flags = netagent_get_flags(*netagent_uuid);
        if (!(flags & NETAGENT_FLAG_REGISTERED)) {
                // Unregistered agents never apply
                return (applies);
@@ -1279,6 +1627,17 @@ necp_netagent_applies_to_client(__unused struct necp_client *client,
                return (applies);
        }
 
+       if (uuid_compare(client->failed_trigger_agent.netagent_uuid, *netagent_uuid) == 0) {
+               if (client->failed_trigger_agent.generation == netagent_get_generation(*netagent_uuid)) {
+                       // If this agent was triggered, and failed, and hasn't changed, keep hiding it
+                       return (applies);
+               } else {
+                       // Mismatch generation, clear out old trigger
+                       uuid_clear(client->failed_trigger_agent.netagent_uuid);
+                       client->failed_trigger_agent.generation = 0;
+               }
+       }
+
        if (flags & NETAGENT_FLAG_SPECIFIC_USE_ONLY) {
                // Specific use agents only apply when required
                bool required = FALSE;
@@ -1288,7 +1647,7 @@ necp_netagent_applies_to_client(__unused struct necp_client *client,
                                if (uuid_is_null(parameters->required_netagents[i])) {
                                        break;
                                }
-                               if (uuid_compare(parameters->required_netagents[i], netagent_uuid) == 0) {
+                               if (uuid_compare(parameters->required_netagents[i], *netagent_uuid) == 0) {
                                        required = TRUE;
                                        break;
                                }
@@ -1309,7 +1668,7 @@ necp_netagent_applies_to_client(__unused struct necp_client *client,
                                        }
 
                                        if (!fetched_type) {
-                                               if (netagent_get_agent_domain_and_type(netagent_uuid, netagent_domain, netagent_type)) {
+                                               if (netagent_get_agent_domain_and_type(*netagent_uuid, netagent_domain, netagent_type)) {
                                                        fetched_type = TRUE;
                                                } else {
                                                        break;
@@ -1337,18 +1696,18 @@ necp_netagent_applies_to_client(__unused struct necp_client *client,
 }
 
 static void
-necp_client_add_agent_flows_for_interface(struct necp_client *client,
-                                                                                 const struct necp_client_parsed_parameters *parsed_parameters,
-                                                                                 ifnet_t ifp)
+necp_client_add_agent_interface_options(struct necp_client *client,
+                                                                               const struct necp_client_parsed_parameters *parsed_parameters,
+                                                                               ifnet_t ifp)
 {
        if (ifp != NULL && ifp->if_agentids != NULL) {
                for (u_int32_t i = 0; i < ifp->if_agentcount; i++) {
                        if (uuid_is_null(ifp->if_agentids[i])) {
                                continue;
                        }
-                       u_int16_t if_flags = nstat_ifnet_to_flags(ifp);
                        // Relies on the side effect that nexus agents that apply will create flows
-                       (void)necp_netagent_applies_to_client(client, parsed_parameters, ifp->if_agentids[i], TRUE, ifp->if_index, if_flags);
+                       (void)necp_netagent_applies_to_client(client, parsed_parameters, &ifp->if_agentids[i], TRUE,
+                                                                                                 ifp->if_index, ifnet_get_generation(ifp));
                }
        }
 }
@@ -1378,9 +1737,11 @@ necp_client_parse_parameters(u_int8_t *parameters,
        u_int32_t num_required_agents = 0;
        u_int32_t num_prohibited_agents = 0;
        u_int32_t num_preferred_agents = 0;
+       u_int32_t num_avoided_agents = 0;
        u_int32_t num_required_agent_types = 0;
        u_int32_t num_prohibited_agent_types = 0;
        u_int32_t num_preferred_agent_types = 0;
+       u_int32_t num_avoided_agent_types = 0;
 
        if (parsed_parameters == NULL) {
                return (EINVAL);
@@ -1536,6 +1897,17 @@ necp_client_parse_parameters(u_int8_t *parameters,
                                                }
                                                break;
                                        }
+                                       case NECP_CLIENT_PARAMETER_AVOID_AGENT: {
+                                               if (num_avoided_agents >= NECP_MAX_PARSED_PARAMETERS) {
+                                                       break;
+                                               }
+                                               if (length >= sizeof(uuid_t)) {
+                                                       memcpy(&parsed_parameters->avoided_netagents[num_avoided_agents], value, sizeof(uuid_t));
+                                                       num_avoided_agents++;
+                                                       parsed_parameters->valid_fields |= NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT;
+                                               }
+                                               break;
+                                       }
                                        case NECP_CLIENT_PARAMETER_REQUIRE_AGENT_TYPE: {
                                                if (num_required_agent_types >= NECP_MAX_PARSED_PARAMETERS) {
                                                        break;
@@ -1569,9 +1941,20 @@ necp_client_parse_parameters(u_int8_t *parameters,
                                                }
                                                break;
                                        }
-                                       case NECP_CLIENT_PARAMETER_FLAGS: {
-                                               if (length >= sizeof(u_int32_t)) {
-                                                       memcpy(&parsed_parameters->flags, value, sizeof(parsed_parameters->flags));
+                                       case NECP_CLIENT_PARAMETER_AVOID_AGENT_TYPE: {
+                                               if (num_avoided_agent_types >= NECP_MAX_PARSED_PARAMETERS) {
+                                                       break;
+                                               }
+                                               if (length >= sizeof(struct necp_client_parameter_netagent_type)) {
+                                                       memcpy(&parsed_parameters->avoided_netagent_types[num_avoided_agent_types], value, sizeof(struct necp_client_parameter_netagent_type));
+                                                       num_avoided_agent_types++;
+                                                       parsed_parameters->valid_fields |= NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE;
+                                               }
+                                               break;
+                                       }
+                                       case NECP_CLIENT_PARAMETER_FLAGS: {
+                                               if (length >= sizeof(u_int32_t)) {
+                                                       memcpy(&parsed_parameters->flags, value, sizeof(parsed_parameters->flags));
                                                        parsed_parameters->valid_fields |= NECP_PARSED_PARAMETERS_FIELD_FLAGS;
                                                }
                                                break;
@@ -1621,8 +2004,10 @@ static int
 necp_client_parse_result(u_int8_t *result,
                                                 u_int32_t result_size,
                                                 union necp_sockaddr_union *local_address,
-                                                union necp_sockaddr_union *remote_address)
+                                                union necp_sockaddr_union *remote_address,
+                                                void **flow_stats)
 {
+#pragma unused(flow_stats)
        int error = 0;
        size_t offset = 0;
 
@@ -1665,37 +2050,109 @@ necp_client_parse_result(u_int8_t *result,
        return (error);
 }
 
+static struct necp_client_flow_registration *
+necp_client_create_flow_registration(struct necp_fd_data *fd_data, struct necp_client *client)
+{
+       NECP_FD_ASSERT_LOCKED(fd_data);
+       NECP_CLIENT_ASSERT_LOCKED(client);
+
+       struct necp_client_flow_registration *new_registration = mcache_alloc(necp_flow_registration_cache, MCR_SLEEP);
+       if (new_registration == NULL) {
+               return NULL;
+       }
+
+       memset(new_registration, 0, sizeof(*new_registration));
+
+       new_registration->last_interface_details = combine_interface_details(IFSCOPE_NONE, NSTAT_IFNET_IS_UNKNOWN_TYPE);
+
+       necp_generate_client_id(new_registration->registration_id, true);
+       LIST_INIT(&new_registration->flow_list);
+
+       // Add registration to client list
+       RB_INSERT(_necp_client_flow_tree, &client->flow_registrations, new_registration);
+
+       // Add registration to fd list
+       RB_INSERT(_necp_fd_flow_tree, &fd_data->flows, new_registration);
+
+       // Add registration to global tree for lookup
+       NECP_FLOW_TREE_LOCK_EXCLUSIVE();
+       RB_INSERT(_necp_client_flow_global_tree, &necp_client_flow_global_tree, new_registration);
+       NECP_FLOW_TREE_UNLOCK();
+
+       new_registration->client = client;
+
+       // Start out assuming there is nothing to read from the flow
+       new_registration->flow_result_read = true;
+
+       return new_registration;
+}
+
+static void
+necp_client_add_socket_flow(struct necp_client_flow_registration *flow_registration,
+                                                       struct inpcb *inp)
+{
+       struct necp_client_flow *new_flow = mcache_alloc(necp_flow_cache, MCR_SLEEP);
+       if (new_flow == NULL) {
+               NECPLOG0(LOG_ERR, "Failed to allocate socket flow");
+               return;
+       }
+
+       memset(new_flow, 0, sizeof(*new_flow));
+
+       new_flow->socket = TRUE;
+       new_flow->u.socket_handle = inp;
+       new_flow->u.cb = inp->necp_cb;
+
+       OSIncrementAtomic(&necp_socket_flow_count);
+
+       LIST_INSERT_HEAD(&flow_registration->flow_list, new_flow, flow_chain);
+}
+
 int
 necp_client_register_socket_flow(pid_t pid, uuid_t client_id, struct inpcb *inp)
 {
        int error = 0;
+       struct necp_fd_data *client_fd = NULL;
        bool found_client = FALSE;
 
-       NECP_CLIENT_TREE_LOCK_SHARED();
-
-       struct necp_client find;
-       uuid_copy(find.client_id, client_id);
-       struct necp_client *client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
-       if (client != NULL) {
-               NECP_CLIENT_LOCK(client);
+       NECP_FD_LIST_LOCK_SHARED();
+       LIST_FOREACH(client_fd, &necp_fd_list, chain) {
+               NECP_FD_LOCK(client_fd);
+               struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
+               if (client != NULL) {
+                       if (!pid || client->proc_pid == pid) {
+                               struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                               if (flow_registration != NULL) {
+                                       // Found the right client and flow registration, add a new flow
+                                       found_client = TRUE;
+                                       necp_client_add_socket_flow(flow_registration, inp);
+                               } else if (RB_EMPTY(&client->flow_registrations) && !necp_client_id_is_flow(client_id)) {
+                                       // No flows yet on this client, add a new registration
+                                       flow_registration = necp_client_create_flow_registration(client_fd, client);
+                                       if (flow_registration == NULL) {
+                                               error = ENOMEM;
+                                       } else {
+                                               // Add a new flow
+                                               found_client = TRUE;
+                                               necp_client_add_socket_flow(flow_registration, inp);
+                                       }
+                               }
+                       }
 
-               if (!pid || client->proc_pid == pid) {
-                       // Found the right client!
-                       found_client = TRUE;
-                       necp_client_add_socket_flow(client, inp);
+                       NECP_CLIENT_UNLOCK(client);
                }
+               NECP_FD_UNLOCK(client_fd);
 
-               NECP_CLIENT_UNLOCK(client);
+               if (found_client) {
+                       break;
+               }
        }
-
-       NECP_CLIENT_TREE_UNLOCK();
+       NECP_FD_LIST_UNLOCK();
 
        if (!found_client) {
                error = ENOENT;
        } else {
-               /*
-                * Count the sockets that have the NECP client UUID set
-                */
+               // Count the sockets that have the NECP client UUID set
                struct socket *so = inp->inp_socket;
                if (!(so->so_flags1 & SOF1_HAS_NECP_CLIENT_UUID)) {
                        so->so_flags1 |= SOF1_HAS_NECP_CLIENT_UUID;
@@ -1707,62 +2164,192 @@ necp_client_register_socket_flow(pid_t pid, uuid_t client_id, struct inpcb *inp)
 }
 
 static void
-necp_client_add_multipath_cb(struct necp_client *client, struct mppcb *mpp)
+necp_client_add_multipath_interface_flows(struct necp_client_flow_registration *flow_registration,
+                                                                                 struct necp_client *client,
+                                                                                 struct mppcb *mpp)
 {
-       struct necp_client_flow *flow = NULL;
+       flow_registration->interface_handle = mpp;
+       flow_registration->interface_cb = mpp->necp_cb;
 
-       client->interface_handle = mpp;
-       client->interface_cb = mpp->necp_cb;
+       proc_t proc = proc_find(client->proc_pid);
+       if (proc == PROC_NULL) {
+               return;
+       }
 
-       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-               if (flow->nexus || flow->socket) {
+       // Traverse all interfaces and add a tracking flow if needed
+       necp_flow_add_interface_flows(proc, client, flow_registration, true);
+
+       proc_rele(proc);
+       proc = PROC_NULL;
+}
+
+int
+necp_client_register_multipath_cb(pid_t pid, uuid_t client_id, struct mppcb *mpp)
+{
+       int error = 0;
+       struct necp_fd_data *client_fd = NULL;
+       bool found_client = FALSE;
+
+       NECP_FD_LIST_LOCK_SHARED();
+       LIST_FOREACH(client_fd, &necp_fd_list, chain) {
+               NECP_FD_LOCK(client_fd);
+               struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
+               if (client != NULL) {
+                       if (!pid || client->proc_pid == pid) {
+                               struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                               if (flow_registration != NULL) {
+                                       // Found the right client and flow registration, add a new flow
+                                       found_client = TRUE;
+                                       necp_client_add_multipath_interface_flows(flow_registration, client, mpp);
+                               } else if (RB_EMPTY(&client->flow_registrations) && !necp_client_id_is_flow(client_id)) {
+                                       // No flows yet on this client, add a new registration
+                                       flow_registration = necp_client_create_flow_registration(client_fd, client);
+                                       if (flow_registration == NULL) {
+                                               error = ENOMEM;
+                                       } else {
+                                               // Add a new flow
+                                               found_client = TRUE;
+                                               necp_client_add_multipath_interface_flows(flow_registration, client, mpp);
+                                       }
+                               }
+                       }
+
+                       NECP_CLIENT_UNLOCK(client);
+               }
+               NECP_FD_UNLOCK(client_fd);
+
+               if (found_client) {
+                       break;
+               }
+       }
+       NECP_FD_LIST_UNLOCK();
+
+       if (!found_client && error == 0) {
+               error = ENOENT;
+       }
+
+       return (error);
+}
+
+#define        NETAGENT_DOMAIN_RADIO_MANAGER   "WirelessRadioManager"
+#define        NETAGENT_TYPE_RADIO_MANAGER     "WirelessRadioManager:BB Manager"
+
+static int
+necp_client_lookup_bb_radio_manager(struct necp_client *client,
+                                   uuid_t netagent_uuid)
+{
+       char netagent_domain[NETAGENT_DOMAINSIZE];
+       char netagent_type[NETAGENT_TYPESIZE];
+       struct necp_aggregate_result result;
+       proc_t proc;
+       int error;
+
+       proc = proc_find(client->proc_pid);
+       if (proc == PROC_NULL) {
+               return ESRCH;
+       }
+
+       error = necp_application_find_policy_match_internal(proc, client->parameters, (u_int32_t)client->parameters_length,
+                               &result, NULL, 0, NULL, NULL, NULL, true);
+
+       proc_rele(proc);
+       proc = PROC_NULL;
+
+       if (error) {
+               return error;
+       }
+
+       for (int i = 0; i < NECP_MAX_NETAGENTS; i++) {
+               if (uuid_is_null(result.netagents[i])) {
+                       // Passed end of valid agents
+                       break;
+               }
+
+               memset(&netagent_domain, 0, NETAGENT_DOMAINSIZE);
+               memset(&netagent_type, 0, NETAGENT_TYPESIZE);
+               if (netagent_get_agent_domain_and_type(result.netagents[i], netagent_domain, netagent_type) == FALSE) {
+                       continue;
+               }
+
+               if (strncmp(netagent_domain, NETAGENT_DOMAIN_RADIO_MANAGER, NETAGENT_DOMAINSIZE) != 0) {
                        continue;
                }
 
-               flow->u.socket_handle = mpp;
-               flow->u.cb = mpp->necp_cb;
+               if (strncmp(netagent_type, NETAGENT_TYPE_RADIO_MANAGER, NETAGENT_TYPESIZE) != 0) {
+                       continue;
+               }
+
+               uuid_copy(netagent_uuid, result.netagents[i]);
 
-               if (flow->viable && flow->u.cb) {
-                       flow->u.cb(mpp, NECP_CLIENT_CBACTION_INITIAL, flow);
+               break;
+       }
+
+       return 0;
+}
+
+static int
+necp_client_assert_bb_radio_manager_common(struct necp_client *client, bool assert)
+{
+       uuid_t netagent_uuid;
+       uint8_t assert_type;
+       int error;
+
+       error = necp_client_lookup_bb_radio_manager(client, netagent_uuid);
+       if (error) {
+               NECPLOG0(LOG_ERR, "BB radio manager agent not found");
+               return error;
+       }
+
+       // Before unasserting, verify that the assertion was already taken
+       if (assert == FALSE) {
+               assert_type = NETAGENT_MESSAGE_TYPE_CLIENT_UNASSERT;
+
+               if (!necp_client_remove_assertion(client, netagent_uuid)) {
+                       return EINVAL;
                }
+       } else {
+               assert_type = NETAGENT_MESSAGE_TYPE_CLIENT_ASSERT;
+       }
+
+       error = netagent_client_message(netagent_uuid, client->client_id, client->proc_pid, client->agent_handle, assert_type);
+       if (error) {
+               NECPLOG0(LOG_ERR, "netagent_client_message failed");
+               return error;
        }
+
+       // Only save the assertion if the action succeeded
+       if (assert == TRUE) {
+               necp_client_add_assertion(client, netagent_uuid);
+       }
+
+       return 0;
 }
 
 int
-necp_client_register_multipath_cb(pid_t pid, uuid_t client_id, struct mppcb *mpp)
+necp_client_assert_bb_radio_manager(uuid_t client_id, bool assert)
 {
+       struct necp_client *client;
        int error = 0;
-       bool found_client = FALSE;
 
        NECP_CLIENT_TREE_LOCK_SHARED();
 
-       struct necp_client find;
-       uuid_copy(find.client_id, client_id);
-       struct necp_client *client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
-       if (client != NULL) {
-               NECP_CLIENT_LOCK(client);
+       client = necp_find_client_and_lock(client_id);
 
-               if (!pid || client->proc_pid == pid) {
-                       // Found the right client!
-                       found_client = TRUE;
-                       necp_client_add_multipath_cb(client, mpp);
-               }
+       if (client) {
+               // Found the right client!
+               error = necp_client_assert_bb_radio_manager_common(client, assert);
 
                NECP_CLIENT_UNLOCK(client);
+       } else {
+               NECPLOG0(LOG_ERR, "Couldn't find client");
+               error = ENOENT;
        }
 
        NECP_CLIENT_TREE_UNLOCK();
 
-       if (!found_client) {
-               error = ENOENT;
-       }
-
        return (error);
 }
 
-#define        NETAGENT_DOMAIN_NETEXT  "NetworkExtension"
-#define        NETAGENT_TYPE_PATHCTRL  "PathController"
-
 static int
 necp_client_unregister_socket_flow(uuid_t client_id, void *handle)
 {
@@ -1777,22 +2364,26 @@ necp_client_unregister_socket_flow(uuid_t client_id, void *handle)
 
                struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
                if (client != NULL) {
-                       // Found the right client!
-                       found_client = TRUE;
+                       struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                       if (flow_registration != NULL) {
+                               // Found the right client and flow!
+                               found_client = TRUE;
 
-                       // Remove flow assignment
-                       struct necp_client_flow *search_flow = NULL;
-                       struct necp_client_flow *temp_flow = NULL;
-                       LIST_FOREACH_SAFE(search_flow, &client->flow_list, flow_chain, temp_flow) {
-                               if (search_flow->socket && search_flow->u.socket_handle == handle) {
-                                       if (search_flow->assigned_results != NULL) {
-                                               FREE(search_flow->assigned_results, M_NETAGENT);
-                                               search_flow->assigned_results = NULL;
+                               // Remove flow assignment
+                               struct necp_client_flow *search_flow = NULL;
+                               struct necp_client_flow *temp_flow = NULL;
+                               LIST_FOREACH_SAFE(search_flow, &flow_registration->flow_list, flow_chain, temp_flow) {
+                                       if (search_flow->socket && search_flow->u.socket_handle == handle) {
+                                               if (search_flow->assigned_results != NULL) {
+                                                       FREE(search_flow->assigned_results, M_NETAGENT);
+                                                       search_flow->assigned_results = NULL;
+                                               }
+                                               client_updated = TRUE;
+                                               flow_registration->flow_result_read = FALSE;
+                                               LIST_REMOVE(search_flow, flow_chain);
+                                               OSDecrementAtomic(&necp_socket_flow_count);
+                                               mcache_free(necp_flow_cache, search_flow);
                                        }
-                                       client_updated = TRUE;
-                                       LIST_REMOVE(search_flow, flow_chain);
-                                       OSDecrementAtomic(&necp_socket_flow_count);
-                                       mcache_free(necp_flow_cache, search_flow);
                                }
                        }
 
@@ -1800,7 +2391,6 @@ necp_client_unregister_socket_flow(uuid_t client_id, void *handle)
                }
 
                if (client_updated) {
-                       client->flow_result_read = FALSE;
                        necp_fd_notify(client_fd, true);
                }
                NECP_FD_UNLOCK(client_fd);
@@ -1826,28 +2416,27 @@ necp_client_unregister_multipath_cb(uuid_t client_id, void *handle)
 
        NECP_CLIENT_TREE_LOCK_SHARED();
 
-       struct necp_client find;
-       uuid_copy(find.client_id, client_id);
-       struct necp_client *client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
+       struct necp_client *client = necp_find_client_and_lock(client_id);
        if (client != NULL) {
-               NECP_CLIENT_LOCK(client);
-
-               // Found the right client!
-               found_client = TRUE;
+               struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+               if (flow_registration != NULL) {
+                       // Found the right client and flow!
+                       found_client = TRUE;
 
-               // Remove flow assignment
-               struct necp_client_flow *search_flow = NULL;
-               struct necp_client_flow *temp_flow = NULL;
-               LIST_FOREACH_SAFE(search_flow, &client->flow_list, flow_chain, temp_flow) {
-                       if (!search_flow->socket && !search_flow->nexus &&
-                               search_flow->u.socket_handle == handle) {
-                               search_flow->u.socket_handle = NULL;
-                               search_flow->u.cb = NULL;
+                       // Remove flow assignment
+                       struct necp_client_flow *search_flow = NULL;
+                       struct necp_client_flow *temp_flow = NULL;
+                       LIST_FOREACH_SAFE(search_flow, &flow_registration->flow_list, flow_chain, temp_flow) {
+                               if (!search_flow->socket && !search_flow->nexus &&
+                                       search_flow->u.socket_handle == handle) {
+                                       search_flow->u.socket_handle = NULL;
+                                       search_flow->u.cb = NULL;
+                               }
                        }
-               }
 
-               client->interface_handle = NULL;
-               client->interface_cb = NULL;
+                       flow_registration->interface_handle = NULL;
+                       flow_registration->interface_cb = NULL;
+               }
 
                NECP_CLIENT_UNLOCK(client);
        }
@@ -1884,58 +2473,68 @@ necp_client_assign_from_socket(pid_t pid, uuid_t client_id, struct inpcb *inp)
 
                struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
                if (client != NULL) {
-                       // Found the right client!
-                       found_client = TRUE;
-
-                       struct necp_client_flow *flow = NULL;
-                       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-                               if (flow->socket && flow->u.socket_handle == inp) {
-                                       // Release prior results and route
-                                       if (flow->assigned_results != NULL) {
-                                               FREE(flow->assigned_results, M_NETAGENT);
-                                               flow->assigned_results = NULL;
-                                       }
+                       struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                       if (flow_registration == NULL && RB_EMPTY(&client->flow_registrations) && !necp_client_id_is_flow(client_id)) {
+                               // No flows yet on this client, add a new registration
+                               flow_registration = necp_client_create_flow_registration(client_fd, client);
+                               if (flow_registration == NULL) {
+                                       error = ENOMEM;
+                               }
+                       }
+                       if (flow_registration != NULL) {
+                               // Found the right client and flow!
+                               found_client = TRUE;
 
-                                       ifnet_t ifp = NULL;
-                                       if ((inp->inp_flags & INP_BOUND_IF) && inp->inp_boundifp) {
-                                               ifp = inp->inp_boundifp;
-                                       } else {
-                                               ifp = inp->inp_last_outifp;
-                                       }
+                               struct necp_client_flow *flow = NULL;
+                               LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
+                                       if (flow->socket && flow->u.socket_handle == inp) {
+                                               // Release prior results and route
+                                               if (flow->assigned_results != NULL) {
+                                                       FREE(flow->assigned_results, M_NETAGENT);
+                                                       flow->assigned_results = NULL;
+                                               }
 
-                                       if (ifp != NULL) {
-                                               flow->interface_index = ifp->if_index;
-                                       } else {
-                                               flow->interface_index = IFSCOPE_NONE;
-                                       }
+                                               ifnet_t ifp = NULL;
+                                               if ((inp->inp_flags & INP_BOUND_IF) && inp->inp_boundifp) {
+                                                       ifp = inp->inp_boundifp;
+                                               } else {
+                                                       ifp = inp->inp_last_outifp;
+                                               }
 
-                                       if (inp->inp_vflag & INP_IPV4) {
-                                               flow->local_addr.sin.sin_family = AF_INET;
-                                               flow->local_addr.sin.sin_len = sizeof(struct sockaddr_in);
-                                               flow->local_addr.sin.sin_port = inp->inp_lport;
-                                               memcpy(&flow->local_addr.sin.sin_addr, &inp->inp_laddr, sizeof(struct in_addr));
-
-                                               flow->remote_addr.sin.sin_family = AF_INET;
-                                               flow->remote_addr.sin.sin_len = sizeof(struct sockaddr_in);
-                                               flow->remote_addr.sin.sin_port = inp->inp_fport;
-                                               memcpy(&flow->remote_addr.sin.sin_addr, &inp->inp_faddr, sizeof(struct in_addr));
-                                       } else if (inp->inp_vflag & INP_IPV6) {
-                                               in6_ip6_to_sockaddr(&inp->in6p_laddr, inp->inp_lport, &flow->local_addr.sin6, sizeof(flow->local_addr));
-                                               in6_ip6_to_sockaddr(&inp->in6p_faddr, inp->inp_fport, &flow->remote_addr.sin6, sizeof(flow->remote_addr));
-                                       }
+                                               if (ifp != NULL) {
+                                                       flow->interface_index = ifp->if_index;
+                                               } else {
+                                                       flow->interface_index = IFSCOPE_NONE;
+                                               }
 
-                                       flow->viable = necp_client_flow_is_viable(proc, client, flow);
+                                               if (inp->inp_vflag & INP_IPV4) {
+                                                       flow->local_addr.sin.sin_family = AF_INET;
+                                                       flow->local_addr.sin.sin_len = sizeof(struct sockaddr_in);
+                                                       flow->local_addr.sin.sin_port = inp->inp_lport;
+                                                       memcpy(&flow->local_addr.sin.sin_addr, &inp->inp_laddr, sizeof(struct in_addr));
+
+                                                       flow->remote_addr.sin.sin_family = AF_INET;
+                                                       flow->remote_addr.sin.sin_len = sizeof(struct sockaddr_in);
+                                                       flow->remote_addr.sin.sin_port = inp->inp_fport;
+                                                       memcpy(&flow->remote_addr.sin.sin_addr, &inp->inp_faddr, sizeof(struct in_addr));
+                                               } else if (inp->inp_vflag & INP_IPV6) {
+                                                       in6_ip6_to_sockaddr(&inp->in6p_laddr, inp->inp_lport, &flow->local_addr.sin6, sizeof(flow->local_addr));
+                                                       in6_ip6_to_sockaddr(&inp->in6p_faddr, inp->inp_fport, &flow->remote_addr.sin6, sizeof(flow->remote_addr));
+                                               }
 
-                                       uuid_t empty_uuid;
-                                       uuid_clear(empty_uuid);
-                                       flow->assigned = TRUE;
-                                       flow->assigned_results = necp_create_nexus_assign_message(empty_uuid, 0, NULL, 0,
-                                                                                                                                                         (struct necp_client_endpoint *)&flow->local_addr,
-                                                                                                                                                         (struct necp_client_endpoint *)&flow->remote_addr,
-                                                                                                                                                         0, &flow->assigned_results_length);
-                                       client->flow_result_read = FALSE;
-                                       client_updated = TRUE;
-                                       break;
+                                               flow->viable = necp_client_flow_is_viable(proc, client, flow);
+
+                                               uuid_t empty_uuid;
+                                               uuid_clear(empty_uuid);
+                                               flow->assigned = TRUE;
+                                               flow->assigned_results = necp_create_nexus_assign_message(empty_uuid, 0, NULL, 0,
+                                                                                                                                                                 (struct necp_client_endpoint *)&flow->local_addr,
+                                                                                                                                                                 (struct necp_client_endpoint *)&flow->remote_addr,
+                                                                                                                                                                 0, NULL, &flow->assigned_results_length);
+                                               flow_registration->flow_result_read = FALSE;
+                                               client_updated = TRUE;
+                                               break;
+                                       }
                                }
                        }
 
@@ -1955,10 +2554,12 @@ necp_client_assign_from_socket(pid_t pid, uuid_t client_id, struct inpcb *inp)
        }
        NECP_FD_LIST_UNLOCK();
 
-       if (!found_client) {
-               error = ENOENT;
-       } else if (!client_updated) {
-               error = EINVAL;
+       if (error == 0) {
+               if (!found_client) {
+                       error = ENOENT;
+               } else if (!client_updated) {
+                       error = EINVAL;
+               }
        }
 
        return (error);
@@ -1985,22 +2586,25 @@ necp_update_flow_protoctl_event(uuid_t netagent_uuid, uuid_t client_id,
 
                struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
                if (client != NULL) {
-                       /* Found the right client! */
-                       found_client = TRUE;
+                       struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                       if (flow_registration != NULL) {
+                               // Found the right client and flow!
+                               found_client = TRUE;
 
-                       struct necp_client_flow *flow = NULL;
-                       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-                               // Verify that the client nexus agent matches
-                               if (flow->nexus &&
-                                   uuid_compare(flow->u.nexus_agent,
-                                   netagent_uuid) == 0) {
-                                       flow->has_protoctl_event = TRUE;
-                                       flow->protoctl_event.protoctl_event_code = protoctl_event_code;
-                                       flow->protoctl_event.protoctl_event_val = protoctl_event_val;
-                                       flow->protoctl_event.protoctl_event_tcp_seq_num = protoctl_event_tcp_seq_number;
-                                       client->flow_result_read = FALSE;
-                                       client_updated = TRUE;
-                                       break;
+                               struct necp_client_flow *flow = NULL;
+                               LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
+                                       // Verify that the client nexus agent matches
+                                       if (flow->nexus &&
+                                               uuid_compare(flow->u.nexus_agent,
+                                               netagent_uuid) == 0) {
+                                               flow->has_protoctl_event = TRUE;
+                                               flow->protoctl_event.protoctl_event_code = protoctl_event_code;
+                                               flow->protoctl_event.protoctl_event_val = protoctl_event_val;
+                                               flow->protoctl_event.protoctl_event_tcp_seq_num = protoctl_event_tcp_seq_number;
+                                               flow_registration->flow_result_read = FALSE;
+                                               client_updated = TRUE;
+                                               break;
+                                       }
                                }
                        }
 
@@ -2033,6 +2637,7 @@ static bool
 necp_assign_client_result_locked(struct proc *proc,
                                                                 struct necp_fd_data *client_fd,
                                                                 struct necp_client *client,
+                                                                struct necp_client_flow_registration *flow_registration,
                                                                 uuid_t netagent_uuid,
                                                                 u_int8_t *assigned_results,
                                                                 size_t assigned_results_length,
@@ -2044,7 +2649,7 @@ necp_assign_client_result_locked(struct proc *proc,
        NECP_CLIENT_ASSERT_LOCKED(client);
 
        struct necp_client_flow *flow = NULL;
-       LIST_FOREACH(flow, &client->flow_list, flow_chain) {
+       LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
                // Verify that the client nexus agent matches
                if (flow->nexus &&
                        uuid_compare(flow->u.nexus_agent, netagent_uuid) == 0) {
@@ -2054,9 +2659,10 @@ necp_assign_client_result_locked(struct proc *proc,
                                flow->assigned_results = NULL;
                        }
 
+                       void *nexus_stats = NULL;
                        if (assigned_results != NULL && assigned_results_length > 0) {
                                int error = necp_client_parse_result(assigned_results, (u_int32_t)assigned_results_length,
-                                                                                                &flow->local_addr, &flow->remote_addr);
+                                                                                                &flow->local_addr, &flow->remote_addr, &nexus_stats);
                                VERIFY(error == 0);
                        }
 
@@ -2065,7 +2671,7 @@ necp_assign_client_result_locked(struct proc *proc,
                        flow->assigned = TRUE;
                        flow->assigned_results = assigned_results;
                        flow->assigned_results_length = assigned_results_length;
-                       client->flow_result_read = FALSE;
+                       flow_registration->flow_result_read = FALSE;
                        client_updated = TRUE;
                        break;
                }
@@ -2099,12 +2705,14 @@ necp_assign_client_result(uuid_t netagent_uuid, uuid_t client_id,
                NECP_FD_LOCK(client_fd);
                struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
                if (client != NULL) {
-                       // Found the right client!
-                       found_client = TRUE;
-
-                       if (necp_assign_client_result_locked(proc, client_fd, client, netagent_uuid,
-                                                                                                assigned_results, assigned_results_length, true)) {
-                               client_updated = TRUE;
+                       struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+                       if (flow_registration != NULL) {
+                               // Found the right client and flow!
+                               found_client = TRUE;
+                               if (necp_assign_client_result_locked(proc, client_fd, client, flow_registration, netagent_uuid,
+                                                                                                        assigned_results, assigned_results_length, true)) {
+                                       client_updated = TRUE;
+                               }
                        }
 
                        NECP_CLIENT_UNLOCK(client);
@@ -2114,36 +2722,251 @@ necp_assign_client_result(uuid_t netagent_uuid, uuid_t client_id,
                proc_rele(proc);
                proc = PROC_NULL;
 
-               if (found_client) {
-                       break;
+               if (found_client) {
+                       break;
+               }
+       }
+
+       NECP_FD_LIST_UNLOCK();
+
+       // upon error, client must free assigned_results
+       if (!found_client) {
+               error = ENOENT;
+       } else if (!client_updated) {
+               error = EINVAL;
+       }
+
+       return (error);
+}
+
+/// Client updating
+
+static bool
+necp_update_parsed_parameters(struct necp_client_parsed_parameters *parsed_parameters,
+                                                         struct necp_aggregate_result *result)
+{
+       if (parsed_parameters == NULL ||
+               result == NULL) {
+               return (false);
+       }
+
+       bool updated = false;
+       for (int i = 0; i < NECP_MAX_NETAGENTS; i++) {
+               if (uuid_is_null(result->netagents[i])) {
+                       // Passed end of valid agents
+                       break;
+               }
+
+               if (!(result->netagent_use_flags[i] & NECP_AGENT_USE_FLAG_SCOPE)) {
+                       // Not a scoped agent, ignore
+                       continue;
+               }
+
+               // This is a scoped agent. Add it to the required agents.
+               if (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT) {
+                       // Already some required agents, add this at the end
+                       for (int j = 0; j < NECP_MAX_PARSED_PARAMETERS; j++) {
+                               if (uuid_compare(parsed_parameters->required_netagents[j], result->netagents[i]) == 0) {
+                                       // Already required, break
+                                       break;
+                               }
+                               if (uuid_is_null(parsed_parameters->required_netagents[j])) {
+                                       // Add here
+                                       memcpy(&parsed_parameters->required_netagents[j], result->netagents[i], sizeof(uuid_t));
+                                       updated = true;
+                                       break;
+                               }
+                       }
+               } else {
+                       // No required agents yet, add this one
+                       parsed_parameters->valid_fields |= NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT;
+                       memcpy(&parsed_parameters->required_netagents[0], result->netagents[i], sizeof(uuid_t));
+                       updated = true;
+               }
+
+               // Remove requirements for agents of the same type
+               if (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE) {
+                       char remove_agent_domain[NETAGENT_DOMAINSIZE] = { 0 };
+                       char remove_agent_type[NETAGENT_TYPESIZE] = { 0 };
+                       if (netagent_get_agent_domain_and_type(result->netagents[i], remove_agent_domain, remove_agent_type)) {
+                               for (int j = 0; j < NECP_MAX_PARSED_PARAMETERS; j++) {
+                                       if (strlen(parsed_parameters->required_netagent_types[j].netagent_domain) == 0 &&
+                                               strlen(parsed_parameters->required_netagent_types[j].netagent_type) == 0) {
+                                               break;
+                                       }
+
+                                       if (strncmp(parsed_parameters->required_netagent_types[j].netagent_domain, remove_agent_domain, NETAGENT_DOMAINSIZE) == 0 &&
+                                               strncmp(parsed_parameters->required_netagent_types[j].netagent_type, remove_agent_type, NETAGENT_TYPESIZE) == 0) {
+
+                                               updated = true;
+
+                                               if (j == NECP_MAX_PARSED_PARAMETERS - 1) {
+                                                       // Last field, just clear and break
+                                                       memset(&parsed_parameters->required_netagent_types[NECP_MAX_PARSED_PARAMETERS - 1], 0, sizeof(struct necp_client_parameter_netagent_type));
+                                                       break;
+                                               } else {
+                                                       // Move the parameters down, clear the last entry
+                                                       memmove(&parsed_parameters->required_netagent_types[j],
+                                                                       &parsed_parameters->required_netagent_types[j + 1],
+                                                                       sizeof(struct necp_client_parameter_netagent_type) * (NECP_MAX_PARSED_PARAMETERS - (j + 1)));
+                                                       memset(&parsed_parameters->required_netagent_types[NECP_MAX_PARSED_PARAMETERS - 1], 0, sizeof(struct necp_client_parameter_netagent_type));
+                                                       // Continue, don't increment but look at the new shifted item instead
+                                                       continue;
+                                               }
+                                       }
+
+                                       // Increment j to look at the next agent type parameter
+                                       j++;
+                               }
+                       }
+               }
+       }
+
+       if (updated &&
+               parsed_parameters->required_interface_index != IFSCOPE_NONE &&
+               (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IF) == 0) {
+               // A required interface index was added after the fact. Clear it.
+               parsed_parameters->required_interface_index = IFSCOPE_NONE;
+       }
+
+
+       return (updated);
+}
+
+static inline bool
+necp_agent_types_match(const char *agent_domain1, const char *agent_type1,
+                                          const char *agent_domain2, const char *agent_type2)
+{
+       return ((strlen(agent_domain1) == 0 ||
+                        strncmp(agent_domain2, agent_domain1, NETAGENT_DOMAINSIZE) == 0) &&
+                       (strlen(agent_type1) == 0 ||
+                        strncmp(agent_type2, agent_type1, NETAGENT_TYPESIZE) == 0));
+}
+
+static inline bool
+necp_calculate_client_result(proc_t proc,
+                                                        struct necp_client *client,
+                                                        struct necp_client_parsed_parameters *parsed_parameters,
+                                                        struct necp_aggregate_result *result,
+                                                        u_int32_t *flags)
+{
+       struct rtentry *route = NULL;
+
+       // Check parameters to find best interface
+       bool validate_agents = false;
+       u_int matching_if_index = 0;
+       if (necp_find_matching_interface_index(parsed_parameters, &matching_if_index, &validate_agents)) {
+               if (matching_if_index != 0) {
+                       parsed_parameters->required_interface_index = matching_if_index;
+               }
+               // Interface found or not needed, match policy.
+               memset(result, 0, sizeof(*result));
+               int error = necp_application_find_policy_match_internal(proc, client->parameters,
+                                                                                                                               (u_int32_t)client->parameters_length,
+                                                                                                                               result, flags, matching_if_index,
+                                                                                                                               NULL, NULL, &route, false);
+               if (error != 0) {
+                       if (route != NULL) {
+                               rtfree(route);
+                       }
+                       return (FALSE);
+               }
+
+               if (validate_agents) {
+                       bool requirement_failed = FALSE;
+                       if (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT) {
+                               for (int i = 0; i < NECP_MAX_PARSED_PARAMETERS; i++) {
+                                       if (uuid_is_null(parsed_parameters->required_netagents[i])) {
+                                               break;
+                                       }
+
+                                       bool requirement_found = FALSE;
+                                       for (int j = 0; j < NECP_MAX_NETAGENTS; j++) {
+                                               if (uuid_is_null(result->netagents[j])) {
+                                                       break;
+                                               }
+
+                                               if (uuid_compare(parsed_parameters->required_netagents[i], result->netagents[j]) == 0) {
+                                                       requirement_found = TRUE;
+                                                       break;
+                                               }
+                                       }
+
+                                       if (!requirement_found) {
+                                               requirement_failed = TRUE;
+                                               break;
+                                       }
+                               }
+                       }
+
+                       if (!requirement_failed && parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE) {
+                               for (int i = 0; i < NECP_MAX_PARSED_PARAMETERS; i++) {
+                                       if (strlen(parsed_parameters->required_netagent_types[i].netagent_domain) == 0 &&
+                                               strlen(parsed_parameters->required_netagent_types[i].netagent_type) == 0) {
+                                               break;
+                                       }
+
+                                       bool requirement_found = FALSE;
+                                       for (int j = 0; j < NECP_MAX_NETAGENTS; j++) {
+                                               if (uuid_is_null(result->netagents[j])) {
+                                                       break;
+                                               }
+
+                                               char policy_agent_domain[NETAGENT_DOMAINSIZE] = { 0 };
+                                               char policy_agent_type[NETAGENT_TYPESIZE] = { 0 };
+
+                                               if (netagent_get_agent_domain_and_type(result->netagents[j], policy_agent_domain, policy_agent_type)) {
+                                                       if (necp_agent_types_match(parsed_parameters->required_netagent_types[i].netagent_domain,
+                                                                                                          parsed_parameters->required_netagent_types[i].netagent_type,
+                                                                                                          policy_agent_domain, policy_agent_type)) {
+                                                               requirement_found = TRUE;
+                                                               break;
+                                                       }
+                                               }
+                                       }
+
+                                       if (!requirement_found) {
+                                               requirement_failed = TRUE;
+                                               break;
+                                       }
+                               }
+                       }
+
+                       if (requirement_failed) {
+                               // Agent requirement failed. Clear out the whole result, make everything fail.
+                               memset(result, 0, sizeof(*result));
+                               if (route != NULL) {
+                                       rtfree(route);
+                               }
+                               return (TRUE);
+                       }
                }
-       }
-
-       NECP_FD_LIST_UNLOCK();
 
-       // upon error, client must free assigned_results
-       if (!found_client) {
-               error = ENOENT;
-       } else if (!client_updated) {
-               error = EINVAL;
+               // Reset current route
+               NECP_CLIENT_ROUTE_LOCK(client);
+               if (client->current_route != NULL) {
+                       rtfree(client->current_route);
+               }
+               client->current_route = route;
+               NECP_CLIENT_ROUTE_UNLOCK(client);
+       } else {
+               // Interface not found. Clear out the whole result, make everything fail.
+               memset(result, 0, sizeof(*result));
        }
 
-       return (error);
+       return (TRUE);
 }
 
-/// Client updating
-
 static bool
 necp_update_client_result(proc_t proc,
                                                  struct necp_fd_data *client_fd,
                                                  struct necp_client *client,
-                                                 struct _necp_client_defunct_list *defunct_list)
+                                                 struct _necp_flow_defunct_list *defunct_list)
 {
        struct necp_client_result_netagent netagent;
        struct necp_aggregate_result result;
        struct necp_client_parsed_parameters *parsed_parameters = NULL;
        u_int32_t flags = 0;
-       struct rtentry *route = NULL;
 
        NECP_CLIENT_ASSERT_LOCKED(client);
 
@@ -2165,35 +2988,18 @@ necp_update_client_result(proc_t proc,
        // Update saved IP protocol
        client->ip_protocol = parsed_parameters->ip_protocol;
 
-       // Check parameters to find best interface
-       u_int matching_if_index = 0;
-       if (necp_find_matching_interface_index(parsed_parameters, &matching_if_index)) {
-               if (matching_if_index != 0) {
-                       parsed_parameters->required_interface_index = matching_if_index;
-               }
-               // Interface found or not needed, match policy.
-               error = necp_application_find_policy_match_internal(proc, client->parameters,
-                                                                                                                       (u_int32_t)client->parameters_length,
-                                                                                                                       &result, &flags, matching_if_index,
-                                                                                                                       NULL, NULL, &route, false);
-               if (error != 0) {
-                       if (route != NULL) {
-                               rtfree(route);
-                       }
+       // Calculate the policy result
+       if (!necp_calculate_client_result(proc, client, parsed_parameters, &result, &flags)) {
+               FREE(parsed_parameters, M_NECP);
+               return (FALSE);
+       }
+
+       if (necp_update_parsed_parameters(parsed_parameters, &result)) {
+               // Changed the parameters based on result, try again (only once)
+               if (!necp_calculate_client_result(proc, client, parsed_parameters, &result, &flags)) {
                        FREE(parsed_parameters, M_NECP);
                        return (FALSE);
                }
-
-               // Reset current route
-               NECP_CLIENT_ROUTE_LOCK(client);
-               if (client->current_route != NULL) {
-                       rtfree(client->current_route);
-               }
-               client->current_route = route;
-               NECP_CLIENT_ROUTE_UNLOCK(client);
-       } else {
-               // Interface not found. Clear out the whole result, make everything fail.
-               memset(&result, 0, sizeof(result));
        }
 
        // Save the last policy id on the client
@@ -2223,9 +3029,6 @@ necp_update_client_result(proc_t proc,
        }
 
        // Recalculate flags
-       if (client->defunct) {
-               flags |= NECP_CLIENT_RESULT_FLAG_DEFUNCT;
-       }
        if (parsed_parameters->flags & NECP_CLIENT_PARAMETER_FLAG_LISTENER) {
                // Listeners are valid as long as they aren't dropped
                if (result.routing_result != NECP_KERNEL_POLICY_RESULT_DROP) {
@@ -2303,7 +3106,7 @@ necp_update_client_result(proc_t proc,
                }
                uuid_copy(netagent.netagent_uuid, result.netagents[i]);
                netagent.generation = netagent_get_generation(netagent.netagent_uuid);
-               if (necp_netagent_applies_to_client(client, parsed_parameters, netagent.netagent_uuid, TRUE, 0, 0)) {
+               if (necp_netagent_applies_to_client(client, parsed_parameters, &netagent.netagent_uuid, TRUE, 0, 0)) {
                        cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_NETAGENT, sizeof(netagent), &netagent, &updated,
                                                                                                                client->result, sizeof(client->result));
                }
@@ -2374,10 +3177,11 @@ necp_update_client_result(proc_t proc,
                TAILQ_FOREACH(multi_interface, &ifnet_ordered_head, if_ordered_link) {
                        if (necp_ifnet_matches_parameters(multi_interface, parsed_parameters, NULL, true)) {
                                // Add multipath interface flows for kernel MPTCP
-                               necp_client_add_interface_flow_if_needed(client, multi_interface->if_index);
+                               necp_client_add_interface_option_if_needed(client, multi_interface->if_index,
+                                                                                                                  ifnet_get_generation(multi_interface), NULL);
 
                                // Add nexus agents for multipath
-                               necp_client_add_agent_flows_for_interface(client, parsed_parameters, multi_interface);
+                               necp_client_add_agent_interface_options(client, parsed_parameters, multi_interface);
                        }
                }
        } else if ((parsed_parameters->flags & NECP_CLIENT_PARAMETER_FLAG_LISTENER) &&
@@ -2385,9 +3189,9 @@ necp_update_client_result(proc_t proc,
                // Get listener interface options from global list
                struct ifnet *listen_interface = NULL;
                TAILQ_FOREACH(listen_interface, &ifnet_head, if_link) {
-                       if (necp_ifnet_matches_parameters(listen_interface, parsed_parameters, NULL, false)) {
+                       if (necp_ifnet_matches_parameters(listen_interface, parsed_parameters, NULL, true)) {
                                // Add nexus agents for listeners
-                               necp_client_add_agent_flows_for_interface(client, parsed_parameters, listen_interface);
+                               necp_client_add_agent_interface_options(client, parsed_parameters, listen_interface);
                        }
                }
        }
@@ -2400,10 +3204,10 @@ necp_update_client_result(proc_t proc,
                                if (uuid_is_null(original_scoped_interface->if_agentids[i])) {
                                        continue;
                                }
-                               u_int16_t if_flags = nstat_ifnet_to_flags(original_scoped_interface);
                                uuid_copy(netagent.netagent_uuid, original_scoped_interface->if_agentids[i]);
                                netagent.generation = netagent_get_generation(netagent.netagent_uuid);
-                               if (necp_netagent_applies_to_client(client, parsed_parameters, netagent.netagent_uuid, FALSE, original_scoped_interface->if_index, if_flags)) {
+                               if (necp_netagent_applies_to_client(client, parsed_parameters, &netagent.netagent_uuid, FALSE,
+                                                                                                       original_scoped_interface->if_index, ifnet_get_generation(original_scoped_interface))) {
                                        cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_NETAGENT, sizeof(netagent), &netagent, &updated,
                                                                                                                                client->result, sizeof(client->result));
                                }
@@ -2418,10 +3222,10 @@ necp_update_client_result(proc_t proc,
                                if (uuid_is_null(direct_interface->if_agentids[i])) {
                                        continue;
                                }
-                               u_int16_t if_flags = nstat_ifnet_to_flags(direct_interface);
                                uuid_copy(netagent.netagent_uuid, direct_interface->if_agentids[i]);
                                netagent.generation = netagent_get_generation(netagent.netagent_uuid);
-                               if (necp_netagent_applies_to_client(client, parsed_parameters, netagent.netagent_uuid, TRUE, direct_interface->if_index, if_flags)) {
+                               if (necp_netagent_applies_to_client(client, parsed_parameters, &netagent.netagent_uuid, TRUE,
+                                                                                                       direct_interface->if_index, ifnet_get_generation(direct_interface))) {
                                        cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_NETAGENT, sizeof(netagent), &netagent, &updated,
                                                                                                                                client->result, sizeof(client->result));
                                }
@@ -2436,10 +3240,10 @@ necp_update_client_result(proc_t proc,
                                if (uuid_is_null(delegate_interface->if_agentids[i])) {
                                        continue;
                                }
-                               u_int16_t if_flags = nstat_ifnet_to_flags(delegate_interface);
                                uuid_copy(netagent.netagent_uuid, delegate_interface->if_agentids[i]);
                                netagent.generation = netagent_get_generation(netagent.netagent_uuid);
-                               if (necp_netagent_applies_to_client(client, parsed_parameters, netagent.netagent_uuid, FALSE, delegate_interface->if_index, if_flags)) {
+                               if (necp_netagent_applies_to_client(client, parsed_parameters, &netagent.netagent_uuid, FALSE,
+                                                                                                       delegate_interface->if_index, ifnet_get_generation(delegate_interface))) {
                                        cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_NETAGENT, sizeof(netagent), &netagent, &updated,
                                                                                                                                client->result, sizeof(client->result));
                                }
@@ -2449,6 +3253,19 @@ necp_update_client_result(proc_t proc,
        }
        ifnet_head_done();
 
+       // Add interface options
+       for (u_int32_t option_i = 0; option_i < client->interface_option_count; option_i++) {
+               if (option_i < NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT) {
+                       struct necp_client_interface_option *option = &client->interface_options[option_i];
+                       cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_INTERFACE_OPTION, sizeof(*option), option, &updated,
+                                                                                                               client->result, sizeof(client->result));
+               } else {
+                       struct necp_client_interface_option *option = &client->extra_interface_options[option_i - NECP_CLIENT_INTERFACE_OPTION_STATIC_COUNT];
+                       cursor = necp_buffer_write_tlv_if_different(cursor, NECP_CLIENT_RESULT_INTERFACE_OPTION, sizeof(*option), option, &updated,
+                                                                                                               client->result, sizeof(client->result));
+               }
+       }
+
        size_t new_result_length = (cursor - client->result);
        if (new_result_length != client->result_length) {
                client->result_length = new_result_length;
@@ -2456,14 +3273,8 @@ necp_update_client_result(proc_t proc,
        }
 
        // Update flow viability/flags
-       bool defuncted_by_flow = FALSE;
-       if (necp_client_update_flows(proc, client, defunct_list, &defuncted_by_flow)) {
+       if (necp_client_update_flows(proc, client, defunct_list)) {
                updated = TRUE;
-               if (defuncted_by_flow && client->defunct) {
-                       // Reset initial TLV
-                       flags |= NECP_CLIENT_RESULT_FLAG_DEFUNCT;
-                       (void)necp_buffer_write_tlv_if_different(client->result, NECP_CLIENT_RESULT_FLAGS, sizeof(flags), &flags, &updated, client->result, sizeof(client->result));
-               }
        }
 
        if (updated) {
@@ -2476,7 +3287,7 @@ necp_update_client_result(proc_t proc,
 }
 
 static inline void
-necp_defunct_client_fd_locked(struct necp_fd_data *client_fd, struct _necp_client_defunct_list *defunct_list, struct proc *proc)
+necp_defunct_client_fd_locked(struct necp_fd_data *client_fd, struct _necp_flow_defunct_list *defunct_list, struct proc *proc)
 {
 #pragma unused(proc)
        bool updated_result = FALSE;
@@ -2485,27 +3296,34 @@ necp_defunct_client_fd_locked(struct necp_fd_data *client_fd, struct _necp_clien
        NECP_FD_ASSERT_LOCKED(client_fd);
 
        RB_FOREACH(client, _necp_client_tree, &client_fd->clients) {
+               struct necp_client_flow_registration *flow_registration = NULL;
+
                NECP_CLIENT_LOCK(client);
-               if (!client->defunct) {
-                       updated_result = necp_set_client_defunct(client);
 
-                       // Prepare close events to be sent to the nexus to effectively remove the flows
-                       struct necp_client_flow *search_flow = NULL;
-                       LIST_FOREACH(search_flow, &client->flow_list, flow_chain) {
+               // Prepare close events to be sent to the nexus to effectively remove the flows
+               struct necp_client_flow *search_flow = NULL;
+               RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+                       LIST_FOREACH(search_flow, &flow_registration->flow_list, flow_chain) {
                                if (search_flow->nexus &&
-                                       !uuid_is_null(search_flow->u.nexus_agent) &&
-                                       search_flow->requested_nexus) {
+                                       !uuid_is_null(search_flow->u.nexus_agent)) {
 
-                                       struct necp_client_defunct *client_defunct;
+                                       struct necp_flow_defunct *flow_defunct;
 
                                        // Sleeping alloc won't fail; copy only what's necessary
-                                       client_defunct = _MALLOC(sizeof (struct necp_client_defunct), M_NECP, M_WAITOK | M_ZERO);
-                                       uuid_copy(client_defunct->nexus_agent, search_flow->u.nexus_agent);
-                                       uuid_copy(client_defunct->client_id, client->client_id);
-                                       client_defunct->proc_pid = client->proc_pid;
+                                       flow_defunct = _MALLOC(sizeof (struct necp_flow_defunct), M_NECP, M_WAITOK | M_ZERO);
+                                       uuid_copy(flow_defunct->nexus_agent, search_flow->u.nexus_agent);
+                                       uuid_copy(flow_defunct->flow_id, ((flow_registration->flags & NECP_CLIENT_FLOW_FLAGS_USE_CLIENT_ID) ?
+                                                                                                         client->client_id :
+                                                                                                         flow_registration->registration_id));
+                                       flow_defunct->proc_pid = client->proc_pid;
+                                       flow_defunct->agent_handle = client->agent_handle;
 
                                        // Add to the list provided by caller
-                                       LIST_INSERT_HEAD(defunct_list, client_defunct, chain);
+                                       LIST_INSERT_HEAD(defunct_list, flow_defunct, chain);
+
+                                       flow_registration->defunct = true;
+                                       flow_registration->flow_result_read = false;
+                                       updated_result = true;
                                }
                        }
                }
@@ -2521,7 +3339,7 @@ necp_defunct_client_fd_locked(struct necp_fd_data *client_fd, struct _necp_clien
 static inline void
 necp_update_client_fd_locked(struct necp_fd_data *client_fd,
                                                         proc_t proc,
-                                                        struct _necp_client_defunct_list *defunct_list)
+                                                        struct _necp_flow_defunct_list *defunct_list)
 {
        struct necp_client *client = NULL;
        bool updated_result = FALSE;
@@ -2545,7 +3363,7 @@ necp_update_all_clients_callout(__unused thread_call_param_t dummy,
 {
        struct necp_fd_data *client_fd = NULL;
 
-       struct _necp_client_defunct_list defunct_list;
+       struct _necp_flow_defunct_list defunct_list;
        LIST_INIT(&defunct_list);
 
        NECP_FD_LIST_LOCK_SHARED();
@@ -2569,25 +3387,26 @@ necp_update_all_clients_callout(__unused thread_call_param_t dummy,
 
        // Handle the case in which some clients became newly defunct
        if (!LIST_EMPTY(&defunct_list)) {
-               struct necp_client_defunct *client_defunct = NULL;
-               struct necp_client_defunct *temp_client_defunct = NULL;
+               struct necp_flow_defunct *flow_defunct = NULL;
+               struct necp_flow_defunct *temp_flow_defunct = NULL;
 
                // For each newly defunct client, send a message to the nexus to remove the flow
-               LIST_FOREACH_SAFE(client_defunct, &defunct_list, chain, temp_client_defunct) {
-                       if (!uuid_is_null(client_defunct->nexus_agent)) {
-                               int netagent_error = netagent_client_message(client_defunct->nexus_agent,
-                                                                                                                        client_defunct->client_id,
-                                                                                                                        client_defunct->proc_pid,
+               LIST_FOREACH_SAFE(flow_defunct, &defunct_list, chain, temp_flow_defunct) {
+                       if (!uuid_is_null(flow_defunct->nexus_agent)) {
+                               int netagent_error = netagent_client_message(flow_defunct->nexus_agent,
+                                                                                                                        flow_defunct->flow_id,
+                                                                                                                        flow_defunct->proc_pid,
+                                                                                                                        flow_defunct->agent_handle,
                                                                                                                         NETAGENT_MESSAGE_TYPE_ABORT_NEXUS);
                                if (netagent_error != 0) {
                                        char namebuf[MAXCOMLEN+1];
                                        (void) strlcpy(namebuf, "unknown", sizeof (namebuf));
-                                       proc_name(client_defunct->proc_pid, namebuf, sizeof (namebuf));
-                                       NECPLOG((netagent_error == ENOENT ? LOG_DEBUG : LOG_ERR), "necp_update_client abort nexus error (%d) for pid %d %s", netagent_error, client_defunct->proc_pid, namebuf);
+                                       proc_name(flow_defunct->proc_pid, namebuf, sizeof (namebuf));
+                                       NECPLOG((netagent_error == ENOENT ? LOG_DEBUG : LOG_ERR), "necp_update_client abort nexus error (%d) for pid %d %s", netagent_error, flow_defunct->proc_pid, namebuf);
                                }
                        }
-                       LIST_REMOVE(client_defunct, chain);
-                       FREE(client_defunct, M_NECP);
+                       LIST_REMOVE(flow_defunct, chain);
+                       FREE(flow_defunct, M_NECP);
                }
        }
        ASSERT(LIST_EMPTY(&defunct_list));
@@ -2646,11 +3465,14 @@ necp_set_client_as_background(proc_t proc,
                NECP_CLIENT_LOCK(client);
 
                bool has_assigned_flow = FALSE;
+               struct necp_client_flow_registration *flow_registration = NULL;
                struct necp_client_flow *search_flow = NULL;
-               LIST_FOREACH(search_flow, &client->flow_list, flow_chain) {
-                       if (search_flow->assigned) {
-                               has_assigned_flow = TRUE;
-                               break;
+               RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+                       LIST_FOREACH(search_flow, &flow_registration->flow_list, flow_chain) {
+                               if (search_flow->assigned) {
+                                       has_assigned_flow = TRUE;
+                                       break;
+                               }
                        }
                }
 
@@ -2683,7 +3505,7 @@ necp_fd_memstatus(proc_t proc, uint32_t status,
 void
 necp_fd_defunct(proc_t proc, struct necp_fd_data *client_fd)
 {
-       struct _necp_client_defunct_list defunct_list;
+       struct _necp_flow_defunct_list defunct_list;
 
        ASSERT(proc != PROC_NULL);
        ASSERT(client_fd != NULL);
@@ -2702,22 +3524,23 @@ necp_fd_defunct(proc_t proc, struct necp_fd_data *client_fd)
        NECP_FD_UNLOCK(client_fd);
 
        if (!LIST_EMPTY(&defunct_list)) {
-               struct necp_client_defunct *client_defunct = NULL;
-               struct necp_client_defunct *temp_client_defunct = NULL;
+               struct necp_flow_defunct *flow_defunct = NULL;
+               struct necp_flow_defunct *temp_flow_defunct = NULL;
 
                // For each defunct client, remove flow from the nexus
-               LIST_FOREACH_SAFE(client_defunct, &defunct_list, chain, temp_client_defunct) {
-                       if (!uuid_is_null(client_defunct->nexus_agent)) {
-                               int netagent_error = netagent_client_message(client_defunct->nexus_agent,
-                                                                                                                        client_defunct->client_id,
-                                                                                                                        client_defunct->proc_pid,
+               LIST_FOREACH_SAFE(flow_defunct, &defunct_list, chain, temp_flow_defunct) {
+                       if (!uuid_is_null(flow_defunct->nexus_agent)) {
+                               int netagent_error = netagent_client_message(flow_defunct->nexus_agent,
+                                                                                                                        flow_defunct->flow_id,
+                                                                                                                        flow_defunct->proc_pid,
+                                                                                                                        flow_defunct->agent_handle,
                                                                                                                         NETAGENT_MESSAGE_TYPE_ABORT_NEXUS);
                                if (netagent_error != 0) {
                                        NECPLOG((netagent_error == ENOENT ? LOG_DEBUG : LOG_ERR), "necp_defunct_client abort nexus error (%d)", netagent_error);
                                }
                        }
-                       LIST_REMOVE(client_defunct, chain);
-                       FREE(client_defunct, M_NECP);
+                       LIST_REMOVE(flow_defunct, chain);
+                       FREE(flow_defunct, M_NECP);
                }
        }
        ASSERT(LIST_EMPTY(&defunct_list));
@@ -2757,7 +3580,7 @@ necp_client_remove_agent_from_result(struct necp_client *client, uuid_t netagent
 }
 
 void
-necp_force_update_client(uuid_t client_id, uuid_t remove_netagent_uuid)
+necp_force_update_client(uuid_t client_id, uuid_t remove_netagent_uuid, u_int32_t agent_generation)
 {
        struct necp_fd_data *client_fd = NULL;
 
@@ -2768,10 +3591,12 @@ necp_force_update_client(uuid_t client_id, uuid_t remove_netagent_uuid)
                NECP_FD_LOCK(client_fd);
                struct necp_client *client = necp_client_fd_find_client_and_lock(client_fd, client_id);
                if (client != NULL) {
+                       client->failed_trigger_agent.generation = agent_generation;
+                       uuid_copy(client->failed_trigger_agent.netagent_uuid, remove_netagent_uuid);
                        if (!uuid_is_null(remove_netagent_uuid)) {
                                necp_client_remove_agent_from_result(client, remove_netagent_uuid);
                        }
-                       client->flow_result_read = FALSE;
+                       client->result_read = FALSE;
                        // Found the client, break
                        updated_result = TRUE;
                        NECP_CLIENT_UNLOCK(client);
@@ -2799,19 +3624,28 @@ necp_force_update_client(uuid_t client_id, uuid_t remove_netagent_uuid)
                                                                                                                 NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT |                  \
                                                                                                                 NECP_PARSED_PARAMETERS_FIELD_PROHIBITED_AGENT |                \
                                                                                                                 NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT |                 \
+                                                                                                                NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT |                   \
                                                                                                                 NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE |             \
                                                                                                                 NECP_PARSED_PARAMETERS_FIELD_PROHIBITED_AGENT_TYPE |   \
-                                                                                                                NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE)
-
-#define NECP_PARSED_PARAMETERS_SCOPED_IFNET_FIELDS (NECP_PARSED_PARAMETERS_FIELD_LOCAL_ADDR |                  \
-                                                                                                       NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IFTYPE |          \
-                                                                                                       NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT |           \
-                                                                                                       NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT |          \
-                                                                                                       NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE |      \
-                                                                                                       NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE)
-
-#define NECP_PARSED_PARAMETERS_PREFERRED_IFNET_FIELDS (NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT | \
-                                                                                                          NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE)
+                                                                                                                NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE |    \
+                                                                                                                NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE)
+
+#define NECP_PARSED_PARAMETERS_SCOPED_FIELDS (NECP_PARSED_PARAMETERS_FIELD_LOCAL_ADDR |                        \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IFTYPE |                \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT |         \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT |                \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT |          \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_REQUIRED_AGENT_TYPE |    \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE |   \
+                                                                                         NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE)
+
+#define NECP_PARSED_PARAMETERS_SCOPED_IFNET_FIELDS (NECP_PARSED_PARAMETERS_FIELD_LOCAL_ADDR |          \
+                                                                                                       NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IFTYPE)
+
+#define NECP_PARSED_PARAMETERS_PREFERRED_FIELDS (NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT |                \
+                                                                                                NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT |                   \
+                                                                                                NECP_PARSED_PARAMETERS_FIELD_PREFERRED_AGENT_TYPE |    \
+                                                                                                NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE)
 
 static bool
 necp_ifnet_matches_type(struct ifnet *ifp, u_int8_t interface_type, bool check_delegates)
@@ -2872,7 +3706,7 @@ necp_ifnet_matches_agent(struct ifnet *ifp, uuid_t *agent_uuid, bool check_deleg
 }
 
 static bool
-necp_necp_ifnet_matches_agent_type(struct ifnet *ifp, const char *agent_domain, const char *agent_type, bool check_delegates)
+necp_ifnet_matches_agent_type(struct ifnet *ifp, const char *agent_domain, const char *agent_type, bool check_delegates)
 {
        struct ifnet *check_ifp = ifp;
 
@@ -2888,13 +3722,10 @@ necp_necp_ifnet_matches_agent_type(struct ifnet *ifp, const char *agent_domain,
                                char if_agent_type[NETAGENT_TYPESIZE] = { 0 };
 
                                if (netagent_get_agent_domain_and_type(check_ifp->if_agentids[index], if_agent_domain, if_agent_type)) {
-                                       if ((strlen(agent_domain) == 0 ||
-                                                strncmp(if_agent_domain, agent_domain, NETAGENT_DOMAINSIZE) == 0) &&
-                                               (strlen(agent_type) == 0 ||
-                                                strncmp(if_agent_type, agent_type, NETAGENT_TYPESIZE) == 0)) {
-                                                       ifnet_lock_done(check_ifp);
-                                                       return (TRUE);
-                                               }
+                                       if (necp_agent_types_match(agent_domain, agent_type, if_agent_domain, if_agent_type)) {
+                                               ifnet_lock_done(check_ifp);
+                                               return (TRUE);
+                                       }
                                }
                        }
                }
@@ -2952,10 +3783,13 @@ necp_interface_type_is_primary_eligible(u_int8_t interface_type)
 
 #define NECP_IFP_IS_ON_ORDERED_LIST(_ifp) ((_ifp)->if_ordered_link.tqe_next != NULL || (_ifp)->if_ordered_link.tqe_prev != NULL)
 
+// Secondary interface flag indicates that the interface is being
+// used for multipath or a listener as an extra path
 static bool
 necp_ifnet_matches_parameters(struct ifnet *ifp,
                                                          struct necp_client_parsed_parameters *parsed_parameters,
-                                                         u_int32_t *preferred_count, bool ignore_require_if)
+                                                         u_int32_t *preferred_count,
+                                                         bool secondary_interface)
 {
        if (preferred_count) {
                *preferred_count = 0;
@@ -2974,7 +3808,9 @@ necp_ifnet_matches_parameters(struct ifnet *ifp,
                }
        }
 
-       if (!ignore_require_if &&
+       if ((!secondary_interface || // Enforce interface type if this is the primary interface
+                !(parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_FLAGS) || // or if there are no flags
+                !(parsed_parameters->flags & NECP_CLIENT_PARAMETER_FLAG_ONLY_PRIMARY_REQUIRES_TYPE)) && // or if the flags don't give an exception
            (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IFTYPE) &&
                !necp_ifnet_matches_type(ifp, parsed_parameters->required_interface_type, FALSE)) {
                return (FALSE);
@@ -3035,7 +3871,7 @@ necp_ifnet_matches_parameters(struct ifnet *ifp,
                                break;
                        }
 
-                       if (!necp_necp_ifnet_matches_agent_type(ifp, parsed_parameters->required_netagent_types[i].netagent_domain, parsed_parameters->required_netagent_types[i].netagent_type, FALSE)) {
+                       if (!necp_ifnet_matches_agent_type(ifp, parsed_parameters->required_netagent_types[i].netagent_domain, parsed_parameters->required_netagent_types[i].netagent_type, FALSE)) {
                                return (FALSE);
                        }
                }
@@ -3048,7 +3884,7 @@ necp_ifnet_matches_parameters(struct ifnet *ifp,
                                break;
                        }
 
-                       if (necp_necp_ifnet_matches_agent_type(ifp, parsed_parameters->prohibited_netagent_types[i].netagent_domain, parsed_parameters->prohibited_netagent_types[i].netagent_type, TRUE)) {
+                       if (necp_ifnet_matches_agent_type(ifp, parsed_parameters->prohibited_netagent_types[i].netagent_domain, parsed_parameters->prohibited_netagent_types[i].netagent_type, TRUE)) {
                                return (FALSE);
                        }
                }
@@ -3075,7 +3911,33 @@ necp_ifnet_matches_parameters(struct ifnet *ifp,
                                        break;
                                }
 
-                               if (necp_necp_ifnet_matches_agent_type(ifp, parsed_parameters->preferred_netagent_types[i].netagent_domain, parsed_parameters->preferred_netagent_types[i].netagent_type, TRUE)) {
+                               if (necp_ifnet_matches_agent_type(ifp, parsed_parameters->preferred_netagent_types[i].netagent_domain, parsed_parameters->preferred_netagent_types[i].netagent_type, TRUE)) {
+                                       (*preferred_count)++;
+                               }
+                       }
+               }
+
+               if (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT) {
+                       for (int i = 0; i < NECP_MAX_PARSED_PARAMETERS; i++) {
+                               if (uuid_is_null(parsed_parameters->avoided_netagents[i])) {
+                                       break;
+                               }
+
+                               if (!necp_ifnet_matches_agent(ifp, &parsed_parameters->avoided_netagents[i], TRUE)) {
+                                       (*preferred_count)++;
+                               }
+                       }
+               }
+
+               if (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_AVOIDED_AGENT_TYPE) {
+                       for (int i = 0; i < NECP_MAX_PARSED_PARAMETERS; i++) {
+                               if (strlen(parsed_parameters->avoided_netagent_types[i].netagent_domain) == 0 &&
+                                       strlen(parsed_parameters->avoided_netagent_types[i].netagent_type) == 0) {
+                                       break;
+                               }
+
+                               if (!necp_ifnet_matches_agent_type(ifp, parsed_parameters->avoided_netagent_types[i].netagent_domain,
+                                                                                                               parsed_parameters->avoided_netagent_types[i].netagent_type, TRUE)) {
                                        (*preferred_count)++;
                                }
                        }
@@ -3086,7 +3948,8 @@ necp_ifnet_matches_parameters(struct ifnet *ifp,
 }
 
 static bool
-necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_parameters, u_int *return_ifindex)
+necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_parameters,
+                                                                  u_int *return_ifindex, bool *validate_agents)
 {
        struct ifnet *ifp = NULL;
        u_int32_t best_preferred_count = 0;
@@ -3102,12 +3965,12 @@ necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_
                return (TRUE);
        }
 
-       has_preferred_fields = (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_PREFERRED_IFNET_FIELDS);
+       has_preferred_fields = (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_PREFERRED_FIELDS);
 
        // We have interesting parameters to parse and find a matching interface
        ifnet_head_lock_shared();
 
-       if (!(parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_SCOPED_IFNET_FIELDS)) {
+       if (!(parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_SCOPED_FIELDS)) {
                // We do have fields to match, but they are only prohibitory
                // If the first interface in the list matches, or there are no ordered interfaces, we don't need to scope
                ifp = TAILQ_FIRST(&ifnet_ordered_head);
@@ -3137,7 +4000,7 @@ necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_
        }
 
        // Then check the remaining interfaces
-       if ((parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_SCOPED_IFNET_FIELDS) &&
+       if ((parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_SCOPED_FIELDS) &&
                ((!(parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_FIELD_REQUIRED_IFTYPE)) ||
                 !necp_interface_type_is_primary_eligible(parsed_parameters->required_interface_type)) &&
                *return_ifindex == 0) {
@@ -3165,13 +4028,21 @@ necp_find_matching_interface_index(struct necp_client_parsed_parameters *parsed_
 
        ifnet_head_done();
 
-       if ((parsed_parameters->valid_fields == (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_PREFERRED_IFNET_FIELDS)) &&
+       if ((parsed_parameters->valid_fields == (parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_PREFERRED_FIELDS)) &&
                best_preferred_count == 0) {
                // If only has preferred fields, and nothing was found, clear the interface index and return TRUE
                *return_ifindex = 0;
                return (TRUE);
        }
 
+       if (*return_ifindex == 0 &&
+               !(parsed_parameters->valid_fields & NECP_PARSED_PARAMETERS_SCOPED_IFNET_FIELDS)) {
+               // Has required fields, but not including specific interface fields. Pass for now, and check
+               // to see if agents are satisfied by policy.
+               *validate_agents = TRUE;
+               return (TRUE);
+       }
+
        return (*return_ifindex != 0);
 }
 
@@ -3194,7 +4065,8 @@ necp_open(struct proc *p, struct necp_open_args *uap, int *retval)
        struct fileproc *fp = NULL;
        int fd = -1;
 
-       if (uap->flags & NECP_OPEN_FLAG_OBSERVER) {
+       if (uap->flags & NECP_OPEN_FLAG_OBSERVER ||
+               uap->flags & NECP_OPEN_FLAG_PUSH_OBSERVER) {
                if (necp_skywalk_priv_check_cred(p, kauth_cred_get()) != 0 &&
                        priv_check_cred(kauth_cred_get(), PRIV_NET_PRIVILEGED_NETWORK_STATISTICS, 0) != 0) {
                        NECPLOG0(LOG_ERR, "Client does not hold necessary entitlement to observe other NECP clients");
@@ -3218,6 +4090,7 @@ necp_open(struct proc *p, struct necp_open_args *uap, int *retval)
        fd_data->necp_fd_type = necp_fd_type_client;
        fd_data->flags = uap->flags;
        RB_INIT(&fd_data->clients);
+       RB_INIT(&fd_data->flows);
        TAILQ_INIT(&fd_data->update_list);
        lck_mtx_init(&fd_data->fd_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
        klist_init(&fd_data->si.si_note);
@@ -3309,11 +4182,12 @@ necp_client_add(struct proc *p, struct necp_fd_data *fd_data, struct necp_client
 
        client->parameters_length = uap->buffer_size;
        client->proc_pid = fd_data->proc_pid; // Save off proc pid in case the client will persist past fd
+       client->agent_handle = (void *)fd_data;
        client->platform_binary = ((csproc_get_platform_binary(p) == 0) ? 0 : 1);
 
-       uuid_generate_random(client->client_id);
+       necp_generate_client_id(client->client_id, false);
        LIST_INIT(&client->assertion_list);
-       LIST_INIT(&client->flow_list);
+       RB_INIT(&client->flow_registrations);
 
        error = copyout(client->client_id, uap->client_id, sizeof(uuid_t));
        if (error) {
@@ -3351,8 +4225,6 @@ static int
 necp_client_remove(struct necp_fd_data *fd_data, struct necp_client_action_args *uap, int *retval)
 {
        int error = 0;
-       struct necp_client *client = NULL;
-       struct necp_client find = {};
        uuid_t client_id = {};
        struct ifnet_stats_per_flow flow_ifnet_stats = {};
 
@@ -3382,15 +4254,27 @@ necp_client_remove(struct necp_fd_data *fd_data, struct necp_client_action_args
        NECP_FD_LOCK(fd_data);
 
        pid_t pid = fd_data->proc_pid;
-       uuid_copy(find.client_id, client_id);
-       client = RB_FIND(_necp_client_tree, &fd_data->clients, &find);
+       struct necp_client *client = necp_client_fd_find_client_unlocked(fd_data, client_id);
        if (client != NULL) {
+               // Remove any flow registrations that match
+               struct necp_client_flow_registration *flow_registration = NULL;
+               struct necp_client_flow_registration *temp_flow_registration = NULL;
+               RB_FOREACH_SAFE(flow_registration, _necp_fd_flow_tree, &fd_data->flows, temp_flow_registration) {
+                       if (flow_registration->client == client) {
+                               NECP_FLOW_TREE_LOCK_EXCLUSIVE();
+                               RB_REMOVE(_necp_client_flow_global_tree, &necp_client_flow_global_tree, flow_registration);
+                               NECP_FLOW_TREE_UNLOCK();
+                               RB_REMOVE(_necp_fd_flow_tree, &fd_data->flows, flow_registration);
+                       }
+               }
+               // Remove client from lists
                NECP_CLIENT_TREE_LOCK_EXCLUSIVE();
                RB_REMOVE(_necp_client_global_tree, &necp_client_global_tree, client);
                NECP_CLIENT_TREE_UNLOCK();
                RB_REMOVE(_necp_client_tree, &fd_data->clients, client);
        }
 
+
        NECP_FD_UNLOCK(fd_data);
 
        if (client != NULL) {
@@ -3400,13 +4284,13 @@ necp_client_remove(struct necp_fd_data *fd_data, struct necp_client_action_args
                error = ENOENT;
                NECPLOG(LOG_ERR, "necp_client_remove invalid client_id (%d)", error);
        }
-
 done:
        *retval = error;
 
        return (error);
 }
 
+
 static int
 necp_client_check_tcp_heuristics(struct necp_client *client, struct necp_client_flow *flow, u_int32_t *flags, u_int8_t *tfo_cookie, u_int8_t *tfo_cookie_len)
 {
@@ -3492,188 +4376,278 @@ do_unlock:
        return (error);
 }
 
+static size_t
+necp_client_calculate_flow_tlv_size(struct necp_client_flow_registration *flow_registration)
+{
+       size_t assigned_results_size = 0;
+       struct necp_client_flow *flow = NULL;
+       LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
+               if (flow->assigned) {
+                       size_t header_length = 0;
+                       if (flow->nexus) {
+                               header_length = sizeof(struct necp_client_nexus_flow_header);
+                       } else {
+                               header_length = sizeof(struct necp_client_flow_header);
+                       }
+                       assigned_results_size += (header_length + flow->assigned_results_length);
+
+                       if (flow->has_protoctl_event) {
+                               assigned_results_size += sizeof(struct necp_client_flow_protoctl_event_header);
+                       }
+               }
+       }
+       return assigned_results_size;
+}
+
+static int
+necp_client_fillout_flow_tlvs(struct necp_client *client,
+                                                         bool client_is_observed,
+                                                         struct necp_client_flow_registration *flow_registration,
+                                                         struct necp_client_action_args *uap,
+                                                         size_t *assigned_results_cursor)
+{
+       int error = 0;
+       struct necp_client_flow *flow = NULL;
+       LIST_FOREACH(flow, &flow_registration->flow_list, flow_chain) {
+               if (flow->assigned) {
+                       // Write TLV headers
+                       struct necp_client_nexus_flow_header header = {};
+                       u_int32_t length = 0;
+                       u_int32_t flags = 0;
+                       u_int8_t tfo_cookie_len = 0;
+                       u_int8_t type = 0;
+
+                       type = NECP_CLIENT_RESULT_FLOW_ID;
+                       length = sizeof(header.flow_header.flow_id);
+                       memcpy(&header.flow_header.flow_id_tlv_header.type, &type, sizeof(type));
+                       memcpy(&header.flow_header.flow_id_tlv_header.length, &length, sizeof(length));
+                       uuid_copy(header.flow_header.flow_id, flow_registration->registration_id);
+
+                       if (flow->nexus) {
+                               if (flow->check_tcp_heuristics) {
+                                       u_int8_t tfo_cookie[NECP_TFO_COOKIE_LEN_MAX];
+                                       tfo_cookie_len = NECP_TFO_COOKIE_LEN_MAX;
+
+                                       if (necp_client_check_tcp_heuristics(client, flow, &flags,
+                                                                                                                tfo_cookie, &tfo_cookie_len) != 0) {
+                                               tfo_cookie_len = 0;
+                                       } else {
+                                               flow->check_tcp_heuristics = FALSE;
+
+                                               if (tfo_cookie_len != 0) {
+                                                       type = NECP_CLIENT_RESULT_TFO_COOKIE;
+                                                       length = tfo_cookie_len;
+                                                       memcpy(&header.tfo_cookie_tlv_header.type, &type, sizeof(type));
+                                                       memcpy(&header.tfo_cookie_tlv_header.length, &length, sizeof(length));
+                                                       memcpy(&header.tfo_cookie_value, tfo_cookie, tfo_cookie_len);
+                                               }
+                                       }
+                               }
+                       }
+
+                       size_t header_length = 0;
+                       if (flow->nexus) {
+                               if (tfo_cookie_len != 0) {
+                                       header_length = sizeof(struct necp_client_nexus_flow_header) - (NECP_TFO_COOKIE_LEN_MAX - tfo_cookie_len);
+                               } else {
+                                       header_length = sizeof(struct necp_client_nexus_flow_header) - sizeof(struct necp_tlv_header) - NECP_TFO_COOKIE_LEN_MAX;
+                               }
+                       } else {
+                               header_length = sizeof(struct necp_client_flow_header);
+                       }
+
+                       type = NECP_CLIENT_RESULT_FLAGS;
+                       length = sizeof(header.flow_header.flags_value);
+                       memcpy(&header.flow_header.flags_tlv_header.type, &type, sizeof(type));
+                       memcpy(&header.flow_header.flags_tlv_header.length, &length, sizeof(length));
+                       if (flow->assigned) {
+                               flags |= NECP_CLIENT_RESULT_FLAG_FLOW_ASSIGNED;
+                       }
+                       if (flow->viable) {
+                               flags |= NECP_CLIENT_RESULT_FLAG_FLOW_VIABLE;
+                       }
+                       if (flow_registration->defunct) {
+                               flags |= NECP_CLIENT_RESULT_FLAG_DEFUNCT;
+                       }
+                       flags |= flow->necp_flow_flags;
+                       memcpy(&header.flow_header.flags_value, &flags, sizeof(flags));
+
+                       type = NECP_CLIENT_RESULT_INTERFACE;
+                       length = sizeof(header.flow_header.interface_value);
+                       memcpy(&header.flow_header.interface_tlv_header.type, &type, sizeof(type));
+                       memcpy(&header.flow_header.interface_tlv_header.length, &length, sizeof(length));
+
+                       struct necp_client_result_interface interface_struct;
+                       interface_struct.generation = 0;
+                       interface_struct.index = flow->interface_index;
+
+                       memcpy(&header.flow_header.interface_value, &interface_struct, sizeof(interface_struct));
+                       if (flow->nexus) {
+                               type = NECP_CLIENT_RESULT_NETAGENT;
+                               length = sizeof(header.agent_value);
+                               memcpy(&header.agent_tlv_header.type, &type, sizeof(type));
+                               memcpy(&header.agent_tlv_header.length, &length, sizeof(length));
+
+                               struct necp_client_result_netagent agent_struct;
+                               agent_struct.generation = 0;
+                               uuid_copy(agent_struct.netagent_uuid, flow->u.nexus_agent);
+
+                               memcpy(&header.agent_value, &agent_struct, sizeof(agent_struct));
+                       }
+
+                       // Don't include outer TLV header in length field
+                       type = NECP_CLIENT_RESULT_FLOW;
+                       length = (header_length - sizeof(struct necp_tlv_header) + flow->assigned_results_length);
+                       if (flow->has_protoctl_event) {
+                               length += sizeof(struct necp_client_flow_protoctl_event_header);
+                       }
+                       memcpy(&header.flow_header.outer_header.type, &type, sizeof(type));
+                       memcpy(&header.flow_header.outer_header.length, &length, sizeof(length));
+
+                       error = copyout(&header, uap->buffer + client->result_length + *assigned_results_cursor, header_length);
+                       if (error) {
+                               NECPLOG(LOG_ERR, "necp_client_copy assigned results tlv_header copyout error (%d)", error);
+                               return (error);
+                       }
+                       *assigned_results_cursor += header_length;
+
+                       if (flow->assigned_results && flow->assigned_results_length) {
+                               // Write inner TLVs
+                               error = copyout(flow->assigned_results, uap->buffer + client->result_length + *assigned_results_cursor,
+                                                               flow->assigned_results_length);
+                               if (error) {
+                                       NECPLOG(LOG_ERR, "necp_client_copy assigned results copyout error (%d)", error);
+                                       return (error);
+                               }
+                       }
+                       *assigned_results_cursor += flow->assigned_results_length;
+
+                       /* Read the protocol event and reset it */
+                       if (flow->has_protoctl_event) {
+                               struct necp_client_flow_protoctl_event_header protoctl_event_header = {};
+
+                               type = NECP_CLIENT_RESULT_PROTO_CTL_EVENT;
+                               length = sizeof(protoctl_event_header.protoctl_event);
+
+                               memcpy(&protoctl_event_header.protoctl_tlv_header.type, &type, sizeof(type));
+                               memcpy(&protoctl_event_header.protoctl_tlv_header.length, &length, sizeof(length));
+                               memcpy(&protoctl_event_header.protoctl_event, &flow->protoctl_event,
+                                          sizeof(flow->protoctl_event));
+
+                               error = copyout(&protoctl_event_header, uap->buffer + client->result_length + *assigned_results_cursor,
+                                                               sizeof(protoctl_event_header));
+
+                               if (error) {
+                                       NECPLOG(LOG_ERR, "necp_client_copy protocol control event results"
+                                                       " tlv_header copyout error (%d)", error);
+                                       return (error);
+                               }
+                               *assigned_results_cursor += sizeof(protoctl_event_header);
+                               flow->has_protoctl_event = FALSE;
+                               flow->protoctl_event.protoctl_event_code = 0;
+                               flow->protoctl_event.protoctl_event_val = 0;
+                               flow->protoctl_event.protoctl_event_tcp_seq_num = 0;
+                       }
+               }
+       }
+       if (!client_is_observed) {
+               flow_registration->flow_result_read = TRUE;
+       }
+       return (0);
+}
+
 static int
-necp_client_copy_internal(struct necp_client *client, bool client_is_observed, struct necp_client_action_args *uap, int *retval)
+necp_client_copy_internal(struct necp_client *client, uuid_t client_id, bool client_is_observed, struct necp_client_action_args *uap, int *retval)
 {
+       NECP_CLIENT_ASSERT_LOCKED(client);
        int error = 0;
        // Copy results out
        if (uap->action == NECP_CLIENT_ACTION_COPY_PARAMETERS) {
                if (uap->buffer_size < client->parameters_length) {
-                       error = EINVAL;
-                       goto done;
+                       return (EINVAL);
                }
                error = copyout(client->parameters, uap->buffer, client->parameters_length);
                if (error) {
                        NECPLOG(LOG_ERR, "necp_client_copy parameters copyout error (%d)", error);
-                       goto done;
+                       return (error);
                }
                *retval = client->parameters_length;
        } else if (uap->action == NECP_CLIENT_ACTION_COPY_UPDATED_RESULT &&
-                          client->result_read && client->flow_result_read) {
+                          client->result_read && !necp_client_has_unread_flows(client)) {
                // Copy updates only, but nothing to read
                // Just return 0 for bytes read
                *retval = 0;
        } else if (uap->action == NECP_CLIENT_ACTION_COPY_RESULT ||
                           uap->action == NECP_CLIENT_ACTION_COPY_UPDATED_RESULT) {
                size_t assigned_results_size = 0;
-               struct necp_client_flow *flow = NULL;
-               LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-                       if (flow->nexus || (flow->socket && flow->assigned)) {
-                               size_t header_length = 0;
-                               if (flow->nexus) {
-                                       header_length = sizeof(struct necp_client_nexus_flow_header);
-                               } else {
-                                       header_length = sizeof(struct necp_client_flow_header);
-                               }
-                               assigned_results_size += (header_length + flow->assigned_results_length);
 
-                               if (flow->has_protoctl_event) {
-                                       assigned_results_size += sizeof(struct necp_client_flow_protoctl_event_header);
+               bool some_flow_is_defunct = false;
+               struct necp_client_flow_registration *single_flow_registration = NULL;
+               if (necp_client_id_is_flow(client_id)) {
+                       single_flow_registration = necp_client_find_flow(client, client_id);
+                       if (single_flow_registration != NULL) {
+                               assigned_results_size += necp_client_calculate_flow_tlv_size(single_flow_registration);
+                       }
+               } else {
+                       // This request is for the client, so copy everything
+                       struct necp_client_flow_registration *flow_registration = NULL;
+                       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+                               if (flow_registration->defunct) {
+                                       some_flow_is_defunct = true;
                                }
+                               assigned_results_size += necp_client_calculate_flow_tlv_size(flow_registration);
                        }
                }
                if (uap->buffer_size < (client->result_length + assigned_results_size)) {
-                       error = EINVAL;
-                       goto done;
+                       return (EINVAL);
+               }
+
+               u_int32_t original_flags = 0;
+               bool flags_updated = false;
+               if (some_flow_is_defunct && client->legacy_client_is_flow) {
+                       // If our client expects the defunct flag in the client, add it now
+                       u_int32_t client_flags = 0;
+                       u_int32_t value_size = 0;
+                       u_int8_t *flags_pointer = necp_buffer_get_tlv_value(client->result, 0, &value_size);
+                       if (flags_pointer != NULL && value_size == sizeof(client_flags)) {
+                               memcpy(&client_flags, flags_pointer, value_size);
+                               original_flags = client_flags;
+                               client_flags |= NECP_CLIENT_RESULT_FLAG_DEFUNCT;
+                               (void)necp_buffer_write_tlv_if_different(client->result, NECP_CLIENT_RESULT_FLAGS,
+                                                                                                                sizeof(client_flags), &client_flags, &flags_updated,
+                                                                                                                client->result, sizeof(client->result));
+                       }
                }
+
                error = copyout(client->result, uap->buffer, client->result_length);
+
+               if (flags_updated) {
+                       // Revert stored flags
+                       (void)necp_buffer_write_tlv_if_different(client->result, NECP_CLIENT_RESULT_FLAGS,
+                                                                                                        sizeof(original_flags), &original_flags, &flags_updated,
+                                                                                                        client->result, sizeof(client->result));
+               }
+
                if (error) {
                        NECPLOG(LOG_ERR, "necp_client_copy result copyout error (%d)", error);
-                       goto done;
+                       return (error);
                }
 
                size_t assigned_results_cursor = 0;
-
-               flow = NULL;
-               LIST_FOREACH(flow, &client->flow_list, flow_chain) {
-                       if (flow->nexus || (flow->socket && flow->assigned)) {
-                               // Write TLV headers
-                               struct necp_client_nexus_flow_header header = {};
-                               u_int32_t length = 0;
-                               u_int32_t flags = 0;
-                               u_int8_t tfo_cookie_len = 0;
-                               u_int8_t type = 0;
-
-                               if (flow->nexus) {
-                                       if (flow->check_tcp_heuristics) {
-                                               u_int8_t tfo_cookie[NECP_TFO_COOKIE_LEN_MAX];
-                                               tfo_cookie_len = NECP_TFO_COOKIE_LEN_MAX;
-
-                                               if (necp_client_check_tcp_heuristics(client, flow, &flags,
-                                                                                   tfo_cookie, &tfo_cookie_len) != 0) {
-                                                       tfo_cookie_len = 0;
-                                               } else {
-                                                       flow->check_tcp_heuristics = FALSE;
-
-                                                       if (tfo_cookie_len != 0) {
-                                                               type = NECP_CLIENT_RESULT_TFO_COOKIE;
-                                                               length = tfo_cookie_len;
-                                                               memcpy(&header.tfo_cookie_tlv_header.type, &type, sizeof(type));
-                                                               memcpy(&header.tfo_cookie_tlv_header.length, &length, sizeof(length));
-                                                               memcpy(&header.tfo_cookie_value, tfo_cookie, tfo_cookie_len);
-                                                       }
-                                               }
-                                       }
-                               }
-
-                               size_t header_length = 0;
-                               if (flow->nexus) {
-                                       if (tfo_cookie_len != 0) {
-                                               header_length = sizeof(struct necp_client_nexus_flow_header) - (NECP_TFO_COOKIE_LEN_MAX - tfo_cookie_len);
-                                       } else {
-                                               header_length = sizeof(struct necp_client_nexus_flow_header) - sizeof(struct necp_tlv_header) - NECP_TFO_COOKIE_LEN_MAX;
-                                       }
-                               } else {
-                                       header_length = sizeof(struct necp_client_flow_header);
-                               }
-
-                               type = NECP_CLIENT_RESULT_FLAGS;
-                               length = sizeof(header.flow_header.flags_value);
-                               memcpy(&header.flow_header.flags_tlv_header.type, &type, sizeof(type));
-                               memcpy(&header.flow_header.flags_tlv_header.length, &length, sizeof(length));
-                               if (flow->assigned) {
-                                       flags |= NECP_CLIENT_RESULT_FLAG_FLOW_ASSIGNED;
-                               }
-                               if (flow->viable) {
-                                       flags |= NECP_CLIENT_RESULT_FLAG_FLOW_VIABLE;
-                               }
-                               memcpy(&header.flow_header.flags_value, &flags, sizeof(flags));
-
-                               type = NECP_CLIENT_RESULT_INTERFACE;
-                               length = sizeof(header.flow_header.interface_value);
-                               memcpy(&header.flow_header.interface_tlv_header.type, &type, sizeof(type));
-                               memcpy(&header.flow_header.interface_tlv_header.length, &length, sizeof(length));
-
-                               struct necp_client_result_interface interface_struct;
-                               interface_struct.generation = 0;
-                               interface_struct.index = flow->interface_index;
-
-                               memcpy(&header.flow_header.interface_value, &interface_struct, sizeof(interface_struct));
-                               if (flow->nexus) {
-                                       type = NECP_CLIENT_RESULT_NETAGENT;
-                                       length = sizeof(header.agent_value);
-                                       memcpy(&header.agent_tlv_header.type, &type, sizeof(type));
-                                       memcpy(&header.agent_tlv_header.length, &length, sizeof(length));
-
-                                       struct necp_client_result_netagent agent_struct;
-                                       agent_struct.generation = 0;
-                                       uuid_copy(agent_struct.netagent_uuid, flow->u.nexus_agent);
-
-                                       memcpy(&header.agent_value, &agent_struct, sizeof(agent_struct));
+               if (necp_client_id_is_flow(client_id)) {
+                       if (single_flow_registration != NULL) {
+                               error = necp_client_fillout_flow_tlvs(client, client_is_observed, single_flow_registration, uap, &assigned_results_cursor);
+                               if (error != 0) {
+                                       return (error);
                                }
-
-                               // Don't include outer TLV header in length field
-                               type = NECP_CLIENT_RESULT_FLOW;
-                               length = (header_length - sizeof(struct necp_tlv_header) + flow->assigned_results_length);
-                               if (flow->has_protoctl_event) {
-                                       length += sizeof(struct necp_client_flow_protoctl_event_header);
-                               }
-                               memcpy(&header.flow_header.outer_header.type, &type, sizeof(type));
-                               memcpy(&header.flow_header.outer_header.length, &length, sizeof(length));
-
-                               error = copyout(&header, uap->buffer + client->result_length + assigned_results_cursor, header_length);
-                               if (error) {
-                                       NECPLOG(LOG_ERR, "necp_client_copy assigned results tlv_header copyout error (%d)", error);
-                                       goto done;
-                               }
-                               assigned_results_cursor += header_length;
-
-                               if (flow->assigned_results && flow->assigned_results_length) {
-                                       // Write inner TLVs
-                                       error = copyout(flow->assigned_results, uap->buffer + client->result_length + assigned_results_cursor,
-                                                                       flow->assigned_results_length);
-                                       if (error) {
-                                               NECPLOG(LOG_ERR, "necp_client_copy assigned results copyout error (%d)", error);
-                                               goto done;
-                                       }
-                               }
-                               assigned_results_cursor += flow->assigned_results_length;
-
-                               /* Read the protocol event and reset it */
-                               if (flow->has_protoctl_event) {
-                                       struct necp_client_flow_protoctl_event_header protoctl_event_header = {};
-
-                                       type = NECP_CLIENT_RESULT_PROTO_CTL_EVENT;
-                                       length = sizeof(protoctl_event_header.protoctl_event);
-
-                                       memcpy(&protoctl_event_header.protoctl_tlv_header.type, &type, sizeof(type));
-                                       memcpy(&protoctl_event_header.protoctl_tlv_header.length, &length, sizeof(length));
-                                       memcpy(&protoctl_event_header.protoctl_event, &flow->protoctl_event,
-                                           sizeof(flow->protoctl_event));
-
-                                       error = copyout(&protoctl_event_header, uap->buffer + client->result_length + assigned_results_cursor,
-                                           sizeof(protoctl_event_header));
-
-                                       if (error) {
-                                               NECPLOG(LOG_ERR, "necp_client_copy protocol control event results"
-                                                   " tlv_header copyout error (%d)", error);
-                                               goto done;
-                                       }
-                                       assigned_results_cursor += sizeof(protoctl_event_header);
-                                       flow->has_protoctl_event = FALSE;
-                                       flow->protoctl_event.protoctl_event_code = 0;
-                                       flow->protoctl_event.protoctl_event_val = 0;
-                                       flow->protoctl_event.protoctl_event_tcp_seq_num = 0;
+                       }
+               } else {
+                       // This request is for the client, so copy everything
+                       struct necp_client_flow_registration *flow_registration = NULL;
+                       RB_FOREACH(flow_registration, _necp_client_flow_tree, &client->flow_registrations) {
+                               error = necp_client_fillout_flow_tlvs(client, client_is_observed, flow_registration, uap, &assigned_results_cursor);
+                               if (error != 0) {
+                                       return (error);
                                }
                        }
                }
@@ -3682,12 +4656,10 @@ necp_client_copy_internal(struct necp_client *client, bool client_is_observed, s
 
                if (!client_is_observed) {
                        client->result_read = TRUE;
-                       client->flow_result_read = TRUE;
                }
        }
 
-done:
-       return (error);
+       return (0);
 }
 
 static int
@@ -3701,28 +4673,25 @@ necp_client_copy(struct necp_fd_data *fd_data, struct necp_client_action_args *u
        *retval = 0;
 
        if (uap->buffer_size == 0 || uap->buffer == 0) {
-               error = EINVAL;
-               goto done;
+               return (EINVAL);
        }
 
        if (uap->action != NECP_CLIENT_ACTION_COPY_PARAMETERS &&
                uap->action != NECP_CLIENT_ACTION_COPY_RESULT &&
                uap->action != NECP_CLIENT_ACTION_COPY_UPDATED_RESULT) {
-               error = EINVAL;
-               goto done;
+               return (EINVAL);
        }
 
        if (uap->client_id) {
                if (uap->client_id_len != sizeof(uuid_t)) {
                        NECPLOG(LOG_ERR, "Incorrect length (got %d, expected %d)", uap->client_id_len, sizeof(uuid_t));
-                       error = ERANGE;
-                       goto done;
+                       return (ERANGE);
                }
 
                error = copyin(uap->client_id, client_id, sizeof(uuid_t));
                if (error) {
                        NECPLOG(LOG_ERR, "necp_client_copy client_id copyin error (%d)", error);
-                       goto done;
+                       return (error);
                }
        }
 
@@ -3735,7 +4704,7 @@ necp_client_copy(struct necp_fd_data *fd_data, struct necp_client_action_args *u
                        struct necp_client *find_client = NULL;
                        RB_FOREACH(find_client, _necp_client_tree, &fd_data->clients) {
                                NECP_CLIENT_LOCK(find_client);
-                               if (!find_client->result_read || !find_client->flow_result_read) {
+                               if (!find_client->result_read || necp_client_has_unread_flows(find_client)) {
                                        client = find_client;
                                        // Leave the client locked, and break
                                        break;
@@ -3749,7 +4718,7 @@ necp_client_copy(struct necp_fd_data *fd_data, struct necp_client_action_args *u
 
        if (client != NULL) {
                // If client is set, it is locked
-               error = necp_client_copy_internal(client, FALSE, uap, retval);
+               error = necp_client_copy_internal(client, client_id, FALSE, uap, retval);
                NECP_CLIENT_UNLOCK(client);
        }
 
@@ -3765,16 +4734,11 @@ necp_client_copy(struct necp_fd_data *fd_data, struct necp_client_action_args *u
 
                        bool found_client = FALSE;
 
-                       struct necp_client find;
-                       uuid_copy(find.client_id, client_id);
-                       client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
+                       client = necp_find_client_and_lock(client_id);
                        if (client != NULL) {
-                               NECP_CLIENT_LOCK(client);
-
                                // Matched, copy out data
                                found_client = TRUE;
-                               error = necp_client_copy_internal(client, TRUE, uap, retval);
-
+                               error = necp_client_copy_internal(client, client_id, TRUE, uap, retval);
                                NECP_CLIENT_UNLOCK(client);
                        }
 
@@ -3783,17 +4747,14 @@ necp_client_copy(struct necp_fd_data *fd_data, struct necp_client_action_args *u
 
                        // No client found, fail
                        if (!found_client) {
-                               error = ENOENT;
-                               goto done;
+                               return (ENOENT);
                        }
                } else {
                        // No client found, and not allowed to search other fds, fail
-                       error = ENOENT;
-                       goto done;
+                       return (ENOENT);
                }
        }
 
-done:
        return (error);
 }
 
@@ -3856,7 +4817,8 @@ necp_client_copy_client_update(struct necp_fd_data *fd_data, struct necp_client_
 }
 
 static int
-necp_client_copy_parameters_locked(struct necp_client *client, struct necp_client_nexus_parameters *parameters)
+necp_client_copy_parameters_locked(struct necp_client *client,
+                                                                  struct necp_client_nexus_parameters *parameters)
 {
        VERIFY(parameters != NULL);
 
@@ -3890,44 +4852,6 @@ necp_client_copy_parameters_locked(struct necp_client *client, struct necp_clien
        return (error);
 }
 
-int
-necp_client_copy_parameters(uuid_t client_id, struct necp_client_nexus_parameters *parameters)
-{
-       int error = 0;
-       struct necp_client *client = NULL;
-
-       if (parameters == NULL) {
-               return EINVAL;
-       }
-
-       // Lock tree
-       NECP_CLIENT_TREE_LOCK_SHARED();
-
-       bool found_client = FALSE;
-       struct necp_client find;
-       uuid_copy(find.client_id, client_id);
-       client = RB_FIND(_necp_client_global_tree, &necp_client_global_tree, &find);
-       if (client != NULL) {
-               NECP_CLIENT_LOCK(client);
-
-               // Matched, parse parameters
-               found_client = TRUE;
-               error = necp_client_copy_parameters_locked(client, parameters);
-
-               NECP_CLIENT_UNLOCK(client);
-       }
-
-       // Unlock tree
-       NECP_CLIENT_TREE_UNLOCK();
-
-       // No client found, fail
-       if (!found_client) {
-               return ENOENT;
-       }
-
-       return error;
-}
-
 static int
 necp_client_list(struct necp_fd_data *fd_data, struct necp_client_action_args *uap, int *retval)
 {
@@ -4145,6 +5069,7 @@ necp_client_agent_action(struct necp_fd_data *fd_data, struct necp_client_action
                                                error = netagent_client_message_with_params(agent_uuid,
                                                                                                                                        client_id,
                                                                                                                                        fd_data->proc_pid,
+                                                                                                                                       client->agent_handle,
                                                                                                                                        netagent_message_type,
                                                                                                                                        &parsed_parameters,
                                                                                                                                        NULL, NULL);
@@ -4316,6 +5241,12 @@ necp_client_copy_interface(__unused struct necp_fd_data *fd_data, struct necp_cl
                if ((interface->if_eflags & IFEF_NOACKPRI) == IFEF_NOACKPRI) {
                        interface_details.flags |= NECP_INTERFACE_FLAG_NOACKPRI;
                }
+               if ((interface->if_eflags & IFEF_3CA) == IFEF_3CA) {
+                       interface_details.flags |= NECP_INTERFACE_FLAG_3CARRIERAGG;
+               }
+               if (IFNET_IS_LOW_POWER(interface)) {
+                       interface_details.flags |= NECP_INTERFACE_FLAG_IS_LOW_POWER;
+               }
                interface_details.mtu = interface->if_mtu;
 
                u_int8_t ipv4_signature_len = sizeof(interface_details.ipv4_signature.signature);
@@ -4439,9 +5370,17 @@ necp_client_update_cache(struct necp_fd_data *fd_data, struct necp_client_action
                goto done;
        }
 
+       struct necp_client_flow_registration *flow_registration = necp_client_find_flow(client, client_id);
+       if (flow_registration == NULL) {
+               NECP_CLIENT_UNLOCK(client);
+               NECP_FD_UNLOCK(fd_data);
+               error = ENOENT;
+               goto done;
+       }
+
        NECP_CLIENT_ROUTE_LOCK(client);
        // This needs to be changed when TFO/ECN is supported by multiple flows
-       struct necp_client_flow *flow = LIST_FIRST(&client->flow_list);
+       struct necp_client_flow *flow = LIST_FIRST(&flow_registration->flow_list);
        if (flow == NULL ||
                (flow->remote_addr.sa.sa_family != AF_INET &&
                 flow->remote_addr.sa.sa_family != AF_INET6) ||
@@ -4608,7 +5547,7 @@ necp_match_policy(struct proc *p, struct necp_match_policy_args *uap, int32_t *r
 {
 #pragma unused(retval)
        u_int8_t *parameters = NULL;
-       struct necp_aggregate_result returned_result = {};
+       struct necp_aggregate_result returned_result;
        int error = 0;
 
        if (uap == NULL) {
@@ -4821,7 +5760,7 @@ done:
 void *
 necp_create_nexus_assign_message(uuid_t nexus_instance, u_int32_t nexus_port, void *key, uint32_t key_length,
                                                                 struct necp_client_endpoint *local_endpoint, struct necp_client_endpoint *remote_endpoint,
-                                                                u_int32_t flow_adv_index, size_t *message_length)
+                                                                u_int32_t flow_adv_index, void *flow_stats, size_t *message_length)
 {
        u_int8_t *buffer = NULL;
        u_int8_t *cursor = NULL;
@@ -4846,6 +5785,9 @@ necp_create_nexus_assign_message(uuid_t nexus_instance, u_int32_t nexus_port, vo
        if (remote_endpoint != NULL) {
                valsize += sizeof(struct necp_tlv_header) + sizeof(struct necp_client_endpoint);
        }
+       if (flow_stats != NULL) {
+               valsize += sizeof(struct necp_tlv_header) + sizeof(void *);
+       }
        if (valsize == 0) {
                return (NULL);
        }
@@ -4872,6 +5814,9 @@ necp_create_nexus_assign_message(uuid_t nexus_instance, u_int32_t nexus_port, vo
        if (remote_endpoint != NULL) {
                cursor = necp_buffer_write_tlv(cursor, NECP_CLIENT_RESULT_REMOTE_ENDPOINT, sizeof(struct necp_client_endpoint), remote_endpoint, buffer, valsize);
        }
+       if (flow_stats != NULL) {
+               cursor = necp_buffer_write_tlv(cursor, NECP_CLIENT_RESULT_NEXUS_FLOW_STATS, sizeof(void *), &flow_stats, buffer, valsize);
+       }
 
        *message_length = valsize;
 
@@ -4949,6 +5894,13 @@ necp_client_init(void)
                /* NOTREACHED */
        }
 
+       necp_flow_registration_size = sizeof(struct necp_client_flow_registration);
+       necp_flow_registration_cache = mcache_create(NECP_FLOW_REGISTRATION_ZONE_NAME, necp_flow_registration_size, sizeof (uint64_t), 0, MCR_SLEEP);
+       if (necp_flow_registration_cache == NULL) {
+               panic("mcache_create(necp_client_flow_registration) failed\n");
+               /* NOTREACHED */
+       }
+
        necp_client_update_tcall = thread_call_allocate_with_options(necp_update_all_clients_callout, NULL,
                                                                                                                                 THREAD_CALL_PRIORITY_KERNEL, THREAD_CALL_OPTIONS_ONCE);
        VERIFY(necp_client_update_tcall != NULL);
@@ -4956,13 +5908,15 @@ necp_client_init(void)
        lck_rw_init(&necp_fd_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
        lck_rw_init(&necp_observer_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
        lck_rw_init(&necp_client_tree_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
+    lck_rw_init(&necp_flow_tree_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
        lck_rw_init(&necp_collect_stats_list_lock, necp_fd_mtx_grp, necp_fd_mtx_attr);
 
        LIST_INIT(&necp_fd_list);
        LIST_INIT(&necp_fd_observer_list);
-       LIST_INIT(&necp_collect_stats_client_list);
+       LIST_INIT(&necp_collect_stats_flow_list);
 
        RB_INIT(&necp_client_global_tree);
+       RB_INIT(&necp_client_flow_global_tree);
 
        return (0);
 }
@@ -4971,5 +5925,6 @@ void
 necp_client_reap_caches(boolean_t purge)
 {
        mcache_reap_now(necp_flow_cache, purge);
+       mcache_reap_now(necp_flow_registration_cache, purge);
 }