Merged in sandervrijders/ouroboros/be-gam (pull request #345)

Be gam

12 files changed, 796 insertions, 105 deletions

diff --git a/src/ipcpd/normal/CMakeLists.txt b/src/ipcpd/normal/CMakeLists.txt
index bdcb78ae..157baa9e 100644
--- a/src/ipcpd/normal/CMakeLists.txt
+++ b/src/ipcpd/normal/CMakeLists.txt
@@ -29,12 +29,14 @@ set(SOURCE_FILES
   dir.c
   fmgr.c
   frct.c
+  gam.c
   main.c
   pathname.c
   pff.c
   ribmgr.c
   shm_pci.c
   # Add policies last
+  pol/complete.c
   pol/flat.c
   )
 
diff --git a/src/ipcpd/normal/fmgr.c b/src/ipcpd/normal/fmgr.c
index a419e9f5..6fe6fb60 100644
--- a/src/ipcpd/normal/fmgr.c
+++ b/src/ipcpd/normal/fmgr.c
@@ -28,6 +28,7 @@
 #include <ouroboros/ipcp-dev.h>
 #include <ouroboros/fqueue.h>
 #include <ouroboros/errno.h>
+#include <ouroboros/cacep.h>
 
 #include <stdlib.h>
 #include <stdbool.h>
@@ -42,15 +43,24 @@
 #include "dir.h"
 #include "pathname.h"
 #include "ro.h"
+#include "gam.h"
 
 #include "flow_alloc.pb-c.h"
 typedef FlowAllocMsg flow_alloc_msg_t;
 
 #define FD_UPDATE_TIMEOUT 100000 /* nanoseconds */
 
+struct nm1_flow {
+        struct list_head    next;
+        int                 fd;
+        qosspec_t           qs;
+        struct cacep_info * info;
+};
+
 struct {
         flow_set_t *       nm1_set[QOS_CUBE_MAX];
         fqueue_t *         nm1_fqs[QOS_CUBE_MAX];
+        struct list_head   nm1_flows;
         pthread_rwlock_t   nm1_flows_lock;
 
         flow_set_t *       np1_set[QOS_CUBE_MAX];
@@ -60,21 +70,23 @@ struct {
         cep_id_t           np1_fd_to_cep_id[AP_MAX_FLOWS];
         int                np1_cep_id_to_fd[IPCPD_MAX_CONNS];
 
-        pthread_t          nm1_sdu_reader;
         pthread_t          np1_sdu_reader;
+        pthread_t          nm1_sdu_reader;
+        pthread_t          nm1_flow_wait;
 
         /* FIXME: Replace with PFF */
         int fd;
+
+        struct gam *       gam;
 } fmgr;
 
 static void * fmgr_np1_sdu_reader(void * o)
 {
         struct shm_du_buff * sdb;
-        struct timespec timeout = {0, FD_UPDATE_TIMEOUT};
-        int fd;
-
-        int i = 0;
-        int ret;
+        struct timespec      timeout = {0, FD_UPDATE_TIMEOUT};
+        int                  fd;
+        int                  i = 0;
+        int                  ret;
 
         (void) o;
 
@@ -118,12 +130,12 @@ static void * fmgr_np1_sdu_reader(void * o)
 
 void * fmgr_nm1_sdu_reader(void * o)
 {
-        struct timespec timeout = {0, FD_UPDATE_TIMEOUT};
+        struct timespec      timeout = {0, FD_UPDATE_TIMEOUT};
         struct shm_du_buff * sdb;
-        struct pci * pci;
-        int fd;
-        int i = 0;
-        int ret;
+        struct pci *         pci;
+        int                  fd;
+        int                  i = 0;
+        int                  ret;
 
         (void) o;
 
@@ -202,6 +214,49 @@ void * fmgr_nm1_sdu_reader(void * o)
         return (void *) 0;
 }
 
+static void * fmgr_nm1_flow_wait(void * o)
+{
+        qoscube_t           cube;
+        struct cacep_info * info;
+        int                 fd;
+        qosspec_t           qs;
+        struct nm1_flow *   flow;
+
+        (void) o;
+
+        while (true) {
+                if (gam_flow_wait(fmgr.gam, &fd, &info, &qs)) {
+                        LOG_ERR("Failed to get next flow descriptor.");
+                        continue;
+                }
+
+                ipcp_flow_get_qoscube(fd, &cube);
+                flow_set_add(fmgr.nm1_set[cube], fd);
+
+                /* FIXME: Temporary, until we have a PFF */
+                fmgr.fd = fd;
+
+                pthread_rwlock_wrlock(&fmgr.nm1_flows_lock);
+                flow = malloc(sizeof(*flow));
+                if (flow == NULL) {
+                        free(info);
+                        pthread_rwlock_unlock(&fmgr.nm1_flows_lock);
+                        continue;
+                }
+
+                flow->info = info;
+                flow->fd = fd;
+                flow->qs = qs;
+
+                list_head_init(&flow->next);
+                list_add(&flow->next, &fmgr.nm1_flows);
+
+                pthread_rwlock_unlock(&fmgr.nm1_flows_lock);
+        }
+
+        return (void *) 0;
+}
+
 static void fmgr_destroy_flows(void)
 {
         int i;
@@ -224,9 +279,6 @@ int fmgr_init()
         for (i = 0; i < IPCPD_MAX_CONNS; ++i)
                 fmgr.np1_cep_id_to_fd[i] = -1;
 
-        pthread_rwlock_init(&fmgr.nm1_flows_lock, NULL);
-        pthread_rwlock_init(&fmgr.np1_flows_lock, NULL);
-
         for (i = 0; i < QOS_CUBE_MAX; ++i) {
                 fmgr.np1_set[i] = flow_set_create();
                 if (fmgr.np1_set[i] == NULL) {
@@ -253,29 +305,55 @@ int fmgr_init()
                 }
         }
 
+        fmgr.gam = gam_create(ribmgr_dt_gam(), DT_AE);
+        if (fmgr.gam == NULL) {
+                LOG_ERR("Failed to create graph adjacency manager.");
+                fmgr_destroy_flows();
+                return -1;
+        }
+
+        list_head_init(&fmgr.nm1_flows);
+
+        pthread_rwlock_init(&fmgr.nm1_flows_lock, NULL);
+        pthread_rwlock_init(&fmgr.np1_flows_lock, NULL);
+
         pthread_create(&fmgr.np1_sdu_reader, NULL, fmgr_np1_sdu_reader, NULL);
         pthread_create(&fmgr.nm1_sdu_reader, NULL, fmgr_nm1_sdu_reader, NULL);
+        pthread_create(&fmgr.nm1_flow_wait, NULL, fmgr_nm1_flow_wait, NULL);
 
         return 0;
 }
 
 int fmgr_fini()
 {
-        int i;
-        int j;
+        struct list_head * pos = NULL;
+        struct list_head * n = NULL;
+        qoscube_t          cube;
 
         pthread_cancel(fmgr.np1_sdu_reader);
         pthread_cancel(fmgr.nm1_sdu_reader);
+        pthread_cancel(fmgr.nm1_flow_wait);
 
         pthread_join(fmgr.np1_sdu_reader, NULL);
         pthread_join(fmgr.nm1_sdu_reader, NULL);
+        pthread_join(fmgr.nm1_flow_wait, NULL);
 
-        for (i = 0; i < AP_MAX_FLOWS; ++i)
-                for (j = 0; j < QOS_CUBE_MAX; ++j)
-                        if (flow_set_has(fmgr.nm1_set[j], i)) {
-                                flow_dealloc(i);
-                                flow_set_del(fmgr.nm1_set[j], i);
-                        }
+        gam_destroy(fmgr.gam);
+
+        pthread_rwlock_wrlock(&fmgr.nm1_flows_lock);
+
+        list_for_each_safe(pos, n, &fmgr.nm1_flows) {
+                struct nm1_flow * flow =
+                        list_entry(pos, struct nm1_flow, next);
+                list_del(&flow->next);
+                flow_dealloc(flow->fd);
+                ipcp_flow_get_qoscube(flow->fd, &cube);
+                flow_set_del(fmgr.nm1_set[cube], flow->fd);
+                free(flow->info);
+                free(flow);
+        }
+
+        pthread_rwlock_unlock(&fmgr.nm1_flows_lock);
 
         pthread_rwlock_destroy(&fmgr.nm1_flows_lock);
         pthread_rwlock_destroy(&fmgr.np1_flows_lock);
@@ -290,12 +368,12 @@ int fmgr_np1_alloc(int       fd,
                    char *    src_ae_name,
                    qoscube_t cube)
 {
-        cep_id_t cep_id;
-        buffer_t buf;
+        cep_id_t         cep_id;
+        buffer_t         buf;
         flow_alloc_msg_t msg = FLOW_ALLOC_MSG__INIT;
-        char * path;
-        uint8_t * ro_data;
-        uint64_t addr;
+        char *           path;
+        uint8_t *        ro_data;
+        uint64_t         addr;
 
         path = pathname_create(RO_DIR);
         if (path == NULL)
@@ -359,9 +437,9 @@ int fmgr_np1_alloc(int       fd,
 static int np1_flow_dealloc(int fd)
 {
         flow_alloc_msg_t msg = FLOW_ALLOC_MSG__INIT;
-        buffer_t buf;
-        int ret;
-        qoscube_t cube;
+        buffer_t         buf;
+        int              ret;
+        qoscube_t        cube;
 
         ipcp_flow_get_qoscube(fd, &cube);
         flow_set_del(fmgr.np1_set[cube], fd);
@@ -388,10 +466,11 @@ static int np1_flow_dealloc(int fd)
         return ret;
 }
 
-int fmgr_np1_alloc_resp(int fd, int response)
+int fmgr_np1_alloc_resp(int fd,
+                        int response)
 {
         flow_alloc_msg_t msg = FLOW_ALLOC_MSG__INIT;
-        buffer_t buf;
+        buffer_t         buf;
 
         msg.code = FLOW_ALLOC_CODE__FLOW_REPLY;
         msg.response = response;
@@ -443,7 +522,8 @@ int fmgr_np1_dealloc(int fd)
         return ret;
 }
 
-int fmgr_np1_post_buf(cep_id_t cep_id, buffer_t * buf)
+int fmgr_np1_post_buf(cep_id_t   cep_id,
+                      buffer_t * buf)
 {
         int ret = 0;
         int fd;
@@ -512,7 +592,8 @@ int fmgr_np1_post_buf(cep_id_t cep_id, buffer_t * buf)
         return ret;
 }
 
-int fmgr_np1_post_sdu(cep_id_t cep_id, struct shm_du_buff * sdb)
+int fmgr_np1_post_sdu(cep_id_t             cep_id,
+                      struct shm_du_buff * sdb)
 {
         int fd;
 
@@ -530,52 +611,21 @@ int fmgr_np1_post_sdu(cep_id_t cep_id, struct shm_du_buff * sdb)
         return 0;
 }
 
-/* FIXME: do this in a topologymanager instance */
-int fmgr_nm1_add_flow(int fd)
+int fmgr_nm1_flow_arr(int       fd,
+                      qosspec_t qs)
 {
-        qoscube_t qos;
+        assert(fmgr.gam);
 
-        if (flow_alloc_resp(fd, 0) < 0) {
-                LOG_ERR("Could not respond to new flow.");
+        if (gam_flow_arr(fmgr.gam, fd, qs)) {
+                LOG_ERR("Failed to hand to graph adjacency manager.");
                 return -1;
         }
 
-        ipcp_flow_get_qoscube(fd, &qos);
-        flow_set_add(fmgr.nm1_set[qos], fd);
-
-        /* FIXME: Temporary, until we have a PFF */
-        fmgr.fd = fd;
-
-        return 0;
-}
-
-int fmgr_nm1_dt_flow(char * dst_name, qoscube_t qos)
-{
-        int fd;
-        int result;
-
-        /* FIXME: Map qos cube on correct QoS. */
-        fd = flow_alloc(dst_name, DT_AE, NULL);
-        if (fd < 0) {
-                LOG_ERR("Failed to allocate flow to %s.", dst_name);
-                return -1;
-        }
-
-        result = flow_alloc_res(fd);
-        if (result < 0) {
-                LOG_ERR("Allocate flow to %s result %d.", dst_name, result);
-                return -1;
-        }
-
-        flow_set_add(fmgr.nm1_set[qos], fd);
-
-        /* FIXME: Temporary, until we have a PFF */
-        fmgr.fd = fd;
-
         return 0;
 }
 
-int fmgr_nm1_write_sdu(struct pci * pci, struct shm_du_buff * sdb)
+int fmgr_nm1_write_sdu(struct pci *         pci,
+                       struct shm_du_buff * sdb)
 {
         if (pci == NULL || sdb == NULL)
                 return -1;
@@ -595,7 +645,8 @@ int fmgr_nm1_write_sdu(struct pci * pci, struct shm_du_buff * sdb)
         return 0;
 }
 
-int fmgr_nm1_write_buf(struct pci * pci, buffer_t * buf)
+int fmgr_nm1_write_buf(struct pci * pci,
+                       buffer_t *   buf)
 {
         buffer_t * buffer;
 
diff --git a/src/ipcpd/normal/fmgr.h b/src/ipcpd/normal/fmgr.h
index 85731081..ae5c8ea8 100644
--- a/src/ipcpd/normal/fmgr.h
+++ b/src/ipcpd/normal/fmgr.h
@@ -23,6 +23,7 @@
 #define OUROBOROS_IPCPD_NORMAL_FMGR_H
 
 #include <ouroboros/shared.h>
+#include <ouroboros/qos.h>
 
 #include "ae.h"
 #include "frct.h"
@@ -47,15 +48,14 @@ int fmgr_np1_post_buf(cep_id_t   id,
 int fmgr_np1_post_sdu(cep_id_t             id,
                       struct shm_du_buff * sdb);
 
-int fmgr_nm1_add_flow(int fd);
-
-int fmgr_nm1_dt_flow(char *    dst_name,
-                     qoscube_t qos);
-
 int fmgr_nm1_write_sdu(struct pci *         pci,
                        struct shm_du_buff * sdb);
 
 int fmgr_nm1_write_buf(struct pci * pci,
                        buffer_t *   buf);
 
+int fmgr_nm1_flow_arr(int       fd,
+                      qosspec_t qs);
+
+
 #endif /* OUROBOROS_IPCPD_NORMAL_FMGR_H */
diff --git a/src/ipcpd/normal/gam.c b/src/ipcpd/normal/gam.c
new file mode 100644
index 00000000..0e626115
--- /dev/null
+++ b/src/ipcpd/normal/gam.c
@@ -0,0 +1,285 @@
+/*
+ * Ouroboros - Copyright (C) 2016 - 2017
+ *
+ * Graph adjacency manager for IPC Process components
+ *
+ *    Dimitri Staessens <[email protected]>
+ *    Sander Vrijders   <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#define OUROBOROS_PREFIX "graph-adjacency-manager"
+
+#include <ouroboros/config.h>
+#include <ouroboros/dev.h>
+#include <ouroboros/logs.h>
+#include <ouroboros/list.h>
+#include <ouroboros/errno.h>
+
+#include "ribmgr.h"
+#include "ipcp.h"
+#include "gam.h"
+#include "pol-gam-ops.h"
+#include "pol/complete.h"
+
+#include <assert.h>
+#include <stdlib.h>
+#include <pthread.h>
+#include <string.h>
+
+struct ga {
+        struct list_head    next;
+
+        qosspec_t           qs;
+        int                 fd;
+        struct cacep_info * info;
+};
+
+struct gam {
+        struct list_head     gas;
+        pthread_mutex_t      gas_lock;
+        pthread_cond_t       gas_cond;
+
+        char *               ae_name;
+
+        struct pol_gam_ops * ops;
+        void *               ops_o;
+};
+
+struct gam * gam_create(enum pol_gam gam_type,
+                        const char * ae_name)
+{
+        struct gam * tmp;
+
+        tmp = malloc(sizeof(*tmp));
+        if (tmp == NULL)
+                return NULL;
+
+        switch (gam_type) {
+        case COMPLETE:
+                tmp->ops = &complete_ops;
+                break;
+        default:
+                free(tmp);
+                return NULL;
+        }
+
+        list_head_init(&tmp->gas);
+
+        tmp->ae_name = strdup(ae_name);
+        if (tmp->ae_name == NULL) {
+                free(tmp);
+                return NULL;
+        }
+
+        if (pthread_mutex_init(&tmp->gas_lock, NULL)) {
+                free(tmp->ae_name);
+                free(tmp);
+                return NULL;
+        }
+
+        if (pthread_cond_init(&tmp->gas_cond, NULL)) {
+                pthread_mutex_destroy(&tmp->gas_lock);
+                free(tmp->ae_name);
+                free(tmp);
+                return NULL;
+        }
+
+        tmp->ops_o = tmp->ops->create(tmp);
+        if (tmp->ops_o == NULL) {
+                pthread_cond_destroy(&tmp->gas_cond);
+                pthread_mutex_destroy(&tmp->gas_lock);
+                free(tmp->ae_name);
+                free(tmp);
+                return NULL;
+        }
+
+        return tmp;
+}
+
+void gam_destroy(struct gam * instance)
+{
+        struct list_head * p = NULL;
+        struct list_head * n = NULL;
+
+        assert(instance);
+
+        instance->ops->destroy(instance->ops_o);
+
+        pthread_mutex_destroy(&instance->gas_lock);
+        pthread_cond_destroy(&instance->gas_cond);
+
+        list_for_each_safe(p, n, &instance->gas) {
+                struct ga * e = list_entry(p, struct ga, next);
+                list_del(&e->next);
+                free(e->info);
+                free(e);
+        }
+
+        free(instance->ae_name);
+        free(instance);
+}
+
+static int add_ga(struct gam *        instance,
+                  int                 fd,
+                  qosspec_t           qs,
+                  struct cacep_info * info)
+{
+        struct ga * ga;
+
+        ga = malloc(sizeof(*ga));
+        if (ga == NULL)
+                return -ENOMEM;
+
+        ga->fd = fd;
+        ga->info = info;
+        ga->qs = qs;
+
+        list_head_init(&ga->next);
+
+        pthread_mutex_lock(&instance->gas_lock);
+        list_add(&ga->next, &instance->gas);
+        pthread_cond_signal(&instance->gas_cond);
+        pthread_mutex_unlock(&instance->gas_lock);
+
+        return 0;
+}
+
+int gam_flow_arr(struct gam * instance,
+                 int          fd,
+                 qosspec_t    qs)
+{
+        struct cacep *      cacep;
+        struct cacep_info * info;
+
+        if (flow_alloc_resp(fd, instance->ops->accept_new_flow(instance->ops_o))
+            < 0) {
+                LOG_ERR("Could not respond to new flow.");
+                return -1;
+        }
+
+        cacep = cacep_create(fd, ipcpi.name, ribmgr_address());
+        if (cacep == NULL) {
+                LOG_ERR("Failed to create CACEP instance.");
+                return -1;
+        }
+
+        info = cacep_auth_wait(cacep);
+        if (info == NULL) {
+                LOG_ERR("Other side failed to authenticate.");
+                cacep_destroy(cacep);
+                return -1;
+        }
+
+        cacep_destroy(cacep);
+
+        if (instance->ops->accept_flow(instance->ops_o, qs, info)) {
+                flow_dealloc(fd);
+                free(info);
+                return 0;
+        }
+
+        if (add_ga(instance, fd, qs, info)) {
+                LOG_ERR("Failed to add ga to graph adjacency manager list.");
+                free(info);
+                return -1;
+        }
+
+        return 0;
+}
+
+int gam_flow_alloc(struct gam * instance,
+                   char *       dst_name,
+                   qosspec_t    qs)
+{
+        struct cacep *      cacep;
+        struct cacep_info * info;
+        int                 fd;
+
+        fd = flow_alloc(dst_name, instance->ae_name, NULL);
+        if (fd < 0) {
+                LOG_ERR("Failed to allocate flow to %s.", dst_name);
+                return -1;
+        }
+
+        if (flow_alloc_res(fd)) {
+                LOG_ERR("Flow allocation to %s failed.", dst_name);
+                flow_dealloc(fd);
+                return -1;
+        }
+
+        cacep = cacep_create(fd, ipcpi.name, ribmgr_address());
+        if (cacep == NULL) {
+                LOG_ERR("Failed to create CACEP instance.");
+                return -1;
+        }
+
+        info = cacep_auth(cacep);
+        if (info == NULL) {
+                LOG_ERR("Failed to authenticate.");
+                cacep_destroy(cacep);
+                return -1;
+        }
+
+        cacep_destroy(cacep);
+
+        if (instance->ops->accept_flow(instance->ops_o, qs, info)) {
+                flow_dealloc(fd);
+                free(info);
+                return 0;
+        }
+
+        if (add_ga(instance, fd, qs, info)) {
+                LOG_ERR("Failed to add GA to graph adjacency manager list.");
+                free(info);
+                return -1;
+        }
+
+        return 0;
+}
+
+int gam_flow_wait(struct gam *         instance,
+                  int *                fd,
+                  struct cacep_info ** info,
+                  qosspec_t *          qs)
+{
+        struct ga * ga;
+
+        assert(fd);
+        assert(info);
+        assert(qs);
+
+        pthread_mutex_lock(&instance->gas_lock);
+
+        while (list_is_empty(&instance->gas))
+                pthread_cond_wait(&instance->gas_cond, &instance->gas_lock);
+
+        ga = list_first_entry((&instance->gas), struct ga, next);
+        if (ga == NULL) {
+                pthread_mutex_unlock(&instance->gas_lock);
+                return -1;
+        }
+
+        *fd   = ga->fd;
+        *info = ga->info;
+        *qs   = ga->qs;
+
+        list_del(&ga->next);
+        free(ga);
+
+        pthread_mutex_unlock(&instance->gas_lock);
+
+        return 0;
+}
diff --git a/src/ipcpd/normal/gam.h b/src/ipcpd/normal/gam.h
new file mode 100644
index 00000000..e858114c
--- /dev/null
+++ b/src/ipcpd/normal/gam.h
@@ -0,0 +1,46 @@
+/*
+ * Ouroboros - Copyright (C) 2016 - 2017
+ *
+ * Graph adjacency manager for IPC Process components
+ *
+ *    Dimitri Staessens <[email protected]>
+ *    Sander Vrijders   <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef OUROBOROS_IPCPD_NORMAL_GAM_H
+#define OUROBOROS_IPCPD_NORMAL_GAM_H
+
+#include <ouroboros/cacep.h>
+
+struct gam * gam_create(enum pol_gam gam_type,
+                        const char * ae_name);
+
+void         gam_destroy(struct gam * instance);
+
+int          gam_flow_arr(struct gam * instance,
+                          int          fd,
+                          qosspec_t    qs);
+
+int          gam_flow_alloc(struct gam * instance,
+                            char *       dst_name,
+                            qosspec_t    qs);
+
+int          gam_flow_wait(struct gam *         instance,
+                           int *                fd,
+                           struct cacep_info ** info,
+                           qosspec_t *          qs);
+
+#endif /* OUROBOROS_IPCPD_NORMAL_GAM_H */
diff --git a/src/ipcpd/normal/main.c b/src/ipcpd/normal/main.c
index 8db754aa..85f56ab0 100644
--- a/src/ipcpd/normal/main.c
+++ b/src/ipcpd/normal/main.c
@@ -26,6 +26,7 @@
 #include <ouroboros/dev.h>
 #include <ouroboros/ipcp-dev.h>
 #include <ouroboros/time_utils.h>
+#include <ouroboros/irm.h>
 
 #include "fmgr.h"
 #include "ribmgr.h"
@@ -47,7 +48,9 @@ int irmd_api;
 
 pthread_t acceptor;
 
-void ipcp_sig_handler(int sig, siginfo_t * info, void * c)
+void ipcp_sig_handler(int         sig,
+                      siginfo_t * info,
+                      void *      c)
 {
         (void) c;
 
@@ -55,7 +58,7 @@ void ipcp_sig_handler(int sig, siginfo_t * info, void * c)
         case SIGINT:
         case SIGTERM:
         case SIGHUP:
-                if (info->si_pid == irmd_api) {
+                if (info->si_pid == ipcpi.irmd_api) {
                         pthread_rwlock_wrlock(&ipcpi.state_lock);
 
                         if (ipcp_get_state() == IPCP_INIT)
@@ -101,7 +104,7 @@ static void * flow_acceptor(void * o)
                 if (strcmp(ae_name, MGMT_AE) == 0) {
                         ribmgr_add_nm1_flow(fd);
                 } else if (strcmp(ae_name, DT_AE) == 0) {
-                        fmgr_nm1_add_flow(fd);
+                        fmgr_nm1_flow_arr(fd, qs);
                 } else {
                         LOG_DBG("Flow allocation request for unknown AE %s.",
                                 ae_name);
@@ -195,12 +198,6 @@ static int normal_ipcp_enroll(char * dst_name)
 
         pthread_rwlock_unlock(&ipcpi.state_lock);
 
-        /* FIXME: Remove once we obtain neighbors during enrollment */
-        if (fmgr_nm1_dt_flow(dst_name, QOS_CUBE_BE)) {
-                LOG_ERR("Failed to establish data transfer flow.");
-                return -1;
-        }
-
         LOG_DBG("Enrolled with %s.", dst_name);
 
         return 0;
@@ -296,10 +293,11 @@ static struct ipcp_ops normal_ops = {
         .ipcp_flow_dealloc    = fmgr_np1_dealloc
 };
 
-int main(int argc, char * argv[])
+int main(int    argc,
+         char * argv[])
 {
         struct sigaction sig_act;
-        sigset_t sigset;
+        sigset_t         sigset;
 
         if (ap_init(argv[0])) {
                 LOG_ERR("Failed to init AP");
@@ -317,8 +315,11 @@ int main(int argc, char * argv[])
                 exit(EXIT_FAILURE);
         }
 
-        /* store the process id of the irmd */
-        irmd_api = atoi(argv[1]);
+        if (irm_bind_api(getpid(), ipcpi.name)) {
+                LOG_ERR("Failed to bind AP name.");
+                close_logfile();
+                exit(EXIT_FAILURE);
+        }
 
         /* init sig_act */
         memset(&sig_act, 0, sizeof(sig_act));
diff --git a/src/ipcpd/normal/pol-gam-ops.h b/src/ipcpd/normal/pol-gam-ops.h
new file mode 100644
index 00000000..eeece8d9
--- /dev/null
+++ b/src/ipcpd/normal/pol-gam-ops.h
@@ -0,0 +1,40 @@
+/*
+ * Ouroboros - Copyright (C) 2016 - 2017
+ *
+ * Graph adjacency manager policy ops
+ *
+ *    Dimitri Staessens <[email protected]>
+ *    Sander Vrijders   <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef OUROBOROS_IPCPD_NORMAL_POL_GAM_OPS_H
+#define OUROBOROS_IPCPD_NORMAL_POL_GAM_OPS_H
+
+#include <ouroboros/cacep.h>
+
+struct pol_gam_ops {
+        void * (* create)(struct gam * instance);
+
+        void   (* destroy)(void * o);
+
+        int    (* accept_new_flow)(void * o);
+
+        int    (* accept_flow)(void *                    o,
+                               qosspec_t                 qs,
+                               const struct cacep_info * info);
+};
+
+#endif /* OUROBOROS_IPCPD_NORMAL_POL_GAM_OPS_H */
diff --git a/src/ipcpd/normal/pol/complete.c b/src/ipcpd/normal/pol/complete.c
new file mode 100644
index 00000000..89e1b91f
--- /dev/null
+++ b/src/ipcpd/normal/pol/complete.c
@@ -0,0 +1,208 @@
+/*
+ * Ouroboros - Copyright (C) 2016 - 2017
+ *
+ * Graph adjacency manager for IPC Process components
+ *
+ *    Dimitri Staessens <[email protected]>
+ *    Sander Vrijders   <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#define OUROBOROS_PREFIX "complete-graph-adjacency-manager"
+
+#include <ouroboros/config.h>
+#include <ouroboros/logs.h>
+#include <ouroboros/list.h>
+#include <ouroboros/qos.h>
+
+#include "pathname.h"
+#include "ro.h"
+#include "ipcp.h"
+#include "gam.h"
+
+#include <string.h>
+#include <stdlib.h>
+
+#define RO_DIR "neighbors"
+
+struct neighbor {
+        struct list_head next;
+        char *           neighbor;
+};
+
+struct complete {
+        struct list_head neighbors;
+        pthread_mutex_t  neighbors_lock;
+
+        pthread_t        allocator;
+
+        struct gam *     gam;
+};
+
+static void * allocator(void * o)
+{
+        qosspec_t         qs;
+        ssize_t           len;
+        char **           children;
+        int               i;
+        char *            ro_name;
+        struct complete * complete = (struct complete *) o;
+
+        qs.delay = 0;
+        qs.jitter = 0;
+
+        ro_name = pathname_create(RO_DIR);
+        if (ro_name == NULL)
+                return (void *) -1;
+
+        len = ro_children(ro_name, &children);
+        if (len > 0) {
+                for (i = 0; i < len; i++) {
+                        if (strcmp(children[i], ipcpi.name) == 0)
+                                continue;
+                        gam_flow_alloc(complete->gam, children[i], qs);
+                }
+        }
+
+        pathname_destroy(ro_name);
+
+        return (void *) 0;
+}
+
+void * complete_create(struct gam * gam)
+{
+        struct ro_attr    attr;
+        char *            ro_name;
+        struct complete * complete;
+
+        ro_attr_init(&attr);
+        attr.enrol_sync = true;
+        attr.recv_set = ALL_MEMBERS;
+
+        complete = malloc(sizeof(*complete));
+        if (complete == NULL)
+                return NULL;
+
+        ro_name = pathname_create(RO_DIR);
+        if (ro_name == NULL) {
+                free(complete);
+                return NULL;
+        }
+
+        if (!ro_exists(RO_DIR)) {
+                if (ro_create(ro_name, &attr, NULL, 0)) {
+                        free(complete);
+                        pathname_destroy(ro_name);
+                        return NULL;
+                }
+        }
+
+        ro_name = pathname_append(ro_name, ipcpi.name);
+        if (ro_name == NULL) {
+                free(complete);
+                pathname_destroy(ro_name);
+                return NULL;
+        }
+
+        if (ro_create(ro_name, &attr, NULL, 0)) {
+                free(complete);
+                pathname_destroy(ro_name);
+                return NULL;
+        }
+        pathname_destroy(ro_name);
+
+        list_head_init(&complete->neighbors);
+        complete->gam = gam;
+
+        if (pthread_mutex_init(&complete->neighbors_lock, NULL)) {
+                free(complete);
+                return NULL;
+        }
+
+        if (pthread_create(&complete->allocator, NULL,
+                           allocator, (void *) complete)) {
+                free(complete);
+                pthread_mutex_destroy(&complete->neighbors_lock);
+                return NULL;
+        }
+
+        return (void *) complete;
+}
+
+void complete_destroy(void * o)
+{
+        struct list_head * p = NULL;
+        struct list_head * n = NULL;
+        struct complete * complete = (struct complete *) o;
+
+        pthread_cancel(complete->allocator);
+        pthread_join(complete->allocator, NULL);
+
+        list_for_each_safe(p, n, &complete->neighbors) {
+                struct neighbor * e = list_entry(p, struct neighbor, next);
+                list_del(&e->next);
+                free(e->neighbor);
+                free(e);
+        }
+}
+
+int complete_accept_new_flow(void * o)
+{
+        (void) o;
+
+        return 0;
+}
+
+int complete_accept_flow(void *                    o,
+                         qosspec_t                 qs,
+                         const struct cacep_info * info)
+{
+        struct list_head * pos = NULL;
+        struct neighbor * n;
+        struct complete * complete = (struct complete *) o;
+
+        (void) qs;
+
+        pthread_mutex_lock(&complete->neighbors_lock);
+
+        list_for_each(pos, &complete->neighbors) {
+                struct neighbor * e = list_entry(pos, struct neighbor, next);
+                if (strcmp(e->neighbor, info->name) == 0) {
+                        pthread_mutex_unlock(&complete->neighbors_lock);
+                        return -1;
+                }
+        }
+
+        n = malloc(sizeof(*n));
+        if (n == NULL) {
+                pthread_mutex_unlock(&complete->neighbors_lock);
+                return -1;
+        }
+
+        list_head_init(&n->next);
+
+        n->neighbor = strdup(info->name);
+        if (n->neighbor == NULL) {
+                pthread_mutex_unlock(&complete->neighbors_lock);
+                free(n);
+                return -1;
+        }
+
+        list_add(&n->next, &complete->neighbors);
+
+        pthread_mutex_unlock(&complete->neighbors_lock);
+
+        return 0;
+}
diff --git a/src/ipcpd/normal/pol/complete.h b/src/ipcpd/normal/pol/complete.h
new file mode 100644
index 00000000..8fcc87ba
--- /dev/null
+++ b/src/ipcpd/normal/pol/complete.h
@@ -0,0 +1,46 @@
+/*
+ * Ouroboros - Copyright (C) 2016 - 2017
+ *
+ * Graph adjacency manager for IPC Process components
+ *
+ *    Dimitri Staessens <[email protected]>
+ *    Sander Vrijders   <[email protected]>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+#ifndef OUROBOROS_IPCPD_NORMAL_POL_COMPLETE_H
+#define OUROBOROS_IPCPD_NORMAL_POL_COMPLETE_H
+
+#include "gam.h"
+#include "pol-gam-ops.h"
+
+void * complete_create(struct gam * instance);
+
+void   complete_destroy(void * o);
+
+int    complete_accept_new_flow(void * o);
+
+int    complete_accept_flow(void *                    o,
+                            qosspec_t                 qs,
+                            const struct cacep_info * info);
+
+struct pol_gam_ops complete_ops = {
+        .create          = complete_create,
+        .destroy         = complete_destroy,
+        .accept_new_flow = complete_accept_new_flow,
+        .accept_flow     = complete_accept_flow
+};
+
+#endif /* OUROBOROS_IPCPD_NORMAL_POL_COMPLETE_H */
diff --git a/src/ipcpd/normal/ribmgr.c b/src/ipcpd/normal/ribmgr.c
index 6356d48c..993fe62a 100644
--- a/src/ipcpd/normal/ribmgr.c
+++ b/src/ipcpd/normal/ribmgr.c
@@ -139,6 +139,8 @@ struct {
         struct addr_auth *  addr_auth;
         enum pol_addr_auth  addr_auth_type;
 
+        enum pol_gam        dt_gam_type;
+
         enum ribmgr_state   state;
         pthread_cond_t      state_cond;
         pthread_mutex_t     state_lock;
@@ -168,6 +170,7 @@ void ribmgr_ro_created(const char * name,
                 rib.dtc.min_pdu_size = stat_msg->min_pdu_size;
                 rib.dtc.max_pdu_size = stat_msg->max_pdu_size;
                 rib.addr_auth_type = stat_msg->addr_auth_type;
+                rib.dt_gam_type = stat_msg->dt_gam_type;
 
                 static_info_msg__free_unpacked(stat_msg, NULL);
         }
@@ -1262,6 +1265,7 @@ int ribmgr_bootstrap(struct dif_config * conf)
         stat_info.min_pdu_size = rib.dtc.min_pdu_size = conf->min_pdu_size;
         stat_info.max_pdu_size = rib.dtc.max_pdu_size = conf->max_pdu_size;
         stat_info.addr_auth_type = rib.addr_auth_type = conf->addr_auth_type;
+        stat_info.dt_gam_type = rib.dt_gam_type = conf->dt_gam_type;
 
         len = static_info_msg__get_packed_size(&stat_info);
         if (len == 0) {
@@ -1354,16 +1358,21 @@ int ribmgr_start_policies(void)
         return 0;
 }
 
-struct dt_const * ribmgr_dt_const()
+struct dt_const * ribmgr_dt_const(void)
 {
         return &(rib.dtc);
 }
 
-uint64_t ribmgr_address()
+uint64_t ribmgr_address(void)
 {
         return rib.address;
 }
 
+enum pol_gam ribmgr_dt_gam(void)
+{
+        return rib.dt_gam_type;
+}
+
 static int send_neighbors_ro(char * name, ro_msg_t * msg, enum cdap_opcode code)
 {
         struct list_head * p = NULL;
@@ -1577,8 +1586,8 @@ ssize_t ro_children(const char * name, char *** children)
         }
         child = node->child;
 
-        **children = malloc(len);
-        if (**children == NULL) {
+        *children = malloc(len);
+        if (*children == NULL) {
                 pthread_mutex_unlock(&rib.ro_lock);
                 return -1;
         }
@@ -1590,7 +1599,7 @@ ssize_t ro_children(const char * name, char *** children)
                                 free((*children)[i]);
                                 i--;
                         }
-                        free(**children);
+                        free(*children);
                         pthread_mutex_unlock(&rib.ro_lock);
                         return -1;
                 }
diff --git a/src/ipcpd/normal/ribmgr.h b/src/ipcpd/normal/ribmgr.h
index ddc98e6e..22212de9 100644
--- a/src/ipcpd/normal/ribmgr.h
+++ b/src/ipcpd/normal/ribmgr.h
@@ -45,4 +45,6 @@ struct dt_const * ribmgr_dt_const(void);
 
 uint64_t          ribmgr_address(void);
 
+enum pol_gam      ribmgr_dt_gam(void);
+
 #endif /* OUROBOROS_IPCPD_NORMAL_RIBMGR_H */
diff --git a/src/ipcpd/normal/static_info.proto b/src/ipcpd/normal/static_info.proto
index 18f02e36..bb6f8c4e 100644
--- a/src/ipcpd/normal/static_info.proto
+++ b/src/ipcpd/normal/static_info.proto
@@ -23,13 +23,14 @@
 syntax = "proto2";
 
 message static_info_msg {
-        required uint32 addr_size       = 1;
-        required uint32 cep_id_size     = 2;
-        required uint32 pdu_length_size = 3;
-        required uint32 seqno_size      = 4;
-        required bool   has_ttl         = 5;
-        required bool   has_chk         = 6;
-        required uint32 min_pdu_size    = 7;
-        required uint32 max_pdu_size    = 8;
-        required uint32 addr_auth_type  = 9;
+        required uint32 addr_size       =  1;
+        required uint32 cep_id_size     =  2;
+        required uint32 pdu_length_size =  3;
+        required uint32 seqno_size      =  4;
+        required bool   has_ttl         =  5;
+        required bool   has_chk         =  6;
+        required uint32 min_pdu_size    =  7;
+        required uint32 max_pdu_size    =  8;
+        required uint32 addr_auth_type  =  9;
+        required uint32 dt_gam_type     = 10;
 }
\ No newline at end of file