Adding upstream version 8.4.4.upstream/8.4.4upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

1 files changed, 2503 insertions, 0 deletions

diff --git a/lib/northbound.c b/lib/northbound.c
new file mode 100644
index 0000000..ccfc384
--- /dev/null
+++ b/lib/northbound.c
@@ -0,0 +1,2503 @@
+/*
+ * Copyright (C) 2018  NetDEF, Inc.
+ *                     Renato Westphal
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the Free
+ * Software Foundation; either version 2 of the License, or (at your option)
+ * any later version.
+ *
+ * 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; see the file COPYING; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include <zebra.h>
+
+#include "libfrr.h"
+#include "log.h"
+#include "lib_errors.h"
+#include "hash.h"
+#include "command.h"
+#include "debug.h"
+#include "db.h"
+#include "frr_pthread.h"
+#include "northbound.h"
+#include "northbound_cli.h"
+#include "northbound_db.h"
+#include "frrstr.h"
+
+DEFINE_MTYPE_STATIC(LIB, NB_NODE, "Northbound Node");
+DEFINE_MTYPE_STATIC(LIB, NB_CONFIG, "Northbound Configuration");
+DEFINE_MTYPE_STATIC(LIB, NB_CONFIG_ENTRY, "Northbound Configuration Entry");
+
+/* Running configuration - shouldn't be modified directly. */
+struct nb_config *running_config;
+
+/* Hash table of user pointers associated with configuration entries. */
+static struct hash *running_config_entries;
+
+/* Management lock for the running configuration. */
+static struct {
+	/* Mutex protecting this structure. */
+	pthread_mutex_t mtx;
+
+	/* Actual lock. */
+	bool locked;
+
+	/* Northbound client who owns this lock. */
+	enum nb_client owner_client;
+
+	/* Northbound user who owns this lock. */
+	const void *owner_user;
+} running_config_mgmt_lock;
+
+/* Knob to record config transaction */
+static bool nb_db_enabled;
+/*
+ * Global lock used to prevent multiple configuration transactions from
+ * happening concurrently.
+ */
+static bool transaction_in_progress;
+
+static int nb_callback_pre_validate(struct nb_context *context,
+				    const struct nb_node *nb_node,
+				    const struct lyd_node *dnode, char *errmsg,
+				    size_t errmsg_len);
+static int nb_callback_configuration(struct nb_context *context,
+				     const enum nb_event event,
+				     struct nb_config_change *change,
+				     char *errmsg, size_t errmsg_len);
+static struct nb_transaction *
+nb_transaction_new(struct nb_context *context, struct nb_config *config,
+		   struct nb_config_cbs *changes, const char *comment,
+		   char *errmsg, size_t errmsg_len);
+static void nb_transaction_free(struct nb_transaction *transaction);
+static int nb_transaction_process(enum nb_event event,
+				  struct nb_transaction *transaction,
+				  char *errmsg, size_t errmsg_len);
+static void nb_transaction_apply_finish(struct nb_transaction *transaction,
+					char *errmsg, size_t errmsg_len);
+static int nb_oper_data_iter_node(const struct lysc_node *snode,
+				  const char *xpath, const void *list_entry,
+				  const struct yang_list_keys *list_keys,
+				  struct yang_translator *translator,
+				  bool first, uint32_t flags,
+				  nb_oper_data_cb cb, void *arg);
+
+static int nb_node_check_config_only(const struct lysc_node *snode, void *arg)
+{
+	bool *config_only = arg;
+
+	if (CHECK_FLAG(snode->flags, LYS_CONFIG_R)) {
+		*config_only = false;
+		return YANG_ITER_STOP;
+	}
+
+	return YANG_ITER_CONTINUE;
+}
+
+static int nb_node_new_cb(const struct lysc_node *snode, void *arg)
+{
+	struct nb_node *nb_node;
+	struct lysc_node *sparent, *sparent_list;
+
+	nb_node = XCALLOC(MTYPE_NB_NODE, sizeof(*nb_node));
+	yang_snode_get_path(snode, YANG_PATH_DATA, nb_node->xpath,
+			    sizeof(nb_node->xpath));
+	nb_node->priority = NB_DFLT_PRIORITY;
+	sparent = yang_snode_real_parent(snode);
+	if (sparent)
+		nb_node->parent = sparent->priv;
+	sparent_list = yang_snode_parent_list(snode);
+	if (sparent_list)
+		nb_node->parent_list = sparent_list->priv;
+
+	/* Set flags. */
+	if (CHECK_FLAG(snode->nodetype, LYS_CONTAINER | LYS_LIST)) {
+		bool config_only = true;
+
+		(void)yang_snodes_iterate_subtree(snode, NULL,
+						  nb_node_check_config_only, 0,
+						  &config_only);
+		if (config_only)
+			SET_FLAG(nb_node->flags, F_NB_NODE_CONFIG_ONLY);
+	}
+	if (CHECK_FLAG(snode->nodetype, LYS_LIST)) {
+		if (yang_snode_num_keys(snode) == 0)
+			SET_FLAG(nb_node->flags, F_NB_NODE_KEYLESS_LIST);
+	}
+
+	/*
+	 * Link the northbound node and the libyang schema node with one
+	 * another.
+	 */
+	nb_node->snode = snode;
+	assert(snode->priv == NULL);
+	((struct lysc_node *)snode)->priv = nb_node;
+
+	return YANG_ITER_CONTINUE;
+}
+
+static int nb_node_del_cb(const struct lysc_node *snode, void *arg)
+{
+	struct nb_node *nb_node;
+
+	nb_node = snode->priv;
+	if (nb_node) {
+		((struct lysc_node *)snode)->priv = NULL;
+		XFREE(MTYPE_NB_NODE, nb_node);
+	}
+
+	return YANG_ITER_CONTINUE;
+}
+
+void nb_nodes_create(void)
+{
+	yang_snodes_iterate(NULL, nb_node_new_cb, 0, NULL);
+}
+
+void nb_nodes_delete(void)
+{
+	yang_snodes_iterate(NULL, nb_node_del_cb, 0, NULL);
+}
+
+struct nb_node *nb_node_find(const char *path)
+{
+	const struct lysc_node *snode;
+
+	/*
+	 * Use libyang to find the schema node associated to the path and get
+	 * the northbound node from there (snode private pointer).
+	 */
+	snode = lys_find_path(ly_native_ctx, NULL, path, 0);
+	if (!snode)
+		return NULL;
+
+	return snode->priv;
+}
+
+void nb_node_set_dependency_cbs(const char *dependency_xpath,
+				const char *dependant_xpath,
+				struct nb_dependency_callbacks *cbs)
+{
+	struct nb_node *dependency = nb_node_find(dependency_xpath);
+	struct nb_node *dependant = nb_node_find(dependant_xpath);
+
+	if (!dependency || !dependant)
+		return;
+
+	dependency->dep_cbs.get_dependant_xpath = cbs->get_dependant_xpath;
+	dependant->dep_cbs.get_dependency_xpath = cbs->get_dependency_xpath;
+}
+
+bool nb_node_has_dependency(struct nb_node *node)
+{
+	return node->dep_cbs.get_dependency_xpath != NULL;
+}
+
+static int nb_node_validate_cb(const struct nb_node *nb_node,
+			       enum nb_operation operation,
+			       int callback_implemented, bool optional)
+{
+	bool valid;
+
+	valid = nb_operation_is_valid(operation, nb_node->snode);
+
+	/*
+	 * Add an exception for operational data callbacks. A rw list usually
+	 * doesn't need any associated operational data callbacks. But if this
+	 * rw list is augmented by another module which adds state nodes under
+	 * it, then this list will need to have the 'get_next()', 'get_keys()'
+	 * and 'lookup_entry()' callbacks. As such, never log a warning when
+	 * these callbacks are implemented when they are not needed, since this
+	 * depends on context (e.g. some daemons might augment "frr-interface"
+	 * while others don't).
+	 */
+	if (!valid && callback_implemented && operation != NB_OP_GET_NEXT
+	    && operation != NB_OP_GET_KEYS && operation != NB_OP_LOOKUP_ENTRY)
+		flog_warn(EC_LIB_NB_CB_UNNEEDED,
+			  "unneeded '%s' callback for '%s'",
+			  nb_operation_name(operation), nb_node->xpath);
+
+	if (!optional && valid && !callback_implemented) {
+		flog_err(EC_LIB_NB_CB_MISSING, "missing '%s' callback for '%s'",
+			 nb_operation_name(operation), nb_node->xpath);
+		return 1;
+	}
+
+	return 0;
+}
+
+/*
+ * Check if the required callbacks were implemented for the given northbound
+ * node.
+ */
+static unsigned int nb_node_validate_cbs(const struct nb_node *nb_node)
+
+{
+	unsigned int error = 0;
+
+	error += nb_node_validate_cb(nb_node, NB_OP_CREATE,
+				     !!nb_node->cbs.create, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_MODIFY,
+				     !!nb_node->cbs.modify, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_DESTROY,
+				     !!nb_node->cbs.destroy, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_MOVE, !!nb_node->cbs.move,
+				     false);
+	error += nb_node_validate_cb(nb_node, NB_OP_PRE_VALIDATE,
+				     !!nb_node->cbs.pre_validate, true);
+	error += nb_node_validate_cb(nb_node, NB_OP_APPLY_FINISH,
+				     !!nb_node->cbs.apply_finish, true);
+	error += nb_node_validate_cb(nb_node, NB_OP_GET_ELEM,
+				     !!nb_node->cbs.get_elem, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_GET_NEXT,
+				     !!nb_node->cbs.get_next, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_GET_KEYS,
+				     !!nb_node->cbs.get_keys, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_LOOKUP_ENTRY,
+				     !!nb_node->cbs.lookup_entry, false);
+	error += nb_node_validate_cb(nb_node, NB_OP_RPC, !!nb_node->cbs.rpc,
+				     false);
+
+	return error;
+}
+
+static unsigned int nb_node_validate_priority(const struct nb_node *nb_node)
+{
+	/* Top-level nodes can have any priority. */
+	if (!nb_node->parent)
+		return 0;
+
+	if (nb_node->priority < nb_node->parent->priority) {
+		flog_err(EC_LIB_NB_CB_INVALID_PRIO,
+			 "node has higher priority than its parent [xpath %s]",
+			 nb_node->xpath);
+		return 1;
+	}
+
+	return 0;
+}
+
+static int nb_node_validate(const struct lysc_node *snode, void *arg)
+{
+	struct nb_node *nb_node = snode->priv;
+	unsigned int *errors = arg;
+
+	/* Validate callbacks and priority. */
+	if (nb_node) {
+		*errors += nb_node_validate_cbs(nb_node);
+		*errors += nb_node_validate_priority(nb_node);
+	}
+
+	return YANG_ITER_CONTINUE;
+}
+
+struct nb_config *nb_config_new(struct lyd_node *dnode)
+{
+	struct nb_config *config;
+
+	config = XCALLOC(MTYPE_NB_CONFIG, sizeof(*config));
+	if (dnode)
+		config->dnode = dnode;
+	else
+		config->dnode = yang_dnode_new(ly_native_ctx, true);
+	config->version = 0;
+
+	return config;
+}
+
+void nb_config_free(struct nb_config *config)
+{
+	if (config->dnode)
+		yang_dnode_free(config->dnode);
+	XFREE(MTYPE_NB_CONFIG, config);
+}
+
+struct nb_config *nb_config_dup(const struct nb_config *config)
+{
+	struct nb_config *dup;
+
+	dup = XCALLOC(MTYPE_NB_CONFIG, sizeof(*dup));
+	dup->dnode = yang_dnode_dup(config->dnode);
+	dup->version = config->version;
+
+	return dup;
+}
+
+int nb_config_merge(struct nb_config *config_dst, struct nb_config *config_src,
+		    bool preserve_source)
+{
+	int ret;
+
+	ret = lyd_merge_siblings(&config_dst->dnode, config_src->dnode, 0);
+	if (ret != 0)
+		flog_warn(EC_LIB_LIBYANG, "%s: lyd_merge() failed", __func__);
+
+	if (!preserve_source)
+		nb_config_free(config_src);
+
+	return (ret == 0) ? NB_OK : NB_ERR;
+}
+
+void nb_config_replace(struct nb_config *config_dst,
+		       struct nb_config *config_src, bool preserve_source)
+{
+	/* Update version. */
+	if (config_src->version != 0)
+		config_dst->version = config_src->version;
+
+	/* Update dnode. */
+	if (config_dst->dnode)
+		yang_dnode_free(config_dst->dnode);
+	if (preserve_source) {
+		config_dst->dnode = yang_dnode_dup(config_src->dnode);
+	} else {
+		config_dst->dnode = config_src->dnode;
+		config_src->dnode = NULL;
+		nb_config_free(config_src);
+	}
+}
+
+/* Generate the nb_config_cbs tree. */
+static inline int nb_config_cb_compare(const struct nb_config_cb *a,
+				       const struct nb_config_cb *b)
+{
+	/* Sort by priority first. */
+	if (a->nb_node->priority < b->nb_node->priority)
+		return -1;
+	if (a->nb_node->priority > b->nb_node->priority)
+		return 1;
+
+	/*
+	 * Preserve the order of the configuration changes as told by libyang.
+	 */
+	if (a->seq < b->seq)
+		return -1;
+	if (a->seq > b->seq)
+		return 1;
+
+	/*
+	 * All 'apply_finish' callbacks have their sequence number set to zero.
+	 * In this case, compare them using their dnode pointers (the order
+	 * doesn't matter for callbacks that have the same priority).
+	 */
+	if (a->dnode < b->dnode)
+		return -1;
+	if (a->dnode > b->dnode)
+		return 1;
+
+	return 0;
+}
+RB_GENERATE(nb_config_cbs, nb_config_cb, entry, nb_config_cb_compare);
+
+static void nb_config_diff_add_change(struct nb_config_cbs *changes,
+				      enum nb_operation operation,
+				      uint32_t *seq,
+				      const struct lyd_node *dnode)
+{
+	struct nb_config_change *change;
+
+	/* Ignore unimplemented nodes. */
+	if (!dnode->schema->priv)
+		return;
+
+	change = XCALLOC(MTYPE_TMP, sizeof(*change));
+	change->cb.operation = operation;
+	change->cb.seq = *seq;
+	*seq = *seq + 1;
+	change->cb.nb_node = dnode->schema->priv;
+	change->cb.dnode = dnode;
+
+	RB_INSERT(nb_config_cbs, changes, &change->cb);
+}
+
+static void nb_config_diff_del_changes(struct nb_config_cbs *changes)
+{
+	while (!RB_EMPTY(nb_config_cbs, changes)) {
+		struct nb_config_change *change;
+
+		change = (struct nb_config_change *)RB_ROOT(nb_config_cbs,
+							    changes);
+		RB_REMOVE(nb_config_cbs, changes, &change->cb);
+		XFREE(MTYPE_TMP, change);
+	}
+}
+
+/*
+ * Helper function used when calculating the delta between two different
+ * configurations. Given a new subtree, calculate all new YANG data nodes,
+ * excluding default leafs and leaf-lists. This is a recursive function.
+ */
+static void nb_config_diff_created(const struct lyd_node *dnode, uint32_t *seq,
+				   struct nb_config_cbs *changes)
+{
+	enum nb_operation operation;
+	struct lyd_node *child;
+
+	/* Ignore unimplemented nodes. */
+	if (!dnode->schema->priv)
+		return;
+
+	switch (dnode->schema->nodetype) {
+	case LYS_LEAF:
+	case LYS_LEAFLIST:
+		if (lyd_is_default(dnode))
+			break;
+
+		if (nb_operation_is_valid(NB_OP_CREATE, dnode->schema))
+			operation = NB_OP_CREATE;
+		else if (nb_operation_is_valid(NB_OP_MODIFY, dnode->schema))
+			operation = NB_OP_MODIFY;
+		else
+			return;
+
+		nb_config_diff_add_change(changes, operation, seq, dnode);
+		break;
+	case LYS_CONTAINER:
+	case LYS_LIST:
+		if (nb_operation_is_valid(NB_OP_CREATE, dnode->schema))
+			nb_config_diff_add_change(changes, NB_OP_CREATE, seq,
+						  dnode);
+
+		/* Process child nodes recursively. */
+		LY_LIST_FOR (lyd_child(dnode), child) {
+			nb_config_diff_created(child, seq, changes);
+		}
+		break;
+	default:
+		break;
+	}
+}
+
+static void nb_config_diff_deleted(const struct lyd_node *dnode, uint32_t *seq,
+				   struct nb_config_cbs *changes)
+{
+	/* Ignore unimplemented nodes. */
+	if (!dnode->schema->priv)
+		return;
+
+	if (nb_operation_is_valid(NB_OP_DESTROY, dnode->schema))
+		nb_config_diff_add_change(changes, NB_OP_DESTROY, seq, dnode);
+	else if (CHECK_FLAG(dnode->schema->nodetype, LYS_CONTAINER)) {
+		struct lyd_node *child;
+
+		/*
+		 * Non-presence containers need special handling since they
+		 * don't have "destroy" callbacks. In this case, what we need to
+		 * do is to call the "destroy" callbacks of their child nodes
+		 * when applicable (i.e. optional nodes).
+		 */
+		LY_LIST_FOR (lyd_child(dnode), child) {
+			nb_config_diff_deleted(child, seq, changes);
+		}
+	}
+}
+
+static int nb_lyd_diff_get_op(const struct lyd_node *dnode)
+{
+	const struct lyd_meta *meta;
+	LY_LIST_FOR (dnode->meta, meta) {
+		if (strcmp(meta->name, "operation")
+		    || strcmp(meta->annotation->module->name, "yang"))
+			continue;
+		return lyd_get_meta_value(meta)[0];
+	}
+	return 'n';
+}
+
+#if 0 /* Used below in nb_config_diff inside normally disabled code */
+static inline void nb_config_diff_dnode_log_path(const char *context,
+						 const char *path,
+						 const struct lyd_node *dnode)
+{
+	if (dnode->schema->nodetype & LYD_NODE_TERM)
+		zlog_debug("nb_config_diff: %s: %s: %s", context, path,
+			   lyd_get_value(dnode));
+	else
+		zlog_debug("nb_config_diff: %s: %s", context, path);
+}
+
+static inline void nb_config_diff_dnode_log(const char *context,
+					    const struct lyd_node *dnode)
+{
+	if (!dnode) {
+		zlog_debug("nb_config_diff: %s: NULL", context);
+		return;
+	}
+
+	char *path = lyd_path(dnode, LYD_PATH_STD, NULL, 0);
+	nb_config_diff_dnode_log_path(context, path, dnode);
+	free(path);
+}
+#endif
+
+/* Calculate the delta between two different configurations. */
+static void nb_config_diff(const struct nb_config *config1,
+			   const struct nb_config *config2,
+			   struct nb_config_cbs *changes)
+{
+	struct lyd_node *diff = NULL;
+	const struct lyd_node *root, *dnode;
+	struct lyd_node *target;
+	int op;
+	LY_ERR err;
+	char *path;
+
+#if 0 /* Useful (noisy) when debugging diff code, and for improving later */
+	if (DEBUG_MODE_CHECK(&nb_dbg_cbs_config, DEBUG_MODE_ALL)) {
+		LY_LIST_FOR(config1->dnode, root) {
+			LYD_TREE_DFS_BEGIN(root, dnode) {
+				nb_config_diff_dnode_log("from", dnode);
+				LYD_TREE_DFS_END(root, dnode);
+			}
+		}
+		LY_LIST_FOR(config2->dnode, root) {
+			LYD_TREE_DFS_BEGIN(root, dnode) {
+				nb_config_diff_dnode_log("to", dnode);
+				LYD_TREE_DFS_END(root, dnode);
+			}
+		}
+	}
+#endif
+
+	err = lyd_diff_siblings(config1->dnode, config2->dnode,
+				LYD_DIFF_DEFAULTS, &diff);
+	assert(!err);
+
+	if (diff && DEBUG_MODE_CHECK(&nb_dbg_cbs_config, DEBUG_MODE_ALL)) {
+		char *s;
+
+		if (!lyd_print_mem(&s, diff, LYD_JSON,
+				   LYD_PRINT_WITHSIBLINGS | LYD_PRINT_WD_ALL)) {
+			zlog_debug("%s: %s", __func__, s);
+			free(s);
+		}
+	}
+
+	uint32_t seq = 0;
+
+	LY_LIST_FOR (diff, root) {
+		LYD_TREE_DFS_BEGIN (root, dnode) {
+			op = nb_lyd_diff_get_op(dnode);
+
+			path = lyd_path(dnode, LYD_PATH_STD, NULL, 0);
+
+#if 0 /* Useful (noisy) when debugging diff code, and for improving later */
+			if (DEBUG_MODE_CHECK(&nb_dbg_cbs_config, DEBUG_MODE_ALL)) {
+				char context[80];
+				snprintf(context, sizeof(context),
+					 "iterating diff: oper: %c seq: %u", op, seq);
+				nb_config_diff_dnode_log_path(context, path, dnode);
+			}
+#endif
+			switch (op) {
+			case 'c': /* create */
+				  /*
+				   * This is rather inefficient, but when we use
+				   * dnode from the diff instead of the
+				   * candidate config node we get failures when
+				   * looking up default values, etc, based on
+				   * the diff tree.
+				   */
+				target = yang_dnode_get(config2->dnode, path);
+				assert(target);
+				nb_config_diff_created(target, &seq, changes);
+
+				/* Skip rest of sub-tree, move to next sibling
+				 */
+				LYD_TREE_DFS_continue = 1;
+				break;
+			case 'd': /* delete */
+				target = yang_dnode_get(config1->dnode, path);
+				assert(target);
+				nb_config_diff_deleted(target, &seq, changes);
+
+				/* Skip rest of sub-tree, move to next sibling
+				 */
+				LYD_TREE_DFS_continue = 1;
+				break;
+			case 'r': /* replace */
+				/* either moving an entry or changing a value */
+				target = yang_dnode_get(config2->dnode, path);
+				assert(target);
+				nb_config_diff_add_change(changes, NB_OP_MODIFY,
+							  &seq, target);
+				break;
+			case 'n': /* none */
+			default:
+				break;
+			}
+			free(path);
+			LYD_TREE_DFS_END(root, dnode);
+		}
+	}
+
+	lyd_free_all(diff);
+}
+
+int nb_candidate_edit(struct nb_config *candidate,
+		      const struct nb_node *nb_node,
+		      enum nb_operation operation, const char *xpath,
+		      const struct yang_data *previous,
+		      const struct yang_data *data)
+{
+	struct lyd_node *dnode, *dep_dnode;
+	char xpath_edit[XPATH_MAXLEN];
+	char dep_xpath[XPATH_MAXLEN];
+	LY_ERR err;
+
+	/* Use special notation for leaf-lists (RFC 6020, section 9.13.5). */
+	if (nb_node->snode->nodetype == LYS_LEAFLIST)
+		snprintf(xpath_edit, sizeof(xpath_edit), "%s[.='%s']", xpath,
+			 data->value);
+	else
+		strlcpy(xpath_edit, xpath, sizeof(xpath_edit));
+
+	switch (operation) {
+	case NB_OP_CREATE:
+	case NB_OP_MODIFY:
+		err = lyd_new_path(candidate->dnode, ly_native_ctx, xpath_edit,
+				   (void *)data->value, LYD_NEW_PATH_UPDATE,
+				   &dnode);
+		if (err) {
+			flog_warn(EC_LIB_LIBYANG,
+				  "%s: lyd_new_path(%s) failed: %d", __func__,
+				  xpath_edit, err);
+			return NB_ERR;
+		} else if (dnode) {
+			/* Create default nodes */
+			LY_ERR err = lyd_new_implicit_tree(
+				dnode, LYD_IMPLICIT_NO_STATE, NULL);
+			if (err) {
+				flog_warn(EC_LIB_LIBYANG,
+					  "%s: lyd_new_implicit_all failed: %d",
+					  __func__, err);
+			}
+			/*
+			 * create dependency
+			 *
+			 * dnode returned by the lyd_new_path may be from a
+			 * different schema, so we need to update the nb_node
+			 */
+			nb_node = dnode->schema->priv;
+			if (nb_node->dep_cbs.get_dependency_xpath) {
+				nb_node->dep_cbs.get_dependency_xpath(
+					dnode, dep_xpath);
+
+				err = lyd_new_path(candidate->dnode,
+						   ly_native_ctx, dep_xpath,
+						   NULL, LYD_NEW_PATH_UPDATE,
+						   &dep_dnode);
+				/* Create default nodes */
+				if (!err && dep_dnode)
+					err = lyd_new_implicit_tree(
+						dep_dnode,
+						LYD_IMPLICIT_NO_STATE, NULL);
+				if (err) {
+					flog_warn(
+						EC_LIB_LIBYANG,
+						"%s: dependency: lyd_new_path(%s) failed: %d",
+						__func__, dep_xpath, err);
+					return NB_ERR;
+				}
+			}
+		}
+		break;
+	case NB_OP_DESTROY:
+		dnode = yang_dnode_get(candidate->dnode, xpath_edit);
+		if (!dnode)
+			/*
+			 * Return a special error code so the caller can choose
+			 * whether to ignore it or not.
+			 */
+			return NB_ERR_NOT_FOUND;
+		/* destroy dependant */
+		if (nb_node->dep_cbs.get_dependant_xpath) {
+			nb_node->dep_cbs.get_dependant_xpath(dnode, dep_xpath);
+
+			dep_dnode = yang_dnode_get(candidate->dnode, dep_xpath);
+			if (dep_dnode)
+				lyd_free_tree(dep_dnode);
+		}
+		lyd_free_tree(dnode);
+		break;
+	case NB_OP_MOVE:
+		/* TODO: update configuration. */
+		break;
+	default:
+		flog_warn(EC_LIB_DEVELOPMENT,
+			  "%s: unknown operation (%u) [xpath %s]", __func__,
+			  operation, xpath_edit);
+		return NB_ERR;
+	}
+
+	return NB_OK;
+}
+
+bool nb_candidate_needs_update(const struct nb_config *candidate)
+{
+	if (candidate->version < running_config->version)
+		return true;
+
+	return false;
+}
+
+int nb_candidate_update(struct nb_config *candidate)
+{
+	struct nb_config *updated_config;
+
+	updated_config = nb_config_dup(running_config);
+	if (nb_config_merge(updated_config, candidate, true) != NB_OK)
+		return NB_ERR;
+
+	nb_config_replace(candidate, updated_config, false);
+
+	return NB_OK;
+}
+
+/*
+ * Perform YANG syntactic and semantic validation.
+ *
+ * WARNING: lyd_validate() can change the configuration as part of the
+ * validation process.
+ */
+static int nb_candidate_validate_yang(struct nb_config *candidate, char *errmsg,
+				      size_t errmsg_len)
+{
+	if (lyd_validate_all(&candidate->dnode, ly_native_ctx,
+			     LYD_VALIDATE_NO_STATE, NULL)
+	    != 0) {
+		yang_print_errors(ly_native_ctx, errmsg, errmsg_len);
+		return NB_ERR_VALIDATION;
+	}
+
+	return NB_OK;
+}
+
+/* Perform code-level validation using the northbound callbacks. */
+static int nb_candidate_validate_code(struct nb_context *context,
+				      struct nb_config *candidate,
+				      struct nb_config_cbs *changes,
+				      char *errmsg, size_t errmsg_len)
+{
+	struct nb_config_cb *cb;
+	struct lyd_node *root, *child;
+	int ret;
+
+	/* First validate the candidate as a whole. */
+	LY_LIST_FOR (candidate->dnode, root) {
+		LYD_TREE_DFS_BEGIN (root, child) {
+			struct nb_node *nb_node;
+
+			nb_node = child->schema->priv;
+			if (!nb_node || !nb_node->cbs.pre_validate)
+				goto next;
+
+			ret = nb_callback_pre_validate(context, nb_node, child,
+						       errmsg, errmsg_len);
+			if (ret != NB_OK)
+				return NB_ERR_VALIDATION;
+
+		next:
+			LYD_TREE_DFS_END(root, child);
+		}
+	}
+
+	/* Now validate the configuration changes. */
+	RB_FOREACH (cb, nb_config_cbs, changes) {
+		struct nb_config_change *change = (struct nb_config_change *)cb;
+
+		ret = nb_callback_configuration(context, NB_EV_VALIDATE, change,
+						errmsg, errmsg_len);
+		if (ret != NB_OK)
+			return NB_ERR_VALIDATION;
+	}
+
+	return NB_OK;
+}
+
+int nb_candidate_validate(struct nb_context *context,
+			  struct nb_config *candidate, char *errmsg,
+			  size_t errmsg_len)
+{
+	struct nb_config_cbs changes;
+	int ret;
+
+	if (nb_candidate_validate_yang(candidate, errmsg, errmsg_len) != NB_OK)
+		return NB_ERR_VALIDATION;
+
+	RB_INIT(nb_config_cbs, &changes);
+	nb_config_diff(running_config, candidate, &changes);
+	ret = nb_candidate_validate_code(context, candidate, &changes, errmsg,
+					 errmsg_len);
+	nb_config_diff_del_changes(&changes);
+
+	return ret;
+}
+
+int nb_candidate_commit_prepare(struct nb_context *context,
+				struct nb_config *candidate,
+				const char *comment,
+				struct nb_transaction **transaction,
+				char *errmsg, size_t errmsg_len)
+{
+	struct nb_config_cbs changes;
+
+	if (nb_candidate_validate_yang(candidate, errmsg, errmsg_len)
+	    != NB_OK) {
+		flog_warn(EC_LIB_NB_CANDIDATE_INVALID,
+			  "%s: failed to validate candidate configuration",
+			  __func__);
+		return NB_ERR_VALIDATION;
+	}
+
+	RB_INIT(nb_config_cbs, &changes);
+	nb_config_diff(running_config, candidate, &changes);
+	if (RB_EMPTY(nb_config_cbs, &changes)) {
+		snprintf(
+			errmsg, errmsg_len,
+			"No changes to apply were found during preparation phase");
+		return NB_ERR_NO_CHANGES;
+	}
+
+	if (nb_candidate_validate_code(context, candidate, &changes, errmsg,
+				       errmsg_len)
+	    != NB_OK) {
+		flog_warn(EC_LIB_NB_CANDIDATE_INVALID,
+			  "%s: failed to validate candidate configuration",
+			  __func__);
+		nb_config_diff_del_changes(&changes);
+		return NB_ERR_VALIDATION;
+	}
+
+	*transaction = nb_transaction_new(context, candidate, &changes, comment,
+					  errmsg, errmsg_len);
+	if (*transaction == NULL) {
+		flog_warn(EC_LIB_NB_TRANSACTION_CREATION_FAILED,
+			  "%s: failed to create transaction: %s", __func__,
+			  errmsg);
+		nb_config_diff_del_changes(&changes);
+		return NB_ERR_LOCKED;
+	}
+
+	return nb_transaction_process(NB_EV_PREPARE, *transaction, errmsg,
+				      errmsg_len);
+}
+
+void nb_candidate_commit_abort(struct nb_transaction *transaction, char *errmsg,
+			       size_t errmsg_len)
+{
+	(void)nb_transaction_process(NB_EV_ABORT, transaction, errmsg,
+				     errmsg_len);
+	nb_transaction_free(transaction);
+}
+
+void nb_candidate_commit_apply(struct nb_transaction *transaction,
+			       bool save_transaction, uint32_t *transaction_id,
+			       char *errmsg, size_t errmsg_len)
+{
+	(void)nb_transaction_process(NB_EV_APPLY, transaction, errmsg,
+				     errmsg_len);
+	nb_transaction_apply_finish(transaction, errmsg, errmsg_len);
+
+	/* Replace running by candidate. */
+	transaction->config->version++;
+	nb_config_replace(running_config, transaction->config, true);
+
+	/* Record transaction. */
+	if (save_transaction && nb_db_enabled
+	    && nb_db_transaction_save(transaction, transaction_id) != NB_OK)
+		flog_warn(EC_LIB_NB_TRANSACTION_RECORD_FAILED,
+			  "%s: failed to record transaction", __func__);
+
+	nb_transaction_free(transaction);
+}
+
+int nb_candidate_commit(struct nb_context *context, struct nb_config *candidate,
+			bool save_transaction, const char *comment,
+			uint32_t *transaction_id, char *errmsg,
+			size_t errmsg_len)
+{
+	struct nb_transaction *transaction = NULL;
+	int ret;
+
+	ret = nb_candidate_commit_prepare(context, candidate, comment,
+					  &transaction, errmsg, errmsg_len);
+	/*
+	 * Apply the changes if the preparation phase succeeded. Otherwise abort
+	 * the transaction.
+	 */
+	if (ret == NB_OK)
+		nb_candidate_commit_apply(transaction, save_transaction,
+					  transaction_id, errmsg, errmsg_len);
+	else if (transaction != NULL)
+		nb_candidate_commit_abort(transaction, errmsg, errmsg_len);
+
+	return ret;
+}
+
+int nb_running_lock(enum nb_client client, const void *user)
+{
+	int ret = -1;
+
+	frr_with_mutex (&running_config_mgmt_lock.mtx) {
+		if (!running_config_mgmt_lock.locked) {
+			running_config_mgmt_lock.locked = true;
+			running_config_mgmt_lock.owner_client = client;
+			running_config_mgmt_lock.owner_user = user;
+			ret = 0;
+		}
+	}
+
+	return ret;
+}
+
+int nb_running_unlock(enum nb_client client, const void *user)
+{
+	int ret = -1;
+
+	frr_with_mutex (&running_config_mgmt_lock.mtx) {
+		if (running_config_mgmt_lock.locked
+		    && running_config_mgmt_lock.owner_client == client
+		    && running_config_mgmt_lock.owner_user == user) {
+			running_config_mgmt_lock.locked = false;
+			running_config_mgmt_lock.owner_client = NB_CLIENT_NONE;
+			running_config_mgmt_lock.owner_user = NULL;
+			ret = 0;
+		}
+	}
+
+	return ret;
+}
+
+int nb_running_lock_check(enum nb_client client, const void *user)
+{
+	int ret = -1;
+
+	frr_with_mutex (&running_config_mgmt_lock.mtx) {
+		if (!running_config_mgmt_lock.locked
+		    || (running_config_mgmt_lock.owner_client == client
+			&& running_config_mgmt_lock.owner_user == user))
+			ret = 0;
+	}
+
+	return ret;
+}
+
+static void nb_log_config_callback(const enum nb_event event,
+				   enum nb_operation operation,
+				   const struct lyd_node *dnode)
+{
+	const char *value;
+	char xpath[XPATH_MAXLEN];
+
+	if (!DEBUG_MODE_CHECK(&nb_dbg_cbs_config, DEBUG_MODE_ALL))
+		return;
+
+	yang_dnode_get_path(dnode, xpath, sizeof(xpath));
+	if (yang_snode_is_typeless_data(dnode->schema))
+		value = "(none)";
+	else
+		value = yang_dnode_get_string(dnode, NULL);
+
+	zlog_debug(
+		"northbound callback: event [%s] op [%s] xpath [%s] value [%s]",
+		nb_event_name(event), nb_operation_name(operation), xpath,
+		value);
+}
+
+static int nb_callback_create(struct nb_context *context,
+			      const struct nb_node *nb_node,
+			      enum nb_event event, const struct lyd_node *dnode,
+			      union nb_resource *resource, char *errmsg,
+			      size_t errmsg_len)
+{
+	struct nb_cb_create_args args = {};
+	bool unexpected_error = false;
+	int ret;
+
+	nb_log_config_callback(event, NB_OP_CREATE, dnode);
+
+	args.context = context;
+	args.event = event;
+	args.dnode = dnode;
+	args.resource = resource;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	ret = nb_node->cbs.create(&args);
+
+	/* Detect and log unexpected errors. */
+	switch (ret) {
+	case NB_OK:
+	case NB_ERR:
+		break;
+	case NB_ERR_VALIDATION:
+		if (event != NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_RESOURCE:
+		if (event != NB_EV_PREPARE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_INCONSISTENCY:
+		if (event == NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	default:
+		unexpected_error = true;
+		break;
+	}
+	if (unexpected_error)
+		DEBUGD(&nb_dbg_cbs_config,
+		       "northbound callback: unexpected return value: %s",
+		       nb_err_name(ret));
+
+	return ret;
+}
+
+static int nb_callback_modify(struct nb_context *context,
+			      const struct nb_node *nb_node,
+			      enum nb_event event, const struct lyd_node *dnode,
+			      union nb_resource *resource, char *errmsg,
+			      size_t errmsg_len)
+{
+	struct nb_cb_modify_args args = {};
+	bool unexpected_error = false;
+	int ret;
+
+	nb_log_config_callback(event, NB_OP_MODIFY, dnode);
+
+	args.context = context;
+	args.event = event;
+	args.dnode = dnode;
+	args.resource = resource;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	ret = nb_node->cbs.modify(&args);
+
+	/* Detect and log unexpected errors. */
+	switch (ret) {
+	case NB_OK:
+	case NB_ERR:
+		break;
+	case NB_ERR_VALIDATION:
+		if (event != NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_RESOURCE:
+		if (event != NB_EV_PREPARE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_INCONSISTENCY:
+		if (event == NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	default:
+		unexpected_error = true;
+		break;
+	}
+	if (unexpected_error)
+		DEBUGD(&nb_dbg_cbs_config,
+		       "northbound callback: unexpected return value: %s",
+		       nb_err_name(ret));
+
+	return ret;
+}
+
+static int nb_callback_destroy(struct nb_context *context,
+			       const struct nb_node *nb_node,
+			       enum nb_event event,
+			       const struct lyd_node *dnode, char *errmsg,
+			       size_t errmsg_len)
+{
+	struct nb_cb_destroy_args args = {};
+	bool unexpected_error = false;
+	int ret;
+
+	nb_log_config_callback(event, NB_OP_DESTROY, dnode);
+
+	args.context = context;
+	args.event = event;
+	args.dnode = dnode;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	ret = nb_node->cbs.destroy(&args);
+
+	/* Detect and log unexpected errors. */
+	switch (ret) {
+	case NB_OK:
+	case NB_ERR:
+		break;
+	case NB_ERR_VALIDATION:
+		if (event != NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_INCONSISTENCY:
+		if (event == NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	default:
+		unexpected_error = true;
+		break;
+	}
+	if (unexpected_error)
+		DEBUGD(&nb_dbg_cbs_config,
+		       "northbound callback: unexpected return value: %s",
+		       nb_err_name(ret));
+
+	return ret;
+}
+
+static int nb_callback_move(struct nb_context *context,
+			    const struct nb_node *nb_node, enum nb_event event,
+			    const struct lyd_node *dnode, char *errmsg,
+			    size_t errmsg_len)
+{
+	struct nb_cb_move_args args = {};
+	bool unexpected_error = false;
+	int ret;
+
+	nb_log_config_callback(event, NB_OP_MOVE, dnode);
+
+	args.context = context;
+	args.event = event;
+	args.dnode = dnode;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	ret = nb_node->cbs.move(&args);
+
+	/* Detect and log unexpected errors. */
+	switch (ret) {
+	case NB_OK:
+	case NB_ERR:
+		break;
+	case NB_ERR_VALIDATION:
+		if (event != NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	case NB_ERR_INCONSISTENCY:
+		if (event == NB_EV_VALIDATE)
+			unexpected_error = true;
+		break;
+	default:
+		unexpected_error = true;
+		break;
+	}
+	if (unexpected_error)
+		DEBUGD(&nb_dbg_cbs_config,
+		       "northbound callback: unexpected return value: %s",
+		       nb_err_name(ret));
+
+	return ret;
+}
+
+static int nb_callback_pre_validate(struct nb_context *context,
+				    const struct nb_node *nb_node,
+				    const struct lyd_node *dnode, char *errmsg,
+				    size_t errmsg_len)
+{
+	struct nb_cb_pre_validate_args args = {};
+	bool unexpected_error = false;
+	int ret;
+
+	nb_log_config_callback(NB_EV_VALIDATE, NB_OP_PRE_VALIDATE, dnode);
+
+	args.dnode = dnode;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	ret = nb_node->cbs.pre_validate(&args);
+
+	/* Detect and log unexpected errors. */
+	switch (ret) {
+	case NB_OK:
+	case NB_ERR_VALIDATION:
+		break;
+	default:
+		unexpected_error = true;
+		break;
+	}
+	if (unexpected_error)
+		DEBUGD(&nb_dbg_cbs_config,
+		       "northbound callback: unexpected return value: %s",
+		       nb_err_name(ret));
+
+	return ret;
+}
+
+static void nb_callback_apply_finish(struct nb_context *context,
+				     const struct nb_node *nb_node,
+				     const struct lyd_node *dnode, char *errmsg,
+				     size_t errmsg_len)
+{
+	struct nb_cb_apply_finish_args args = {};
+
+	nb_log_config_callback(NB_EV_APPLY, NB_OP_APPLY_FINISH, dnode);
+
+	args.context = context;
+	args.dnode = dnode;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	nb_node->cbs.apply_finish(&args);
+}
+
+struct yang_data *nb_callback_get_elem(const struct nb_node *nb_node,
+				       const char *xpath,
+				       const void *list_entry)
+{
+	struct nb_cb_get_elem_args args = {};
+
+	DEBUGD(&nb_dbg_cbs_state,
+	       "northbound callback (get_elem): xpath [%s] list_entry [%p]",
+	       xpath, list_entry);
+
+	args.xpath = xpath;
+	args.list_entry = list_entry;
+	return nb_node->cbs.get_elem(&args);
+}
+
+const void *nb_callback_get_next(const struct nb_node *nb_node,
+				 const void *parent_list_entry,
+				 const void *list_entry)
+{
+	struct nb_cb_get_next_args args = {};
+
+	DEBUGD(&nb_dbg_cbs_state,
+	       "northbound callback (get_next): node [%s] parent_list_entry [%p] list_entry [%p]",
+	       nb_node->xpath, parent_list_entry, list_entry);
+
+	args.parent_list_entry = parent_list_entry;
+	args.list_entry = list_entry;
+	return nb_node->cbs.get_next(&args);
+}
+
+int nb_callback_get_keys(const struct nb_node *nb_node, const void *list_entry,
+			 struct yang_list_keys *keys)
+{
+	struct nb_cb_get_keys_args args = {};
+
+	DEBUGD(&nb_dbg_cbs_state,
+	       "northbound callback (get_keys): node [%s] list_entry [%p]",
+	       nb_node->xpath, list_entry);
+
+	args.list_entry = list_entry;
+	args.keys = keys;
+	return nb_node->cbs.get_keys(&args);
+}
+
+const void *nb_callback_lookup_entry(const struct nb_node *nb_node,
+				     const void *parent_list_entry,
+				     const struct yang_list_keys *keys)
+{
+	struct nb_cb_lookup_entry_args args = {};
+
+	DEBUGD(&nb_dbg_cbs_state,
+	       "northbound callback (lookup_entry): node [%s] parent_list_entry [%p]",
+	       nb_node->xpath, parent_list_entry);
+
+	args.parent_list_entry = parent_list_entry;
+	args.keys = keys;
+	return nb_node->cbs.lookup_entry(&args);
+}
+
+int nb_callback_rpc(const struct nb_node *nb_node, const char *xpath,
+		    const struct list *input, struct list *output, char *errmsg,
+		    size_t errmsg_len)
+{
+	struct nb_cb_rpc_args args = {};
+
+	DEBUGD(&nb_dbg_cbs_rpc, "northbound RPC: %s", xpath);
+
+	args.xpath = xpath;
+	args.input = input;
+	args.output = output;
+	args.errmsg = errmsg;
+	args.errmsg_len = errmsg_len;
+	return nb_node->cbs.rpc(&args);
+}
+
+/*
+ * Call the northbound configuration callback associated to a given
+ * configuration change.
+ */
+static int nb_callback_configuration(struct nb_context *context,
+				     const enum nb_event event,
+				     struct nb_config_change *change,
+				     char *errmsg, size_t errmsg_len)
+{
+	enum nb_operation operation = change->cb.operation;
+	char xpath[XPATH_MAXLEN];
+	const struct nb_node *nb_node = change->cb.nb_node;
+	const struct lyd_node *dnode = change->cb.dnode;
+	union nb_resource *resource;
+	int ret = NB_ERR;
+
+	if (event == NB_EV_VALIDATE)
+		resource = NULL;
+	else
+		resource = &change->resource;
+
+	switch (operation) {
+	case NB_OP_CREATE:
+		ret = nb_callback_create(context, nb_node, event, dnode,
+					 resource, errmsg, errmsg_len);
+		break;
+	case NB_OP_MODIFY:
+		ret = nb_callback_modify(context, nb_node, event, dnode,
+					 resource, errmsg, errmsg_len);
+		break;
+	case NB_OP_DESTROY:
+		ret = nb_callback_destroy(context, nb_node, event, dnode,
+					  errmsg, errmsg_len);
+		break;
+	case NB_OP_MOVE:
+		ret = nb_callback_move(context, nb_node, event, dnode, errmsg,
+				       errmsg_len);
+		break;
+	default:
+		yang_dnode_get_path(dnode, xpath, sizeof(xpath));
+		flog_err(EC_LIB_DEVELOPMENT,
+			 "%s: unknown operation (%u) [xpath %s]", __func__,
+			 operation, xpath);
+		exit(1);
+	}
+
+	if (ret != NB_OK) {
+		yang_dnode_get_path(dnode, xpath, sizeof(xpath));
+
+		switch (event) {
+		case NB_EV_VALIDATE:
+			flog_warn(EC_LIB_NB_CB_CONFIG_VALIDATE,
+				  "error processing configuration change: error [%s] event [%s] operation [%s] xpath [%s]%s%s",
+				  nb_err_name(ret), nb_event_name(event),
+				  nb_operation_name(operation), xpath,
+				  errmsg[0] ? " message: " : "", errmsg);
+			break;
+		case NB_EV_PREPARE:
+			flog_warn(EC_LIB_NB_CB_CONFIG_PREPARE,
+				  "error processing configuration change: error [%s] event [%s] operation [%s] xpath [%s]%s%s",
+				  nb_err_name(ret), nb_event_name(event),
+				  nb_operation_name(operation), xpath,
+				  errmsg[0] ? " message: " : "", errmsg);
+			break;
+		case NB_EV_ABORT:
+			flog_warn(EC_LIB_NB_CB_CONFIG_ABORT,
+				  "error processing configuration change: error [%s] event [%s] operation [%s] xpath [%s]%s%s",
+				  nb_err_name(ret), nb_event_name(event),
+				  nb_operation_name(operation), xpath,
+				  errmsg[0] ? " message: " : "", errmsg);
+			break;
+		case NB_EV_APPLY:
+			flog_err(EC_LIB_NB_CB_CONFIG_APPLY,
+				 "error processing configuration change: error [%s] event [%s] operation [%s] xpath [%s]%s%s",
+				 nb_err_name(ret), nb_event_name(event),
+				 nb_operation_name(operation), xpath,
+				 errmsg[0] ? " message: " : "", errmsg);
+			break;
+		default:
+			flog_err(EC_LIB_DEVELOPMENT,
+				 "%s: unknown event (%u) [xpath %s]", __func__,
+				 event, xpath);
+			exit(1);
+		}
+	}
+
+	return ret;
+}
+
+static struct nb_transaction *
+nb_transaction_new(struct nb_context *context, struct nb_config *config,
+		   struct nb_config_cbs *changes, const char *comment,
+		   char *errmsg, size_t errmsg_len)
+{
+	struct nb_transaction *transaction;
+
+	if (nb_running_lock_check(context->client, context->user)) {
+		strlcpy(errmsg,
+			"running configuration is locked by another client",
+			errmsg_len);
+		return NULL;
+	}
+
+	if (transaction_in_progress) {
+		strlcpy(errmsg,
+			"there's already another transaction in progress",
+			errmsg_len);
+		return NULL;
+	}
+	transaction_in_progress = true;
+
+	transaction = XCALLOC(MTYPE_TMP, sizeof(*transaction));
+	transaction->context = context;
+	if (comment)
+		strlcpy(transaction->comment, comment,
+			sizeof(transaction->comment));
+	transaction->config = config;
+	transaction->changes = *changes;
+
+	return transaction;
+}
+
+static void nb_transaction_free(struct nb_transaction *transaction)
+{
+	nb_config_diff_del_changes(&transaction->changes);
+	XFREE(MTYPE_TMP, transaction);
+	transaction_in_progress = false;
+}
+
+/* Process all configuration changes associated to a transaction. */
+static int nb_transaction_process(enum nb_event event,
+				  struct nb_transaction *transaction,
+				  char *errmsg, size_t errmsg_len)
+{
+	struct nb_config_cb *cb;
+
+	RB_FOREACH (cb, nb_config_cbs, &transaction->changes) {
+		struct nb_config_change *change = (struct nb_config_change *)cb;
+		int ret;
+
+		/*
+		 * Only try to release resources that were allocated
+		 * successfully.
+		 */
+		if (event == NB_EV_ABORT && !change->prepare_ok)
+			break;
+
+		/* Call the appropriate callback. */
+		ret = nb_callback_configuration(transaction->context, event,
+						change, errmsg, errmsg_len);
+		switch (event) {
+		case NB_EV_PREPARE:
+			if (ret != NB_OK)
+				return ret;
+			change->prepare_ok = true;
+			break;
+		case NB_EV_ABORT:
+		case NB_EV_APPLY:
+			/*
+			 * At this point it's not possible to reject the
+			 * transaction anymore, so any failure here can lead to
+			 * inconsistencies and should be treated as a bug.
+			 * Operations prone to errors, like validations and
+			 * resource allocations, should be performed during the
+			 * 'prepare' phase.
+			 */
+			break;
+		default:
+			break;
+		}
+	}
+
+	return NB_OK;
+}
+
+static struct nb_config_cb *
+nb_apply_finish_cb_new(struct nb_config_cbs *cbs, const struct nb_node *nb_node,
+		       const struct lyd_node *dnode)
+{
+	struct nb_config_cb *cb;
+
+	cb = XCALLOC(MTYPE_TMP, sizeof(*cb));
+	cb->nb_node = nb_node;
+	cb->dnode = dnode;
+	RB_INSERT(nb_config_cbs, cbs, cb);
+
+	return cb;
+}
+
+static struct nb_config_cb *
+nb_apply_finish_cb_find(struct nb_config_cbs *cbs,
+			const struct nb_node *nb_node,
+			const struct lyd_node *dnode)
+{
+	struct nb_config_cb s;
+
+	s.seq = 0;
+	s.nb_node = nb_node;
+	s.dnode = dnode;
+	return RB_FIND(nb_config_cbs, cbs, &s);
+}
+
+/* Call the 'apply_finish' callbacks. */
+static void nb_transaction_apply_finish(struct nb_transaction *transaction,
+					char *errmsg, size_t errmsg_len)
+{
+	struct nb_config_cbs cbs;
+	struct nb_config_cb *cb;
+
+	/* Initialize tree of 'apply_finish' callbacks. */
+	RB_INIT(nb_config_cbs, &cbs);
+
+	/* Identify the 'apply_finish' callbacks that need to be called. */
+	RB_FOREACH (cb, nb_config_cbs, &transaction->changes) {
+		struct nb_config_change *change = (struct nb_config_change *)cb;
+		const struct lyd_node *dnode = change->cb.dnode;
+
+		/*
+		 * Iterate up to the root of the data tree. When a node is being
+		 * deleted, skip its 'apply_finish' callback if one is defined
+		 * (the 'apply_finish' callbacks from the node ancestors should
+		 * be called though).
+		 */
+		if (change->cb.operation == NB_OP_DESTROY) {
+			char xpath[XPATH_MAXLEN];
+
+			dnode = lyd_parent(dnode);
+			if (!dnode)
+				break;
+
+			/*
+			 * The dnode from 'delete' callbacks point to elements
+			 * from the running configuration. Use yang_dnode_get()
+			 * to get the corresponding dnode from the candidate
+			 * configuration that is being committed.
+			 */
+			yang_dnode_get_path(dnode, xpath, sizeof(xpath));
+			dnode = yang_dnode_get(transaction->config->dnode,
+					       xpath);
+		}
+		while (dnode) {
+			struct nb_node *nb_node;
+
+			nb_node = dnode->schema->priv;
+			if (!nb_node || !nb_node->cbs.apply_finish)
+				goto next;
+
+			/*
+			 * Don't call the callback more than once for the same
+			 * data node.
+			 */
+			if (nb_apply_finish_cb_find(&cbs, nb_node, dnode))
+				goto next;
+
+			nb_apply_finish_cb_new(&cbs, nb_node, dnode);
+
+		next:
+			dnode = lyd_parent(dnode);
+		}
+	}
+
+	/* Call the 'apply_finish' callbacks, sorted by their priorities. */
+	RB_FOREACH (cb, nb_config_cbs, &cbs)
+		nb_callback_apply_finish(transaction->context, cb->nb_node,
+					 cb->dnode, errmsg, errmsg_len);
+
+	/* Release memory. */
+	while (!RB_EMPTY(nb_config_cbs, &cbs)) {
+		cb = RB_ROOT(nb_config_cbs, &cbs);
+		RB_REMOVE(nb_config_cbs, &cbs, cb);
+		XFREE(MTYPE_TMP, cb);
+	}
+}
+
+static int nb_oper_data_iter_children(const struct lysc_node *snode,
+				      const char *xpath, const void *list_entry,
+				      const struct yang_list_keys *list_keys,
+				      struct yang_translator *translator,
+				      bool first, uint32_t flags,
+				      nb_oper_data_cb cb, void *arg)
+{
+	const struct lysc_node *child;
+
+	LY_LIST_FOR (lysc_node_child(snode), child) {
+		int ret;
+
+		ret = nb_oper_data_iter_node(child, xpath, list_entry,
+					     list_keys, translator, false,
+					     flags, cb, arg);
+		if (ret != NB_OK)
+			return ret;
+	}
+
+	return NB_OK;
+}
+
+static int nb_oper_data_iter_leaf(const struct nb_node *nb_node,
+				  const char *xpath, const void *list_entry,
+				  const struct yang_list_keys *list_keys,
+				  struct yang_translator *translator,
+				  uint32_t flags, nb_oper_data_cb cb, void *arg)
+{
+	struct yang_data *data;
+
+	if (CHECK_FLAG(nb_node->snode->flags, LYS_CONFIG_W))
+		return NB_OK;
+
+	/* Ignore list keys. */
+	if (lysc_is_key(nb_node->snode))
+		return NB_OK;
+
+	data = nb_callback_get_elem(nb_node, xpath, list_entry);
+	if (data == NULL)
+		/* Leaf of type "empty" is not present. */
+		return NB_OK;
+
+	return (*cb)(nb_node->snode, translator, data, arg);
+}
+
+static int nb_oper_data_iter_container(const struct nb_node *nb_node,
+				       const char *xpath,
+				       const void *list_entry,
+				       const struct yang_list_keys *list_keys,
+				       struct yang_translator *translator,
+				       uint32_t flags, nb_oper_data_cb cb,
+				       void *arg)
+{
+	const struct lysc_node *snode = nb_node->snode;
+
+	if (CHECK_FLAG(nb_node->flags, F_NB_NODE_CONFIG_ONLY))
+		return NB_OK;
+
+	/* Read-only presence containers. */
+	if (nb_node->cbs.get_elem) {
+		struct yang_data *data;
+		int ret;
+
+		data = nb_callback_get_elem(nb_node, xpath, list_entry);
+		if (data == NULL)
+			/* Presence container is not present. */
+			return NB_OK;
+
+		ret = (*cb)(snode, translator, data, arg);
+		if (ret != NB_OK)
+			return ret;
+	}
+
+	/* Read-write presence containers. */
+	if (CHECK_FLAG(snode->flags, LYS_CONFIG_W)) {
+		struct lysc_node_container *scontainer;
+
+		scontainer = (struct lysc_node_container *)snode;
+		if (CHECK_FLAG(scontainer->flags, LYS_PRESENCE)
+		    && !yang_dnode_get(running_config->dnode, xpath))
+			return NB_OK;
+	}
+
+	/* Iterate over the child nodes. */
+	return nb_oper_data_iter_children(snode, xpath, list_entry, list_keys,
+					  translator, false, flags, cb, arg);
+}
+
+static int
+nb_oper_data_iter_leaflist(const struct nb_node *nb_node, const char *xpath,
+			   const void *parent_list_entry,
+			   const struct yang_list_keys *parent_list_keys,
+			   struct yang_translator *translator, uint32_t flags,
+			   nb_oper_data_cb cb, void *arg)
+{
+	const void *list_entry = NULL;
+
+	if (CHECK_FLAG(nb_node->snode->flags, LYS_CONFIG_W))
+		return NB_OK;
+
+	do {
+		struct yang_data *data;
+		int ret;
+
+		list_entry = nb_callback_get_next(nb_node, parent_list_entry,
+						  list_entry);
+		if (!list_entry)
+			/* End of the list. */
+			break;
+
+		data = nb_callback_get_elem(nb_node, xpath, list_entry);
+		if (data == NULL)
+			continue;
+
+		ret = (*cb)(nb_node->snode, translator, data, arg);
+		if (ret != NB_OK)
+			return ret;
+	} while (list_entry);
+
+	return NB_OK;
+}
+
+static int nb_oper_data_iter_list(const struct nb_node *nb_node,
+				  const char *xpath_list,
+				  const void *parent_list_entry,
+				  const struct yang_list_keys *parent_list_keys,
+				  struct yang_translator *translator,
+				  uint32_t flags, nb_oper_data_cb cb, void *arg)
+{
+	const struct lysc_node *snode = nb_node->snode;
+	const void *list_entry = NULL;
+	uint32_t position = 1;
+
+	if (CHECK_FLAG(nb_node->flags, F_NB_NODE_CONFIG_ONLY))
+		return NB_OK;
+
+	/* Iterate over all list entries. */
+	do {
+		const struct lysc_node_leaf *skey;
+		struct yang_list_keys list_keys;
+		char xpath[XPATH_MAXLEN * 2];
+		int ret;
+
+		/* Obtain list entry. */
+		list_entry = nb_callback_get_next(nb_node, parent_list_entry,
+						  list_entry);
+		if (!list_entry)
+			/* End of the list. */
+			break;
+
+		if (!CHECK_FLAG(nb_node->flags, F_NB_NODE_KEYLESS_LIST)) {
+			/* Obtain the list entry keys. */
+			if (nb_callback_get_keys(nb_node, list_entry,
+						 &list_keys)
+			    != NB_OK) {
+				flog_warn(EC_LIB_NB_CB_STATE,
+					  "%s: failed to get list keys",
+					  __func__);
+				return NB_ERR;
+			}
+
+			/* Build XPath of the list entry. */
+			strlcpy(xpath, xpath_list, sizeof(xpath));
+			unsigned int i = 0;
+			LY_FOR_KEYS (snode, skey) {
+				assert(i < list_keys.num);
+				snprintf(xpath + strlen(xpath),
+					 sizeof(xpath) - strlen(xpath),
+					 "[%s='%s']", skey->name,
+					 list_keys.key[i]);
+				i++;
+			}
+			assert(i == list_keys.num);
+		} else {
+			/*
+			 * Keyless list - build XPath using a positional index.
+			 */
+			snprintf(xpath, sizeof(xpath), "%s[%u]", xpath_list,
+				 position);
+			position++;
+		}
+
+		/* Iterate over the child nodes. */
+		ret = nb_oper_data_iter_children(
+			nb_node->snode, xpath, list_entry, &list_keys,
+			translator, false, flags, cb, arg);
+		if (ret != NB_OK)
+			return ret;
+	} while (list_entry);
+
+	return NB_OK;
+}
+
+static int nb_oper_data_iter_node(const struct lysc_node *snode,
+				  const char *xpath_parent,
+				  const void *list_entry,
+				  const struct yang_list_keys *list_keys,
+				  struct yang_translator *translator,
+				  bool first, uint32_t flags,
+				  nb_oper_data_cb cb, void *arg)
+{
+	struct nb_node *nb_node;
+	char xpath[XPATH_MAXLEN];
+	int ret = NB_OK;
+
+	if (!first && CHECK_FLAG(flags, NB_OPER_DATA_ITER_NORECURSE)
+	    && CHECK_FLAG(snode->nodetype, LYS_CONTAINER | LYS_LIST))
+		return NB_OK;
+
+	/* Update XPath. */
+	strlcpy(xpath, xpath_parent, sizeof(xpath));
+	if (!first && snode->nodetype != LYS_USES) {
+		struct lysc_node *parent;
+
+		/* Get the real parent. */
+		parent = snode->parent;
+
+		/*
+		 * When necessary, include the namespace of the augmenting
+		 * module.
+		 */
+		if (parent && parent->module != snode->module)
+			snprintf(xpath + strlen(xpath),
+				 sizeof(xpath) - strlen(xpath), "/%s:%s",
+				 snode->module->name, snode->name);
+		else
+			snprintf(xpath + strlen(xpath),
+				 sizeof(xpath) - strlen(xpath), "/%s",
+				 snode->name);
+	}
+
+	nb_node = snode->priv;
+	switch (snode->nodetype) {
+	case LYS_CONTAINER:
+		ret = nb_oper_data_iter_container(nb_node, xpath, list_entry,
+						  list_keys, translator, flags,
+						  cb, arg);
+		break;
+	case LYS_LEAF:
+		ret = nb_oper_data_iter_leaf(nb_node, xpath, list_entry,
+					     list_keys, translator, flags, cb,
+					     arg);
+		break;
+	case LYS_LEAFLIST:
+		ret = nb_oper_data_iter_leaflist(nb_node, xpath, list_entry,
+						 list_keys, translator, flags,
+						 cb, arg);
+		break;
+	case LYS_LIST:
+		ret = nb_oper_data_iter_list(nb_node, xpath, list_entry,
+					     list_keys, translator, flags, cb,
+					     arg);
+		break;
+	case LYS_USES:
+		ret = nb_oper_data_iter_children(snode, xpath, list_entry,
+						 list_keys, translator, false,
+						 flags, cb, arg);
+		break;
+	default:
+		break;
+	}
+
+	return ret;
+}
+
+int nb_oper_data_iterate(const char *xpath, struct yang_translator *translator,
+			 uint32_t flags, nb_oper_data_cb cb, void *arg)
+{
+	struct nb_node *nb_node;
+	const void *list_entry = NULL;
+	struct yang_list_keys list_keys;
+	struct list *list_dnodes;
+	struct lyd_node *dnode, *dn;
+	struct listnode *ln;
+	int ret;
+
+	nb_node = nb_node_find(xpath);
+	if (!nb_node) {
+		flog_warn(EC_LIB_YANG_UNKNOWN_DATA_PATH,
+			  "%s: unknown data path: %s", __func__, xpath);
+		return NB_ERR;
+	}
+
+	/* For now this function works only with containers and lists. */
+	if (!CHECK_FLAG(nb_node->snode->nodetype, LYS_CONTAINER | LYS_LIST)) {
+		flog_warn(
+			EC_LIB_NB_OPERATIONAL_DATA,
+			"%s: can't iterate over YANG leaf or leaf-list [xpath %s]",
+			__func__, xpath);
+		return NB_ERR;
+	}
+
+	/*
+	 * Create a data tree from the XPath so that we can parse the keys of
+	 * all YANG lists (if any).
+	 */
+
+	LY_ERR err = lyd_new_path(NULL, ly_native_ctx, xpath, NULL,
+				  LYD_NEW_PATH_UPDATE, &dnode);
+	if (err || !dnode) {
+		const char *errmsg =
+			err ? ly_errmsg(ly_native_ctx) : "node not found";
+		flog_warn(EC_LIB_LIBYANG, "%s: lyd_new_path() failed %s",
+			  __func__, errmsg);
+		return NB_ERR;
+	}
+
+	/*
+	 * Create a linked list to sort the data nodes starting from the root.
+	 */
+	list_dnodes = list_new();
+	for (dn = dnode; dn; dn = lyd_parent(dn)) {
+		if (dn->schema->nodetype != LYS_LIST || !lyd_child(dn))
+			continue;
+		listnode_add_head(list_dnodes, dn);
+	}
+	/*
+	 * Use the northbound callbacks to find list entry pointer corresponding
+	 * to the given XPath.
+	 */
+	for (ALL_LIST_ELEMENTS_RO(list_dnodes, ln, dn)) {
+		struct lyd_node *child;
+		struct nb_node *nn;
+		unsigned int n = 0;
+
+		/* Obtain the list entry keys. */
+		memset(&list_keys, 0, sizeof(list_keys));
+		LY_LIST_FOR (lyd_child(dn), child) {
+			if (!lysc_is_key(child->schema))
+				break;
+			strlcpy(list_keys.key[n],
+				yang_dnode_get_string(child, NULL),
+				sizeof(list_keys.key[n]));
+			n++;
+		}
+		list_keys.num = n;
+		if (list_keys.num != yang_snode_num_keys(dn->schema)) {
+			list_delete(&list_dnodes);
+			yang_dnode_free(dnode);
+			return NB_ERR_NOT_FOUND;
+		}
+
+		/* Find the list entry pointer. */
+		nn = dn->schema->priv;
+		if (!nn->cbs.lookup_entry) {
+			flog_warn(
+				EC_LIB_NB_OPERATIONAL_DATA,
+				"%s: data path doesn't support iteration over operational data: %s",
+				__func__, xpath);
+			list_delete(&list_dnodes);
+			yang_dnode_free(dnode);
+			return NB_ERR;
+		}
+
+		list_entry =
+			nb_callback_lookup_entry(nn, list_entry, &list_keys);
+		if (list_entry == NULL) {
+			list_delete(&list_dnodes);
+			yang_dnode_free(dnode);
+			return NB_ERR_NOT_FOUND;
+		}
+	}
+
+	/* If a list entry was given, iterate over that list entry only. */
+	if (dnode->schema->nodetype == LYS_LIST && lyd_child(dnode))
+		ret = nb_oper_data_iter_children(
+			nb_node->snode, xpath, list_entry, &list_keys,
+			translator, true, flags, cb, arg);
+	else
+		ret = nb_oper_data_iter_node(nb_node->snode, xpath, list_entry,
+					     &list_keys, translator, true,
+					     flags, cb, arg);
+
+	list_delete(&list_dnodes);
+	yang_dnode_free(dnode);
+
+	return ret;
+}
+
+bool nb_operation_is_valid(enum nb_operation operation,
+			   const struct lysc_node *snode)
+{
+	struct nb_node *nb_node = snode->priv;
+	struct lysc_node_container *scontainer;
+	struct lysc_node_leaf *sleaf;
+
+	switch (operation) {
+	case NB_OP_CREATE:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_LEAF:
+			sleaf = (struct lysc_node_leaf *)snode;
+			if (sleaf->type->basetype != LY_TYPE_EMPTY)
+				return false;
+			break;
+		case LYS_CONTAINER:
+			scontainer = (struct lysc_node_container *)snode;
+			if (!CHECK_FLAG(scontainer->flags, LYS_PRESENCE))
+				return false;
+			break;
+		case LYS_LIST:
+		case LYS_LEAFLIST:
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_MODIFY:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_LEAF:
+			sleaf = (struct lysc_node_leaf *)snode;
+			if (sleaf->type->basetype == LY_TYPE_EMPTY)
+				return false;
+
+			/* List keys can't be modified. */
+			if (lysc_is_key(sleaf))
+				return false;
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_DESTROY:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_LEAF:
+			sleaf = (struct lysc_node_leaf *)snode;
+
+			/* List keys can't be deleted. */
+			if (lysc_is_key(sleaf))
+				return false;
+
+			/*
+			 * Only optional leafs can be deleted, or leafs whose
+			 * parent is a case statement.
+			 */
+			if (snode->parent->nodetype == LYS_CASE)
+				return true;
+			if (sleaf->when)
+				return true;
+			if (CHECK_FLAG(sleaf->flags, LYS_MAND_TRUE)
+			    || sleaf->dflt)
+				return false;
+			break;
+		case LYS_CONTAINER:
+			scontainer = (struct lysc_node_container *)snode;
+			if (!CHECK_FLAG(scontainer->flags, LYS_PRESENCE))
+				return false;
+			break;
+		case LYS_LIST:
+		case LYS_LEAFLIST:
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_MOVE:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_LIST:
+		case LYS_LEAFLIST:
+			if (!CHECK_FLAG(snode->flags, LYS_ORDBY_USER))
+				return false;
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_PRE_VALIDATE:
+	case NB_OP_APPLY_FINISH:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+			return false;
+		return true;
+	case NB_OP_GET_ELEM:
+		if (!CHECK_FLAG(snode->flags, LYS_CONFIG_R))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_LEAF:
+		case LYS_LEAFLIST:
+			break;
+		case LYS_CONTAINER:
+			scontainer = (struct lysc_node_container *)snode;
+			if (!CHECK_FLAG(scontainer->flags, LYS_PRESENCE))
+				return false;
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_GET_NEXT:
+		switch (snode->nodetype) {
+		case LYS_LIST:
+			if (CHECK_FLAG(nb_node->flags, F_NB_NODE_CONFIG_ONLY))
+				return false;
+			break;
+		case LYS_LEAFLIST:
+			if (CHECK_FLAG(snode->flags, LYS_CONFIG_W))
+				return false;
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_GET_KEYS:
+	case NB_OP_LOOKUP_ENTRY:
+		switch (snode->nodetype) {
+		case LYS_LIST:
+			if (CHECK_FLAG(nb_node->flags, F_NB_NODE_CONFIG_ONLY))
+				return false;
+			if (CHECK_FLAG(nb_node->flags, F_NB_NODE_KEYLESS_LIST))
+				return false;
+			break;
+		default:
+			return false;
+		}
+		return true;
+	case NB_OP_RPC:
+		if (CHECK_FLAG(snode->flags, LYS_CONFIG_W | LYS_CONFIG_R))
+			return false;
+
+		switch (snode->nodetype) {
+		case LYS_RPC:
+		case LYS_ACTION:
+			break;
+		default:
+			return false;
+		}
+		return true;
+	default:
+		return false;
+	}
+}
+
+DEFINE_HOOK(nb_notification_send, (const char *xpath, struct list *arguments),
+	    (xpath, arguments));
+
+int nb_notification_send(const char *xpath, struct list *arguments)
+{
+	int ret;
+
+	DEBUGD(&nb_dbg_notif, "northbound notification: %s", xpath);
+
+	ret = hook_call(nb_notification_send, xpath, arguments);
+	if (arguments)
+		list_delete(&arguments);
+
+	return ret;
+}
+
+/* Running configuration user pointers management. */
+struct nb_config_entry {
+	char xpath[XPATH_MAXLEN];
+	void *entry;
+};
+
+static bool running_config_entry_cmp(const void *value1, const void *value2)
+{
+	const struct nb_config_entry *c1 = value1;
+	const struct nb_config_entry *c2 = value2;
+
+	return strmatch(c1->xpath, c2->xpath);
+}
+
+static unsigned int running_config_entry_key_make(const void *value)
+{
+	return string_hash_make(value);
+}
+
+static void *running_config_entry_alloc(void *p)
+{
+	struct nb_config_entry *new, *key = p;
+
+	new = XCALLOC(MTYPE_NB_CONFIG_ENTRY, sizeof(*new));
+	strlcpy(new->xpath, key->xpath, sizeof(new->xpath));
+
+	return new;
+}
+
+static void running_config_entry_free(void *arg)
+{
+	XFREE(MTYPE_NB_CONFIG_ENTRY, arg);
+}
+
+void nb_running_set_entry(const struct lyd_node *dnode, void *entry)
+{
+	struct nb_config_entry *config, s;
+
+	yang_dnode_get_path(dnode, s.xpath, sizeof(s.xpath));
+	config = hash_get(running_config_entries, &s,
+			  running_config_entry_alloc);
+	config->entry = entry;
+}
+
+void nb_running_move_tree(const char *xpath_from, const char *xpath_to)
+{
+	struct nb_config_entry *entry;
+	struct list *entries = hash_to_list(running_config_entries);
+	struct listnode *ln;
+
+	for (ALL_LIST_ELEMENTS_RO(entries, ln, entry)) {
+		if (!frrstr_startswith(entry->xpath, xpath_from))
+			continue;
+
+		hash_release(running_config_entries, entry);
+
+		char *newpath =
+			frrstr_replace(entry->xpath, xpath_from, xpath_to);
+		strlcpy(entry->xpath, newpath, sizeof(entry->xpath));
+		XFREE(MTYPE_TMP, newpath);
+
+		(void)hash_get(running_config_entries, entry,
+			       hash_alloc_intern);
+	}
+
+	list_delete(&entries);
+}
+
+static void *nb_running_unset_entry_helper(const struct lyd_node *dnode)
+{
+	struct nb_config_entry *config, s;
+	struct lyd_node *child;
+	void *entry = NULL;
+
+	yang_dnode_get_path(dnode, s.xpath, sizeof(s.xpath));
+	config = hash_release(running_config_entries, &s);
+	if (config) {
+		entry = config->entry;
+		running_config_entry_free(config);
+	}
+
+	/* Unset user pointers from the child nodes. */
+	if (CHECK_FLAG(dnode->schema->nodetype, LYS_LIST | LYS_CONTAINER)) {
+		LY_LIST_FOR (lyd_child(dnode), child) {
+			(void)nb_running_unset_entry_helper(child);
+		}
+	}
+
+	return entry;
+}
+
+void *nb_running_unset_entry(const struct lyd_node *dnode)
+{
+	void *entry;
+
+	entry = nb_running_unset_entry_helper(dnode);
+	assert(entry);
+
+	return entry;
+}
+
+static void *nb_running_get_entry_worker(const struct lyd_node *dnode,
+					 const char *xpath,
+					 bool abort_if_not_found,
+					 bool rec_search)
+{
+	const struct lyd_node *orig_dnode = dnode;
+	char xpath_buf[XPATH_MAXLEN];
+	bool rec_flag = true;
+
+	assert(dnode || xpath);
+
+	if (!dnode)
+		dnode = yang_dnode_get(running_config->dnode, xpath);
+
+	while (rec_flag && dnode) {
+		struct nb_config_entry *config, s;
+
+		yang_dnode_get_path(dnode, s.xpath, sizeof(s.xpath));
+		config = hash_lookup(running_config_entries, &s);
+		if (config)
+			return config->entry;
+
+		rec_flag = rec_search;
+
+		dnode = lyd_parent(dnode);
+	}
+
+	if (!abort_if_not_found)
+		return NULL;
+
+	yang_dnode_get_path(orig_dnode, xpath_buf, sizeof(xpath_buf));
+	flog_err(EC_LIB_YANG_DNODE_NOT_FOUND,
+		 "%s: failed to find entry [xpath %s]", __func__, xpath_buf);
+	zlog_backtrace(LOG_ERR);
+	abort();
+}
+
+void *nb_running_get_entry(const struct lyd_node *dnode, const char *xpath,
+			   bool abort_if_not_found)
+{
+	return nb_running_get_entry_worker(dnode, xpath, abort_if_not_found,
+					   true);
+}
+
+void *nb_running_get_entry_non_rec(const struct lyd_node *dnode,
+				   const char *xpath, bool abort_if_not_found)
+{
+	return nb_running_get_entry_worker(dnode, xpath, abort_if_not_found,
+					   false);
+}
+
+/* Logging functions. */
+const char *nb_event_name(enum nb_event event)
+{
+	switch (event) {
+	case NB_EV_VALIDATE:
+		return "validate";
+	case NB_EV_PREPARE:
+		return "prepare";
+	case NB_EV_ABORT:
+		return "abort";
+	case NB_EV_APPLY:
+		return "apply";
+	default:
+		return "unknown";
+	}
+}
+
+const char *nb_operation_name(enum nb_operation operation)
+{
+	switch (operation) {
+	case NB_OP_CREATE:
+		return "create";
+	case NB_OP_MODIFY:
+		return "modify";
+	case NB_OP_DESTROY:
+		return "destroy";
+	case NB_OP_MOVE:
+		return "move";
+	case NB_OP_PRE_VALIDATE:
+		return "pre_validate";
+	case NB_OP_APPLY_FINISH:
+		return "apply_finish";
+	case NB_OP_GET_ELEM:
+		return "get_elem";
+	case NB_OP_GET_NEXT:
+		return "get_next";
+	case NB_OP_GET_KEYS:
+		return "get_keys";
+	case NB_OP_LOOKUP_ENTRY:
+		return "lookup_entry";
+	case NB_OP_RPC:
+		return "rpc";
+	default:
+		return "unknown";
+	}
+}
+
+const char *nb_err_name(enum nb_error error)
+{
+	switch (error) {
+	case NB_OK:
+		return "ok";
+	case NB_ERR:
+		return "generic error";
+	case NB_ERR_NO_CHANGES:
+		return "no changes";
+	case NB_ERR_NOT_FOUND:
+		return "element not found";
+	case NB_ERR_LOCKED:
+		return "resource is locked";
+	case NB_ERR_VALIDATION:
+		return "validation";
+	case NB_ERR_RESOURCE:
+		return "failed to allocate resource";
+	case NB_ERR_INCONSISTENCY:
+		return "internal inconsistency";
+	default:
+		return "unknown";
+	}
+}
+
+const char *nb_client_name(enum nb_client client)
+{
+	switch (client) {
+	case NB_CLIENT_CLI:
+		return "CLI";
+	case NB_CLIENT_CONFD:
+		return "ConfD";
+	case NB_CLIENT_SYSREPO:
+		return "Sysrepo";
+	case NB_CLIENT_GRPC:
+		return "gRPC";
+	default:
+		return "unknown";
+	}
+}
+
+static void nb_load_callbacks(const struct frr_yang_module_info *module)
+{
+	for (size_t i = 0; module->nodes[i].xpath; i++) {
+		struct nb_node *nb_node;
+		uint32_t priority;
+
+		if (i > YANG_MODULE_MAX_NODES) {
+			zlog_err(
+				"%s: %s.yang has more than %u nodes. Please increase YANG_MODULE_MAX_NODES to fix this problem.",
+				__func__, module->name, YANG_MODULE_MAX_NODES);
+			exit(1);
+		}
+
+		nb_node = nb_node_find(module->nodes[i].xpath);
+		if (!nb_node) {
+			flog_warn(EC_LIB_YANG_UNKNOWN_DATA_PATH,
+				  "%s: unknown data path: %s", __func__,
+				  module->nodes[i].xpath);
+			continue;
+		}
+
+		nb_node->cbs = module->nodes[i].cbs;
+		priority = module->nodes[i].priority;
+		if (priority != 0)
+			nb_node->priority = priority;
+	}
+}
+
+void nb_validate_callbacks(void)
+{
+	unsigned int errors = 0;
+
+	yang_snodes_iterate(NULL, nb_node_validate, 0, &errors);
+	if (errors > 0) {
+		flog_err(
+			EC_LIB_NB_CBS_VALIDATION,
+			"%s: failed to validate northbound callbacks: %u error(s)",
+			__func__, errors);
+		exit(1);
+	}
+}
+
+
+void nb_init(struct thread_master *tm,
+	     const struct frr_yang_module_info *const modules[],
+	     size_t nmodules, bool db_enabled)
+{
+	struct yang_module *loaded[nmodules], **loadedp = loaded;
+	bool explicit_compile;
+
+	/*
+	 * Currently using this explicit compile feature in libyang2 leads to
+	 * incorrect behavior in FRR. The functionality suppresses the compiling
+	 * of modules until they have all been loaded into the context. This
+	 * avoids multiple recompiles of the same modules as they are
+	 * imported/augmented etc.
+	 */
+	explicit_compile = false;
+
+	nb_db_enabled = db_enabled;
+
+	yang_init(true, explicit_compile);
+
+	/* Load YANG modules and their corresponding northbound callbacks. */
+	for (size_t i = 0; i < nmodules; i++) {
+		DEBUGD(&nb_dbg_events, "northbound: loading %s.yang",
+		       modules[i]->name);
+		*loadedp++ = yang_module_load(modules[i]->name);
+	}
+
+	if (explicit_compile)
+		yang_init_loading_complete();
+
+	/* Initialize the compiled nodes with northbound data */
+	for (size_t i = 0; i < nmodules; i++) {
+		yang_snodes_iterate(loaded[i]->info, nb_node_new_cb, 0, NULL);
+		nb_load_callbacks(modules[i]);
+	}
+
+	/* Validate northbound callbacks. */
+	nb_validate_callbacks();
+
+	/* Create an empty running configuration. */
+	running_config = nb_config_new(NULL);
+	running_config_entries = hash_create(running_config_entry_key_make,
+					     running_config_entry_cmp,
+					     "Running Configuration Entries");
+	pthread_mutex_init(&running_config_mgmt_lock.mtx, NULL);
+
+	/* Initialize the northbound CLI. */
+	nb_cli_init(tm);
+}
+
+void nb_terminate(void)
+{
+	/* Terminate the northbound CLI. */
+	nb_cli_terminate();
+
+	/* Delete all nb_node's from all YANG modules. */
+	nb_nodes_delete();
+
+	/* Delete the running configuration. */
+	hash_clean(running_config_entries, running_config_entry_free);
+	hash_free(running_config_entries);
+	nb_config_free(running_config);
+	pthread_mutex_destroy(&running_config_mgmt_lock.mtx);
+}