lib/pipeline/rte_swx_ctl.c

/* SPDX-License-Identifier: BSD-3-Clause
 * Copyright(c) 2020 Intel Corporation
 */
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <sys/queue.h>
#include <unistd.h>

#include <rte_common.h>
#include <rte_byteorder.h>

#include <rte_swx_table_selector.h>

#include "rte_swx_ctl.h"

#define CHECK(condition, err_code)                                             \
do {                                                                           \
	if (!(condition))                                                      \
		return -(err_code);                                            \
} while (0)

#define ntoh64(x) rte_be_to_cpu_64(x)
#define hton64(x) rte_cpu_to_be_64(x)

#if RTE_BYTE_ORDER == RTE_LITTLE_ENDIAN
#define field_ntoh(val, n_bits) (ntoh64((val) << (64 - n_bits)))
#define field_hton(val, n_bits) (hton64((val) << (64 - n_bits)))
#else
#define field_ntoh(val, n_bits) (val)
#define field_hton(val, n_bits) (val)
#endif

struct action {
	struct rte_swx_ctl_action_info info;
	struct rte_swx_ctl_action_arg_info *args;
	uint32_t data_size;
};

struct table {
	struct rte_swx_ctl_table_info info;
	struct rte_swx_ctl_table_match_field_info *mf;

	/* Match field with the smallest offset. */
	struct rte_swx_ctl_table_match_field_info *mf_first;

	/* Match field with the biggest offset. */
	struct rte_swx_ctl_table_match_field_info *mf_last;

	struct rte_swx_ctl_table_action_info *actions;
	struct rte_swx_table_ops ops;
	struct rte_swx_table_params params;

	/* Set of "stable" keys: these keys are currently part of the table;
	 * these keys will be preserved with no action data changes after the
	 * next commit.
	 */
	struct rte_swx_table_entry_list entries;

	/* Set of new keys: these keys are currently NOT part of the table;
	 * these keys will be added to the table on the next commit, if
	 * the commit operation is successful.
	 */
	struct rte_swx_table_entry_list pending_add;

	/* Set of keys to be modified: these keys are currently part of the
	 * table; these keys are still going to be part of the table after the
	 * next commit, but their action data will be modified if the commit
	 * operation is successful. The modify0 list contains the keys with the
	 * current action data, the modify1 list contains the keys with the
	 * modified action data.
	 */
	struct rte_swx_table_entry_list pending_modify0;
	struct rte_swx_table_entry_list pending_modify1;

	/* Set of keys to be deleted: these keys are currently part of the
	 * table; these keys are to be deleted from the table on the next
	 * commit, if the commit operation is successful.
	 */
	struct rte_swx_table_entry_list pending_delete;

	/* The pending default action: this is NOT the current default action;
	 * this will be the new default action after the next commit, if the
	 * next commit operation is successful.
	 */
	struct rte_swx_table_entry *pending_default;

	int is_stub;
	uint32_t n_add;
	uint32_t n_modify;
	uint32_t n_delete;
};

struct selector {
	/* Selector table info. */
	struct rte_swx_ctl_selector_info info;

	/* group_id field. */
	struct rte_swx_ctl_table_match_field_info group_id_field;

	/* selector fields. */
	struct rte_swx_ctl_table_match_field_info *selector_fields;

	/* member_id field. */
	struct rte_swx_ctl_table_match_field_info member_id_field;

	/* Current selector table. Array of info.n_groups_max elements.*/
	struct rte_swx_table_selector_group **groups;

	/* Pending selector table subject to the next commit. Array of info.n_groups_max elements.
	 */
	struct rte_swx_table_selector_group **pending_groups;

	/* Valid flag per group. Array of n_groups_max elements. */
	int *groups_added;

	/* Pending delete flag per group. Group deletion is subject to the next commit. Array of
	 * info.n_groups_max elements.
	 */
	int *groups_pending_delete;

	/* Params. */
	struct rte_swx_table_selector_params params;
};

struct rte_swx_ctl_pipeline {
	struct rte_swx_ctl_pipeline_info info;
	struct rte_swx_pipeline *p;
	struct action *actions;
	struct table *tables;
	struct selector *selectors;
	struct rte_swx_table_state *ts;
	struct rte_swx_table_state *ts_next;
	int numa_node;
};

static struct action *
action_find(struct rte_swx_ctl_pipeline *ctl, const char *action_name)
{
	uint32_t i;

	for (i = 0; i < ctl->info.n_actions; i++) {
		struct action *a = &ctl->actions[i];

		if (!strcmp(action_name, a->info.name))
			return a;
	}

	return NULL;
}

static void
action_free(struct rte_swx_ctl_pipeline *ctl)
{
	uint32_t i;

	if (!ctl->actions)
		return;

	for (i = 0; i < ctl->info.n_actions; i++) {
		struct action *action = &ctl->actions[i];

		free(action->args);
	}

	free(ctl->actions);
	ctl->actions = NULL;
}

static struct table *
table_find(struct rte_swx_ctl_pipeline *ctl, const char *table_name)
{
	uint32_t i;

	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *table = &ctl->tables[i];

		if (!strcmp(table_name, table->info.name))
			return table;
	}

	return NULL;
}

static int
table_params_get(struct rte_swx_ctl_pipeline *ctl, uint32_t table_id)
{
	struct table *table = &ctl->tables[table_id];
	struct rte_swx_ctl_table_match_field_info *first = NULL, *last = NULL;
	uint8_t *key_mask = NULL;
	enum rte_swx_table_match_type match_type = RTE_SWX_TABLE_MATCH_WILDCARD;
	uint32_t key_size = 0, key_offset = 0, action_data_size = 0, i;

	if (table->info.n_match_fields) {
		uint32_t n_match_fields_em = 0, i;

		/* Find first (smallest offset) and last (biggest offset) match fields. */
		first = &table->mf[0];
		last = &table->mf[0];

		for (i = 1; i < table->info.n_match_fields; i++) {
			struct rte_swx_ctl_table_match_field_info *f = &table->mf[i];

			if (f->offset < first->offset)
				first = f;

			if (f->offset > last->offset)
				last = f;
		}

		/* match_type. */
		for (i = 0; i < table->info.n_match_fields; i++) {
			struct rte_swx_ctl_table_match_field_info *f = &table->mf[i];

			if (f->match_type == RTE_SWX_TABLE_MATCH_EXACT)
				n_match_fields_em++;
		}

		if (n_match_fields_em == table->info.n_match_fields)
			match_type = RTE_SWX_TABLE_MATCH_EXACT;

		/* key_offset. */
		key_offset = first->offset / 8;

		/* key_size. */
		key_size = (last->offset + last->n_bits - first->offset) / 8;

		/* key_mask. */
		key_mask = calloc(1, key_size);
		CHECK(key_mask, ENOMEM);

		for (i = 0; i < table->info.n_match_fields; i++) {
			struct rte_swx_ctl_table_match_field_info *f = &table->mf[i];
			uint32_t start;
			size_t size;

			start = (f->offset - first->offset) / 8;
			size = f->n_bits / 8;

			memset(&key_mask[start], 0xFF, size);
		}
	}

	/* action_data_size. */
	for (i = 0; i < table->info.n_actions; i++) {
		uint32_t action_id = table->actions[i].action_id;
		struct action *a = &ctl->actions[action_id];

		if (a->data_size > action_data_size)
			action_data_size = a->data_size;
	}

	/* Fill in. */
	table->params.match_type = match_type;
	table->params.key_size = key_size;
	table->params.key_offset = key_offset;
	table->params.key_mask0 = key_mask;
	table->params.action_data_size = action_data_size;
	table->params.n_keys_max = table->info.size;

	table->mf_first = first;
	table->mf_last = last;

	return 0;
}

static void
table_entry_free(struct rte_swx_table_entry *entry)
{
	if (!entry)
		return;

	free(entry->key);
	free(entry->key_mask);
	free(entry->action_data);
	free(entry);
}

static struct rte_swx_table_entry *
table_entry_alloc(struct table *table)
{
	struct rte_swx_table_entry *entry;

	entry = calloc(1, sizeof(struct rte_swx_table_entry));
	if (!entry)
		goto error;

	/* key, key_mask. */
	if (!table->is_stub) {
		entry->key = calloc(1, table->params.key_size);
		if (!entry->key)
			goto error;

		if (table->params.match_type != RTE_SWX_TABLE_MATCH_EXACT) {
			entry->key_mask = calloc(1, table->params.key_size);
			if (!entry->key_mask)
				goto error;
		}
	}

	/* action_data. */
	if (table->params.action_data_size) {
		entry->action_data = calloc(1, table->params.action_data_size);
		if (!entry->action_data)
			goto error;
	}

	return entry;

error:
	table_entry_free(entry);
	return NULL;
}

static int
table_entry_key_check_em(struct table *table, struct rte_swx_table_entry *entry)
{
	uint8_t *key_mask0 = table->params.key_mask0;
	uint32_t key_size = table->params.key_size, i;

	if (!entry->key_mask)
		return 0;

	for (i = 0; i < key_size; i++) {
		uint8_t km0 = key_mask0[i];
		uint8_t km = entry->key_mask[i];

		if ((km & km0) != km0)
			return -EINVAL;
	}

	return 0;
}

static int
table_entry_check(struct rte_swx_ctl_pipeline *ctl,
		  uint32_t table_id,
		  struct rte_swx_table_entry *entry,
		  int key_check,
		  int data_check)
{
	struct table *table = &ctl->tables[table_id];
	int status;

	CHECK(entry, EINVAL);

	if (key_check && !table->is_stub) {
		/* key. */
		CHECK(entry->key, EINVAL);

		/* key_mask. */
		if (table->params.match_type == RTE_SWX_TABLE_MATCH_EXACT) {
			status = table_entry_key_check_em(table, entry);
			if (status)
				return status;
		}
	}

	if (data_check) {
		struct action *a;
		uint32_t i;

		/* action_id. */
		for (i = 0; i < table->info.n_actions; i++)
			if (entry->action_id == table->actions[i].action_id)
				break;

		CHECK(i < table->info.n_actions, EINVAL);

		/* action_data. */
		a = &ctl->actions[entry->action_id];
		CHECK(!(a->data_size && !entry->action_data), EINVAL);
	}

	return 0;
}

static struct rte_swx_table_entry *
table_entry_duplicate(struct rte_swx_ctl_pipeline *ctl,
		      uint32_t table_id,
		      struct rte_swx_table_entry *entry,
		      int key_duplicate,
		      int data_duplicate)
{
	struct table *table = &ctl->tables[table_id];
	struct rte_swx_table_entry *new_entry = NULL;

	if (!entry)
		goto error;

	new_entry = calloc(1, sizeof(struct rte_swx_table_entry));
	if (!new_entry)
		goto error;

	if (key_duplicate && !table->is_stub) {
		/* key. */
		if (!entry->key)
			goto error;

		new_entry->key = malloc(table->params.key_size);
		if (!new_entry->key)
			goto error;

		memcpy(new_entry->key, entry->key, table->params.key_size);

		/* key_signature. */
		new_entry->key_signature = entry->key_signature;

		/* key_mask. */
		if (entry->key_mask) {
			new_entry->key_mask = malloc(table->params.key_size);
			if (!new_entry->key_mask)
				goto error;

			memcpy(new_entry->key_mask,
			       entry->key_mask,
			       table->params.key_size);
		}

		/* key_priority. */
		new_entry->key_priority = entry->key_priority;
	}

	if (data_duplicate) {
		struct action *a;
		uint32_t i;

		/* action_id. */
		for (i = 0; i < table->info.n_actions; i++)
			if (entry->action_id == table->actions[i].action_id)
				break;

		if (i >= table->info.n_actions)
			goto error;

		new_entry->action_id = entry->action_id;

		/* action_data. */
		a = &ctl->actions[entry->action_id];
		if (a->data_size && !entry->action_data)
			goto error;

		/* The table layer provisions a constant action data size per
		 * entry, which should be the largest data size for all the
		 * actions enabled for the current table, and attempts to copy
		 * this many bytes each time a table entry is added, even if the
		 * specific action requires less data or even no data at all,
		 * hence we always have to allocate the max.
		 */
		new_entry->action_data = calloc(1, table->params.action_data_size);
		if (!new_entry->action_data)
			goto error;

		if (a->data_size)
			memcpy(new_entry->action_data,
			       entry->action_data,
			       a->data_size);
	}

	return new_entry;

error:
	table_entry_free(new_entry);
	return NULL;
}

static int
table_entry_keycmp(struct table *table,
		   struct rte_swx_table_entry *e0,
		   struct rte_swx_table_entry *e1)
{
	uint32_t key_size = table->params.key_size;
	uint32_t i;

	for (i = 0; i < key_size; i++) {
		uint8_t *key_mask0 = table->params.key_mask0;
		uint8_t km0, km[2], k[2];

		km0 = key_mask0 ? key_mask0[i] : 0xFF;

		km[0] = e0->key_mask ? e0->key_mask[i] : 0xFF;
		km[1] = e1->key_mask ? e1->key_mask[i] : 0xFF;

		k[0] = e0->key[i];
		k[1] = e1->key[i];

		/* Mask comparison. */
		if ((km[0] & km0) != (km[1] & km0))
			return 1; /* Not equal. */

		/* Value comparison. */
		if ((k[0] & km[0] & km0) != (k[1] & km[1] & km0))
			return 1; /* Not equal. */
	}

	return 0; /* Equal. */
}

static struct rte_swx_table_entry *
table_entries_find(struct table *table, struct rte_swx_table_entry *entry)
{
	struct rte_swx_table_entry *e;

	TAILQ_FOREACH(e, &table->entries, node)
		if (!table_entry_keycmp(table, entry, e))
			return e; /* Found. */

	return NULL; /* Not found. */
}

static void
table_entries_free(struct table *table)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(&table->entries);
		if (!entry)
			break;

		TAILQ_REMOVE(&table->entries, entry, node);
		table_entry_free(entry);
	}
}

static struct rte_swx_table_entry *
table_pending_add_find(struct table *table, struct rte_swx_table_entry *entry)
{
	struct rte_swx_table_entry *e;

	TAILQ_FOREACH(e, &table->pending_add, node)
		if (!table_entry_keycmp(table, entry, e))
			return e; /* Found. */

	return NULL; /* Not found. */
}

static void
table_pending_add_admit(struct table *table)
{
	TAILQ_CONCAT(&table->entries, &table->pending_add, node);
}

static void
table_pending_add_free(struct table *table)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(&table->pending_add);
		if (!entry)
			break;

		TAILQ_REMOVE(&table->pending_add, entry, node);
		table_entry_free(entry);
	}
}

static struct rte_swx_table_entry *
table_pending_modify0_find(struct table *table,
			   struct rte_swx_table_entry *entry)
{
	struct rte_swx_table_entry *e;

	TAILQ_FOREACH(e, &table->pending_modify0, node)
		if (!table_entry_keycmp(table, entry, e))
			return e; /* Found. */

	return NULL; /* Not found. */
}

static void
table_pending_modify0_admit(struct table *table)
{
	TAILQ_CONCAT(&table->entries, &table->pending_modify0, node);
}

static void
table_pending_modify0_free(struct table *table)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(&table->pending_modify0);
		if (!entry)
			break;

		TAILQ_REMOVE(&table->pending_modify0, entry, node);
		table_entry_free(entry);
	}
}

static struct rte_swx_table_entry *
table_pending_modify1_find(struct table *table,
			   struct rte_swx_table_entry *entry)
{
	struct rte_swx_table_entry *e;

	TAILQ_FOREACH(e, &table->pending_modify1, node)
		if (!table_entry_keycmp(table, entry, e))
			return e; /* Found. */

	return NULL; /* Not found. */
}

static void
table_pending_modify1_admit(struct table *table)
{
	TAILQ_CONCAT(&table->entries, &table->pending_modify1, node);
}

static void
table_pending_modify1_free(struct table *table)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(&table->pending_modify1);
		if (!entry)
			break;

		TAILQ_REMOVE(&table->pending_modify1, entry, node);
		table_entry_free(entry);
	}
}

static struct rte_swx_table_entry *
table_pending_delete_find(struct table *table,
			  struct rte_swx_table_entry *entry)
{
	struct rte_swx_table_entry *e;

	TAILQ_FOREACH(e, &table->pending_delete, node)
		if (!table_entry_keycmp(table, entry, e))
			return e; /* Found. */

	return NULL; /* Not found. */
}

static void
table_pending_delete_admit(struct table *table)
{
	TAILQ_CONCAT(&table->entries, &table->pending_delete, node);
}

static void
table_pending_delete_free(struct table *table)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(&table->pending_delete);
		if (!entry)
			break;

		TAILQ_REMOVE(&table->pending_delete, entry, node);
		table_entry_free(entry);
	}
}

static void
table_pending_default_free(struct table *table)
{
	if (!table->pending_default)
		return;

	free(table->pending_default->action_data);
	free(table->pending_default);
	table->pending_default = NULL;
}

static int
table_is_update_pending(struct table *table, int consider_pending_default)
{
	struct rte_swx_table_entry *e;
	uint32_t n = 0;

	/* Pending add. */
	TAILQ_FOREACH(e, &table->pending_add, node)
		n++;

	/* Pending modify. */
	TAILQ_FOREACH(e, &table->pending_modify1, node)
		n++;

	/* Pending delete. */
	TAILQ_FOREACH(e, &table->pending_delete, node)
		n++;

	/* Pending default. */
	if (consider_pending_default && table->pending_default)
		n++;

	return n;
}

static void
table_free(struct rte_swx_ctl_pipeline *ctl)
{
	uint32_t i;

	if (!ctl->tables)
		return;

	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *table = &ctl->tables[i];

		free(table->mf);
		free(table->actions);
		free(table->params.key_mask0);

		table_entries_free(table);
		table_pending_add_free(table);
		table_pending_modify0_free(table);
		table_pending_modify1_free(table);
		table_pending_delete_free(table);
		table_pending_default_free(table);
	}

	free(ctl->tables);
	ctl->tables = NULL;
}

static void
selector_group_members_free(struct selector *s, uint32_t group_id)
{
	struct rte_swx_table_selector_group *group = s->groups[group_id];

	if (!group)
		return;

	for ( ; ; ) {
		struct rte_swx_table_selector_member *m;

		m = TAILQ_FIRST(&group->members);
		if (!m)
			break;

		TAILQ_REMOVE(&group->members, m, node);
		free(m);
	}

	free(group);
	s->groups[group_id] = NULL;
}

static void
selector_pending_group_members_free(struct selector *s, uint32_t group_id)
{
	struct rte_swx_table_selector_group *group = s->pending_groups[group_id];

	if (!group)
		return;

	for ( ; ; ) {
		struct rte_swx_table_selector_member *m;

		m = TAILQ_FIRST(&group->members);
		if (!m)
			break;

		TAILQ_REMOVE(&group->members, m, node);
		free(m);
	}

	free(group);
	s->pending_groups[group_id] = NULL;
}

static int
selector_group_duplicate_to_pending(struct selector *s, uint32_t group_id)
{
	struct rte_swx_table_selector_group *g, *gp;
	struct rte_swx_table_selector_member *m;

	selector_pending_group_members_free(s, group_id);

	g = s->groups[group_id];
	gp = s->pending_groups[group_id];

	if (!gp) {
		gp = calloc(1, sizeof(struct rte_swx_table_selector_group));
		if (!gp)
			goto error;

		TAILQ_INIT(&gp->members);

		s->pending_groups[group_id] = gp;
	}

	if (!g)
		return 0;

	TAILQ_FOREACH(m, &g->members, node) {
		struct rte_swx_table_selector_member *mp;

		mp = calloc(1, sizeof(struct rte_swx_table_selector_member));
		if (!mp)
			goto error;

		memcpy(mp, m, sizeof(struct rte_swx_table_selector_member));

		TAILQ_INSERT_TAIL(&gp->members, mp, node);
	}

	return 0;

error:
	selector_pending_group_members_free(s, group_id);
	return -ENOMEM;
}

static void
selector_free(struct rte_swx_ctl_pipeline *ctl)
{
	uint32_t i;

	if (!ctl->selectors)
		return;

	for (i = 0; i < ctl->info.n_selectors; i++) {
		struct selector *s = &ctl->selectors[i];
		uint32_t i;

		/* selector_fields. */
		free(s->selector_fields);

		/* groups. */
		if (s->groups)
			for (i = 0; i < s->info.n_groups_max; i++)
				selector_group_members_free(s, i);

		free(s->groups);

		/* pending_groups. */
		if (s->pending_groups)
			for (i = 0; i < s->info.n_groups_max; i++)
				selector_pending_group_members_free(s, i);

		free(s->pending_groups);

		/* groups_added. */
		free(s->groups_added);

		/* groups_pending_delete. */
		free(s->groups_pending_delete);

		/* params. */
		free(s->params.selector_mask);
	}

	free(ctl->selectors);
	ctl->selectors = NULL;
}

static struct selector *
selector_find(struct rte_swx_ctl_pipeline *ctl, const char *selector_name)
{
	uint32_t i;

	for (i = 0; i < ctl->info.n_selectors; i++) {
		struct selector *s = &ctl->selectors[i];

		if (!strcmp(selector_name, s->info.name))
			return s;
	}

	return NULL;
}

static int
selector_params_get(struct rte_swx_ctl_pipeline *ctl, uint32_t selector_id)
{
	struct selector *s = &ctl->selectors[selector_id];
	struct rte_swx_ctl_table_match_field_info *first = NULL, *last = NULL;
	uint8_t *selector_mask = NULL;
	uint32_t selector_size = 0, selector_offset = 0, i;

	/* Find first (smallest offset) and last (biggest offset) match fields. */
	first = &s->selector_fields[0];
	last = &s->selector_fields[0];

	for (i = 1; i < s->info.n_selector_fields; i++) {
		struct rte_swx_ctl_table_match_field_info *f = &s->selector_fields[i];

		if (f->offset < first->offset)
			first = f;

		if (f->offset > last->offset)
			last = f;
	}

	/* selector_offset. */
	selector_offset = first->offset / 8;

	/* selector_size. */
	selector_size = (last->offset + last->n_bits - first->offset) / 8;

	/* selector_mask. */
	selector_mask = calloc(1, selector_size);
	if (!selector_mask)
		return -ENOMEM;

	for (i = 0; i < s->info.n_selector_fields; i++) {
		struct rte_swx_ctl_table_match_field_info *f = &s->selector_fields[i];
		uint32_t start;
		size_t size;

		start = (f->offset - first->offset) / 8;
		size = f->n_bits / 8;

		memset(&selector_mask[start], 0xFF, size);
	}

	/* Fill in. */
	s->params.group_id_offset = s->group_id_field.offset / 8;
	s->params.selector_size = selector_size;
	s->params.selector_offset = selector_offset;
	s->params.selector_mask = selector_mask;
	s->params.member_id_offset = s->member_id_field.offset / 8;
	s->params.n_groups_max = s->info.n_groups_max;
	s->params.n_members_per_group_max = s->info.n_members_per_group_max;

	return 0;
}

static void
table_state_free(struct rte_swx_ctl_pipeline *ctl)
{
	uint32_t i;

	if (!ctl->ts_next)
		return;

	/* For each table, free its table state. */
	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *table = &ctl->tables[i];
		struct rte_swx_table_state *ts = &ctl->ts_next[i];

		/* Default action data. */
		free(ts->default_action_data);

		/* Table object. */
		if (!table->is_stub && table->ops.free && ts->obj)
			table->ops.free(ts->obj);
	}

	/* For each selector table, free its table state. */
	for (i = 0; i < ctl->info.n_selectors; i++) {
		struct rte_swx_table_state *ts = &ctl->ts_next[i];

		/* Table object. */
		if (ts->obj)
			rte_swx_table_selector_free(ts->obj);
	}

	free(ctl->ts_next);
	ctl->ts_next = NULL;
}

static int
table_state_create(struct rte_swx_ctl_pipeline *ctl)
{
	int status = 0;
	uint32_t i;

	ctl->ts_next = calloc(ctl->info.n_tables + ctl->info.n_selectors,
			      sizeof(struct rte_swx_table_state));
	if (!ctl->ts_next) {
		status = -ENOMEM;
		goto error;
	}

	/* Tables. */
	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *table = &ctl->tables[i];
		struct rte_swx_table_state *ts = &ctl->ts[i];
		struct rte_swx_table_state *ts_next = &ctl->ts_next[i];

		/* Table object. */
		if (!table->is_stub && table->ops.add) {
			ts_next->obj = table->ops.create(&table->params,
							 &table->entries,
							 table->info.args,
							 ctl->numa_node);
			if (!ts_next->obj) {
				status = -ENODEV;
				goto error;
			}
		}

		if (!table->is_stub && !table->ops.add)
			ts_next->obj = ts->obj;

		/* Default action data: duplicate from current table state. */
		ts_next->default_action_data =
			malloc(table->params.action_data_size);
		if (!ts_next->default_action_data) {
			status = -ENOMEM;
			goto error;
		}

		memcpy(ts_next->default_action_data,
		       ts->default_action_data,
		       table->params.action_data_size);

		ts_next->default_action_id = ts->default_action_id;
	}

	/* Selector tables. */
	for (i = 0; i < ctl->info.n_selectors; i++) {
		struct selector *s = &ctl->selectors[i];
		struct rte_swx_table_state *ts_next = &ctl->ts_next[ctl->info.n_tables + i];

		/* Table object. */
		ts_next->obj = rte_swx_table_selector_create(&s->params, NULL, ctl->numa_node);
		if (!ts_next->obj) {
			status = -ENODEV;
			goto error;
		}
	}

	return 0;

error:
	table_state_free(ctl);
	return status;
}

void
rte_swx_ctl_pipeline_free(struct rte_swx_ctl_pipeline *ctl)
{
	if (!ctl)
		return;

	action_free(ctl);

	table_state_free(ctl);

	selector_free(ctl);

	table_free(ctl);

	free(ctl);
}

struct rte_swx_ctl_pipeline *
rte_swx_ctl_pipeline_create(struct rte_swx_pipeline *p)
{
	struct rte_swx_ctl_pipeline *ctl = NULL;
	uint32_t i;
	int status;

	if (!p)
		goto error;

	ctl = calloc(1, sizeof(struct rte_swx_ctl_pipeline));
	if (!ctl)
		goto error;

	/* info. */
	status = rte_swx_ctl_pipeline_info_get(p, &ctl->info);
	if (status)
		goto error;

	/* numa_node. */
	status = rte_swx_ctl_pipeline_numa_node_get(p, &ctl->numa_node);
	if (status)
		goto error;

	/* p. */
	ctl->p = p;

	/* actions. */
	ctl->actions = calloc(ctl->info.n_actions, sizeof(struct action));
	if (!ctl->actions)
		goto error;

	for (i = 0; i < ctl->info.n_actions; i++) {
		struct action *a = &ctl->actions[i];
		uint32_t j;

		/* info. */
		status = rte_swx_ctl_action_info_get(p, i, &a->info);
		if (status)
			goto error;

		/* args. */
		a->args = calloc(a->info.n_args,
				 sizeof(struct rte_swx_ctl_action_arg_info));
		if (!a->args)
			goto error;

		for (j = 0; j < a->info.n_args; j++) {
			status = rte_swx_ctl_action_arg_info_get(p,
								 i,
								 j,
								 &a->args[j]);
			if (status)
				goto error;
		}

		/* data_size. */
		for (j = 0; j < a->info.n_args; j++) {
			struct rte_swx_ctl_action_arg_info *info = &a->args[j];

			a->data_size += info->n_bits;
		}

		a->data_size = (a->data_size + 7) / 8;
	}

	/* tables. */
	ctl->tables = calloc(ctl->info.n_tables, sizeof(struct table));
	if (!ctl->tables)
		goto error;

	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *t = &ctl->tables[i];

		TAILQ_INIT(&t->entries);
		TAILQ_INIT(&t->pending_add);
		TAILQ_INIT(&t->pending_modify0);
		TAILQ_INIT(&t->pending_modify1);
		TAILQ_INIT(&t->pending_delete);
	}

	for (i = 0; i < ctl->info.n_tables; i++) {
		struct table *t = &ctl->tables[i];
		uint32_t j;

		/* info. */
		status = rte_swx_ctl_table_info_get(p, i, &t->info);
		if (status)
			goto error;

		/* mf. */
		t->mf = calloc(t->info.n_match_fields,
			sizeof(struct rte_swx_ctl_table_match_field_info));
		if (!t->mf)
			goto error;

		for (j = 0; j < t->info.n_match_fields; j++) {
			status = rte_swx_ctl_table_match_field_info_get(p,
				i,
				j,
				&t->mf[j]);
			if (status)
				goto error;
		}

		/* actions. */
		t->actions = calloc(t->info.n_actions,
			sizeof(struct rte_swx_ctl_table_action_info));
		if (!t->actions)
			goto error;

		for (j = 0; j < t->info.n_actions; j++) {
			status = rte_swx_ctl_table_action_info_get(p,
				i,
				j,
				&t->actions[j]);
			if (status ||
			    t->actions[j].action_id >= ctl->info.n_actions)
				goto error;
		}

		/* ops, is_stub. */
		status = rte_swx_ctl_table_ops_get(p, i, &t->ops, &t->is_stub);
		if (status)
			goto error;

		if ((t->is_stub && t->info.n_match_fields) ||
		    (!t->is_stub && !t->info.n_match_fields))
			goto error;

		/* params. */
		status = table_params_get(ctl, i);
		if (status)
			goto error;
	}

	/* selector tables. */
	ctl->selectors = calloc(ctl->info.n_selectors, sizeof(struct selector));
	if (!ctl->selectors)
		goto error;

	for (i = 0; i < ctl->info.n_selectors; i++) {
		struct selector *s = &ctl->selectors[i];
		uint32_t j;

		/* info. */
		status = rte_swx_ctl_selector_info_get(p, i, &s->info);
		if (status)
			goto error;

		/* group_id field. */
		status = rte_swx_ctl_selector_group_id_field_info_get(p,
			i,
			&s->group_id_field);
		if (status)
			goto error;

		/* selector fields. */
		s->selector_fields = calloc(s->info.n_selector_fields,
			sizeof(struct rte_swx_ctl_table_match_field_info));
		if (!s->selector_fields)
			goto error;

		for (j = 0; j < s->info.n_selector_fields; j++) {
			status = rte_swx_ctl_selector_field_info_get(p,
				i,
				j,
				&s->selector_fields[j]);
			if (status)
				goto error;
		}

		/* member_id field. */
		status = rte_swx_ctl_selector_member_id_field_info_get(p,
			i,
			&s->member_id_field);
		if (status)
			goto error;

		/* groups. */
		s->groups = calloc(s->info.n_groups_max,
			sizeof(struct rte_swx_table_selector_group *));
		if (!s->groups)
			goto error;

		/* pending_groups. */
		s->pending_groups = calloc(s->info.n_groups_max,
			sizeof(struct rte_swx_table_selector_group *));
		if (!s->pending_groups)
			goto error;

		/* groups_added. */
		s->groups_added = calloc(s->info.n_groups_max, sizeof(int));
		if (!s->groups_added)
			goto error;

		/* groups_pending_delete. */
		s->groups_pending_delete = calloc(s->info.n_groups_max, sizeof(int));
		if (!s->groups_pending_delete)
			goto error;

		/* params. */
		status = selector_params_get(ctl, i);
		if (status)
			goto error;
	}

	/* ts. */
	status = rte_swx_pipeline_table_state_get(p, &ctl->ts);
	if (status)
		goto error;

	/* ts_next. */
	status = table_state_create(ctl);
	if (status)
		goto error;

	return ctl;

error:
	rte_swx_ctl_pipeline_free(ctl);
	return NULL;
}

int
rte_swx_ctl_pipeline_table_entry_add(struct rte_swx_ctl_pipeline *ctl,
				     const char *table_name,
				     struct rte_swx_table_entry *entry)
{
	struct table *table;
	struct rte_swx_table_entry *new_entry, *existing_entry;
	uint32_t table_id;

	CHECK(ctl, EINVAL);
	CHECK(table_name && table_name[0], EINVAL);

	table = table_find(ctl, table_name);
	CHECK(table, EINVAL);
	table_id = table - ctl->tables;

	CHECK(entry, EINVAL);
	CHECK(!table_entry_check(ctl, table_id, entry, 1, 1), EINVAL);

	new_entry = table_entry_duplicate(ctl, table_id, entry, 1, 1);
	CHECK(new_entry, ENOMEM);

	/* The new entry is found in the table->entries list:
	 * - Add the new entry to the table->pending_modify1 list;
	 * - Move the existing entry from the table->entries list to the
	 *   table->pending_modify0 list.
	 */
	existing_entry = table_entries_find(table, entry);
	if (existing_entry) {
		TAILQ_INSERT_TAIL(&table->pending_modify1,
				  new_entry,
				  node);

		TAILQ_REMOVE(&table->entries,
			     existing_entry,
			     node);

		TAILQ_INSERT_TAIL(&table->pending_modify0,
				  existing_entry,
				  node);

		return 0;
	}

	/* The new entry is found in the table->pending_add list:
	 * - Replace the entry in the table->pending_add list with the new entry
	 *   (and free the replaced entry).
	 */
	existing_entry = table_pending_add_find(table, entry);
	if (existing_entry) {
		TAILQ_INSERT_AFTER(&table->pending_add,
				   existing_entry,
				   new_entry,
				   node);

		TAILQ_REMOVE(&table->pending_add,
			     existing_entry,
			     node);

		table_entry_free(existing_entry);

		return 0;
	}

	/* The new entry is found in the table->pending_modify1 list:
	 * - Replace the entry in the table->pending_modify1 list with the new
	 *   entry (and free the replaced entry).
	 */
	existing_entry = table_pending_modify1_find(table, entry);
	if (existing_entry) {
		TAILQ_INSERT_AFTER(&table->pending_modify1,
				   existing_entry,
				   new_entry,
				   node);

		TAILQ_REMOVE(&table->pending_modify1,
			     existing_entry,
			     node);

		table_entry_free(existing_entry);

		return 0;
	}

	/* The new entry is found in the table->pending_delete list:
	 * - Add the new entry to the table->pending_modify1 list;
	 * - Move the existing entry from the table->pending_delete list to the
	 *   table->pending_modify0 list.
	 */
	existing_entry = table_pending_delete_find(table, entry);
	if (existing_entry) {
		TAILQ_INSERT_TAIL(&table->pending_modify1,
				  new_entry,
				  node);

		TAILQ_REMOVE(&table->pending_delete,
			     existing_entry,
			     node);

		TAILQ_INSERT_TAIL(&table->pending_modify0,
				  existing_entry,
				  node);

		return 0;
	}

	/* The new entry is not found in any of the above lists:
	 * - Add the new entry to the table->pending_add list.
	 */
	TAILQ_INSERT_TAIL(&table->pending_add, new_entry, node);

	return 0;
}

int
rte_swx_ctl_pipeline_table_entry_delete(struct rte_swx_ctl_pipeline *ctl,
					const char *table_name,
					struct rte_swx_table_entry *entry)
{
	struct table *table;
	struct rte_swx_table_entry *existing_entry;
	uint32_t table_id;

	CHECK(ctl, EINVAL);

	CHECK(table_name && table_name[0], EINVAL);
	table = table_find(ctl, table_name);
	CHECK(table, EINVAL);
	table_id = table - ctl->tables;

	CHECK(entry, EINVAL);
	CHECK(!table_entry_check(ctl, table_id, entry, 1, 0), EINVAL);

	/* The entry is found in the table->entries list:
	 * - Move the existing entry from the table->entries list to to the
	 *   table->pending_delete list.
	 */
	existing_entry = table_entries_find(table, entry);
	if (existing_entry) {
		TAILQ_REMOVE(&table->entries,
			     existing_entry,
			     node);

		TAILQ_INSERT_TAIL(&table->pending_delete,
				  existing_entry,
				  node);

		return 0;
	}

	/* The entry is found in the table->pending_add list:
	 * - Remove the entry from the table->pending_add list and free it.
	 */
	existing_entry = table_pending_add_find(table, entry);
	if (existing_entry) {
		TAILQ_REMOVE(&table->pending_add,
			     existing_entry,
			     node);

		table_entry_free(existing_entry);
	}

	/* The entry is found in the table->pending_modify1 list:
	 * - Free the entry in the table->pending_modify1 list;
	 * - Move the existing entry from the table->pending_modify0 list to the
	 *   table->pending_delete list.
	 */
	existing_entry = table_pending_modify1_find(table, entry);
	if (existing_entry) {
		struct rte_swx_table_entry *real_existing_entry;

		TAILQ_REMOVE(&table->pending_modify1,
			     existing_entry,
			     node);

		table_entry_free(existing_entry);

		real_existing_entry = table_pending_modify0_find(table, entry);
		CHECK(real_existing_entry, EINVAL); /* Coverity. */

		TAILQ_REMOVE(&table->pending_modify0,
			     real_existing_entry,
			     node);

		TAILQ_INSERT_TAIL(&table->pending_delete,
				  real_existing_entry,
				  node);

		return 0;
	}

	/* The entry is found in the table->pending_delete list:
	 * - Do nothing: the existing entry is already in the
	 *   table->pending_delete list, i.e. already marked for delete, so
	 *   simply keep it there as it is.
	 */

	/* The entry is not found in any of the above lists:
	 * - Do nothing: no existing entry to delete.
	 */

	return 0;
}

int
rte_swx_ctl_pipeline_table_default_entry_add(struct rte_swx_ctl_pipeline *ctl,
					     const char *table_name,
					     struct rte_swx_table_entry *entry)
{
	struct table *table;
	struct rte_swx_table_entry *new_entry;
	uint32_t table_id;

	CHECK(ctl, EINVAL);

	CHECK(table_name && table_name[0], EINVAL);
	table = table_find(ctl, table_name);
	CHECK(table, EINVAL);
	table_id = table - ctl->tables;
	CHECK(!table->info.default_action_is_const, EINVAL);

	CHECK(entry, EINVAL);
	CHECK(!table_entry_check(ctl, table_id, entry, 0, 1), EINVAL);

	new_entry = table_entry_duplicate(ctl, table_id, entry, 0, 1);
	CHECK(new_entry, ENOMEM);

	table_pending_default_free(table);

	table->pending_default = new_entry;
	return 0;
}


static void
table_entry_list_free(struct rte_swx_table_entry_list *list)
{
	for ( ; ; ) {
		struct rte_swx_table_entry *entry;

		entry = TAILQ_FIRST(list);
		if (!entry)
			break;

		TAILQ_REMOVE(list, entry, node);
		table_entry_free(entry);
	}
}

static int
table_entry_list_duplicate(struct rte_swx_ctl_pipeline *ctl,
			   uint32_t table_id,
			   struct rte_swx_table_entry_list *dst,
			   struct rte_swx_table_entry_list *src)
{
	struct rte_swx_table_entry *src_entry;

	TAILQ_FOREACH(src_entry, src, node) {
		struct rte_swx_table_entry *dst_entry;

		dst_entry = table_entry_duplicate(ctl, table_id, src_entry, 1, 1);
		if (!dst_entry)
			goto error;

		TAILQ_INSERT_TAIL(dst, dst_entry, node);
	}

	return 0;

error:
	table_entry_list_free(dst);
	return -ENOMEM;
}

/* This commit stage contains all the operations that can fail; in case ANY of
 * them fails for ANY table, ALL of them are rolled back for ALL the tables.
 */
static int
table_rollfwd0(struct rte_swx_ctl_pipeline *ctl,
	       uint32_t table_id,
	       uint32_t after_swap)
{
	struct table *table = &ctl->tables[table_id];
	struct rte_swx_table_state *ts = &ctl->ts[table_id];
	struct rte_swx_table_state *ts_next = &ctl->ts_next[table_id];

	if (table->is_stub || !table_is_update_pending(table, 0))
		return 0;

	/*
	 * Current table supports incremental update.
	 */
	if (table->ops.add) {
		/* Reset counters. */
		table->n_add = 0;
		table->n_modify = 0;
		table->n_delete = 0;

		/* Add pending rules. */
		struct rte_swx_table_entry *entry;

		TAILQ_FOREACH(entry, &table->pending_add, node) {
			int status;

			status = table->ops.add(ts_next->obj, entry);
			if (status)
				return status;

			table->n_add++;
		}

		/* Modify pending rules. */
		TAILQ_FOREACH(entry, &table->pending_modify1, node) {
			int status;

			status = table->ops.add(ts_next->obj, entry);
			if (status)
				return status;

			table->n_modify++;
		}

		/* Delete pending rules. */
		TAILQ_FOREACH(entry, &table->pending_delete, node) {
			int status;

			status = table->ops.del(ts_next->obj, entry);
			if (status)
				return status;

			table->n_delete++;
		}

		return 0;
	}

	/*
	 * Current table does NOT support incremental update.
	 */
	if (!after_swap) {
		struct rte_swx_table_entry_list list;
		int status;

		/* Create updated list of entries included. */
		TAILQ_INIT(&list);

		status = table_entry_list_duplicate(ctl,
						    table_id,
						    &list,
						    &table->entries);
		if (status)
			goto error;

		status = table_entry_list_duplicate(ctl,
						    table_id,
						    &list,
						    &table->pending_add);
		if (status)
			goto error;

		status = table_entry_list_duplicate(ctl,
						    table_id,
						    &list,
						    &table->pending_modify1);
		if (status)
			goto error;

		/* Create new table object with the updates included. */
		ts_next->obj = table->ops.create(&table->params,
						 &list,
						 table->info.args,
						 ctl->numa_node);
		if (!ts_next->obj) {
			status = -ENODEV;
			goto error;
		}

		table_entry_list_free(&list);

		return 0;

error:
		table_entry_list_free(&list);
		return status;
	}

	/* Free the old table object. */
	if (ts_next->obj && table->ops.free)
		table->ops.free(ts_next->obj);

	/* Copy over the new table object. */
	ts_next->obj = ts->obj;

	return 0;
}

/* This commit stage contains all the operations that cannot fail. They are
 * executed only if the previous stage was successful for ALL the tables. Hence,
 * none of these operations has to be rolled back for ANY table.
 */
static void
table_rollfwd1(struct rte_swx_ctl_pipeline *ctl, uint32_t table_id)
{
	struct table *table = &ctl->tables[table_id];
	struct rte_swx_table_state *ts_next = &ctl->ts_next[table_id];
	struct action *a;
	uint8_t *action_data;
	uint64_t action_id;

	/* Copy the pending default entry. */
	if (!table->pending_default)
		return;

	action_id = table->pending_default->action_id;
	action_data = table->pending_default->action_data;
	a = &ctl->actions[action_id];

	memcpy(ts_next->default_action_data,
	       action_data,
	       a->data_size);

	ts_next->default_action_id = action_id;
}

/* This last commit stage is simply finalizing a successful commit operation.
 * This stage is only executed if all the previous stages were successful. This
 * stage cannot fail.
 */
static void
table_rollfwd2(struct rte_swx_ctl_pipeline *ctl, uint32_t table_id)
{
	struct table *table = &ctl->tables[table_id];

	/* Move all the pending add entries to the table, as they are now part
	 * of the table.
	 */
	table_pending_add_admit(table);

	/* Move all the pending modify1 entries to table, are they are now part
	 * of the table. Free up all the pending modify0 entries, as they are no
	 * longer part of the table.
	 */
	table_pending_modify1_admit(table);
	table_pending_modify0_free(table);

	/* Free up all the pending delete entries, as they are no longer part of
	 * the table.
	 */
	table_pending_delete_free(table);

	/* Free up the pending default entry, as it is now part of the table. */
	table_pending_default_free(table);
}

/* The rollback stage is only executed when the commit failed, i.e. ANY of the
 * commit operations that can fail did fail for ANY table. It reverts ALL the
 * tables to their state before the commit started, as if the commit never
 * happened.
 */
static void
table_rollback(struct rte_swx_ctl_pipeline *ctl, uint32_t table_id)
{
	struct table *table = &ctl->tables[table_id];
	struct rte_swx_table_state *ts_next = &ctl->ts_next[table_id];

	if (table->is_stub || !table_is_update_pending(table, 0))
		return;

	if (table->ops.add) {
		struct rte_swx_table_entry *entry;

		/* Add back all the entries that were just deleted. */
		TAILQ_FOREACH(entry, &table->pending_delete, node) {
			if (!table->n_delete)
				break;

			table->ops.add(ts_next->obj, entry);
			table->n_delete--;
		}

		/* Add back the old copy for all the entries that were just
		 * modified.
		 */
		TAILQ_FOREACH(entry, &table->pending_modify0, node) {
			if (!table->n_modify)
				break;

			table->ops.add(ts_next->obj, entry);
			table->n_modify--;
		}

		/* Delete all the entries that were just added. */
		TAILQ_FOREACH(entry, &table->pending_add, node) {
			if (!table->n_add)
				break;

			table->ops.del(ts_next->obj, entry);
			table->n_add--;
		}
	} else {
		struct rte_swx_table_state *ts = &ctl->ts[table_id];

		/* Free the new table object, as update was cancelled. */
		if (ts_next->obj && table->ops.free)
			table->ops.free(ts_next->obj);

		/* Reinstate the old table object. */
		ts_next->obj = ts->obj;
	}
}

/* This stage is conditionally executed (as instructed by the user) after a
 * failed commit operation to remove ALL the pending work for ALL the tables.
 */
static void
table_abort(struct rte_swx_ctl_pipeline *ctl, uint32_t table_id)
{
	struct table *table = &ctl->tables[table_id];

	/* Free up all the pending add entries, as none of them is part of the
	 * table.
	 */
	table_pending_add_free(table);

	/* Free up all the pending modify1 entries, as none of them made it to
	 * the table. Add back all the pending modify0 entries, as none of them
	 * was deleted from the table.
	 */
	table_pending_modify1_free(table);
	table_pending_modify0_admit(table);

	/* Add back all the pending delete entries, as none of them was deleted
	 * from the table.
	 */
	table_pending_delete_admit(table);

	/* Free up the pending default entry, as it is no longer going to be
	 * added to the table.
	 */
	table_pending_default_free(table);
}

int
rte_swx_ctl_pipeline_selector_group_add(struct rte_swx_ctl_pipeline *ctl,
					const char *selector_name,
					uint32_t *group_id)
{
	struct selector *s;
	uint32_t i;

	/* Check input arguments. */
	if (!ctl || !selector_name || !selector_name[0] || !group_id)
		return -EINVAL;

	s = selector_find(ctl, selector_name);
	if (!s)
		return -EINVAL;

	/* Find an unused group. */
	for (i = 0; i < s->info.n_groups_max; i++)
		if (!s->groups_added[i]) {
			*group_id = i;
			s->groups_added[i] = 1;
			return 0;
		}

	return -ENOSPC;
}

int
rte_swx_ctl_pipeline_selector_group_delete(struct rte_swx_ctl_pipeline *ctl,
					   const char *selector_name,
					   uint32_t group_id)
{
	struct selector *s;
	struct rte_swx_table_selector_group *group;

	/* Check input arguments. */
	if (!ctl || !selector_name || !selector_name[0])
		return -EINVAL;

	s = selector_find(ctl, selector_name);
	if (!s ||
	   (group_id >= s->info.n_groups_max) ||
	   !s->groups_added[group_id])
		return -EINVAL;

	/* Check if this group is already scheduled for deletion. */
	if (s->groups_pending_delete[group_id])
		return 0;

	/* Initialize the pending group, if needed. */
	if (!s->pending_groups[group_id]) {
		int status;

		status = selector_group_duplicate_to_pending(s, group_id);
		if (status)
			return status;
	}

	group = s->pending_groups[group_id];

	/* Schedule removal of all the members from the current group. */
	for ( ; ; ) {
		struct rte_swx_table_selector_member *m;

		m = TAILQ_FIRST(&group->members);
		if (!m)
			break;

		TAILQ_REMOVE(&group->members, m, node);
		free(m);
	}

	/* Schedule the group for deletion. */
	s->groups_pending_delete[group_id] = 1;

	return 0;
}

int
rte_swx_ctl_pipeline_selector_group_member_add(struct rte_swx_ctl_pipeline *ctl,
					       const char *selector_name,
					       uint32_t group_id,
					       uint32_t member_id,
					       uint32_t member_weight)
{
	struct selector *s;
	struct rte_swx_table_selector_group *group;
	struct rte_swx_table_selector_member *m;

	if (!member_weight)
		return rte_swx_ctl_pipeline_selector_group_member_delete(ctl,
									 selector_name,
									 group_id,
									 member_id);

	/* Check input arguments. */
	if (!ctl || !selector_name || !selector_name[0])
		return -EINVAL;

	s = selector_find(ctl, selector_name);
	if (!s ||
	   (group_id >= s->info.n_groups_max) ||
	   !s->groups_added[group_id] ||
	   s->groups_pending_delete[group_id])
		return -EINVAL;

	/* Initialize the pending group, if needed. */
	if (!s->pending_groups[group_id]) {
		int status;

		status = selector_group_duplicate_to_pending(s, group_id);
		if (status)
			return status;
	}

	group = s->pending_groups[group_id];

	/* If this member is already in this group, then simply update its weight and return. */
	TAILQ_FOREACH(m, &group->members, node)
		if (m->member_id == member_id) {
			m->member_weight = member_weight;
			return 0;
		}

	/* Add new member to this group. */
	m = calloc(1, sizeof(struct rte_swx_table_selector_member));
	if (!m)
		return -ENOMEM;

	m->member_id = member_id;
	m->member_weight = member_weight;

	TAILQ_INSERT_TAIL(&group->members, m, node);

	return 0;
}

int
rte_swx_ctl_pipeline_selector_group_member_delete(struct rte_swx_ctl_pipeline *ctl,
						  const char *selector_name,
						  uint32_t group_id __rte_unused,
						  uint32_t member_id __rte_unused)
{
	struct selector *s;
	struct rte_swx_table_selector_group *group;
	struct rte_swx_table_selector_member *m;

	/* Check input arguments. */
	if (!ctl || !selector_name || !selector_name[0])
		return -EINVAL;

	s = selector_find(ctl, selector_name);
	if (!s ||
	    (group_id >= s->info.n_groups_max) ||
	    !s->groups_added[group_id] ||
	    s->groups_pending_delete[group_id])
		return -EINVAL;

	/* Initialize the pending group, if needed. */
	if (!s->pending_groups[group_id]) {
		int status;

		status = selector_group_duplicate_to_pending(s, group_id);
		if (status)
			return status;
	}

	group = s->pending_groups[group_id];

	/* Look for this member in the group and remove it, if found. */
	TAILQ_FOREACH(m, &group->members, node)
		if (m->member_id == member_id) {
			TAILQ_REMOVE(&group->members, m, node);
			free(m);
			return 0;
		}

	return 0;
}

static int
selector_rollfwd(struct rte_swx_ctl_pipeline *ctl, uint32_t selector_id)
{
	struct selector *s = &ctl->selectors[selector_id];
	struct rte_swx_table_state *ts_next = &ctl->ts_next[ctl->info.n_tables + selector_id];
	uint32_t group_id;

	/* Push pending group member changes (s->pending_groups[group_id]) to the selector table
	 * mirror copy (ts_next->obj).
	 */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++) {
		struct rte_swx_table_selector_group *group = s->pending_groups[group_id];
		int status;

		/* Skip this group if no change needed. */
		if (!group)
			continue;

		/* Apply the pending changes for the current group. */
		status = rte_swx_table_selector_group_set(ts_next->obj, group_id, group);
		if (status)
			return status;
	}

	return 0;
}

static void
selector_rollfwd_finalize(struct rte_swx_ctl_pipeline *ctl, uint32_t selector_id)
{
	struct selector *s = &ctl->selectors[selector_id];
	uint32_t group_id;

	/* Commit pending group member changes (s->pending_groups[group_id]) to the stable group
	 * records (s->groups[group_id).
	 */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++) {
		struct rte_swx_table_selector_group *g = s->groups[group_id];
		struct rte_swx_table_selector_group *gp = s->pending_groups[group_id];

		/* Skip this group if no change needed. */
		if (!gp)
			continue;

		/* Transition the pending changes to stable. */
		s->groups[group_id] = gp;
		s->pending_groups[group_id] = NULL;

		/* Free the old group member list. */
		if (!g)
			continue;

		for ( ; ; ) {
			struct rte_swx_table_selector_member *m;

			m = TAILQ_FIRST(&g->members);
			if (!m)
				break;

			TAILQ_REMOVE(&g->members, m, node);
			free(m);
		}

		free(g);
	}

	/* Commit pending group validity changes (from s->groups_pending_delete[group_id] to
	 * s->groups_added[group_id].
	 */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++)
		if (s->groups_pending_delete[group_id]) {
			s->groups_added[group_id] = 0;
			s->groups_pending_delete[group_id] = 0;
		}
}

static void
selector_rollback(struct rte_swx_ctl_pipeline *ctl, uint32_t selector_id)
{
	struct selector *s = &ctl->selectors[selector_id];
	struct rte_swx_table_state *ts = &ctl->ts[ctl->info.n_tables + selector_id];
	struct rte_swx_table_state *ts_next = &ctl->ts_next[ctl->info.n_tables + selector_id];
	uint32_t group_id;

	/* Discard any previous changes to the selector table mirror copy (ts_next->obj). */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++) {
		struct rte_swx_table_selector_group *gp = s->pending_groups[group_id];

		if (gp) {
			ts_next->obj = ts->obj;
			break;
		}
	}
}

static void
selector_abort(struct rte_swx_ctl_pipeline *ctl, uint32_t selector_id)
{
	struct selector *s = &ctl->selectors[selector_id];
	uint32_t group_id;

	/* Discard any pending group member changes (s->pending_groups[group_id]). */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++)
		selector_pending_group_members_free(s, group_id);

	/* Discard any pending group deletions. */
	memset(s->groups_pending_delete, 0, s->info.n_groups_max * sizeof(int));
}

int
rte_swx_ctl_pipeline_commit(struct rte_swx_ctl_pipeline *ctl, int abort_on_fail)
{
	struct rte_swx_table_state *ts;
	int status = 0;
	uint32_t i;

	CHECK(ctl, EINVAL);

	/* Operate the changes on the current ts_next before it becomes the new ts. First, operate
	 * all the changes that can fail; if no failure, then operate the changes that cannot fail.
	 */
	for (i = 0; i < ctl->info.n_tables; i++) {
		status = table_rollfwd0(ctl, i, 0);
		if (status)
			goto rollback;
	}

	for (i = 0; i < ctl->info.n_selectors; i++) {
		status = selector_rollfwd(ctl, i);
		if (status)
			goto rollback;
	}

	for (i = 0; i < ctl->info.n_tables; i++)
		table_rollfwd1(ctl, i);

	/* Swap the table state for the data plane. The current ts and ts_next
	 * become the new ts_next and ts, respectively.
	 */
	rte_swx_pipeline_table_state_set(ctl->p, ctl->ts_next);
	usleep(100);
	ts = ctl->ts;
	ctl->ts = ctl->ts_next;
	ctl->ts_next = ts;

	/* Operate the changes on the current ts_next, which is the previous ts, in order to get
	 * the current ts_next in sync with the current ts. Since the changes that can fail did
	 * not fail on the previous ts_next, it is guaranteed that they will not fail on the
	 * current ts_next, hence no error checking is needed.
	 */
	for (i = 0; i < ctl->info.n_tables; i++) {
		table_rollfwd0(ctl, i, 1);
		table_rollfwd1(ctl, i);
		table_rollfwd2(ctl, i);
	}

	for (i = 0; i < ctl->info.n_selectors; i++) {
		selector_rollfwd(ctl, i);
		selector_rollfwd_finalize(ctl, i);
	}

	return 0;

rollback:
	for (i = 0; i < ctl->info.n_tables; i++) {
		table_rollback(ctl, i);
		if (abort_on_fail)
			table_abort(ctl, i);
	}

	for (i = 0; i < ctl->info.n_selectors; i++) {
		selector_rollback(ctl, i);
		if (abort_on_fail)
			selector_abort(ctl, i);
	}

	return status;
}

void
rte_swx_ctl_pipeline_abort(struct rte_swx_ctl_pipeline *ctl)
{
	uint32_t i;

	if (!ctl)
		return;

	for (i = 0; i < ctl->info.n_tables; i++)
		table_abort(ctl, i);

	for (i = 0; i < ctl->info.n_selectors; i++)
		selector_abort(ctl, i);
}

static int
mask_to_prefix(uint64_t mask, uint32_t mask_length, uint32_t *prefix_length)
{
	uint32_t n_trailing_zeros = 0, n_ones = 0, i;

	if (!mask) {
		*prefix_length = 0;
		return 0;
	}

	/* Count trailing zero bits. */
	for (i = 0; i < 64; i++) {
		if (mask & (1LLU << i))
			break;

		n_trailing_zeros++;
	}

	/* Count the one bits that follow. */
	for ( ; i < 64; i++) {
		if (!(mask & (1LLU << i)))
			break;

		n_ones++;
	}

	/* Check that no more one bits are present */
	for ( ; i < 64; i++)
		if (mask & (1LLU << i))
			return -EINVAL;

	/* Check that the input mask is a prefix or the right length. */
	if (n_ones + n_trailing_zeros != mask_length)
		return -EINVAL;

	*prefix_length = n_ones;
	return 0;
}

static int
token_is_comment(const char *token)
{
	if ((token[0] == '#') ||
	    (token[0] == ';') ||
	    ((token[0] == '/') && (token[1] == '/')))
		return 1; /* TRUE. */

	return 0; /* FALSE. */
}

#define RTE_SWX_CTL_ENTRY_TOKENS_MAX 256

struct rte_swx_table_entry *
rte_swx_ctl_pipeline_table_entry_read(struct rte_swx_ctl_pipeline *ctl,
				      const char *table_name,
				      const char *string,
				      int *is_blank_or_comment)
{
	char *token_array[RTE_SWX_CTL_ENTRY_TOKENS_MAX], **tokens;
	struct table *table;
	struct action *action;
	struct rte_swx_table_entry *entry = NULL;
	char *s0 = NULL, *s;
	uint32_t n_tokens = 0, arg_offset = 0, lpm_prefix_length_max = 0, lpm_prefix_length = 0, i;
	int lpm = 0, blank_or_comment = 0;

	/* Check input arguments. */
	if (!ctl)
		goto error;

	if (!table_name || !table_name[0])
		goto error;

	table = table_find(ctl, table_name);
	if (!table)
		goto error;

	if (!string || !string[0])
		goto error;

	/* Memory allocation. */
	s0 = strdup(string);
	if (!s0)
		goto error;

	entry = table_entry_alloc(table);
	if (!entry)
		goto error;

	/* Parse the string into tokens. */
	for (s = s0; ; ) {
		char *token;

		token = strtok_r(s, " \f\n\r\t\v", &s);
		if (!token || token_is_comment(token))
			break;

		if (n_tokens >= RTE_SWX_CTL_ENTRY_TOKENS_MAX)
			goto error;

		token_array[n_tokens] = token;
		n_tokens++;
	}

	if (!n_tokens) {
		blank_or_comment = 1;
		goto error;
	}

	tokens = token_array;

	/*
	 * Match.
	 */
	if (!(n_tokens && !strcmp(tokens[0], "match")))
		goto action;

	if (n_tokens < 1 + table->info.n_match_fields)
		goto error;

	for (i = 0; i < table->info.n_match_fields; i++) {
		struct rte_swx_ctl_table_match_field_info *mf = &table->mf[i];
		char *mf_val = tokens[1 + i], *mf_mask = NULL;
		uint64_t val, mask = UINT64_MAX;
		uint32_t offset = (mf->offset - table->mf_first->offset) / 8;

		/*
		 * Mask.
		 */
		mf_mask = strchr(mf_val, '/');
		if (mf_mask) {
			*mf_mask = 0;
			mf_mask++;

			/* Parse. */
			mask = strtoull(mf_mask, &mf_mask, 0);
			if (mf_mask[0])
				goto error;

			/* LPM. */
			if (mf->match_type == RTE_SWX_TABLE_MATCH_LPM) {
				int status;

				lpm = 1;

				lpm_prefix_length_max = mf->n_bits;

				status = mask_to_prefix(mask, mf->n_bits, &lpm_prefix_length);
				if (status)
					goto error;
			}

			/* Endianness conversion. */
			if (mf->is_header)
				mask = field_hton(mask, mf->n_bits);
		}

		/* Copy to entry. */
		if (entry->key_mask)
			memcpy(&entry->key_mask[offset],
			       (uint8_t *)&mask,
			       mf->n_bits / 8);

		/*
		 * Value.
		 */
		/* Parse. */
		val = strtoull(mf_val, &mf_val, 0);
		if (mf_val[0])
			goto error;

		/* Endianness conversion. */
		if (mf->is_header)
			val = field_hton(val, mf->n_bits);

		/* Copy to entry. */
		memcpy(&entry->key[offset],
		       (uint8_t *)&val,
		       mf->n_bits / 8);
	}

	tokens += 1 + table->info.n_match_fields;
	n_tokens -= 1 + table->info.n_match_fields;

	/*
	 * Match priority.
	 */
	if (n_tokens && !strcmp(tokens[0], "priority")) {
		char *priority = tokens[1];
		uint32_t val;

		if (n_tokens < 2)
			goto error;

		/* Parse. */
		val = strtoul(priority, &priority, 0);
		if (priority[0])
			goto error;

		/* Copy to entry. */
		entry->key_priority = val;

		tokens += 2;
		n_tokens -= 2;
	}

	/* LPM. */
	if (lpm)
		entry->key_priority = lpm_prefix_length_max - lpm_prefix_length;

	/*
	 * Action.
	 */
action:
	if (!(n_tokens && !strcmp(tokens[0], "action")))
		goto other;

	if (n_tokens < 2)
		goto error;

	action = action_find(ctl, tokens[1]);
	if (!action)
		goto error;

	if (n_tokens < 2 + action->info.n_args * 2)
		goto error;

	/* action_id. */
	entry->action_id = action - ctl->actions;

	/* action_data. */
	for (i = 0; i < action->info.n_args; i++) {
		struct rte_swx_ctl_action_arg_info *arg = &action->args[i];
		char *arg_name, *arg_val;
		uint64_t val;

		arg_name = tokens[2 + i * 2];
		arg_val = tokens[2 + i * 2 + 1];

		if (strcmp(arg_name, arg->name))
			goto error;

		val = strtoull(arg_val, &arg_val, 0);
		if (arg_val[0])
			goto error;

		/* Endianness conversion. */
		if (arg->is_network_byte_order)
			val = field_hton(val, arg->n_bits);

		/* Copy to entry. */
		memcpy(&entry->action_data[arg_offset],
		       (uint8_t *)&val,
		       arg->n_bits / 8);

		arg_offset += arg->n_bits / 8;
	}

	tokens += 2 + action->info.n_args * 2;
	n_tokens -= 2 + action->info.n_args * 2;

other:
	if (n_tokens)
		goto error;

	free(s0);
	return entry;

error:
	table_entry_free(entry);
	free(s0);
	if (is_blank_or_comment)
		*is_blank_or_comment = blank_or_comment;
	return NULL;
}

static void
table_entry_printf(FILE *f,
		   struct rte_swx_ctl_pipeline *ctl,
		   struct table *table,
		   struct rte_swx_table_entry *entry)
{
	struct action *action = &ctl->actions[entry->action_id];
	uint32_t i;

	fprintf(f, "match ");
	for (i = 0; i < table->params.key_size; i++)
		fprintf(f, "%02x", entry->key[i]);

	if (entry->key_mask) {
		fprintf(f, "/");
		for (i = 0; i < table->params.key_size; i++)
			fprintf(f, "%02x", entry->key_mask[i]);
	}

	fprintf(f, " priority %u", entry->key_priority);

	fprintf(f, " action %s ", action->info.name);
	for (i = 0; i < action->data_size; i++)
		fprintf(f, "%02x", entry->action_data[i]);

	fprintf(f, "\n");
}

int
rte_swx_ctl_pipeline_table_fprintf(FILE *f,
				   struct rte_swx_ctl_pipeline *ctl,
				   const char *table_name)
{
	struct table *table;
	struct rte_swx_table_entry *entry;
	uint32_t n_entries = 0, i;

	if (!f || !ctl || !table_name || !table_name[0])
		return -EINVAL;

	table = table_find(ctl, table_name);
	if (!table)
		return -EINVAL;

	/* Table. */
	fprintf(f, "# Table %s: key size %u bytes, key offset %u, key mask [",
		table->info.name,
		table->params.key_size,
		table->params.key_offset);

	for (i = 0; i < table->params.key_size; i++)
		fprintf(f, "%02x", table->params.key_mask0[i]);

	fprintf(f, "], action data size %u bytes\n",
		table->params.action_data_size);

	/* Table entries. */
	TAILQ_FOREACH(entry, &table->entries, node) {
		table_entry_printf(f, ctl, table, entry);
		n_entries++;
	}

	TAILQ_FOREACH(entry, &table->pending_modify0, node) {
		table_entry_printf(f, ctl, table, entry);
		n_entries++;
	}

	TAILQ_FOREACH(entry, &table->pending_delete, node) {
		table_entry_printf(f, ctl, table, entry);
		n_entries++;
	}

	fprintf(f, "# Table %s currently has %u entries.\n",
		table_name,
		n_entries);
	return 0;
}

int
rte_swx_ctl_pipeline_selector_fprintf(FILE *f,
				      struct rte_swx_ctl_pipeline *ctl,
				      const char *selector_name)
{
	struct selector *s;
	uint32_t group_id;

	if (!f || !ctl || !selector_name || !selector_name[0])
		return -EINVAL;

	s = selector_find(ctl, selector_name);
	if (!s)
		return -EINVAL;

	/* Selector. */
	fprintf(f, "# Selector %s: max groups %u, max members per group %u\n",
		s->info.name,
		s->info.n_groups_max,
		s->info.n_members_per_group_max);

	/* Groups. */
	for (group_id = 0; group_id < s->info.n_groups_max; group_id++) {
		struct rte_swx_table_selector_group *group = s->groups[group_id];
		struct rte_swx_table_selector_member *m;
		uint32_t n_members = 0;

		fprintf(f, "Group %u = [", group_id);

		/* Non-empty group. */
		if (group)
			TAILQ_FOREACH(m, &group->members, node) {
				fprintf(f, "%u:%u ", m->member_id, m->member_weight);
				n_members++;
			}

		/* Empty group. */
		if (!n_members)
			fprintf(f, "0:1 ");

		fprintf(f, "]\n");
	}

	return 0;
}