2
0
mirror of https://github.com/openvswitch/ovs synced 2025-08-22 01:51:26 +00:00
ovs/ovsdb/condition.c

614 lines
18 KiB
C
Raw Normal View History

/* Copyright (c) 2009, 2010, 2011 Nicira, Inc.
2009-11-04 15:11:44 -08:00
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at:
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <config.h>
#include "condition.h"
#include <limits.h>
#include "column.h"
#include "openvswitch/json.h"
2009-11-04 15:11:44 -08:00
#include "ovsdb-error.h"
#include "row.h"
#include <string.h>
2009-11-04 15:11:44 -08:00
#include "table.h"
#include "util.h"
2009-11-04 15:11:44 -08:00
static struct ovsdb_error *
2009-11-04 15:11:44 -08:00
ovsdb_clause_from_json(const struct ovsdb_table_schema *ts,
const struct json *json,
struct ovsdb_symbol_table *symtab,
2009-11-04 15:11:44 -08:00
struct ovsdb_clause *clause)
{
struct ovsdb_error *error;
const char *function_name;
const char *column_name;
struct ovsdb_type type;
if (json->type == JSON_TRUE || json->type == JSON_FALSE) {
clause->function =
json->type == JSON_TRUE ? OVSDB_F_TRUE : OVSDB_F_FALSE;
/* Column and arg fields are not being used with boolean functions.
* Use dummy values */
const struct ovsdb_column *uuid_column =
ovsdb_table_schema_get_column(ts, "_uuid");
ovs_assert(uuid_column);
clause->column = uuid_column;
clause->index = clause->column->index;
ovsdb_datum_init_default(&clause->arg, &clause->column->type);
return NULL;
}
2009-11-04 15:11:44 -08:00
if (json->type != JSON_ARRAY
|| json_array_size(json) != 3
|| json_array_at(json, 0)->type != JSON_STRING
|| json_array_at(json, 1)->type != JSON_STRING) {
2009-11-04 15:11:44 -08:00
return ovsdb_syntax_error(json, NULL, "Parse error in condition.");
}
column_name = json_string(json_array_at(json, 0));
2009-11-04 15:11:44 -08:00
clause->column = ovsdb_table_schema_get_column(ts, column_name);
if (!clause->column) {
return ovsdb_syntax_error(json, "unknown column",
"No column %s in table %s.",
column_name, ts->name);
}
clause->index = clause->column->index;
2009-11-04 15:11:44 -08:00
type = clause->column->type;
function_name = json_string(json_array_at(json, 1));
2009-11-04 15:11:44 -08:00
error = ovsdb_function_from_string(function_name, &clause->function);
if (error) {
return error;
}
/* Type-check and relax restrictions on 'type' if appropriate. */
switch (clause->function) {
case OVSDB_F_LT:
case OVSDB_F_LE:
case OVSDB_F_GT:
case OVSDB_F_GE:
/* Allow these operators for types with n_min == 0, n_max == 1.
* (They will always be "false" if the value is missing.) */
if (!(ovsdb_type_is_scalar(&type)
|| ovsdb_type_is_optional_scalar(&type))
2010-02-08 14:09:36 -08:00
|| (type.key.type != OVSDB_TYPE_INTEGER
&& type.key.type != OVSDB_TYPE_REAL)) {
2009-11-04 15:11:44 -08:00
char *s = ovsdb_type_to_english(&type);
error = ovsdb_syntax_error(
json, NULL, "Type mismatch: \"%s\" operator may not be "
"applied to column %s of type %s.",
ovsdb_function_to_string(clause->function),
clause->column->name, s);
free(s);
return error;
}
/* Force the argument to be a scalar. */
type.n_min = 1;
2009-11-04 15:11:44 -08:00
break;
2009-11-04 15:11:44 -08:00
case OVSDB_F_EQ:
case OVSDB_F_NE:
break;
case OVSDB_F_EXCLUDES:
if (!ovsdb_type_is_scalar(&type)) {
type.n_min = 0;
type.n_max = UINT_MAX;
}
break;
case OVSDB_F_INCLUDES:
if (!ovsdb_type_is_scalar(&type)) {
type.n_min = 0;
}
break;
case OVSDB_F_TRUE:
case OVSDB_F_FALSE:
OVS_NOT_REACHED();
2009-11-04 15:11:44 -08:00
}
return ovsdb_datum_from_json(&clause->arg, &type,
json_array_at(json, 2), symtab);
2009-11-04 15:11:44 -08:00
}
static void
ovsdb_clause_free(struct ovsdb_clause *clause)
{
ovsdb_datum_destroy(&clause->arg, &clause->column->type);
}
static int
compare_clauses_3way(const void *a_, const void *b_)
{
const struct ovsdb_clause *a = a_;
const struct ovsdb_clause *b = b_;
if (a->function != b->function) {
/* Bring functions to the front based on the fraction of table rows
* that they are (heuristically) expected to leave in the query
* results. Note that "enum ovsdb_function" is intentionally ordered
* to make this trivial. */
return a->function < b->function ? -1 : 1;
} else if (a->column->index != b->column->index) {
if (a->column->index < OVSDB_N_STD_COLUMNS
|| b->column->index < OVSDB_N_STD_COLUMNS) {
/* Bring the standard columns and in particular the UUID column
* (since OVSDB_COL_UUID has value 0) to the front. We have an
* index on the UUID column, so that makes our queries cheaper. */
return a->column->index < b->column->index ? -1 : 1;
} else {
/* Order clauses predictably to make testing easier. */
return strcmp(a->column->name, b->column->name);
}
} else {
return 0;
}
}
static int
compare_clauses_3way_with_data(const void *a_, const void *b_)
{
const struct ovsdb_clause *a = a_;
const struct ovsdb_clause *b = b_;
int res;
res = compare_clauses_3way(a, b);
return res ? res : ovsdb_datum_compare_3way(&a->arg,
&b->arg,
&a->column->type);
}
struct ovsdb_o_column {
const struct ovsdb_column *column;
struct hmap o_clauses;
};
struct ovsdb_o_clause {
struct ovsdb_datum *arg;
struct hmap_node hmap_node;
};
static void
ovsdb_condition_optimize(struct ovsdb_condition *cnd)
{
size_t i;
uint32_t hash;
if (!cnd->optimized) {
return;
}
for(i = 0; i < cnd->n_clauses; i++) {
struct ovsdb_clause *clause = &cnd->clauses[i];
if (clause->function != OVSDB_F_EQ) {
continue;
}
struct ovsdb_o_clause *o_clause = xzalloc(sizeof *o_clause);
struct ovsdb_o_column *o_column =
shash_find_data(&cnd->o_columns, clause->column->name);
if (!o_column) {
o_column = xzalloc(sizeof *o_column);
o_column->column = clause->column;
hmap_init(&o_column->o_clauses);
shash_add(&cnd->o_columns, clause->column->name, o_column);
}
o_clause->arg = &clause->arg;
hash = ovsdb_datum_hash(&clause->arg, &clause->column->type, 0);
hmap_insert(&o_column->o_clauses, &o_clause->hmap_node, hash);
}
}
static void
ovsdb_condition_optimize_destroy(struct ovsdb_condition *cnd)
{
struct shash_node *node;
SHASH_FOR_EACH_SAFE (node, &cnd->o_columns) {
struct ovsdb_o_column *o_column = node->data;
struct ovsdb_o_clause *c;
HMAP_FOR_EACH_SAFE (c, hmap_node, &o_column->o_clauses) {
hmap_remove(&o_column->o_clauses, &c->hmap_node);
free(c);
}
hmap_destroy(&o_column->o_clauses);
shash_delete(&cnd->o_columns, node);
free(o_column);
}
shash_destroy(&cnd->o_columns);
}
2009-11-04 15:11:44 -08:00
struct ovsdb_error *
ovsdb_condition_from_json(const struct ovsdb_table_schema *ts,
const struct json *json,
struct ovsdb_symbol_table *symtab,
2009-11-04 15:11:44 -08:00
struct ovsdb_condition *cnd)
{
size_t i, n = json_array_size(json);
2009-11-04 15:11:44 -08:00
ovsdb_condition_init(cnd);
cnd->clauses = xmalloc(n * sizeof *cnd->clauses);
for (i = 0; i < n; i++) {
2009-11-04 15:11:44 -08:00
struct ovsdb_error *error;
error = ovsdb_clause_from_json(ts, json_array_at(json, i), symtab,
2009-11-04 15:11:44 -08:00
&cnd->clauses[i]);
if (error) {
ovsdb_condition_destroy(cnd);
cnd->clauses = NULL;
cnd->n_clauses = 0;
return error;
}
cnd->n_clauses++;
if (cnd->clauses[i].function > OVSDB_F_EQ) {
cnd->optimized = false;
}
2009-11-04 15:11:44 -08:00
}
/* A real database would have a query optimizer here. */
qsort(cnd->clauses, cnd->n_clauses, sizeof *cnd->clauses,
compare_clauses_3way_with_data);
2009-11-04 15:11:44 -08:00
ovsdb_condition_optimize(cnd);
2009-11-04 15:11:44 -08:00
return NULL;
}
static struct json *
ovsdb_clause_to_json(const struct ovsdb_clause *clause)
{
if (clause->function != OVSDB_F_TRUE &&
clause->function != OVSDB_F_FALSE) {
return json_array_create_3(
json_string_create(clause->column->name),
json_string_create(ovsdb_function_to_string(clause->function)),
ovsdb_datum_to_json(&clause->arg, &clause->column->type));
}
return json_boolean_create(clause->function == OVSDB_F_TRUE);
2009-11-04 15:11:44 -08:00
}
struct json *
ovsdb_condition_to_json(const struct ovsdb_condition *cnd)
{
struct json **clauses;
size_t i;
clauses = xmalloc(cnd->n_clauses * sizeof *clauses);
for (i = 0; i < cnd->n_clauses; i++) {
clauses[i] = ovsdb_clause_to_json(&cnd->clauses[i]);
}
return json_array_create(clauses, cnd->n_clauses);
}
static bool
ovsdb_clause_evaluate(const struct ovsdb_datum *fields,
const struct ovsdb_clause *c,
unsigned int index_map[])
{
const struct ovsdb_datum *field = &fields[index_map ?
index_map[c->column->index] :
c->column->index];
const struct ovsdb_datum *arg = &c->arg;
const struct ovsdb_type *type = &c->column->type;
if (c->function == OVSDB_F_TRUE ||
c->function == OVSDB_F_FALSE) {
return c->function == OVSDB_F_TRUE;
}
if (ovsdb_type_is_optional_scalar(type)
&& field->n == 0
&& (c->function == OVSDB_F_LT ||
c->function == OVSDB_F_LE ||
c->function == OVSDB_F_GT ||
c->function == OVSDB_F_GE)) {
return false;
} else if ((ovsdb_type_is_scalar(type)
|| ovsdb_type_is_optional_scalar(type))
&& field->n == 1
&& arg->n == 1) {
int cmp = ovsdb_atom_compare_3way(&field->keys[0], &arg->keys[0],
type->key.type);
switch (c->function) {
case OVSDB_F_LT:
return cmp < 0;
case OVSDB_F_LE:
return cmp <= 0;
case OVSDB_F_EQ:
case OVSDB_F_INCLUDES:
return cmp == 0;
case OVSDB_F_NE:
case OVSDB_F_EXCLUDES:
return cmp != 0;
case OVSDB_F_GE:
return cmp >= 0;
case OVSDB_F_GT:
return cmp > 0;
case OVSDB_F_TRUE:
case OVSDB_F_FALSE:
OVS_NOT_REACHED();
}
} else {
switch (c->function) {
case OVSDB_F_EQ:
return ovsdb_datum_equals(field, arg, type);
case OVSDB_F_NE:
return !ovsdb_datum_equals(field, arg, type);
case OVSDB_F_INCLUDES:
return ovsdb_datum_includes_all(arg, field, type);
case OVSDB_F_EXCLUDES:
return ovsdb_datum_excludes_all(arg, field, type);
case OVSDB_F_LT:
case OVSDB_F_LE:
case OVSDB_F_GE:
case OVSDB_F_GT:
case OVSDB_F_TRUE:
case OVSDB_F_FALSE:
OVS_NOT_REACHED();
}
}
OVS_NOT_REACHED();
}
static void
ovsdb_clause_clone(struct ovsdb_clause *new, struct ovsdb_clause *old)
{
new->function = old->function;
new->column = old->column;
ovsdb_datum_clone(&new->arg, &old->arg);
}
2009-11-04 15:11:44 -08:00
bool
ovsdb_condition_match_every_clause(const struct ovsdb_row *row,
const struct ovsdb_condition *cnd)
2009-11-04 15:11:44 -08:00
{
size_t i;
for (i = 0; i < cnd->n_clauses; i++) {
if (!ovsdb_clause_evaluate(row->fields, &cnd->clauses[i], NULL)) {
return false;
2009-11-04 15:11:44 -08:00
}
}
return true;
}
static bool
ovsdb_condition_match_any_clause_optimized(const struct ovsdb_datum *row_datum,
const struct ovsdb_condition *cnd,
unsigned int index_map[])
{
if (ovsdb_condition_is_true(cnd)) {
return true;
}
struct shash_node *node;
SHASH_FOR_EACH (node, &cnd->o_columns) {
struct ovsdb_o_column *o_column = node->data;
const struct ovsdb_column *column = o_column->column;
const struct ovsdb_datum *arg = &row_datum[index_map ?
index_map[column->index] :
column->index];
uint32_t hash = ovsdb_datum_hash(arg, &column->type, 0);
struct ovsdb_o_clause *o_clause;
HMAP_FOR_EACH_WITH_HASH(o_clause, hmap_node, hash, &o_column->o_clauses) {
if (ovsdb_datum_equals(arg, o_clause->arg, &column->type)) {
return true;
}
}
}
return false;
}
/* Returns true if condition evaluation of one of the clauses is
* true. index_map[] is an optional array that if exists indicates a mapping
* between indexing row_datum to the indexes in ovsdb_column */
bool
ovsdb_condition_match_any_clause(const struct ovsdb_datum *row_datum,
const struct ovsdb_condition *cnd,
unsigned int index_map[])
{
size_t i;
if (cnd->optimized) {
return ovsdb_condition_match_any_clause_optimized(row_datum, cnd,
index_map);
}
for (i = 0; i < cnd->n_clauses; i++) {
if (ovsdb_clause_evaluate(row_datum, &cnd->clauses[i], index_map)) {
return true;
}
}
return false;
}
2009-11-04 15:11:44 -08:00
void
ovsdb_condition_destroy(struct ovsdb_condition *cnd)
{
size_t i;
for (i = 0; i < cnd->n_clauses; i++) {
ovsdb_clause_free(&cnd->clauses[i]);
}
free(cnd->clauses);
cnd->n_clauses = 0;
ovsdb_condition_optimize_destroy(cnd);
}
void
ovsdb_condition_init(struct ovsdb_condition *cnd)
{
cnd->clauses = NULL;
cnd->n_clauses = 0;
cnd->optimized = true;
shash_init(&cnd->o_columns);
}
bool
ovsdb_condition_empty(const struct ovsdb_condition *cnd)
{
return cnd->n_clauses == 0;
}
int
ovsdb_condition_cmp_3way(const struct ovsdb_condition *a,
const struct ovsdb_condition *b)
{
size_t i;
int res;
if (a->n_clauses != b->n_clauses) {
return a->n_clauses < b->n_clauses ? -1 : 1;
}
/* We assume clauses are sorted */
for (i = 0; i < a->n_clauses; i++) {
res = (compare_clauses_3way_with_data(&a->clauses[i], &b->clauses[i]));
if (res != 0) {
return res;
}
}
return 0;
}
ovsdb: condition: Process condition changes incrementally. In most cases, after the condition change request, the new condition is the same as old one plus minus a few clauses. Today, ovsdb-server will evaluate every database row against all the old clauses and then against all the new clauses in order to tell if an update should be generated. For example, every time a new port is added, ovn-controller adds two new clauses to conditions for a Port_Binding table. And this condition may grow significantly in size making addition of every new port heavier on the server side. The difference between conditions is not larger and, likely, significantly smaller than old and new conditions combined. And if the row doesn't match clauses that are different between old and new conditions, that row should not be part of the update. It either matches both old and new, or it doesn't match either of them. If the row matches some clauses in the difference, then we need to perform a full match against old and new in order to tell if it should be added/removed/modified. This is necessary because different clauses may select same rows. Let's generate the condition difference and use it to avoid evaluation of all the clauses for rows not affected by the condition change. Testing shows 70% reduction in total CPU time in ovn-heater's 120-node density-light test with conditional monitoring. Average CPU usage during the test phase went down from frequent 100% spikes to just 6-8%. Note: This will not help with new connections, or re-connections, or new monitor requests after database conversion. ovsdb-server will still evaluate every database row against every clause in the condition in these cases. So, it's still important to not have too many clauses in conditions for large tables. Reviewed-by: Simon Horman <simon.horman@corigine.com> Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
2023-05-26 19:18:43 +02:00
/* Given conditions 'a' and 'b', composes a new condition 'diff' that contains
* clauses that are present in one of the conditions, but not in the other.
*
* If some data doesn't match the resulted 'diff' condition, that means one of:
* 1. The data matches both 'a' and 'b'.
* 2. The data does not match either 'a' or 'b'.
*
* However, that is not true if one of the original conditions is a trivial
* True or False. In this case the function will currently just return an
* empty (True) condition. */
void
ovsdb_condition_diff(struct ovsdb_condition *diff,
const struct ovsdb_condition *a,
const struct ovsdb_condition *b)
{
size_t i, j;
int cmp;
ovsdb_condition_init(diff);
if (ovsdb_condition_is_trivial(a) || ovsdb_condition_is_trivial(b)) {
return;
}
diff->clauses = xcalloc(a->n_clauses + b->n_clauses,
sizeof *diff->clauses);
/* Clauses are sorted. */
for (i = j = 0; i < a->n_clauses && j < b->n_clauses;) {
cmp = compare_clauses_3way_with_data(&a->clauses[i], &b->clauses[j]);
if (cmp < 0) {
ovsdb_clause_clone(&diff->clauses[diff->n_clauses++],
&a->clauses[i++]);
} else if (cmp > 0) {
ovsdb_clause_clone(&diff->clauses[diff->n_clauses++],
&b->clauses[j++]);
} else {
i++;
j++;
}
}
for (; i < a->n_clauses; i++) {
ovsdb_clause_clone(&diff->clauses[diff->n_clauses++],
&a->clauses[i]);
}
for (; j < b->n_clauses; j++) {
ovsdb_clause_clone(&diff->clauses[diff->n_clauses++],
&b->clauses[j]);
}
if (diff->n_clauses) {
diff->optimized = a->optimized && b->optimized;
if (diff->optimized) {
ovsdb_condition_optimize(diff);
}
} else {
free(diff->clauses);
diff->clauses = NULL;
ovsdb: condition: Process condition changes incrementally. In most cases, after the condition change request, the new condition is the same as old one plus minus a few clauses. Today, ovsdb-server will evaluate every database row against all the old clauses and then against all the new clauses in order to tell if an update should be generated. For example, every time a new port is added, ovn-controller adds two new clauses to conditions for a Port_Binding table. And this condition may grow significantly in size making addition of every new port heavier on the server side. The difference between conditions is not larger and, likely, significantly smaller than old and new conditions combined. And if the row doesn't match clauses that are different between old and new conditions, that row should not be part of the update. It either matches both old and new, or it doesn't match either of them. If the row matches some clauses in the difference, then we need to perform a full match against old and new in order to tell if it should be added/removed/modified. This is necessary because different clauses may select same rows. Let's generate the condition difference and use it to avoid evaluation of all the clauses for rows not affected by the condition change. Testing shows 70% reduction in total CPU time in ovn-heater's 120-node density-light test with conditional monitoring. Average CPU usage during the test phase went down from frequent 100% spikes to just 6-8%. Note: This will not help with new connections, or re-connections, or new monitor requests after database conversion. ovsdb-server will still evaluate every database row against every clause in the condition in these cases. So, it's still important to not have too many clauses in conditions for large tables. Reviewed-by: Simon Horman <simon.horman@corigine.com> Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
2023-05-26 19:18:43 +02:00
}
}
void
ovsdb_condition_clone(struct ovsdb_condition *to,
const struct ovsdb_condition *from)
{
size_t i;
ovsdb_condition_init(to);
to->clauses = xzalloc(from->n_clauses * sizeof *to->clauses);
for (i = 0; i < from->n_clauses; i++) {
ovsdb_clause_clone(&to->clauses[i], &from->clauses[i]);
}
to->n_clauses = from->n_clauses;
to->optimized = from->optimized;
if (to->optimized) {
ovsdb_condition_optimize(to);
}
}
/* Return true if ovsdb_condition_match_any_clause() will return true on
* any row */
bool
ovsdb_condition_is_true(const struct ovsdb_condition *cond)
{
return (!cond->n_clauses ||
(cond->n_clauses >= 1 && (cond->clauses[0].function == OVSDB_F_TRUE)) ||
(cond->n_clauses >= 2 && (cond->clauses[1].function == OVSDB_F_TRUE)));
}
bool
ovsdb_condition_is_false(const struct ovsdb_condition *cond)
{
return ((cond->n_clauses == 1) &&
(cond->clauses[0].function == OVSDB_F_FALSE));
}
const struct ovsdb_column **
ovsdb_condition_get_columns(const struct ovsdb_condition *cond,
size_t *n_columns)
{
const struct ovsdb_column **columns;
size_t i;
columns = xmalloc(cond->n_clauses * sizeof *columns);
for (i = 0; i < cond->n_clauses; i++) {
columns[i] = cond->clauses[i].column;
}
*n_columns = i;
return columns;
2009-11-04 15:11:44 -08:00
}