numam-dpdk/lib/librte_acl/rte_acl.c
Konstantin Ananyev 62945e029e acl: introduce config parameter for performance/space trade-off
If at build phase we don't make any trie splitting,
then temporary build structures and resulting RT structure might be
much bigger than current.
>From other side - having just one trie instead of multiple can speedup
search quite significantly.
>From my measurements on rule-sets with ~10K rules:
RT table up to 8 times bigger, classify() up to 80% faster
than current implementation.
To make it possible for the user to decide about performance/space trade-off -
new parameter for build config structure (max_size) is introduced.
Setting it to the value greater than zero, instructs  rte_acl_build() to:
- make sure that size of RT table wouldn't exceed given value.
- attempt to minimise number of tries in the table.
Setting it to zero maintains current behaviour.
That introduces a minor change in the public API, but I think the possible
performance gain is too big to ignore it.

Signed-off-by: Konstantin Ananyev <konstantin.ananyev@intel.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
2015-01-28 17:11:26 +01:00

550 lines
14 KiB
C

/*-
* BSD LICENSE
*
* Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in
* the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Intel Corporation nor the names of its
* contributors may be used to endorse or promote products derived
* from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include <rte_acl.h>
#include "acl.h"
#define BIT_SIZEOF(x) (sizeof(x) * CHAR_BIT)
TAILQ_HEAD(rte_acl_list, rte_tailq_entry);
/*
* If the compiler doesn't support AVX2 instructions,
* then the dummy one would be used instead for AVX2 classify method.
*/
int __attribute__ ((weak))
rte_acl_classify_avx2(__rte_unused const struct rte_acl_ctx *ctx,
__rte_unused const uint8_t **data,
__rte_unused uint32_t *results,
__rte_unused uint32_t num,
__rte_unused uint32_t categories)
{
return -ENOTSUP;
}
static const rte_acl_classify_t classify_fns[] = {
[RTE_ACL_CLASSIFY_DEFAULT] = rte_acl_classify_scalar,
[RTE_ACL_CLASSIFY_SCALAR] = rte_acl_classify_scalar,
[RTE_ACL_CLASSIFY_SSE] = rte_acl_classify_sse,
[RTE_ACL_CLASSIFY_AVX2] = rte_acl_classify_avx2,
};
/* by default, use always available scalar code path. */
static enum rte_acl_classify_alg rte_acl_default_classify =
RTE_ACL_CLASSIFY_SCALAR;
static void
rte_acl_set_default_classify(enum rte_acl_classify_alg alg)
{
rte_acl_default_classify = alg;
}
extern int
rte_acl_set_ctx_classify(struct rte_acl_ctx *ctx, enum rte_acl_classify_alg alg)
{
if (ctx == NULL || (uint32_t)alg >= RTE_DIM(classify_fns))
return -EINVAL;
ctx->alg = alg;
return 0;
}
/*
* Select highest available classify method as default one.
* Note that CLASSIFY_AVX2 should be set as a default only
* if both conditions are met:
* at build time compiler supports AVX2 and target cpu supports AVX2.
*/
static void __attribute__((constructor))
rte_acl_init(void)
{
enum rte_acl_classify_alg alg = RTE_ACL_CLASSIFY_DEFAULT;
#ifdef CC_AVX2_SUPPORT
if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_AVX2))
alg = RTE_ACL_CLASSIFY_AVX2;
else if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE4_1))
#else
if (rte_cpu_get_flag_enabled(RTE_CPUFLAG_SSE4_1))
#endif
alg = RTE_ACL_CLASSIFY_SSE;
rte_acl_set_default_classify(alg);
}
int
rte_acl_classify_alg(const struct rte_acl_ctx *ctx, const uint8_t **data,
uint32_t *results, uint32_t num, uint32_t categories,
enum rte_acl_classify_alg alg)
{
if (categories != 1 &&
((RTE_ACL_RESULTS_MULTIPLIER - 1) & categories) != 0)
return -EINVAL;
return classify_fns[alg](ctx, data, results, num, categories);
}
int
rte_acl_classify(const struct rte_acl_ctx *ctx, const uint8_t **data,
uint32_t *results, uint32_t num, uint32_t categories)
{
return rte_acl_classify_alg(ctx, data, results, num, categories,
ctx->alg);
}
struct rte_acl_ctx *
rte_acl_find_existing(const char *name)
{
struct rte_acl_ctx *ctx = NULL;
struct rte_acl_list *acl_list;
struct rte_tailq_entry *te;
/* check that we have an initialised tail queue */
acl_list = RTE_TAILQ_LOOKUP_BY_IDX(RTE_TAILQ_ACL, rte_acl_list);
if (acl_list == NULL) {
rte_errno = E_RTE_NO_TAILQ;
return NULL;
}
rte_rwlock_read_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_FOREACH(te, acl_list, next) {
ctx = (struct rte_acl_ctx *) te->data;
if (strncmp(name, ctx->name, sizeof(ctx->name)) == 0)
break;
}
rte_rwlock_read_unlock(RTE_EAL_TAILQ_RWLOCK);
if (te == NULL) {
rte_errno = ENOENT;
return NULL;
}
return ctx;
}
void
rte_acl_free(struct rte_acl_ctx *ctx)
{
struct rte_acl_list *acl_list;
struct rte_tailq_entry *te;
if (ctx == NULL)
return;
/* check that we have an initialised tail queue */
acl_list = RTE_TAILQ_LOOKUP_BY_IDX(RTE_TAILQ_ACL, rte_acl_list);
if (acl_list == NULL) {
rte_errno = E_RTE_NO_TAILQ;
return;
}
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
/* find our tailq entry */
TAILQ_FOREACH(te, acl_list, next) {
if (te->data == (void *) ctx)
break;
}
if (te == NULL) {
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
return;
}
TAILQ_REMOVE(acl_list, te, next);
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
rte_free(ctx->mem);
rte_free(ctx);
rte_free(te);
}
struct rte_acl_ctx *
rte_acl_create(const struct rte_acl_param *param)
{
size_t sz;
struct rte_acl_ctx *ctx;
struct rte_acl_list *acl_list;
struct rte_tailq_entry *te;
char name[sizeof(ctx->name)];
/* check that we have an initialised tail queue */
acl_list = RTE_TAILQ_LOOKUP_BY_IDX(RTE_TAILQ_ACL, rte_acl_list);
if (acl_list == NULL) {
rte_errno = E_RTE_NO_TAILQ;
return NULL;
}
/* check that input parameters are valid. */
if (param == NULL || param->name == NULL) {
rte_errno = EINVAL;
return NULL;
}
snprintf(name, sizeof(name), "ACL_%s", param->name);
/* calculate amount of memory required for pattern set. */
sz = sizeof(*ctx) + param->max_rule_num * param->rule_size;
/* get EAL TAILQ lock. */
rte_rwlock_write_lock(RTE_EAL_TAILQ_RWLOCK);
/* if we already have one with that name */
TAILQ_FOREACH(te, acl_list, next) {
ctx = (struct rte_acl_ctx *) te->data;
if (strncmp(param->name, ctx->name, sizeof(ctx->name)) == 0)
break;
}
/* if ACL with such name doesn't exist, then create a new one. */
if (te == NULL) {
ctx = NULL;
te = rte_zmalloc("ACL_TAILQ_ENTRY", sizeof(*te), 0);
if (te == NULL) {
RTE_LOG(ERR, ACL, "Cannot allocate tailq entry!\n");
goto exit;
}
ctx = rte_zmalloc_socket(name, sz, RTE_CACHE_LINE_SIZE, param->socket_id);
if (ctx == NULL) {
RTE_LOG(ERR, ACL,
"allocation of %zu bytes on socket %d for %s failed\n",
sz, param->socket_id, name);
rte_free(te);
goto exit;
}
/* init new allocated context. */
ctx->rules = ctx + 1;
ctx->max_rules = param->max_rule_num;
ctx->rule_sz = param->rule_size;
ctx->socket_id = param->socket_id;
ctx->alg = rte_acl_default_classify;
snprintf(ctx->name, sizeof(ctx->name), "%s", param->name);
te->data = (void *) ctx;
TAILQ_INSERT_TAIL(acl_list, te, next);
}
exit:
rte_rwlock_write_unlock(RTE_EAL_TAILQ_RWLOCK);
return ctx;
}
static int
acl_add_rules(struct rte_acl_ctx *ctx, const void *rules, uint32_t num)
{
uint8_t *pos;
if (num + ctx->num_rules > ctx->max_rules)
return -ENOMEM;
pos = ctx->rules;
pos += ctx->rule_sz * ctx->num_rules;
memcpy(pos, rules, num * ctx->rule_sz);
ctx->num_rules += num;
return 0;
}
static int
acl_check_rule(const struct rte_acl_rule_data *rd)
{
if ((rd->category_mask & LEN2MASK(RTE_ACL_MAX_CATEGORIES)) == 0 ||
rd->priority > RTE_ACL_MAX_PRIORITY ||
rd->priority < RTE_ACL_MIN_PRIORITY ||
rd->userdata == RTE_ACL_INVALID_USERDATA)
return -EINVAL;
return 0;
}
int
rte_acl_add_rules(struct rte_acl_ctx *ctx, const struct rte_acl_rule *rules,
uint32_t num)
{
const struct rte_acl_rule *rv;
uint32_t i;
int32_t rc;
if (ctx == NULL || rules == NULL || 0 == ctx->rule_sz)
return -EINVAL;
for (i = 0; i != num; i++) {
rv = (const struct rte_acl_rule *)
((uintptr_t)rules + i * ctx->rule_sz);
rc = acl_check_rule(&rv->data);
if (rc != 0) {
RTE_LOG(ERR, ACL, "%s(%s): rule #%u is invalid\n",
__func__, ctx->name, i + 1);
return rc;
}
}
return acl_add_rules(ctx, rules, num);
}
/*
* Reset all rules.
* Note that RT structures are not affected.
*/
void
rte_acl_reset_rules(struct rte_acl_ctx *ctx)
{
if (ctx != NULL)
ctx->num_rules = 0;
}
/*
* Reset all rules and destroys RT structures.
*/
void
rte_acl_reset(struct rte_acl_ctx *ctx)
{
if (ctx != NULL) {
rte_acl_reset_rules(ctx);
rte_acl_build(ctx, &ctx->config);
}
}
/*
* Dump ACL context to the stdout.
*/
void
rte_acl_dump(const struct rte_acl_ctx *ctx)
{
if (!ctx)
return;
printf("acl context <%s>@%p\n", ctx->name, ctx);
printf(" socket_id=%"PRId32"\n", ctx->socket_id);
printf(" alg=%"PRId32"\n", ctx->alg);
printf(" max_rules=%"PRIu32"\n", ctx->max_rules);
printf(" rule_size=%"PRIu32"\n", ctx->rule_sz);
printf(" num_rules=%"PRIu32"\n", ctx->num_rules);
printf(" num_categories=%"PRIu32"\n", ctx->num_categories);
printf(" num_tries=%"PRIu32"\n", ctx->num_tries);
}
/*
* Dump all ACL contexts to the stdout.
*/
void
rte_acl_list_dump(void)
{
struct rte_acl_ctx *ctx;
struct rte_acl_list *acl_list;
struct rte_tailq_entry *te;
/* check that we have an initialised tail queue */
acl_list = RTE_TAILQ_LOOKUP_BY_IDX(RTE_TAILQ_ACL, rte_acl_list);
if (acl_list == NULL) {
rte_errno = E_RTE_NO_TAILQ;
return;
}
rte_rwlock_read_lock(RTE_EAL_TAILQ_RWLOCK);
TAILQ_FOREACH(te, acl_list, next) {
ctx = (struct rte_acl_ctx *) te->data;
rte_acl_dump(ctx);
}
rte_rwlock_read_unlock(RTE_EAL_TAILQ_RWLOCK);
}
/*
* Support for legacy ipv4vlan rules.
*/
RTE_ACL_RULE_DEF(acl_ipv4vlan_rule, RTE_ACL_IPV4VLAN_NUM_FIELDS);
static int
acl_ipv4vlan_check_rule(const struct rte_acl_ipv4vlan_rule *rule)
{
if (rule->src_port_low > rule->src_port_high ||
rule->dst_port_low > rule->dst_port_high ||
rule->src_mask_len > BIT_SIZEOF(rule->src_addr) ||
rule->dst_mask_len > BIT_SIZEOF(rule->dst_addr))
return -EINVAL;
return acl_check_rule(&rule->data);
}
static void
acl_ipv4vlan_convert_rule(const struct rte_acl_ipv4vlan_rule *ri,
struct acl_ipv4vlan_rule *ro)
{
ro->data = ri->data;
ro->field[RTE_ACL_IPV4VLAN_PROTO_FIELD].value.u8 = ri->proto;
ro->field[RTE_ACL_IPV4VLAN_VLAN1_FIELD].value.u16 = ri->vlan;
ro->field[RTE_ACL_IPV4VLAN_VLAN2_FIELD].value.u16 = ri->domain;
ro->field[RTE_ACL_IPV4VLAN_SRC_FIELD].value.u32 = ri->src_addr;
ro->field[RTE_ACL_IPV4VLAN_DST_FIELD].value.u32 = ri->dst_addr;
ro->field[RTE_ACL_IPV4VLAN_SRCP_FIELD].value.u16 = ri->src_port_low;
ro->field[RTE_ACL_IPV4VLAN_DSTP_FIELD].value.u16 = ri->dst_port_low;
ro->field[RTE_ACL_IPV4VLAN_PROTO_FIELD].mask_range.u8 = ri->proto_mask;
ro->field[RTE_ACL_IPV4VLAN_VLAN1_FIELD].mask_range.u16 = ri->vlan_mask;
ro->field[RTE_ACL_IPV4VLAN_VLAN2_FIELD].mask_range.u16 =
ri->domain_mask;
ro->field[RTE_ACL_IPV4VLAN_SRC_FIELD].mask_range.u32 =
ri->src_mask_len;
ro->field[RTE_ACL_IPV4VLAN_DST_FIELD].mask_range.u32 = ri->dst_mask_len;
ro->field[RTE_ACL_IPV4VLAN_SRCP_FIELD].mask_range.u16 =
ri->src_port_high;
ro->field[RTE_ACL_IPV4VLAN_DSTP_FIELD].mask_range.u16 =
ri->dst_port_high;
}
int
rte_acl_ipv4vlan_add_rules(struct rte_acl_ctx *ctx,
const struct rte_acl_ipv4vlan_rule *rules,
uint32_t num)
{
int32_t rc;
uint32_t i;
struct acl_ipv4vlan_rule rv;
if (ctx == NULL || rules == NULL || ctx->rule_sz != sizeof(rv))
return -EINVAL;
/* check input rules. */
for (i = 0; i != num; i++) {
rc = acl_ipv4vlan_check_rule(rules + i);
if (rc != 0) {
RTE_LOG(ERR, ACL, "%s(%s): rule #%u is invalid\n",
__func__, ctx->name, i + 1);
return rc;
}
}
if (num + ctx->num_rules > ctx->max_rules)
return -ENOMEM;
/* perform conversion to the internal format and add to the context. */
for (i = 0, rc = 0; i != num && rc == 0; i++) {
acl_ipv4vlan_convert_rule(rules + i, &rv);
rc = acl_add_rules(ctx, &rv, 1);
}
return rc;
}
static void
acl_ipv4vlan_config(struct rte_acl_config *cfg,
const uint32_t layout[RTE_ACL_IPV4VLAN_NUM],
uint32_t num_categories)
{
static const struct rte_acl_field_def
ipv4_defs[RTE_ACL_IPV4VLAN_NUM_FIELDS] = {
{
.type = RTE_ACL_FIELD_TYPE_BITMASK,
.size = sizeof(uint8_t),
.field_index = RTE_ACL_IPV4VLAN_PROTO_FIELD,
.input_index = RTE_ACL_IPV4VLAN_PROTO,
},
{
.type = RTE_ACL_FIELD_TYPE_BITMASK,
.size = sizeof(uint16_t),
.field_index = RTE_ACL_IPV4VLAN_VLAN1_FIELD,
.input_index = RTE_ACL_IPV4VLAN_VLAN,
},
{
.type = RTE_ACL_FIELD_TYPE_BITMASK,
.size = sizeof(uint16_t),
.field_index = RTE_ACL_IPV4VLAN_VLAN2_FIELD,
.input_index = RTE_ACL_IPV4VLAN_VLAN,
},
{
.type = RTE_ACL_FIELD_TYPE_MASK,
.size = sizeof(uint32_t),
.field_index = RTE_ACL_IPV4VLAN_SRC_FIELD,
.input_index = RTE_ACL_IPV4VLAN_SRC,
},
{
.type = RTE_ACL_FIELD_TYPE_MASK,
.size = sizeof(uint32_t),
.field_index = RTE_ACL_IPV4VLAN_DST_FIELD,
.input_index = RTE_ACL_IPV4VLAN_DST,
},
{
.type = RTE_ACL_FIELD_TYPE_RANGE,
.size = sizeof(uint16_t),
.field_index = RTE_ACL_IPV4VLAN_SRCP_FIELD,
.input_index = RTE_ACL_IPV4VLAN_PORTS,
},
{
.type = RTE_ACL_FIELD_TYPE_RANGE,
.size = sizeof(uint16_t),
.field_index = RTE_ACL_IPV4VLAN_DSTP_FIELD,
.input_index = RTE_ACL_IPV4VLAN_PORTS,
},
};
memcpy(&cfg->defs, ipv4_defs, sizeof(ipv4_defs));
cfg->num_fields = RTE_DIM(ipv4_defs);
cfg->defs[RTE_ACL_IPV4VLAN_PROTO_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_PROTO];
cfg->defs[RTE_ACL_IPV4VLAN_VLAN1_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_VLAN];
cfg->defs[RTE_ACL_IPV4VLAN_VLAN2_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_VLAN] +
cfg->defs[RTE_ACL_IPV4VLAN_VLAN1_FIELD].size;
cfg->defs[RTE_ACL_IPV4VLAN_SRC_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_SRC];
cfg->defs[RTE_ACL_IPV4VLAN_DST_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_DST];
cfg->defs[RTE_ACL_IPV4VLAN_SRCP_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_PORTS];
cfg->defs[RTE_ACL_IPV4VLAN_DSTP_FIELD].offset =
layout[RTE_ACL_IPV4VLAN_PORTS] +
cfg->defs[RTE_ACL_IPV4VLAN_SRCP_FIELD].size;
cfg->num_categories = num_categories;
}
int
rte_acl_ipv4vlan_build(struct rte_acl_ctx *ctx,
const uint32_t layout[RTE_ACL_IPV4VLAN_NUM],
uint32_t num_categories)
{
struct rte_acl_config cfg;
if (ctx == NULL || layout == NULL)
return -EINVAL;
memset(&cfg, 0, sizeof(cfg));
acl_ipv4vlan_config(&cfg, layout, num_categories);
return rte_acl_build(ctx, &cfg);
}