#include <stdint.h>
#include <string.h>
-/* Verbs header. */
-/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */
-#ifdef PEDANTIC
-#pragma GCC diagnostic ignored "-Wpedantic"
-#endif
-#include <infiniband/verbs.h>
-#ifdef PEDANTIC
-#pragma GCC diagnostic error "-Wpedantic"
-#endif
-
#include <rte_common.h>
#include <rte_ether.h>
-#include <rte_ethdev_driver.h>
+#include <ethdev_driver.h>
#include <rte_flow.h>
#include <rte_flow_driver.h>
#include <rte_malloc.h>
#include <rte_ip.h>
-#include "mlx5.h"
+#include <mlx5_glue.h>
+#include <mlx5_prm.h>
+#include <mlx5_malloc.h>
+
#include "mlx5_defs.h"
+#include "mlx5.h"
#include "mlx5_flow.h"
-#include "mlx5_glue.h"
-#include "mlx5_prm.h"
-#include "mlx5_rxtx.h"
+#include "mlx5_rx.h"
#define VERBS_SPEC_INNER(item_flags) \
(!!((item_flags) & MLX5_FLOW_LAYER_TUNNEL) ? IBV_FLOW_SPEC_INNER : 0)
+/* Verbs specification header. */
+struct ibv_spec_header {
+ enum ibv_flow_spec_type type;
+ uint16_t size;
+};
+
+/**
+ * Discover the maximum number of priority available.
+ *
+ * @param[in] dev
+ * Pointer to the Ethernet device structure.
+ * @param[in] vprio
+ * Expected result variants.
+ * @param[in] vprio_n
+ * Number of entries in @p vprio array.
+ * @return
+ * Number of supported flow priority on success, a negative errno
+ * value otherwise and rte_errno is set.
+ */
+static int
+flow_verbs_discover_priorities(struct rte_eth_dev *dev,
+ const uint16_t *vprio, int vprio_n)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct {
+ struct ibv_flow_attr attr;
+ struct ibv_flow_spec_eth eth;
+ struct ibv_flow_spec_action_drop drop;
+ } flow_attr = {
+ .attr = {
+ .num_of_specs = 2,
+ .port = (uint8_t)priv->dev_port,
+ },
+ .eth = {
+ .type = IBV_FLOW_SPEC_ETH,
+ .size = sizeof(struct ibv_flow_spec_eth),
+ },
+ .drop = {
+ .size = sizeof(struct ibv_flow_spec_action_drop),
+ .type = IBV_FLOW_SPEC_ACTION_DROP,
+ },
+ };
+ struct ibv_flow *flow;
+ struct mlx5_hrxq *drop = priv->drop_queue.hrxq;
+ int i;
+ int priority = 0;
+
+#if defined(HAVE_MLX5DV_DR_DEVX_PORT) || defined(HAVE_MLX5DV_DR_DEVX_PORT_V35)
+ /* If DevX supported, driver must support 16 verbs flow priorities. */
+ priority = 16;
+ goto out;
+#endif
+ if (!drop->qp) {
+ rte_errno = ENOTSUP;
+ return -rte_errno;
+ }
+ for (i = 0; i != vprio_n; i++) {
+ flow_attr.attr.priority = vprio[i] - 1;
+ flow = mlx5_glue->create_flow(drop->qp, &flow_attr.attr);
+ if (!flow)
+ break;
+ claim_zero(mlx5_glue->destroy_flow(flow));
+ priority = vprio[i];
+ }
+#if defined(HAVE_MLX5DV_DR_DEVX_PORT) || defined(HAVE_MLX5DV_DR_DEVX_PORT_V35)
+out:
+#endif
+ DRV_LOG(INFO, "port %u supported flow priorities:"
+ " 0-%d for ingress or egress root table,"
+ " 0-%d for non-root table or transfer root table.",
+ dev->data->port_id, priority - 2,
+ MLX5_NON_ROOT_FLOW_MAX_PRIO - 1);
+ return priority;
+}
+
+/**
+ * Get Verbs flow counter by index.
+ *
+ * @param[in] dev
+ * Pointer to the Ethernet device structure.
+ * @param[in] idx
+ * mlx5 flow counter index in the container.
+ * @param[out] ppool
+ * mlx5 flow counter pool in the container,
+ *
+ * @return
+ * A pointer to the counter, NULL otherwise.
+ */
+static struct mlx5_flow_counter *
+flow_verbs_counter_get_by_idx(struct rte_eth_dev *dev,
+ uint32_t idx,
+ struct mlx5_flow_counter_pool **ppool)
+{
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_flow_counter_mng *cmng = &priv->sh->cmng;
+ struct mlx5_flow_counter_pool *pool;
+
+ idx = (idx - 1) & (MLX5_CNT_SHARED_OFFSET - 1);
+ pool = cmng->pools[idx / MLX5_COUNTERS_PER_POOL];
+ MLX5_ASSERT(pool);
+ if (ppool)
+ *ppool = pool;
+ return MLX5_POOL_GET_CNT(pool, idx % MLX5_COUNTERS_PER_POOL);
+}
+
/**
* Create Verbs flow counter with Verbs library.
*
{
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42)
struct mlx5_priv *priv = dev->data->dev_private;
- struct ibv_context *ctx = priv->sh->ctx;
+ struct ibv_context *ctx = priv->sh->cdev->ctx;
struct ibv_counter_set_init_attr init = {
- .counter_set_id = counter->id};
+ .counter_set_id = counter->shared_info.id};
- counter->cs = mlx5_glue->create_counter_set(ctx, &init);
- if (!counter->cs) {
+ counter->dcs_when_free = mlx5_glue->create_counter_set(ctx, &init);
+ if (!counter->dcs_when_free) {
rte_errno = ENOTSUP;
return -ENOTSUP;
}
return 0;
#elif defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
struct mlx5_priv *priv = dev->data->dev_private;
- struct ibv_context *ctx = priv->sh->ctx;
+ struct ibv_context *ctx = priv->sh->cdev->ctx;
struct ibv_counters_init_attr init = {0};
struct ibv_counter_attach_attr attach;
int ret;
memset(&attach, 0, sizeof(attach));
- counter->cs = mlx5_glue->create_counters(ctx, &init);
- if (!counter->cs) {
+ counter->dcs_when_free = mlx5_glue->create_counters(ctx, &init);
+ if (!counter->dcs_when_free) {
rte_errno = ENOTSUP;
return -ENOTSUP;
}
attach.counter_desc = IBV_COUNTER_PACKETS;
attach.index = 0;
- ret = mlx5_glue->attach_counters(counter->cs, &attach, NULL);
+ ret = mlx5_glue->attach_counters(counter->dcs_when_free, &attach, NULL);
if (!ret) {
attach.counter_desc = IBV_COUNTER_BYTES;
attach.index = 1;
ret = mlx5_glue->attach_counters
- (counter->cs, &attach, NULL);
+ (counter->dcs_when_free, &attach, NULL);
}
if (ret) {
- claim_zero(mlx5_glue->destroy_counters(counter->cs));
- counter->cs = NULL;
+ claim_zero(mlx5_glue->destroy_counters(counter->dcs_when_free));
+ counter->dcs_when_free = NULL;
rte_errno = ret;
return -ret;
}
*
* @param[in] dev
* Pointer to the Ethernet device structure.
- * @param[in] shared
- * Indicate if this counter is shared with other flows.
* @param[in] id
* Counter identifier.
*
* @return
- * A pointer to the counter, NULL otherwise and rte_errno is set.
+ * Index to the counter, 0 otherwise and rte_errno is set.
*/
-static struct mlx5_flow_counter *
-flow_verbs_counter_new(struct rte_eth_dev *dev, uint32_t shared, uint32_t id)
+static uint32_t
+flow_verbs_counter_new(struct rte_eth_dev *dev, uint32_t id __rte_unused)
{
struct mlx5_priv *priv = dev->data->dev_private;
- struct mlx5_flow_counter *cnt;
+ struct mlx5_flow_counter_mng *cmng = &priv->sh->cmng;
+ struct mlx5_flow_counter_pool *pool = NULL;
+ struct mlx5_flow_counter *cnt = NULL;
+ uint32_t n_valid = cmng->n_valid;
+ uint32_t pool_idx, cnt_idx;
+ uint32_t i;
int ret;
- if (shared) {
- TAILQ_FOREACH(cnt, &priv->sh->cmng.flow_counters, next) {
- if (cnt->shared && cnt->id == id) {
- cnt->ref_cnt++;
- return cnt;
- }
- }
+ for (pool_idx = 0; pool_idx < n_valid; ++pool_idx) {
+ pool = cmng->pools[pool_idx];
+ if (!pool)
+ continue;
+ cnt = TAILQ_FIRST(&pool->counters[0]);
+ if (cnt)
+ break;
}
- cnt = rte_calloc(__func__, 1, sizeof(*cnt), 0);
if (!cnt) {
- rte_errno = ENOMEM;
- return NULL;
+ struct mlx5_flow_counter_pool **pools;
+ uint32_t size;
+
+ if (n_valid == cmng->n) {
+ /* Resize the container pool array. */
+ size = sizeof(struct mlx5_flow_counter_pool *) *
+ (n_valid + MLX5_CNT_CONTAINER_RESIZE);
+ pools = mlx5_malloc(MLX5_MEM_ZERO, size, 0,
+ SOCKET_ID_ANY);
+ if (!pools)
+ return 0;
+ if (n_valid) {
+ memcpy(pools, cmng->pools,
+ sizeof(struct mlx5_flow_counter_pool *) *
+ n_valid);
+ mlx5_free(cmng->pools);
+ }
+ cmng->pools = pools;
+ cmng->n += MLX5_CNT_CONTAINER_RESIZE;
+ }
+ /* Allocate memory for new pool*/
+ size = sizeof(*pool) + sizeof(*cnt) * MLX5_COUNTERS_PER_POOL;
+ pool = mlx5_malloc(MLX5_MEM_ZERO, size, 0, SOCKET_ID_ANY);
+ if (!pool)
+ return 0;
+ for (i = 0; i < MLX5_COUNTERS_PER_POOL; ++i) {
+ cnt = MLX5_POOL_GET_CNT(pool, i);
+ TAILQ_INSERT_HEAD(&pool->counters[0], cnt, next);
+ }
+ cnt = MLX5_POOL_GET_CNT(pool, 0);
+ cmng->pools[n_valid] = pool;
+ pool_idx = n_valid;
+ cmng->n_valid++;
}
- cnt->id = id;
- cnt->shared = shared;
- cnt->ref_cnt = 1;
- cnt->hits = 0;
- cnt->bytes = 0;
+ TAILQ_REMOVE(&pool->counters[0], cnt, next);
+ i = MLX5_CNT_ARRAY_IDX(pool, cnt);
+ cnt_idx = MLX5_MAKE_CNT_IDX(pool_idx, i);
/* Create counter with Verbs. */
ret = flow_verbs_counter_create(dev, cnt);
if (!ret) {
- TAILQ_INSERT_HEAD(&priv->sh->cmng.flow_counters, cnt, next);
- return cnt;
+ cnt->dcs_when_active = cnt->dcs_when_free;
+ cnt->hits = 0;
+ cnt->bytes = 0;
+ return cnt_idx;
}
+ TAILQ_INSERT_HEAD(&pool->counters[0], cnt, next);
/* Some error occurred in Verbs library. */
- rte_free(cnt);
rte_errno = -ret;
- return NULL;
+ return 0;
}
/**
* @param[in] dev
* Pointer to the Ethernet device structure.
* @param[in] counter
- * Pointer to the counter handler.
+ * Index to the counter handler.
*/
static void
-flow_verbs_counter_release(struct rte_eth_dev *dev,
- struct mlx5_flow_counter *counter)
+flow_verbs_counter_release(struct rte_eth_dev *dev, uint32_t counter)
{
- struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_flow_counter_pool *pool;
+ struct mlx5_flow_counter *cnt;
- if (--counter->ref_cnt == 0) {
+ cnt = flow_verbs_counter_get_by_idx(dev, counter, &pool);
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42)
- claim_zero(mlx5_glue->destroy_counter_set(counter->cs));
+ claim_zero(mlx5_glue->destroy_counter_set
+ ((struct ibv_counter_set *)cnt->dcs_when_active));
#elif defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
- claim_zero(mlx5_glue->destroy_counters(counter->cs));
+ claim_zero(mlx5_glue->destroy_counters
+ ((struct ibv_counters *)cnt->dcs_when_active));
#endif
- TAILQ_REMOVE(&priv->sh->cmng.flow_counters, counter, next);
- rte_free(counter);
- }
+ TAILQ_INSERT_HEAD(&pool->counters[0], cnt, next);
}
/**
{
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42) || \
defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
- if (flow->counter && flow->counter->cs) {
+ if (flow->counter) {
+ struct mlx5_flow_counter_pool *pool;
+ struct mlx5_flow_counter *cnt = flow_verbs_counter_get_by_idx
+ (dev, flow->counter, &pool);
struct rte_flow_query_count *qc = data;
uint64_t counters[2] = {0, 0};
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42)
struct ibv_query_counter_set_attr query_cs_attr = {
- .cs = flow->counter->cs,
+ .dcs_when_free = (struct ibv_counter_set *)
+ cnt->dcs_when_active,
.query_flags = IBV_COUNTER_SET_FORCE_UPDATE,
};
struct ibv_counter_set_data query_out = {
&query_out);
#elif defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
int err = mlx5_glue->query_counters
- (flow->counter->cs, counters,
+ ((struct ibv_counters *)cnt->dcs_when_active, counters,
RTE_DIM(counters),
IBV_READ_COUNTERS_ATTR_PREFER_CACHED);
#endif
"cannot read counter");
qc->hits_set = 1;
qc->bytes_set = 1;
- qc->hits = counters[0] - flow->counter->hits;
- qc->bytes = counters[1] - flow->counter->bytes;
+ qc->hits = counters[0] - cnt->hits;
+ qc->bytes = counters[1] - cnt->bytes;
if (qc->reset) {
- flow->counter->hits = counters[0];
- flow->counter->bytes = counters[1];
+ cnt->hits = counters[0];
+ cnt->bytes = counters[1];
}
return 0;
}
* Size in bytes of the specification to copy.
*/
static void
-flow_verbs_spec_add(struct mlx5_flow_verbs *verbs, void *src, unsigned int size)
+flow_verbs_spec_add(struct mlx5_flow_verbs_workspace *verbs,
+ void *src, unsigned int size)
{
void *dst;
if (!verbs)
return;
- assert(verbs->specs);
+ MLX5_ASSERT(verbs->specs);
dst = (void *)(verbs->specs + verbs->size);
memcpy(dst, src, size);
- ++verbs->attr->num_of_specs;
+ ++verbs->attr.num_of_specs;
verbs->size += size;
}
if (!(item_flags & l2m))
flow_verbs_spec_add(&dev_flow->verbs, ð, size);
else
- flow_verbs_item_vlan_update(dev_flow->verbs.attr, ð);
+ flow_verbs_item_vlan_update(&dev_flow->verbs.attr, ð);
if (!tunnel)
- dev_flow->verbs.vf_vlan.tag =
+ dev_flow->handle->vf_vlan.tag =
rte_be_to_cpu_16(spec->tci) & 0x0fff;
}
ipv6.val.traffic_class = (vtc_flow_val & RTE_IPV6_HDR_TC_MASK) >>
RTE_IPV6_HDR_TC_SHIFT;
ipv6.val.next_hdr = spec->hdr.proto;
- ipv6.val.hop_limit = spec->hdr.hop_limits;
ipv6.mask.flow_label =
rte_cpu_to_be_32((vtc_flow_mask & RTE_IPV6_HDR_FL_MASK) >>
RTE_IPV6_HDR_FL_SHIFT);
ipv6.mask.traffic_class = (vtc_flow_mask & RTE_IPV6_HDR_TC_MASK) >>
RTE_IPV6_HDR_TC_SHIFT;
ipv6.mask.next_hdr = mask->hdr.proto;
- ipv6.mask.hop_limit = mask->hdr.hop_limits;
/* Remove unwanted bits from values. */
for (i = 0; i < RTE_DIM(ipv6.val.src_ip); ++i) {
ipv6.val.src_ip[i] &= ipv6.mask.src_ip[i];
ipv6.val.flow_label &= ipv6.mask.flow_label;
ipv6.val.traffic_class &= ipv6.mask.traffic_class;
ipv6.val.next_hdr &= ipv6.mask.next_hdr;
- ipv6.val.hop_limit &= ipv6.mask.hop_limit;
}
flow_verbs_spec_add(&dev_flow->verbs, &ipv6, size);
}
udp.val.src_port &= udp.mask.src_port;
udp.val.dst_port &= udp.mask.dst_port;
}
+ item++;
+ while (item->type == RTE_FLOW_ITEM_TYPE_VOID)
+ item++;
+ if (!(udp.val.dst_port & udp.mask.dst_port)) {
+ switch ((item)->type) {
+ case RTE_FLOW_ITEM_TYPE_VXLAN:
+ udp.val.dst_port = htons(MLX5_UDP_PORT_VXLAN);
+ udp.mask.dst_port = 0xffff;
+ break;
+ case RTE_FLOW_ITEM_TYPE_VXLAN_GPE:
+ udp.val.dst_port = htons(MLX5_UDP_PORT_VXLAN_GPE);
+ udp.mask.dst_port = 0xffff;
+ break;
+ case RTE_FLOW_ITEM_TYPE_MPLS:
+ udp.val.dst_port = htons(MLX5_UDP_PORT_MPLS);
+ udp.mask.dst_port = 0xffff;
+ break;
+ default:
+ break;
+ }
+ }
+
flow_verbs_spec_add(&dev_flow->verbs, &udp, size);
}
const struct rte_flow_item *item __rte_unused,
uint64_t item_flags)
{
- struct mlx5_flow_verbs *verbs = &dev_flow->verbs;
+ struct mlx5_flow_verbs_workspace *verbs = &dev_flow->verbs;
#ifndef HAVE_IBV_DEVICE_MPLS_SUPPORT
unsigned int size = sizeof(struct ibv_flow_spec_tunnel);
struct ibv_flow_spec_tunnel tunnel = {
.size = size,
};
#else
+ static const struct rte_flow_item_gre empty_gre = {0,};
const struct rte_flow_item_gre *spec = item->spec;
const struct rte_flow_item_gre *mask = item->mask;
unsigned int size = sizeof(struct ibv_flow_spec_gre);
.size = size,
};
- if (!mask)
- mask = &rte_flow_item_gre_mask;
- if (spec) {
- tunnel.val.c_ks_res0_ver = spec->c_rsvd0_ver;
- tunnel.val.protocol = spec->protocol;
- tunnel.mask.c_ks_res0_ver = mask->c_rsvd0_ver;
- tunnel.mask.protocol = mask->protocol;
- /* Remove unwanted bits from values. */
- tunnel.val.c_ks_res0_ver &= tunnel.mask.c_ks_res0_ver;
+ if (!spec) {
+ spec = &empty_gre;
+ mask = &empty_gre;
+ } else {
+ if (!mask)
+ mask = &rte_flow_item_gre_mask;
+ }
+ tunnel.val.c_ks_res0_ver = spec->c_rsvd0_ver;
+ tunnel.val.protocol = spec->protocol;
+ tunnel.mask.c_ks_res0_ver = mask->c_rsvd0_ver;
+ tunnel.mask.protocol = mask->protocol;
+ /* Remove unwanted bits from values. */
+ tunnel.val.c_ks_res0_ver &= tunnel.mask.c_ks_res0_ver;
+ tunnel.val.key &= tunnel.mask.key;
+ if (tunnel.mask.protocol) {
tunnel.val.protocol &= tunnel.mask.protocol;
- tunnel.val.key &= tunnel.mask.key;
+ } else {
+ tunnel.val.protocol = mlx5_translate_tunnel_etypes(item_flags);
+ if (tunnel.val.protocol) {
+ tunnel.mask.protocol = 0xFFFF;
+ tunnel.val.protocol =
+ rte_cpu_to_be_16(tunnel.val.protocol);
+ }
}
#endif
if (item_flags & MLX5_FLOW_LAYER_OUTER_L3_IPV4)
- flow_verbs_item_gre_ip_protocol_update(verbs->attr,
+ flow_verbs_item_gre_ip_protocol_update(&verbs->attr,
IBV_FLOW_SPEC_IPV4_EXT,
IPPROTO_GRE);
else
- flow_verbs_item_gre_ip_protocol_update(verbs->attr,
+ flow_verbs_item_gre_ip_protocol_update(&verbs->attr,
IBV_FLOW_SPEC_IPV6,
IPPROTO_GRE);
flow_verbs_spec_add(verbs, &tunnel, size);
* the input is valid and that there is space to insert the requested action
* into the flow.
*
- * @param[in] dev_flow
- * Pointer to mlx5_flow.
+ * @param[in] rss_desc
+ * Pointer to mlx5_flow_rss_desc.
* @param[in] action
* Action configuration.
*/
static void
-flow_verbs_translate_action_queue(struct mlx5_flow *dev_flow,
+flow_verbs_translate_action_queue(struct mlx5_flow_rss_desc *rss_desc,
const struct rte_flow_action *action)
{
const struct rte_flow_action_queue *queue = action->conf;
- struct rte_flow *flow = dev_flow->flow;
- if (flow->rss.queue)
- (*flow->rss.queue)[0] = queue->index;
- flow->rss.queue_num = 1;
+ rss_desc->queue[0] = queue->index;
+ rss_desc->queue_num = 1;
}
/**
* the input is valid and that there is space to insert the requested action
* into the flow.
*
+ * @param[in] rss_desc
+ * Pointer to mlx5_flow_rss_desc.
* @param[in] action
* Action configuration.
- * @param[in, out] action_flags
- * Pointer to the detected actions.
- * @param[in] dev_flow
- * Pointer to mlx5_flow.
*/
static void
-flow_verbs_translate_action_rss(struct mlx5_flow *dev_flow,
+flow_verbs_translate_action_rss(struct mlx5_flow_rss_desc *rss_desc,
const struct rte_flow_action *action)
{
const struct rte_flow_action_rss *rss = action->conf;
const uint8_t *rss_key;
- struct rte_flow *flow = dev_flow->flow;
- if (flow->rss.queue)
- memcpy((*flow->rss.queue), rss->queue,
- rss->queue_num * sizeof(uint16_t));
- flow->rss.queue_num = rss->queue_num;
+ memcpy(rss_desc->queue, rss->queue, rss->queue_num * sizeof(uint16_t));
+ rss_desc->queue_num = rss->queue_num;
/* NULL RSS key indicates default RSS key. */
rss_key = !rss->key ? rss_hash_default_key : rss->key;
- memcpy(flow->rss.key, rss_key, MLX5_RSS_HASH_KEY_LEN);
+ memcpy(rss_desc->key, rss_key, MLX5_RSS_HASH_KEY_LEN);
/*
* rss->level and rss.types should be set in advance when expanding
* items for RSS.
struct rte_flow *flow = dev_flow->flow;
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42) || \
defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
+ struct mlx5_flow_counter_pool *pool;
+ struct mlx5_flow_counter *cnt = NULL;
unsigned int size = sizeof(struct ibv_flow_spec_counter_action);
struct ibv_flow_spec_counter_action counter = {
.type = IBV_FLOW_SPEC_ACTION_COUNT,
#endif
if (!flow->counter) {
- flow->counter = flow_verbs_counter_new(dev, count->shared,
- count->id);
+ flow->counter = flow_verbs_counter_new(dev, count->id);
if (!flow->counter)
return rte_flow_error_set(error, rte_errno,
RTE_FLOW_ERROR_TYPE_ACTION,
" context.");
}
#if defined(HAVE_IBV_DEVICE_COUNTERS_SET_V42)
- counter.counter_set_handle = flow->counter->cs->handle;
+ cnt = flow_verbs_counter_get_by_idx(dev, flow->counter, &pool);
+ counter.counter_set_handle =
+ ((struct ibv_counter_set *)cnt->dcs_when_active)->handle;
flow_verbs_spec_add(&dev_flow->verbs, &counter, size);
#elif defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45)
- counter.counters = flow->counter->cs;
+ cnt = flow_verbs_counter_get_by_idx(dev, flow->counter, &pool);
+ counter.counters = (struct ibv_counters *)cnt->dcs_when_active;
flow_verbs_spec_add(&dev_flow->verbs, &counter, size);
#endif
return 0;
* Pointer to the list of actions.
* @param[in] external
* This flow rule is created by request external to PMD.
+ * @param[in] hairpin
+ * Number of hairpin TX actions, 0 means classic flow.
* @param[out] error
* Pointer to the error structure.
*
const struct rte_flow_item items[],
const struct rte_flow_action actions[],
bool external __rte_unused,
+ int hairpin __rte_unused,
struct rte_flow_error *error)
{
int ret;
uint64_t last_item = 0;
uint8_t next_protocol = 0xff;
uint16_t ether_type = 0;
+ bool is_empty_vlan = false;
+ uint16_t udp_dport = 0;
if (items == NULL)
return -1;
break;
case RTE_FLOW_ITEM_TYPE_ETH:
ret = mlx5_flow_validate_item_eth(items, item_flags,
- error);
+ false, error);
if (ret < 0)
return ret;
last_item = tunnel ? MLX5_FLOW_LAYER_INNER_L2 :
ether_type &=
((const struct rte_flow_item_eth *)
items->mask)->type;
+ if (ether_type == RTE_BE16(RTE_ETHER_TYPE_VLAN))
+ is_empty_vlan = true;
ether_type = rte_be_to_cpu_16(ether_type);
} else {
ether_type = 0;
} else {
ether_type = 0;
}
+ is_empty_vlan = false;
break;
case RTE_FLOW_ITEM_TYPE_IPV4:
- ret = mlx5_flow_validate_item_ipv4(items, item_flags,
- last_item,
- ether_type, NULL,
- error);
+ ret = mlx5_flow_validate_item_ipv4
+ (items, item_flags,
+ last_item, ether_type, NULL,
+ MLX5_ITEM_RANGE_NOT_ACCEPTED,
+ error);
if (ret < 0)
return ret;
last_item = tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV4 :
ret = mlx5_flow_validate_item_udp(items, item_flags,
next_protocol,
error);
+ const struct rte_flow_item_udp *spec = items->spec;
+ const struct rte_flow_item_udp *mask = items->mask;
+ if (!mask)
+ mask = &rte_flow_item_udp_mask;
+ if (spec != NULL)
+ udp_dport = rte_be_to_cpu_16
+ (spec->hdr.dst_port &
+ mask->hdr.dst_port);
+
if (ret < 0)
return ret;
last_item = tunnel ? MLX5_FLOW_LAYER_INNER_L4_UDP :
MLX5_FLOW_LAYER_OUTER_L4_TCP;
break;
case RTE_FLOW_ITEM_TYPE_VXLAN:
- ret = mlx5_flow_validate_item_vxlan(items, item_flags,
- error);
+ ret = mlx5_flow_validate_item_vxlan(dev, udp_dport,
+ items, item_flags,
+ attr, error);
if (ret < 0)
return ret;
last_item = MLX5_FLOW_LAYER_VXLAN;
return ret;
last_item = MLX5_FLOW_LAYER_MPLS;
break;
+ case RTE_FLOW_ITEM_TYPE_ICMP:
+ case RTE_FLOW_ITEM_TYPE_ICMP6:
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_ITEM,
+ NULL, "ICMP/ICMP6 "
+ "item not supported");
default:
return rte_flow_error_set(error, ENOTSUP,
RTE_FLOW_ERROR_TYPE_ITEM,
}
item_flags |= last_item;
}
+ if (is_empty_vlan)
+ return rte_flow_error_set(error, ENOTSUP,
+ RTE_FLOW_ERROR_TYPE_ITEM, NULL,
+ "VLAN matching without vid specification is not supported");
for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) {
switch (actions->type) {
case RTE_FLOW_ACTION_TYPE_VOID:
"action not supported");
}
}
+ /*
+ * Validate the drop action mutual exclusion with other actions.
+ * Drop action is mutually-exclusive with any other action, except for
+ * Count action.
+ */
+ if ((action_flags & MLX5_FLOW_ACTION_DROP) &&
+ (action_flags & ~(MLX5_FLOW_ACTION_DROP | MLX5_FLOW_ACTION_COUNT)))
+ return rte_flow_error_set(error, EINVAL,
+ RTE_FLOW_ERROR_TYPE_ACTION, NULL,
+ "Drop action is mutually-exclusive "
+ "with any other action, except for "
+ "Count action");
if (!(action_flags & MLX5_FLOW_FATE_ACTIONS))
return rte_flow_error_set(error, EINVAL,
RTE_FLOW_ERROR_TYPE_ACTION, actions,
* The required size is calculate based on the actions and items. This function
* also returns the detected actions and items for later use.
*
+ * @param[in] dev
+ * Pointer to Ethernet device.
* @param[in] attr
* Pointer to the flow attributes.
* @param[in] items
* is set.
*/
static struct mlx5_flow *
-flow_verbs_prepare(const struct rte_flow_attr *attr __rte_unused,
+flow_verbs_prepare(struct rte_eth_dev *dev,
+ const struct rte_flow_attr *attr __rte_unused,
const struct rte_flow_item items[],
const struct rte_flow_action actions[],
struct rte_flow_error *error)
{
- size_t size = sizeof(struct mlx5_flow) + sizeof(struct ibv_flow_attr);
+ size_t size = 0;
+ uint32_t handle_idx = 0;
struct mlx5_flow *dev_flow;
+ struct mlx5_flow_handle *dev_handle;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_flow_workspace *wks = mlx5_flow_get_thread_workspace();
+ MLX5_ASSERT(wks);
size += flow_verbs_get_actions_size(actions);
size += flow_verbs_get_items_size(items);
- dev_flow = rte_calloc(__func__, 1, size, 0);
- if (!dev_flow) {
+ if (size > MLX5_VERBS_MAX_SPEC_ACT_SIZE) {
+ rte_flow_error_set(error, E2BIG,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+ "Verbs spec/action size too large");
+ return NULL;
+ }
+ /* In case of corrupting the memory. */
+ if (wks->flow_idx >= MLX5_NUM_MAX_DEV_FLOWS) {
+ rte_flow_error_set(error, ENOSPC,
+ RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
+ "not free temporary device flow");
+ return NULL;
+ }
+ dev_handle = mlx5_ipool_zmalloc(priv->sh->ipool[MLX5_IPOOL_MLX5_FLOW],
+ &handle_idx);
+ if (!dev_handle) {
rte_flow_error_set(error, ENOMEM,
RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
- "not enough memory to create flow");
+ "not enough memory to create flow handle");
return NULL;
}
- dev_flow->verbs.attr = (void *)(dev_flow + 1);
- dev_flow->verbs.specs = (void *)(dev_flow->verbs.attr + 1);
+ MLX5_ASSERT(wks->flow_idx + 1 < RTE_DIM(wks->flows));
+ dev_flow = &wks->flows[wks->flow_idx++];
+ dev_flow->handle = dev_handle;
+ dev_flow->handle_idx = handle_idx;
+ /* Memcpy is used, only size needs to be cleared to 0. */
+ dev_flow->verbs.size = 0;
+ dev_flow->verbs.attr.num_of_specs = 0;
dev_flow->ingress = attr->ingress;
- dev_flow->transfer = attr->transfer;
+ dev_flow->hash_fields = 0;
+ /* Need to set transfer attribute: not supported in Verbs mode. */
return dev_flow;
}
uint64_t priority = attr->priority;
uint32_t subpriority = 0;
struct mlx5_priv *priv = dev->data->dev_private;
-
- if (priority == MLX5_FLOW_PRIO_RSVD)
- priority = priv->config.flow_prio - 1;
+ struct mlx5_flow_workspace *wks = mlx5_flow_get_thread_workspace();
+ struct mlx5_flow_rss_desc *rss_desc;
+ const struct rte_flow_item *tunnel_item = NULL;
+
+ MLX5_ASSERT(wks);
+ rss_desc = &wks->rss_desc;
+ if (priority == MLX5_FLOW_LOWEST_PRIO_INDICATOR)
+ priority = priv->sh->flow_max_priority - 1;
for (; actions->type != RTE_FLOW_ACTION_TYPE_END; actions++) {
int ret;
case RTE_FLOW_ACTION_TYPE_FLAG:
flow_verbs_translate_action_flag(dev_flow, actions);
action_flags |= MLX5_FLOW_ACTION_FLAG;
+ wks->mark = 1;
break;
case RTE_FLOW_ACTION_TYPE_MARK:
flow_verbs_translate_action_mark(dev_flow, actions);
action_flags |= MLX5_FLOW_ACTION_MARK;
+ wks->mark = 1;
break;
case RTE_FLOW_ACTION_TYPE_DROP:
flow_verbs_translate_action_drop(dev_flow, actions);
action_flags |= MLX5_FLOW_ACTION_DROP;
+ dev_flow->handle->fate_action = MLX5_FLOW_FATE_DROP;
break;
case RTE_FLOW_ACTION_TYPE_QUEUE:
- flow_verbs_translate_action_queue(dev_flow, actions);
+ flow_verbs_translate_action_queue(rss_desc, actions);
action_flags |= MLX5_FLOW_ACTION_QUEUE;
+ dev_flow->handle->fate_action = MLX5_FLOW_FATE_QUEUE;
break;
case RTE_FLOW_ACTION_TYPE_RSS:
- flow_verbs_translate_action_rss(dev_flow, actions);
+ flow_verbs_translate_action_rss(rss_desc, actions);
action_flags |= MLX5_FLOW_ACTION_RSS;
+ dev_flow->handle->fate_action = MLX5_FLOW_FATE_QUEUE;
break;
case RTE_FLOW_ACTION_TYPE_COUNT:
ret = flow_verbs_translate_action_count(dev_flow,
"action not supported");
}
}
- dev_flow->actions = action_flags;
+ dev_flow->act_flags = action_flags;
for (; items->type != RTE_FLOW_ITEM_TYPE_END; items++) {
int tunnel = !!(item_flags & MLX5_FLOW_LAYER_TUNNEL);
subpriority = MLX5_PRIORITY_MAP_L3;
dev_flow->hash_fields |=
mlx5_flow_hashfields_adjust
- (dev_flow, tunnel,
+ (rss_desc, tunnel,
MLX5_IPV4_LAYER_TYPES,
MLX5_IPV4_IBV_RX_HASH);
item_flags |= tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV4 :
subpriority = MLX5_PRIORITY_MAP_L3;
dev_flow->hash_fields |=
mlx5_flow_hashfields_adjust
- (dev_flow, tunnel,
+ (rss_desc, tunnel,
MLX5_IPV6_LAYER_TYPES,
MLX5_IPV6_IBV_RX_HASH);
item_flags |= tunnel ? MLX5_FLOW_LAYER_INNER_L3_IPV6 :
flow_verbs_translate_item_tcp(dev_flow, items,
item_flags);
subpriority = MLX5_PRIORITY_MAP_L4;
- dev_flow->hash_fields |=
- mlx5_flow_hashfields_adjust
- (dev_flow, tunnel, ETH_RSS_TCP,
+ if (dev_flow->hash_fields != 0)
+ dev_flow->hash_fields |=
+ mlx5_flow_hashfields_adjust
+ (rss_desc, tunnel, RTE_ETH_RSS_TCP,
(IBV_RX_HASH_SRC_PORT_TCP |
IBV_RX_HASH_DST_PORT_TCP));
item_flags |= tunnel ? MLX5_FLOW_LAYER_INNER_L4_TCP :
flow_verbs_translate_item_udp(dev_flow, items,
item_flags);
subpriority = MLX5_PRIORITY_MAP_L4;
- dev_flow->hash_fields |=
- mlx5_flow_hashfields_adjust
- (dev_flow, tunnel, ETH_RSS_UDP,
+ if (dev_flow->hash_fields != 0)
+ dev_flow->hash_fields |=
+ mlx5_flow_hashfields_adjust
+ (rss_desc, tunnel, RTE_ETH_RSS_UDP,
(IBV_RX_HASH_SRC_PORT_UDP |
IBV_RX_HASH_DST_PORT_UDP));
item_flags |= tunnel ? MLX5_FLOW_LAYER_INNER_L4_UDP :
case RTE_FLOW_ITEM_TYPE_VXLAN:
flow_verbs_translate_item_vxlan(dev_flow, items,
item_flags);
- subpriority = MLX5_PRIORITY_MAP_L2;
+ subpriority = MLX5_TUNNEL_PRIO_GET(rss_desc);
item_flags |= MLX5_FLOW_LAYER_VXLAN;
break;
case RTE_FLOW_ITEM_TYPE_VXLAN_GPE:
flow_verbs_translate_item_vxlan_gpe(dev_flow, items,
item_flags);
- subpriority = MLX5_PRIORITY_MAP_L2;
+ subpriority = MLX5_TUNNEL_PRIO_GET(rss_desc);
item_flags |= MLX5_FLOW_LAYER_VXLAN_GPE;
break;
case RTE_FLOW_ITEM_TYPE_GRE:
- flow_verbs_translate_item_gre(dev_flow, items,
- item_flags);
- subpriority = MLX5_PRIORITY_MAP_L2;
+ subpriority = MLX5_TUNNEL_PRIO_GET(rss_desc);
item_flags |= MLX5_FLOW_LAYER_GRE;
+ tunnel_item = items;
break;
case RTE_FLOW_ITEM_TYPE_MPLS:
flow_verbs_translate_item_mpls(dev_flow, items,
item_flags);
- subpriority = MLX5_PRIORITY_MAP_L2;
+ subpriority = MLX5_TUNNEL_PRIO_GET(rss_desc);
item_flags |= MLX5_FLOW_LAYER_MPLS;
break;
default:
return rte_flow_error_set(error, ENOTSUP,
RTE_FLOW_ERROR_TYPE_ITEM,
- NULL,
- "item not supported");
+ NULL, "item not supported");
}
}
- dev_flow->layers = item_flags;
- dev_flow->verbs.attr->priority =
+ if (item_flags & MLX5_FLOW_LAYER_GRE)
+ flow_verbs_translate_item_gre(dev_flow, tunnel_item,
+ item_flags);
+ dev_flow->handle->layers = item_flags;
+ /* Other members of attr will be ignored. */
+ dev_flow->verbs.attr.priority =
mlx5_flow_adjust_priority(dev, priority, subpriority);
- dev_flow->verbs.attr->port = (uint8_t)priv->ibv_port;
+ dev_flow->verbs.attr.port = (uint8_t)priv->dev_port;
return 0;
}
static void
flow_verbs_remove(struct rte_eth_dev *dev, struct rte_flow *flow)
{
- struct mlx5_flow_verbs *verbs;
- struct mlx5_flow *dev_flow;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_flow_handle *handle;
+ uint32_t handle_idx;
if (!flow)
return;
- LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
- verbs = &dev_flow->verbs;
- if (verbs->flow) {
- claim_zero(mlx5_glue->destroy_flow(verbs->flow));
- verbs->flow = NULL;
+ SILIST_FOREACH(priv->sh->ipool[MLX5_IPOOL_MLX5_FLOW], flow->dev_handles,
+ handle_idx, handle, next) {
+ if (handle->drv_flow) {
+ claim_zero(mlx5_glue->destroy_flow(handle->drv_flow));
+ handle->drv_flow = NULL;
}
- if (verbs->hrxq) {
- if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
- mlx5_hrxq_drop_release(dev);
- else
- mlx5_hrxq_release(dev, verbs->hrxq);
- verbs->hrxq = NULL;
- }
- if (dev_flow->verbs.vf_vlan.tag &&
- dev_flow->verbs.vf_vlan.created) {
- mlx5_vlan_vmwa_release(dev, &dev_flow->verbs.vf_vlan);
+ /* hrxq is union, don't touch it only the flag is set. */
+ if (handle->rix_hrxq &&
+ handle->fate_action == MLX5_FLOW_FATE_QUEUE) {
+ mlx5_hrxq_release(dev, handle->rix_hrxq);
+ handle->rix_hrxq = 0;
}
+ if (handle->vf_vlan.tag && handle->vf_vlan.created)
+ mlx5_vlan_vmwa_release(dev, &handle->vf_vlan);
}
}
static void
flow_verbs_destroy(struct rte_eth_dev *dev, struct rte_flow *flow)
{
- struct mlx5_flow *dev_flow;
+ struct mlx5_priv *priv = dev->data->dev_private;
+ struct mlx5_flow_handle *handle;
if (!flow)
return;
flow_verbs_remove(dev, flow);
- while (!LIST_EMPTY(&flow->dev_flows)) {
- dev_flow = LIST_FIRST(&flow->dev_flows);
- LIST_REMOVE(dev_flow, next);
- rte_free(dev_flow);
+ while (flow->dev_handles) {
+ uint32_t tmp_idx = flow->dev_handles;
+
+ handle = mlx5_ipool_get(priv->sh->ipool[MLX5_IPOOL_MLX5_FLOW],
+ tmp_idx);
+ if (!handle)
+ return;
+ flow->dev_handles = handle->next.next;
+ mlx5_ipool_free(priv->sh->ipool[MLX5_IPOOL_MLX5_FLOW],
+ tmp_idx);
}
if (flow->counter) {
flow_verbs_counter_release(dev, flow->counter);
- flow->counter = NULL;
+ flow->counter = 0;
}
}
struct rte_flow_error *error)
{
struct mlx5_priv *priv = dev->data->dev_private;
- struct mlx5_flow_verbs *verbs;
+ struct mlx5_flow_handle *handle;
struct mlx5_flow *dev_flow;
+ struct mlx5_hrxq *hrxq;
+ uint32_t dev_handles;
int err;
-
- LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
- verbs = &dev_flow->verbs;
- if (dev_flow->actions & MLX5_FLOW_ACTION_DROP) {
- verbs->hrxq = mlx5_hrxq_drop_new(dev);
- if (!verbs->hrxq) {
- rte_flow_error_set
- (error, errno,
- RTE_FLOW_ERROR_TYPE_UNSPECIFIED, NULL,
- "cannot get drop hash queue");
- goto error;
- }
+ int idx;
+ struct mlx5_flow_workspace *wks = mlx5_flow_get_thread_workspace();
+
+ MLX5_ASSERT(wks);
+ for (idx = wks->flow_idx - 1; idx >= 0; idx--) {
+ dev_flow = &wks->flows[idx];
+ handle = dev_flow->handle;
+ if (handle->fate_action == MLX5_FLOW_FATE_DROP) {
+ MLX5_ASSERT(priv->drop_queue.hrxq);
+ hrxq = priv->drop_queue.hrxq;
} else {
- struct mlx5_hrxq *hrxq;
-
- assert(flow->rss.queue);
- hrxq = mlx5_hrxq_get(dev, flow->rss.key,
- MLX5_RSS_HASH_KEY_LEN,
- dev_flow->hash_fields,
- (*flow->rss.queue),
- flow->rss.queue_num);
- if (!hrxq)
- hrxq = mlx5_hrxq_new(dev, flow->rss.key,
- MLX5_RSS_HASH_KEY_LEN,
- dev_flow->hash_fields,
- (*flow->rss.queue),
- flow->rss.queue_num,
- !!(dev_flow->layers &
- MLX5_FLOW_LAYER_TUNNEL));
+ struct mlx5_flow_rss_desc *rss_desc = &wks->rss_desc;
+
+ MLX5_ASSERT(rss_desc->queue_num);
+ rss_desc->key_len = MLX5_RSS_HASH_KEY_LEN;
+ rss_desc->hash_fields = dev_flow->hash_fields;
+ rss_desc->tunnel = !!(handle->layers &
+ MLX5_FLOW_LAYER_TUNNEL);
+ rss_desc->shared_rss = 0;
+ hrxq = mlx5_hrxq_get(dev, rss_desc);
if (!hrxq) {
rte_flow_error_set
(error, rte_errno,
"cannot get hash queue");
goto error;
}
- verbs->hrxq = hrxq;
+ handle->rix_hrxq = hrxq->idx;
}
- verbs->flow = mlx5_glue->create_flow(verbs->hrxq->qp,
- verbs->attr);
- if (!verbs->flow) {
+ MLX5_ASSERT(hrxq);
+ handle->drv_flow = mlx5_glue->create_flow
+ (hrxq->qp, &dev_flow->verbs.attr);
+ if (!handle->drv_flow) {
rte_flow_error_set(error, errno,
RTE_FLOW_ERROR_TYPE_UNSPECIFIED,
NULL,
goto error;
}
if (priv->vmwa_context &&
- dev_flow->verbs.vf_vlan.tag &&
- !dev_flow->verbs.vf_vlan.created) {
+ handle->vf_vlan.tag && !handle->vf_vlan.created) {
/*
* The rule contains the VLAN pattern.
* For VF we are going to create VLAN
* interface to make hypervisor set correct
* e-Switch vport context.
*/
- mlx5_vlan_vmwa_acquire(dev, &dev_flow->verbs.vf_vlan);
+ mlx5_vlan_vmwa_acquire(dev, &handle->vf_vlan);
}
}
return 0;
error:
err = rte_errno; /* Save rte_errno before cleanup. */
- LIST_FOREACH(dev_flow, &flow->dev_flows, next) {
- verbs = &dev_flow->verbs;
- if (verbs->hrxq) {
- if (dev_flow->actions & MLX5_FLOW_ACTION_DROP)
- mlx5_hrxq_drop_release(dev);
- else
- mlx5_hrxq_release(dev, verbs->hrxq);
- verbs->hrxq = NULL;
- }
- if (dev_flow->verbs.vf_vlan.tag &&
- dev_flow->verbs.vf_vlan.created) {
- mlx5_vlan_vmwa_release(dev, &dev_flow->verbs.vf_vlan);
+ SILIST_FOREACH(priv->sh->ipool[MLX5_IPOOL_MLX5_FLOW], flow->dev_handles,
+ dev_handles, handle, next) {
+ /* hrxq is union, don't touch it only the flag is set. */
+ if (handle->rix_hrxq &&
+ handle->fate_action == MLX5_FLOW_FATE_QUEUE) {
+ mlx5_hrxq_release(dev, handle->rix_hrxq);
+ handle->rix_hrxq = 0;
}
+ if (handle->vf_vlan.tag && handle->vf_vlan.created)
+ mlx5_vlan_vmwa_release(dev, &handle->vf_vlan);
}
rte_errno = err; /* Restore rte_errno. */
return -rte_errno;
return ret;
}
+static int
+flow_verbs_sync_domain(struct rte_eth_dev *dev, uint32_t domains,
+ uint32_t flags)
+{
+ RTE_SET_USED(dev);
+ RTE_SET_USED(domains);
+ RTE_SET_USED(flags);
+
+ return 0;
+}
+
const struct mlx5_flow_driver_ops mlx5_flow_verbs_drv_ops = {
.validate = flow_verbs_validate,
.prepare = flow_verbs_prepare,
.remove = flow_verbs_remove,
.destroy = flow_verbs_destroy,
.query = flow_verbs_query,
+ .sync_domain = flow_verbs_sync_domain,
+ .discover_priorities = flow_verbs_discover_priorities,
};