X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx5%2Fmlx5.c;h=75ada96cc5e95f022f79422c1753f5bb56c107e8;hb=9b080425e31297199dcbfa35d95d287d729c68a0;hp=01a2ef78052e5b8a360bcb13f661d105bf80abf9;hpb=830d209161c1562f894fa2c843043012c5602b13;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5.c b/drivers/net/mlx5/mlx5.c index 01a2ef7805..75ada96cc5 100644 --- a/drivers/net/mlx5/mlx5.c +++ b/drivers/net/mlx5/mlx5.c @@ -6,8 +6,6 @@ #include #include #include -#include -#include #include #include #include @@ -31,21 +29,25 @@ #include #include #include -#include #include #include #include #include #include +#include +#include +#include +#include + +#include "mlx5_defs.h" #include "mlx5.h" #include "mlx5_utils.h" #include "mlx5_rxtx.h" #include "mlx5_autoconf.h" -#include "mlx5_defs.h" -#include "mlx5_glue.h" #include "mlx5_mr.h" #include "mlx5_flow.h" +#include "rte_pmd_mlx5.h" /* Device parameter to enable RX completion queue compression. */ #define MLX5_RXQ_CQE_COMP_EN "rxq_cqe_comp_en" @@ -62,6 +64,9 @@ /* Device parameter to configure log 2 of the number of strides for MPRQ. */ #define MLX5_RX_MPRQ_LOG_STRIDE_NUM "mprq_log_stride_num" +/* Device parameter to configure log 2 of the stride size for MPRQ. */ +#define MLX5_RX_MPRQ_LOG_STRIDE_SIZE "mprq_log_stride_size" + /* Device parameter to limit the size of memcpy'd packet for MPRQ. */ #define MLX5_RX_MPRQ_MAX_MEMCPY_LEN "mprq_max_memcpy_len" @@ -95,6 +100,12 @@ /* Device parameter to enable multi-packet send WQEs. */ #define MLX5_TXQ_MPW_EN "txq_mpw_en" +/* + * Device parameter to force doorbell register mapping + * to non-cahed region eliminating the extra write memory barrier. + */ +#define MLX5_TX_DB_NC "tx_db_nc" + /* * Device parameter to include 2 dsegs in the title WQEBB. * Deprecated, ignored. @@ -125,6 +136,9 @@ /* Activate DV flow steering. */ #define MLX5_DV_FLOW_EN "dv_flow_en" +/* Enable extensive flow metadata support. */ +#define MLX5_DV_XMETA_EN "dv_xmeta_en" + /* Activate Netlink support in VF mode. */ #define MLX5_VF_NL_EN "vf_nl_en" @@ -140,6 +154,12 @@ /* Configure timeout of LRO session (in microseconds). */ #define MLX5_LRO_TIMEOUT_USEC "lro_timeout_usec" +/* + * Device parameter to configure the total data buffer size for a single + * hairpin queue (logarithm value). + */ +#define MLX5_HP_BUF_SIZE "hp_buf_log_sz" + #ifndef HAVE_IBV_MLX5_MOD_MPW #define MLX5DV_CONTEXT_FLAGS_MPW_ALLOWED (1 << 2) #define MLX5DV_CONTEXT_FLAGS_ENHANCED_MPW (1 << 3) @@ -178,17 +198,137 @@ struct mlx5_dev_spawn_data { static LIST_HEAD(, mlx5_ibv_shared) mlx5_ibv_list = LIST_HEAD_INITIALIZER(); static pthread_mutex_t mlx5_ibv_list_mutex = PTHREAD_MUTEX_INITIALIZER; +static struct mlx5_indexed_pool_config mlx5_ipool_cfg[] = { +#ifdef HAVE_IBV_FLOW_DV_SUPPORT + { + .size = sizeof(struct mlx5_flow_dv_encap_decap_resource), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_encap_decap_ipool", + }, + { + .size = sizeof(struct mlx5_flow_dv_push_vlan_action_resource), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_push_vlan_ipool", + }, + { + .size = sizeof(struct mlx5_flow_dv_tag_resource), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_tag_ipool", + }, + { + .size = sizeof(struct mlx5_flow_dv_port_id_action_resource), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_port_id_ipool", + }, + { + .size = sizeof(struct mlx5_flow_tbl_data_entry), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_jump_ipool", + }, +#endif + { + .size = sizeof(struct mlx5_flow_meter), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_meter_ipool", + }, + { + .size = sizeof(struct mlx5_flow_mreg_copy_resource), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_mcp_ipool", + }, + { + .size = (sizeof(struct mlx5_hrxq) + MLX5_RSS_HASH_KEY_LEN), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_hrxq_ipool", + }, + { + .size = sizeof(struct mlx5_flow_handle), + .trunk_size = 64, + .grow_trunk = 3, + .grow_shift = 2, + .need_lock = 0, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "mlx5_flow_handle_ipool", + }, + { + .size = sizeof(struct rte_flow), + .trunk_size = 4096, + .need_lock = 1, + .release_mem_en = 1, + .malloc = rte_malloc_socket, + .free = rte_free, + .type = "rte_flow_ipool", + }, +}; + + #define MLX5_FLOW_MIN_ID_POOL_SIZE 512 #define MLX5_ID_GENERATION_ARRAY_FACTOR 16 +#define MLX5_FLOW_TABLE_HLIST_ARRAY_SIZE 4096 +#define MLX5_TAGS_HLIST_ARRAY_SIZE 8192 + /** * Allocate ID pool structure. * + * @param[in] max_id + * The maximum id can be allocated from the pool. + * * @return * Pointer to pool object, NULL value otherwise. */ struct mlx5_flow_id_pool * -mlx5_flow_id_pool_alloc(void) +mlx5_flow_id_pool_alloc(uint32_t max_id) { struct mlx5_flow_id_pool *pool; void *mem; @@ -211,6 +351,7 @@ mlx5_flow_id_pool_alloc(void) pool->curr = pool->free_arr; pool->last = pool->free_arr + MLX5_FLOW_MIN_ID_POOL_SIZE; pool->base_index = 0; + pool->max_id = max_id; return pool; error: rte_free(pool); @@ -245,7 +386,7 @@ uint32_t mlx5_flow_id_get(struct mlx5_flow_id_pool *pool, uint32_t *id) { if (pool->curr == pool->free_arr) { - if (pool->base_index == UINT32_MAX) { + if (pool->base_index == pool->max_id) { rte_errno = ENOMEM; DRV_LOG(ERR, "no free id"); return -rte_errno; @@ -278,7 +419,7 @@ mlx5_flow_id_release(struct mlx5_flow_id_pool *pool, uint32_t id) if (pool->curr == pool->last) { size = pool->curr - pool->free_arr; size2 = size * MLX5_ID_GENERATION_ARRAY_FACTOR; - assert(size2 > size); + MLX5_ASSERT(size2 > size); mem = rte_malloc("", size2 * sizeof(uint32_t), 0); if (!mem) { DRV_LOG(ERR, "can't allocate mem for id pool"); @@ -368,9 +509,11 @@ mlx5_flow_counters_mng_close(struct mlx5_ibv_shared *sh) claim_zero (mlx5_glue->destroy_flow_action (pool->counters_raw[j].action)); - if (!batch && pool->counters_raw[j].dcs) + if (!batch && MLX5_GET_POOL_CNT_EXT + (pool, j)->dcs) claim_zero(mlx5_devx_cmd_destroy - (pool->counters_raw[j].dcs)); + (MLX5_GET_POOL_CNT_EXT + (pool, j)->dcs)); } TAILQ_REMOVE(&sh->cmng.ccont[i].pool_list, pool, next); @@ -387,6 +530,49 @@ mlx5_flow_counters_mng_close(struct mlx5_ibv_shared *sh) memset(&sh->cmng, 0, sizeof(sh->cmng)); } +/** + * Initialize the flow resources' indexed mempool. + * + * @param[in] sh + * Pointer to mlx5_ibv_shared object. + * @param[in] sh + * Pointer to user dev config. + */ +static void +mlx5_flow_ipool_create(struct mlx5_ibv_shared *sh, + const struct mlx5_dev_config *config __rte_unused) +{ + uint8_t i; + +#ifdef HAVE_IBV_FLOW_DV_SUPPORT + /* + * While DV is supported, user chooses the verbs mode, + * the mlx5 flow handle size is different with the + * MLX5_FLOW_HANDLE_VERBS_SIZE. + */ + if (!config->dv_flow_en) + mlx5_ipool_cfg[MLX5_IPOOL_MLX5_FLOW].size = + MLX5_FLOW_HANDLE_VERBS_SIZE; +#endif + for (i = 0; i < MLX5_IPOOL_MAX; ++i) + sh->ipool[i] = mlx5_ipool_create(&mlx5_ipool_cfg[i]); +} + +/** + * Release the flow resources' indexed mempool. + * + * @param[in] sh + * Pointer to mlx5_ibv_shared object. + */ +static void +mlx5_flow_ipool_destroy(struct mlx5_ibv_shared *sh) +{ + uint8_t i; + + for (i = 0; i < MLX5_IPOOL_MAX; ++i) + mlx5_ipool_destroy(sh->ipool[i]); +} + /** * Extract pdn of PD object using DV API. * @@ -418,6 +604,35 @@ mlx5_get_pdn(struct ibv_pd *pd __rte_unused, uint32_t *pdn __rte_unused) } #endif /* HAVE_IBV_FLOW_DV_SUPPORT */ +static int +mlx5_config_doorbell_mapping_env(const struct mlx5_dev_config *config) +{ + char *env; + int value; + + MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); + /* Get environment variable to store. */ + env = getenv(MLX5_SHUT_UP_BF); + value = env ? !!strcmp(env, "0") : MLX5_ARG_UNSET; + if (config->dbnc == MLX5_ARG_UNSET) + setenv(MLX5_SHUT_UP_BF, MLX5_SHUT_UP_BF_DEFAULT, 1); + else + setenv(MLX5_SHUT_UP_BF, + config->dbnc == MLX5_TXDB_NCACHED ? "1" : "0", 1); + return value; +} + +static void +mlx5_restore_doorbell_mapping_env(int value) +{ + MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); + /* Restore the original environment variable state. */ + if (value == MLX5_ARG_UNSET) + unsetenv(MLX5_SHUT_UP_BF); + else + setenv(MLX5_SHUT_UP_BF, value ? "1" : "0", 1); +} + /** * Allocate shared IB device context. If there is multiport device the * master and representors will share this context, if there is single @@ -431,24 +646,28 @@ mlx5_get_pdn(struct ibv_pd *pd __rte_unused, uint32_t *pdn __rte_unused) * * @param[in] spawn * Pointer to the IB device attributes (name, port, etc). + * @param[in] config + * Pointer to device configuration structure. * * @return * Pointer to mlx5_ibv_shared object on success, * otherwise NULL and rte_errno is set. */ static struct mlx5_ibv_shared * -mlx5_alloc_shared_ibctx(const struct mlx5_dev_spawn_data *spawn) +mlx5_alloc_shared_ibctx(const struct mlx5_dev_spawn_data *spawn, + const struct mlx5_dev_config *config) { struct mlx5_ibv_shared *sh; + int dbmap_env; int err = 0; uint32_t i; #ifdef HAVE_IBV_FLOW_DV_SUPPORT struct mlx5_devx_tis_attr tis_attr = { 0 }; #endif -assert(spawn); + MLX5_ASSERT(spawn); /* Secondary process should not create the shared context. */ - assert(rte_eal_process_type() == RTE_PROC_PRIMARY); + MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); pthread_mutex_lock(&mlx5_ibv_list_mutex); /* Search for IB context by device name. */ LIST_FOREACH(sh, &mlx5_ibv_list, next) { @@ -458,7 +677,7 @@ assert(spawn); } } /* No device found, we have to create new shared context. */ - assert(spawn->max_port); + MLX5_ASSERT(spawn->max_port); sh = rte_zmalloc("ethdev shared ib context", sizeof(struct mlx5_ibv_shared) + spawn->max_port * @@ -469,18 +688,32 @@ assert(spawn); rte_errno = ENOMEM; goto exit; } + /* + * Configure environment variable "MLX5_BF_SHUT_UP" + * before the device creation. The rdma_core library + * checks the variable at device creation and + * stores the result internally. + */ + dbmap_env = mlx5_config_doorbell_mapping_env(config); /* Try to open IB device with DV first, then usual Verbs. */ errno = 0; sh->ctx = mlx5_glue->dv_open_device(spawn->ibv_dev); if (sh->ctx) { sh->devx = 1; DRV_LOG(DEBUG, "DevX is supported"); + /* The device is created, no need for environment. */ + mlx5_restore_doorbell_mapping_env(dbmap_env); } else { + /* The environment variable is still configured. */ sh->ctx = mlx5_glue->open_device(spawn->ibv_dev); - if (!sh->ctx) { - err = errno ? errno : ENODEV; + err = errno ? errno : ENODEV; + /* + * The environment variable is not needed anymore, + * all device creation attempts are completed. + */ + mlx5_restore_doorbell_mapping_env(dbmap_env); + if (!sh->ctx) goto error; - } DRV_LOG(DEBUG, "DevX is NOT supported"); } err = mlx5_glue->query_device_ex(sh->ctx, NULL, &sh->device_attr); @@ -531,6 +764,13 @@ assert(spawn); goto error; } } + sh->flow_id_pool = mlx5_flow_id_pool_alloc + ((1 << HAIRPIN_FLOW_ID_BITS) - 1); + if (!sh->flow_id_pool) { + DRV_LOG(ERR, "can't create flow id pool"); + err = ENOMEM; + goto error; + } #endif /* HAVE_IBV_FLOW_DV_SUPPORT */ /* * Once the device is added to the list of memory event @@ -541,7 +781,7 @@ assert(spawn); * At this point the device is not added to the memory * event list yet, context is just being created. */ - err = mlx5_mr_btree_init(&sh->mr.cache, + err = mlx5_mr_btree_init(&sh->share_cache.cache, MLX5_MR_BTREE_CACHE_N * 2, spawn->pci_dev->device.numa_node); if (err) { @@ -549,6 +789,7 @@ assert(spawn); goto error; } mlx5_flow_counters_mng_init(sh); + mlx5_flow_ipool_create(sh, config); /* Add device to memory callback list. */ rte_rwlock_write_lock(&mlx5_shared_data->mem_event_rwlock); LIST_INSERT_HEAD(&mlx5_shared_data->mem_event_cb_list, @@ -561,7 +802,7 @@ exit: return sh; error: pthread_mutex_unlock(&mlx5_ibv_list_mutex); - assert(sh); + MLX5_ASSERT(sh); if (sh->tis) claim_zero(mlx5_devx_cmd_destroy(sh->tis)); if (sh->td) @@ -570,8 +811,10 @@ error: claim_zero(mlx5_glue->dealloc_pd(sh->pd)); if (sh->ctx) claim_zero(mlx5_glue->close_device(sh->ctx)); + if (sh->flow_id_pool) + mlx5_flow_id_pool_release(sh->flow_id_pool); rte_free(sh); - assert(err > 0); + MLX5_ASSERT(err > 0); rte_errno = err; return NULL; } @@ -587,31 +830,31 @@ static void mlx5_free_shared_ibctx(struct mlx5_ibv_shared *sh) { pthread_mutex_lock(&mlx5_ibv_list_mutex); -#ifndef NDEBUG +#ifdef RTE_LIBRTE_MLX5_DEBUG /* Check the object presence in the list. */ struct mlx5_ibv_shared *lctx; LIST_FOREACH(lctx, &mlx5_ibv_list, next) if (lctx == sh) break; - assert(lctx); + MLX5_ASSERT(lctx); if (lctx != sh) { DRV_LOG(ERR, "Freeing non-existing shared IB context"); goto exit; } #endif - assert(sh); - assert(sh->refcnt); + MLX5_ASSERT(sh); + MLX5_ASSERT(sh->refcnt); /* Secondary process should not free the shared context. */ - assert(rte_eal_process_type() == RTE_PROC_PRIMARY); + MLX5_ASSERT(rte_eal_process_type() == RTE_PROC_PRIMARY); if (--sh->refcnt) goto exit; - /* Release created Memory Regions. */ - mlx5_mr_release(sh); /* Remove from memory callback device list. */ rte_rwlock_write_lock(&mlx5_shared_data->mem_event_rwlock); LIST_REMOVE(sh, mem_event_cb); rte_rwlock_write_unlock(&mlx5_shared_data->mem_event_rwlock); + /* Release created Memory Regions. */ + mlx5_mr_release_cache(&sh->share_cache); /* Remove context from the global device list. */ LIST_REMOVE(sh, next); /* @@ -619,7 +862,8 @@ mlx5_free_shared_ibctx(struct mlx5_ibv_shared *sh) * Only primary process handles async device events. **/ mlx5_flow_counters_mng_close(sh); - assert(!sh->intr_cnt); + mlx5_flow_ipool_destroy(sh); + MLX5_ASSERT(!sh->intr_cnt); if (sh->intr_cnt) mlx5_intr_callback_unregister (&sh->intr_handle, mlx5_dev_interrupt_handler, sh); @@ -641,11 +885,150 @@ mlx5_free_shared_ibctx(struct mlx5_ibv_shared *sh) claim_zero(mlx5_devx_cmd_destroy(sh->td)); if (sh->ctx) claim_zero(mlx5_glue->close_device(sh->ctx)); + if (sh->flow_id_pool) + mlx5_flow_id_pool_release(sh->flow_id_pool); rte_free(sh); exit: pthread_mutex_unlock(&mlx5_ibv_list_mutex); } +/** + * Destroy table hash list and all the root entries per domain. + * + * @param[in] priv + * Pointer to the private device data structure. + */ +static void +mlx5_free_table_hash_list(struct mlx5_priv *priv) +{ + struct mlx5_ibv_shared *sh = priv->sh; + struct mlx5_flow_tbl_data_entry *tbl_data; + union mlx5_flow_tbl_key table_key = { + { + .table_id = 0, + .reserved = 0, + .domain = 0, + .direction = 0, + } + }; + struct mlx5_hlist_entry *pos; + + if (!sh->flow_tbls) + return; + pos = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64); + if (pos) { + tbl_data = container_of(pos, struct mlx5_flow_tbl_data_entry, + entry); + MLX5_ASSERT(tbl_data); + mlx5_hlist_remove(sh->flow_tbls, pos); + rte_free(tbl_data); + } + table_key.direction = 1; + pos = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64); + if (pos) { + tbl_data = container_of(pos, struct mlx5_flow_tbl_data_entry, + entry); + MLX5_ASSERT(tbl_data); + mlx5_hlist_remove(sh->flow_tbls, pos); + rte_free(tbl_data); + } + table_key.direction = 0; + table_key.domain = 1; + pos = mlx5_hlist_lookup(sh->flow_tbls, table_key.v64); + if (pos) { + tbl_data = container_of(pos, struct mlx5_flow_tbl_data_entry, + entry); + MLX5_ASSERT(tbl_data); + mlx5_hlist_remove(sh->flow_tbls, pos); + rte_free(tbl_data); + } + mlx5_hlist_destroy(sh->flow_tbls, NULL, NULL); +} + +/** + * Initialize flow table hash list and create the root tables entry + * for each domain. + * + * @param[in] priv + * Pointer to the private device data structure. + * + * @return + * Zero on success, positive error code otherwise. + */ +static int +mlx5_alloc_table_hash_list(struct mlx5_priv *priv) +{ + struct mlx5_ibv_shared *sh = priv->sh; + char s[MLX5_HLIST_NAMESIZE]; + int err = 0; + + MLX5_ASSERT(sh); + snprintf(s, sizeof(s), "%s_flow_table", priv->sh->ibdev_name); + sh->flow_tbls = mlx5_hlist_create(s, MLX5_FLOW_TABLE_HLIST_ARRAY_SIZE); + if (!sh->flow_tbls) { + DRV_LOG(ERR, "flow tables with hash creation failed.\n"); + err = ENOMEM; + return err; + } +#ifndef HAVE_MLX5DV_DR + /* + * In case we have not DR support, the zero tables should be created + * because DV expect to see them even if they cannot be created by + * RDMA-CORE. + */ + union mlx5_flow_tbl_key table_key = { + { + .table_id = 0, + .reserved = 0, + .domain = 0, + .direction = 0, + } + }; + struct mlx5_flow_tbl_data_entry *tbl_data = rte_zmalloc(NULL, + sizeof(*tbl_data), 0); + + if (!tbl_data) { + err = ENOMEM; + goto error; + } + tbl_data->entry.key = table_key.v64; + err = mlx5_hlist_insert(sh->flow_tbls, &tbl_data->entry); + if (err) + goto error; + rte_atomic32_init(&tbl_data->tbl.refcnt); + rte_atomic32_inc(&tbl_data->tbl.refcnt); + table_key.direction = 1; + tbl_data = rte_zmalloc(NULL, sizeof(*tbl_data), 0); + if (!tbl_data) { + err = ENOMEM; + goto error; + } + tbl_data->entry.key = table_key.v64; + err = mlx5_hlist_insert(sh->flow_tbls, &tbl_data->entry); + if (err) + goto error; + rte_atomic32_init(&tbl_data->tbl.refcnt); + rte_atomic32_inc(&tbl_data->tbl.refcnt); + table_key.direction = 0; + table_key.domain = 1; + tbl_data = rte_zmalloc(NULL, sizeof(*tbl_data), 0); + if (!tbl_data) { + err = ENOMEM; + goto error; + } + tbl_data->entry.key = table_key.v64; + err = mlx5_hlist_insert(sh->flow_tbls, &tbl_data->entry); + if (err) + goto error; + rte_atomic32_init(&tbl_data->tbl.refcnt); + rte_atomic32_inc(&tbl_data->tbl.refcnt); + return err; +error: + mlx5_free_table_hash_list(priv); +#endif /* HAVE_MLX5DV_DR */ + return err; +} + /** * Initialize DR related data within private structure. * Routine checks the reference counter and does actual @@ -660,12 +1043,28 @@ exit: static int mlx5_alloc_shared_dr(struct mlx5_priv *priv) { -#ifdef HAVE_MLX5DV_DR struct mlx5_ibv_shared *sh = priv->sh; + char s[MLX5_HLIST_NAMESIZE]; int err = 0; + + if (!sh->flow_tbls) + err = mlx5_alloc_table_hash_list(priv); + else + DRV_LOG(DEBUG, "sh->flow_tbls[%p] already created, reuse\n", + (void *)sh->flow_tbls); + if (err) + return err; + /* Create tags hash list table. */ + snprintf(s, sizeof(s), "%s_tags", sh->ibdev_name); + sh->tag_table = mlx5_hlist_create(s, MLX5_TAGS_HLIST_ARRAY_SIZE); + if (!sh->tag_table) { + DRV_LOG(ERR, "tags with hash creation failed.\n"); + err = ENOMEM; + goto error; + } +#ifdef HAVE_MLX5DV_DR void *domain; - assert(sh); if (sh->dv_refcnt) { /* Shared DV/DR structures is already initialized. */ sh->dv_refcnt++; @@ -704,12 +1103,12 @@ mlx5_alloc_shared_dr(struct mlx5_priv *priv) } #endif sh->pop_vlan_action = mlx5_glue->dr_create_flow_action_pop_vlan(); +#endif /* HAVE_MLX5DV_DR */ sh->dv_refcnt++; priv->dr_shared = 1; return 0; - error: - /* Rollback the created objects. */ + /* Rollback the created objects. */ if (sh->rx_domain) { mlx5_glue->dr_destroy_domain(sh->rx_domain); sh->rx_domain = NULL; @@ -730,11 +1129,13 @@ error: mlx5_glue->destroy_flow_action(sh->pop_vlan_action); sh->pop_vlan_action = NULL; } + if (sh->tag_table) { + /* tags should be destroyed with flow before. */ + mlx5_hlist_destroy(sh->tag_table, NULL, NULL); + sh->tag_table = NULL; + } + mlx5_free_table_hash_list(priv); return err; -#else - (void)priv; - return 0; -#endif } /** @@ -746,15 +1147,15 @@ error: static void mlx5_free_shared_dr(struct mlx5_priv *priv) { -#ifdef HAVE_MLX5DV_DR struct mlx5_ibv_shared *sh; if (!priv->dr_shared) return; priv->dr_shared = 0; sh = priv->sh; - assert(sh); - assert(sh->dv_refcnt); + MLX5_ASSERT(sh); +#ifdef HAVE_MLX5DV_DR + MLX5_ASSERT(sh->dv_refcnt); if (sh->dv_refcnt && --sh->dv_refcnt) return; if (sh->rx_domain) { @@ -780,9 +1181,13 @@ mlx5_free_shared_dr(struct mlx5_priv *priv) sh->pop_vlan_action = NULL; } pthread_mutex_destroy(&sh->dv_mutex); -#else - (void)priv; -#endif +#endif /* HAVE_MLX5DV_DR */ + if (sh->tag_table) { + /* tags should be destroyed with flow before. */ + mlx5_hlist_destroy(sh->tag_table, NULL, NULL); + sh->tag_table = NULL; + } + mlx5_free_table_hash_list(priv); } /** @@ -809,7 +1214,7 @@ mlx5_init_shared_data(void) SOCKET_ID_ANY, 0); if (mz == NULL) { DRV_LOG(ERR, - "Cannot allocate mlx5 shared data\n"); + "Cannot allocate mlx5 shared data"); ret = -rte_errno; goto error; } @@ -821,7 +1226,7 @@ mlx5_init_shared_data(void) mz = rte_memzone_lookup(MZ_MLX5_PMD_SHARED_DATA); if (mz == NULL) { DRV_LOG(ERR, - "Cannot attach mlx5 shared data\n"); + "Cannot attach mlx5 shared data"); ret = -rte_errno; goto error; } @@ -885,7 +1290,7 @@ mlx5_alloc_verbs_buf(size_t size, void *data) socket = ctrl->socket; } - assert(data != NULL); + MLX5_ASSERT(data != NULL); ret = rte_malloc_socket(__func__, size, alignment, socket); if (!ret && size) rte_errno = ENOMEM; @@ -903,7 +1308,7 @@ mlx5_alloc_verbs_buf(size_t size, void *data) static void mlx5_free_verbs_buf(void *ptr, void *data __rte_unused) { - assert(data != NULL); + MLX5_ASSERT(data != NULL); rte_free(ptr); } @@ -922,7 +1327,7 @@ int mlx5_udp_tunnel_port_add(struct rte_eth_dev *dev __rte_unused, struct rte_eth_udp_tunnel *udp_tunnel) { - assert(udp_tunnel != NULL); + MLX5_ASSERT(udp_tunnel != NULL); if (udp_tunnel->prot_type == RTE_TUNNEL_TYPE_VXLAN && udp_tunnel->udp_port == 4789) return 0; @@ -1001,8 +1406,20 @@ mlx5_dev_close(struct rte_eth_dev *dev) /* In case mlx5_dev_stop() has not been called. */ mlx5_dev_interrupt_handler_uninstall(dev); mlx5_dev_interrupt_handler_devx_uninstall(dev); + /* + * If default mreg copy action is removed at the stop stage, + * the search will return none and nothing will be done anymore. + */ + mlx5_flow_stop_default(dev); mlx5_traffic_disable(dev); - mlx5_flow_flush(dev, NULL); + /* + * If all the flows are already flushed in the device stop stage, + * then this will return directly without any action. + */ + mlx5_flow_list_flush(dev, &priv->flows, true); + mlx5_flow_meter_flush(dev, NULL); + /* Free the intermediate buffers for flow creation. */ + mlx5_flow_free_intermediate(dev); /* Prevent crashes when queues are still in use. */ dev->rx_pkt_burst = removed_rx_burst; dev->tx_pkt_burst = removed_tx_burst; @@ -1026,6 +1443,8 @@ mlx5_dev_close(struct rte_eth_dev *dev) priv->txqs = NULL; } mlx5_proc_priv_uninit(dev); + if (priv->mreg_cp_tbl) + mlx5_hlist_destroy(priv->mreg_cp_tbl, NULL, NULL); mlx5_mprq_free_mp(dev); mlx5_free_shared_dr(priv); if (priv->rss_conf.rss_key != NULL) @@ -1033,23 +1452,15 @@ mlx5_dev_close(struct rte_eth_dev *dev) if (priv->reta_idx != NULL) rte_free(priv->reta_idx); if (priv->config.vf) - mlx5_nl_mac_addr_flush(dev); + mlx5_nl_mac_addr_flush(priv->nl_socket_route, mlx5_ifindex(dev), + dev->data->mac_addrs, + MLX5_MAX_MAC_ADDRESSES, priv->mac_own); if (priv->nl_socket_route >= 0) close(priv->nl_socket_route); if (priv->nl_socket_rdma >= 0) close(priv->nl_socket_rdma); if (priv->vmwa_context) mlx5_vlan_vmwa_exit(priv->vmwa_context); - if (priv->sh) { - /* - * Free the shared context in last turn, because the cleanup - * routines above may use some shared fields, like - * mlx5_nl_mac_addr_flush() uses ibdev_path for retrieveing - * ifindex if Netlink fails. - */ - mlx5_free_shared_ibctx(priv->sh); - priv->sh = NULL; - } ret = mlx5_hrxq_verify(dev); if (ret) DRV_LOG(WARNING, "port %u some hash Rx queue still remain", @@ -1078,6 +1489,16 @@ mlx5_dev_close(struct rte_eth_dev *dev) if (ret) DRV_LOG(WARNING, "port %u some flows still remain", dev->data->port_id); + if (priv->sh) { + /* + * Free the shared context in last turn, because the cleanup + * routines above may use some shared fields, like + * mlx5_nl_mac_addr_flush() uses ibdev_path for retrieveing + * ifindex if Netlink fails. + */ + mlx5_free_shared_ibctx(priv->sh); + priv->sh = NULL; + } if (priv->domain_id != RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID) { unsigned int c = 0; uint16_t port_id; @@ -1150,6 +1571,10 @@ const struct eth_dev_ops mlx5_dev_ops = { .filter_ctrl = mlx5_dev_filter_ctrl, .rx_descriptor_status = mlx5_rx_descriptor_status, .tx_descriptor_status = mlx5_tx_descriptor_status, + .rxq_info_get = mlx5_rxq_info_get, + .txq_info_get = mlx5_txq_info_get, + .rx_burst_mode_get = mlx5_rx_burst_mode_get, + .tx_burst_mode_get = mlx5_tx_burst_mode_get, .rx_queue_count = mlx5_rx_queue_count, .rx_queue_intr_enable = mlx5_rx_intr_enable, .rx_queue_intr_disable = mlx5_rx_intr_disable, @@ -1158,6 +1583,7 @@ const struct eth_dev_ops mlx5_dev_ops = { .get_module_info = mlx5_get_module_info, .get_module_eeprom = mlx5_get_module_eeprom, .hairpin_cap_get = mlx5_hairpin_cap_get, + .mtr_ops_get = mlx5_flow_meter_ops_get, }; /* Available operations from secondary process. */ @@ -1171,6 +1597,10 @@ static const struct eth_dev_ops mlx5_dev_sec_ops = { .dev_infos_get = mlx5_dev_infos_get, .rx_descriptor_status = mlx5_rx_descriptor_status, .tx_descriptor_status = mlx5_tx_descriptor_status, + .rxq_info_get = mlx5_rxq_info_get, + .txq_info_get = mlx5_txq_info_get, + .rx_burst_mode_get = mlx5_rx_burst_mode_get, + .tx_burst_mode_get = mlx5_tx_burst_mode_get, .get_module_info = mlx5_get_module_info, .get_module_eeprom = mlx5_get_module_eeprom, }; @@ -1215,12 +1645,17 @@ const struct eth_dev_ops mlx5_dev_ops_isolate = { .filter_ctrl = mlx5_dev_filter_ctrl, .rx_descriptor_status = mlx5_rx_descriptor_status, .tx_descriptor_status = mlx5_tx_descriptor_status, + .rxq_info_get = mlx5_rxq_info_get, + .txq_info_get = mlx5_txq_info_get, + .rx_burst_mode_get = mlx5_rx_burst_mode_get, + .tx_burst_mode_get = mlx5_tx_burst_mode_get, .rx_queue_intr_enable = mlx5_rx_intr_enable, .rx_queue_intr_disable = mlx5_rx_intr_disable, .is_removed = mlx5_is_removed, .get_module_info = mlx5_get_module_info, .get_module_eeprom = mlx5_get_module_eeprom, .hairpin_cap_get = mlx5_hairpin_cap_get, + .mtr_ops_get = mlx5_flow_meter_ops_get, }; /** @@ -1262,6 +1697,8 @@ mlx5_args_check(const char *key, const char *val, void *opaque) config->mprq.enabled = !!tmp; } else if (strcmp(MLX5_RX_MPRQ_LOG_STRIDE_NUM, key) == 0) { config->mprq.stride_num_n = tmp; + } else if (strcmp(MLX5_RX_MPRQ_LOG_STRIDE_SIZE, key) == 0) { + config->mprq.stride_size_n = tmp; } else if (strcmp(MLX5_RX_MPRQ_MAX_MEMCPY_LEN, key) == 0) { config->mprq.max_memcpy_len = tmp; } else if (strcmp(MLX5_RXQS_MIN_MPRQ, key) == 0) { @@ -1282,6 +1719,16 @@ mlx5_args_check(const char *key, const char *val, void *opaque) DRV_LOG(WARNING, "%s: deprecated parameter, ignored", key); } else if (strcmp(MLX5_TXQ_MPW_EN, key) == 0) { config->mps = !!tmp; + } else if (strcmp(MLX5_TX_DB_NC, key) == 0) { + if (tmp != MLX5_TXDB_CACHED && + tmp != MLX5_TXDB_NCACHED && + tmp != MLX5_TXDB_HEURISTIC) { + DRV_LOG(ERR, "invalid Tx doorbell " + "mapping parameter"); + rte_errno = EINVAL; + return -rte_errno; + } + config->dbnc = tmp; } else if (strcmp(MLX5_TXQ_MPW_HDR_DSEG_EN, key) == 0) { DRV_LOG(WARNING, "%s: deprecated parameter, ignored", key); } else if (strcmp(MLX5_TXQ_MAX_INLINE_LEN, key) == 0) { @@ -1300,12 +1747,26 @@ mlx5_args_check(const char *key, const char *val, void *opaque) config->dv_esw_en = !!tmp; } else if (strcmp(MLX5_DV_FLOW_EN, key) == 0) { config->dv_flow_en = !!tmp; + } else if (strcmp(MLX5_DV_XMETA_EN, key) == 0) { + if (tmp != MLX5_XMETA_MODE_LEGACY && + tmp != MLX5_XMETA_MODE_META16 && + tmp != MLX5_XMETA_MODE_META32) { + DRV_LOG(ERR, "invalid extensive " + "metadata parameter"); + rte_errno = EINVAL; + return -rte_errno; + } + config->dv_xmeta_en = tmp; } else if (strcmp(MLX5_MR_EXT_MEMSEG_EN, key) == 0) { config->mr_ext_memseg_en = !!tmp; } else if (strcmp(MLX5_MAX_DUMP_FILES_NUM, key) == 0) { config->max_dump_files_num = tmp; } else if (strcmp(MLX5_LRO_TIMEOUT_USEC, key) == 0) { config->lro.timeout = tmp; + } else if (strcmp(MLX5_CLASS_ARG_NAME, key) == 0) { + DRV_LOG(DEBUG, "class argument is %s.", val); + } else if (strcmp(MLX5_HP_BUF_SIZE, key) == 0) { + config->log_hp_size = tmp; } else { DRV_LOG(WARNING, "%s: unknown parameter", key); rte_errno = EINVAL; @@ -1334,6 +1795,7 @@ mlx5_args(struct mlx5_dev_config *config, struct rte_devargs *devargs) MLX5_RXQ_PKT_PAD_EN, MLX5_RX_MPRQ_EN, MLX5_RX_MPRQ_LOG_STRIDE_NUM, + MLX5_RX_MPRQ_LOG_STRIDE_SIZE, MLX5_RX_MPRQ_MAX_MEMCPY_LEN, MLX5_RXQS_MIN_MPRQ, MLX5_TXQ_INLINE, @@ -1345,16 +1807,20 @@ mlx5_args(struct mlx5_dev_config *config, struct rte_devargs *devargs) MLX5_TXQ_MPW_EN, MLX5_TXQ_MPW_HDR_DSEG_EN, MLX5_TXQ_MAX_INLINE_LEN, + MLX5_TX_DB_NC, MLX5_TX_VEC_EN, MLX5_RX_VEC_EN, MLX5_L3_VXLAN_EN, MLX5_VF_NL_EN, MLX5_DV_ESW_EN, MLX5_DV_FLOW_EN, + MLX5_DV_XMETA_EN, MLX5_MR_EXT_MEMSEG_EN, MLX5_REPRESENTOR, MLX5_MAX_DUMP_FILES_NUM, MLX5_LRO_TIMEOUT_USEC, + MLX5_CLASS_ARG_NAME, + MLX5_HP_BUF_SIZE, NULL, }; struct rte_kvargs *kvlist; @@ -1407,7 +1873,7 @@ mlx5_init_once(void) if (mlx5_init_shared_data()) return -rte_errno; sd = mlx5_shared_data; - assert(sd); + MLX5_ASSERT(sd); rte_spinlock_lock(&sd->lock); switch (rte_eal_process_type()) { case RTE_PROC_PRIMARY: @@ -1417,7 +1883,8 @@ mlx5_init_once(void) rte_rwlock_init(&sd->mem_event_rwlock); rte_mem_event_callback_register("MLX5_MEM_EVENT_CB", mlx5_mr_mem_event_cb, NULL); - ret = mlx5_mp_init_primary(); + ret = mlx5_mp_init_primary(MLX5_MP_NAME, + mlx5_mp_primary_handle); if (ret) goto out; sd->init_done = true; @@ -1425,7 +1892,8 @@ mlx5_init_once(void) case RTE_PROC_SECONDARY: if (ld->init_done) break; - ret = mlx5_mp_init_secondary(); + ret = mlx5_mp_init_secondary(MLX5_MP_NAME, + mlx5_mp_secondary_handle); if (ret) goto out; ++sd->secondary_cnt; @@ -1447,7 +1915,7 @@ out: * key is specified in devargs * - if DevX is enabled the inline mode is queried from the * device (HCA attributes and NIC vport context if needed). - * - otherwise L2 mode (18 bytes) is assumed for ConnectX-4/4LX + * - otherwise L2 mode (18 bytes) is assumed for ConnectX-4/4 Lx * and none (0 bytes) for other NICs * * @param spawn @@ -1559,6 +2027,60 @@ exit: DRV_LOG(DEBUG, "min tx inline configured: %d", config->txq_inline_min); } +/** + * Configures the metadata mask fields in the shared context. + * + * @param [in] dev + * Pointer to Ethernet device. + */ +static void +mlx5_set_metadata_mask(struct rte_eth_dev *dev) +{ + struct mlx5_priv *priv = dev->data->dev_private; + struct mlx5_ibv_shared *sh = priv->sh; + uint32_t meta, mark, reg_c0; + + reg_c0 = ~priv->vport_meta_mask; + switch (priv->config.dv_xmeta_en) { + case MLX5_XMETA_MODE_LEGACY: + meta = UINT32_MAX; + mark = MLX5_FLOW_MARK_MASK; + break; + case MLX5_XMETA_MODE_META16: + meta = reg_c0 >> rte_bsf32(reg_c0); + mark = MLX5_FLOW_MARK_MASK; + break; + case MLX5_XMETA_MODE_META32: + meta = UINT32_MAX; + mark = (reg_c0 >> rte_bsf32(reg_c0)) & MLX5_FLOW_MARK_MASK; + break; + default: + meta = 0; + mark = 0; + MLX5_ASSERT(false); + break; + } + if (sh->dv_mark_mask && sh->dv_mark_mask != mark) + DRV_LOG(WARNING, "metadata MARK mask mismatche %08X:%08X", + sh->dv_mark_mask, mark); + else + sh->dv_mark_mask = mark; + if (sh->dv_meta_mask && sh->dv_meta_mask != meta) + DRV_LOG(WARNING, "metadata META mask mismatche %08X:%08X", + sh->dv_meta_mask, meta); + else + sh->dv_meta_mask = meta; + if (sh->dv_regc0_mask && sh->dv_regc0_mask != reg_c0) + DRV_LOG(WARNING, "metadata reg_c0 mask mismatche %08X:%08X", + sh->dv_meta_mask, reg_c0); + else + sh->dv_regc0_mask = reg_c0; + DRV_LOG(DEBUG, "metadata mode %u", priv->config.dv_xmeta_en); + DRV_LOG(DEBUG, "metadata MARK mask %08X", sh->dv_mark_mask); + DRV_LOG(DEBUG, "metadata META mask %08X", sh->dv_meta_mask); + DRV_LOG(DEBUG, "metadata reg_c0 mask %08X", sh->dv_regc0_mask); +} + /** * Allocate page of door-bells and register it using DevX API. * @@ -1627,8 +2149,8 @@ mlx5_get_dbr(struct rte_eth_dev *dev, struct mlx5_devx_dbr_page **dbr_page) i++) ; /* Empty. */ /* Find the first clear bit. */ + MLX5_ASSERT(i < MLX5_DBR_BITMAP_SIZE); j = rte_bsf64(~page->dbr_bitmap[i]); - assert(i < (MLX5_DBR_PER_PAGE / 64)); page->dbr_bitmap[i] |= (1 << j); page->dbr_count++; *dbr_page = page; @@ -1679,6 +2201,25 @@ mlx5_release_dbr(struct rte_eth_dev *dev, uint32_t umem_id, uint64_t offset) return ret; } +int +rte_pmd_mlx5_get_dyn_flag_names(char *names[], unsigned int n) +{ + static const char *const dynf_names[] = { + RTE_PMD_MLX5_FINE_GRANULARITY_INLINE, + RTE_MBUF_DYNFLAG_METADATA_NAME + }; + unsigned int i; + + if (n < RTE_DIM(dynf_names)) + return -ENOMEM; + for (i = 0; i < RTE_DIM(dynf_names); i++) { + if (names[i] == NULL) + return -EINVAL; + strcpy(names[i], dynf_names[i]); + } + return RTE_DIM(dynf_names); +} + /** * Check sibling device configurations. * @@ -1702,7 +2243,7 @@ mlx5_dev_check_sibling_config(struct mlx5_priv *priv, struct mlx5_dev_config *sh_conf = NULL; uint16_t port_id; - assert(sh); + MLX5_ASSERT(sh); /* Nothing to compare for the single/first device. */ if (sh->refcnt == 1) return 0; @@ -1724,6 +2265,12 @@ mlx5_dev_check_sibling_config(struct mlx5_priv *priv, rte_errno = EINVAL; return rte_errno; } + if (sh_conf->dv_xmeta_en ^ config->dv_xmeta_en) { + DRV_LOG(ERR, "\"dv_xmeta_en\" configuration mismatch" + " for shared %s context", sh->ibdev_name); + rte_errno = EINVAL; + return rte_errno; + } return 0; } /** @@ -1773,7 +2320,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, uint16_t port_id; unsigned int i; #ifdef HAVE_MLX5DV_DR_DEVX_PORT - struct mlx5dv_devx_port devx_port; + struct mlx5dv_devx_port devx_port = { .comp_mask = 0 }; #endif /* Determine if this port representor is supposed to be spawned. */ @@ -1821,6 +2368,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, } DRV_LOG(DEBUG, "naming Ethernet device \"%s\"", name); if (rte_eal_process_type() == RTE_PROC_SECONDARY) { + struct mlx5_mp_id mp_id; + eth_dev = rte_eth_dev_attach_secondary(name); if (eth_dev == NULL) { DRV_LOG(ERR, "can not attach rte ethdev"); @@ -1832,8 +2381,10 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, err = mlx5_proc_priv_init(eth_dev); if (err) return NULL; + mp_id.port_id = eth_dev->data->port_id; + strlcpy(mp_id.name, MLX5_MP_NAME, RTE_MP_MAX_NAME_LEN); /* Receive command fd from primary process */ - err = mlx5_mp_req_verbs_cmd_fd(eth_dev); + err = mlx5_mp_req_verbs_cmd_fd(&mp_id); if (err < 0) return NULL; /* Remap UAR for Tx queues. */ @@ -1849,7 +2400,20 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, eth_dev->tx_pkt_burst = mlx5_select_tx_function(eth_dev); return eth_dev; } - sh = mlx5_alloc_shared_ibctx(spawn); + /* + * Some parameters ("tx_db_nc" in particularly) are needed in + * advance to create dv/verbs device context. We proceed the + * devargs here to get ones, and later proceed devargs again + * to override some hardware settings. + */ + err = mlx5_args(&config, dpdk_dev->devargs); + if (err) { + err = rte_errno; + DRV_LOG(ERR, "failed to process device arguments: %s", + strerror(rte_errno)); + goto error; + } + sh = mlx5_alloc_shared_ibctx(spawn, &config); if (!sh) return NULL; config.devx = sh->devx; @@ -1913,8 +2477,6 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, mprq_caps.min_single_wqe_log_num_of_strides; mprq_max_stride_num_n = mprq_caps.max_single_wqe_log_num_of_strides; - config.mprq.stride_num_n = RTE_MAX(MLX5_MPRQ_STRIDE_NUM_N, - mprq_min_stride_num_n); } #endif if (RTE_CACHE_LINE_SIZE == 128 && @@ -1933,7 +2495,9 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, tunnel_en = ((dv_attr.tunnel_offloads_caps & MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_VXLAN) && (dv_attr.tunnel_offloads_caps & - MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_GRE)); + MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_GRE) && + (dv_attr.tunnel_offloads_caps & + MLX5DV_RAW_PACKET_CAP_TUNNELED_OFFLOAD_GENEVE)); } DRV_LOG(DEBUG, "tunnel offloading is %ssupported", tunnel_en ? "" : "not "); @@ -1982,6 +2546,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, priv->ibv_port = spawn->ibv_port; priv->pci_dev = spawn->pci_dev; priv->mtu = RTE_ETHER_MTU; + priv->mp_id.port_id = port_id; + strlcpy(priv->mp_id.name, MLX5_MP_NAME, RTE_MP_MAX_NAME_LEN); #ifndef RTE_ARCH_64 /* Initialize UAR access locks for 32bit implementations. */ rte_spinlock_init(&priv->uar_lock_cq); @@ -1991,7 +2557,6 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, /* Some internal functions rely on Netlink sockets, open them now. */ priv->nl_socket_rdma = mlx5_nl_init(NETLINK_RDMA); priv->nl_socket_route = mlx5_nl_init(NETLINK_ROUTE); - priv->nl_sn = 0; priv->representor = !!switch_info->representor; priv->master = !!switch_info->master; priv->domain_id = RTE_ETH_DEV_SWITCH_DOMAIN_ID_INVALID; @@ -2005,36 +2570,41 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, * vport index. The engaged part of metadata register is * defined by mask. */ - devx_port.comp_mask = MLX5DV_DEVX_PORT_VPORT | - MLX5DV_DEVX_PORT_MATCH_REG_C_0; - err = mlx5_glue->devx_port_query(sh->ctx, spawn->ibv_port, &devx_port); - if (err) { - DRV_LOG(WARNING, "can't query devx port %d on device %s\n", - spawn->ibv_port, spawn->ibv_dev->name); - devx_port.comp_mask = 0; + if (switch_info->representor || switch_info->master) { + devx_port.comp_mask = MLX5DV_DEVX_PORT_VPORT | + MLX5DV_DEVX_PORT_MATCH_REG_C_0; + err = mlx5_glue->devx_port_query(sh->ctx, spawn->ibv_port, + &devx_port); + if (err) { + DRV_LOG(WARNING, + "can't query devx port %d on device %s", + spawn->ibv_port, spawn->ibv_dev->name); + devx_port.comp_mask = 0; + } } if (devx_port.comp_mask & MLX5DV_DEVX_PORT_MATCH_REG_C_0) { priv->vport_meta_tag = devx_port.reg_c_0.value; priv->vport_meta_mask = devx_port.reg_c_0.mask; if (!priv->vport_meta_mask) { DRV_LOG(ERR, "vport zero mask for port %d" - " on bonding device %s\n", + " on bonding device %s", spawn->ibv_port, spawn->ibv_dev->name); err = ENOTSUP; goto error; } if (priv->vport_meta_tag & ~priv->vport_meta_mask) { DRV_LOG(ERR, "invalid vport tag for port %d" - " on bonding device %s\n", + " on bonding device %s", spawn->ibv_port, spawn->ibv_dev->name); err = ENOTSUP; goto error; } - } else if (devx_port.comp_mask & MLX5DV_DEVX_PORT_VPORT) { + } + if (devx_port.comp_mask & MLX5DV_DEVX_PORT_VPORT) { priv->vport_id = devx_port.vport_num; } else if (spawn->pf_bond >= 0) { DRV_LOG(ERR, "can't deduce vport index for port %d" - " on bonding device %s\n", + " on bonding device %s", spawn->ibv_port, spawn->ibv_dev->name); err = ENOTSUP; goto error; @@ -2087,13 +2657,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, } own_domain_id = 1; } - err = mlx5_args(&config, dpdk_dev->devargs); - if (err) { - err = rte_errno; - DRV_LOG(ERR, "failed to process device arguments: %s", - strerror(rte_errno)); - goto error; - } + /* Override some values set by hardware configuration. */ + mlx5_args(&config, dpdk_dev->devargs); err = mlx5_dev_check_sibling_config(priv, &config); if (err) goto error; @@ -2105,7 +2670,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, !defined(HAVE_IBV_DEVICE_COUNTERS_SET_V45) DRV_LOG(DEBUG, "counters are not supported"); #endif -#ifndef HAVE_IBV_FLOW_DV_SUPPORT +#if !defined(HAVE_IBV_FLOW_DV_SUPPORT) || !defined(HAVE_MLX5DV_DR) if (config.dv_flow_en) { DRV_LOG(WARNING, "DV flow is not supported"); config.dv_flow_en = 0; @@ -2156,7 +2721,8 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, else config.mps = config.mps ? mps : MLX5_MPW_DISABLED; DRV_LOG(INFO, "%sMPS is %s", - config.mps == MLX5_MPW_ENHANCED ? "enhanced " : "", + config.mps == MLX5_MPW_ENHANCED ? "enhanced " : + config.mps == MLX5_MPW ? "legacy " : "", config.mps != MLX5_MPW_DISABLED ? "enabled" : "disabled"); if (config.cqe_comp && !cqe_comp) { DRV_LOG(WARNING, "Rx CQE compression isn't supported"); @@ -2181,7 +2747,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, priv->counter_fallback = 1; #endif if (priv->counter_fallback) - DRV_LOG(INFO, "Use fall-back DV counter management\n"); + DRV_LOG(INFO, "Use fall-back DV counter management"); /* Check for LRO support. */ if (config.dest_tir && config.hca_attr.lro_cap && config.dv_flow_en) { @@ -2198,19 +2764,60 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, DRV_LOG(DEBUG, "LRO session timeout set to %d usec", config.lro.timeout); } +#if defined(HAVE_MLX5DV_DR) && defined(HAVE_MLX5_DR_CREATE_ACTION_FLOW_METER) + if (config.hca_attr.qos.sup && config.hca_attr.qos.srtcm_sup && + config.dv_flow_en) { + uint8_t reg_c_mask = + config.hca_attr.qos.flow_meter_reg_c_ids; + /* + * Meter needs two REG_C's for color match and pre-sfx + * flow match. Here get the REG_C for color match. + * REG_C_0 and REG_C_1 is reserved for metadata feature. + */ + reg_c_mask &= 0xfc; + if (__builtin_popcount(reg_c_mask) < 1) { + priv->mtr_en = 0; + DRV_LOG(WARNING, "No available register for" + " meter."); + } else { + priv->mtr_color_reg = ffs(reg_c_mask) - 1 + + REG_C_0; + priv->mtr_en = 1; + priv->mtr_reg_share = + config.hca_attr.qos.flow_meter_reg_share; + DRV_LOG(DEBUG, "The REG_C meter uses is %d", + priv->mtr_color_reg); + } + } +#endif } if (config.mprq.enabled && mprq) { - if (config.mprq.stride_num_n > mprq_max_stride_num_n || - config.mprq.stride_num_n < mprq_min_stride_num_n) { + if (config.mprq.stride_num_n && + (config.mprq.stride_num_n > mprq_max_stride_num_n || + config.mprq.stride_num_n < mprq_min_stride_num_n)) { config.mprq.stride_num_n = - RTE_MAX(MLX5_MPRQ_STRIDE_NUM_N, - mprq_min_stride_num_n); + RTE_MIN(RTE_MAX(MLX5_MPRQ_STRIDE_NUM_N, + mprq_min_stride_num_n), + mprq_max_stride_num_n); DRV_LOG(WARNING, "the number of strides" " for Multi-Packet RQ is out of range," " setting default value (%u)", 1 << config.mprq.stride_num_n); } + if (config.mprq.stride_size_n && + (config.mprq.stride_size_n > mprq_max_stride_size_n || + config.mprq.stride_size_n < mprq_min_stride_size_n)) { + config.mprq.stride_size_n = + RTE_MIN(RTE_MAX(MLX5_MPRQ_STRIDE_SIZE_N, + mprq_min_stride_size_n), + mprq_max_stride_size_n); + DRV_LOG(WARNING, + "the size of a stride" + " for Multi-Packet RQ is out of range," + " setting default value (%u)", + 1 << config.mprq.stride_size_n); + } config.mprq.min_stride_size_n = mprq_min_stride_size_n; config.mprq.max_stride_size_n = mprq_max_stride_size_n; } else if (config.mprq.enabled && !mprq) { @@ -2236,7 +2843,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, * is permanent throughout the lifetime of device. So, we may store * the ifindex here and use the cached value further. */ - assert(spawn->ifindex); + MLX5_ASSERT(spawn->ifindex); priv->if_index = spawn->ifindex; eth_dev->data->dev_private = priv; priv->dev_data = eth_dev->data; @@ -2257,7 +2864,7 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2], mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]); -#ifndef NDEBUG +#ifdef RTE_LIBRTE_MLX5_DEBUG { char ifname[IF_NAMESIZE]; @@ -2284,9 +2891,14 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, /* Register MAC address. */ claim_zero(mlx5_mac_addr_add(eth_dev, &mac, 0, 0)); if (config.vf && config.vf_nl_en) - mlx5_nl_mac_addr_sync(eth_dev); - TAILQ_INIT(&priv->flows); - TAILQ_INIT(&priv->ctrl_flows); + mlx5_nl_mac_addr_sync(priv->nl_socket_route, + mlx5_ifindex(eth_dev), + eth_dev->data->mac_addrs, + MLX5_MAX_MAC_ADDRESSES); + priv->flows = 0; + priv->ctrl_flows = 0; + TAILQ_INIT(&priv->flow_meters); + TAILQ_INIT(&priv->flow_meter_profiles); /* Hint libmlx5 to use PMD allocator for data plane resources */ struct mlx5dv_ctx_allocators alctr = { .alloc = &mlx5_alloc_verbs_buf, @@ -2323,6 +2935,17 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, err = mlx5_alloc_shared_dr(priv); if (err) goto error; + /* + * RSS id is shared with meter flow id. Meter flow id can only + * use the 24 MSB of the register. + */ + priv->qrss_id_pool = mlx5_flow_id_pool_alloc(UINT32_MAX >> + MLX5_MTR_COLOR_BITS); + if (!priv->qrss_id_pool) { + DRV_LOG(ERR, "can't create flow id pool"); + err = ENOMEM; + goto error; + } } /* Supported Verbs flow priority number detection. */ err = mlx5_flow_discover_priorities(eth_dev); @@ -2331,9 +2954,60 @@ mlx5_dev_spawn(struct rte_device *dpdk_dev, goto error; } priv->config.flow_prio = err; + if (!priv->config.dv_esw_en && + priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY) { + DRV_LOG(WARNING, "metadata mode %u is not supported " + "(no E-Switch)", priv->config.dv_xmeta_en); + priv->config.dv_xmeta_en = MLX5_XMETA_MODE_LEGACY; + } + mlx5_set_metadata_mask(eth_dev); + if (priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY && + !priv->sh->dv_regc0_mask) { + DRV_LOG(ERR, "metadata mode %u is not supported " + "(no metadata reg_c[0] is available)", + priv->config.dv_xmeta_en); + err = ENOTSUP; + goto error; + } + /* + * Allocate the buffer for flow creating, just once. + * The allocation must be done before any flow creating. + */ + mlx5_flow_alloc_intermediate(eth_dev); + /* Query availibility of metadata reg_c's. */ + err = mlx5_flow_discover_mreg_c(eth_dev); + if (err < 0) { + err = -err; + goto error; + } + if (!mlx5_flow_ext_mreg_supported(eth_dev)) { + DRV_LOG(DEBUG, + "port %u extensive metadata register is not supported", + eth_dev->data->port_id); + if (priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY) { + DRV_LOG(ERR, "metadata mode %u is not supported " + "(no metadata registers available)", + priv->config.dv_xmeta_en); + err = ENOTSUP; + goto error; + } + } + if (priv->config.dv_flow_en && + priv->config.dv_xmeta_en != MLX5_XMETA_MODE_LEGACY && + mlx5_flow_ext_mreg_supported(eth_dev) && + priv->sh->dv_regc0_mask) { + priv->mreg_cp_tbl = mlx5_hlist_create(MLX5_FLOW_MREG_HNAME, + MLX5_FLOW_MREG_HTABLE_SZ); + if (!priv->mreg_cp_tbl) { + err = ENOMEM; + goto error; + } + } return eth_dev; error: if (priv) { + if (priv->mreg_cp_tbl) + mlx5_hlist_destroy(priv->mreg_cp_tbl, NULL, NULL); if (priv->sh) mlx5_free_shared_dr(priv); if (priv->nl_socket_route >= 0) @@ -2342,6 +3016,8 @@ error: close(priv->nl_socket_rdma); if (priv->vmwa_context) mlx5_vlan_vmwa_exit(priv->vmwa_context); + if (priv->qrss_id_pool) + mlx5_flow_id_pool_release(priv->qrss_id_pool); if (own_domain_id) claim_zero(rte_eth_switch_domain_free(priv->domain_id)); rte_free(priv); @@ -2355,7 +3031,7 @@ error: } if (sh) mlx5_free_shared_ibctx(sh); - assert(err > 0); + MLX5_ASSERT(err > 0); rte_errno = err; return NULL; } @@ -2458,7 +3134,7 @@ mlx5_device_bond_pci_match(const struct ibv_device *ibv_dev, if (!file) return -1; /* Use safe format to check maximal buffer length. */ - assert(atol(RTE_STR(IF_NAMESIZE)) == IF_NAMESIZE); + MLX5_ASSERT(atol(RTE_STR(IF_NAMESIZE)) == IF_NAMESIZE); while (fscanf(file, "%" RTE_STR(IF_NAMESIZE) "s", ifname) == 1) { char tmp_str[IF_NAMESIZE + 32]; struct rte_pci_addr pci_addr; @@ -2543,13 +3219,20 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, struct mlx5_dev_config dev_config; int ret; + if (mlx5_class_get(pci_dev->device.devargs) != MLX5_CLASS_NET) { + DRV_LOG(DEBUG, "Skip probing - should be probed by other mlx5" + " driver."); + return 1; + } + if (rte_eal_process_type() == RTE_PROC_PRIMARY) + mlx5_pmd_socket_init(); ret = mlx5_init_once(); if (ret) { DRV_LOG(ERR, "unable to init PMD global data: %s", strerror(rte_errno)); return -rte_errno; } - assert(pci_drv == &mlx5_driver); + MLX5_ASSERT(pci_drv == &mlx5_driver); errno = 0; ibv_list = mlx5_glue->get_device_list(&ret); if (!ibv_list) { @@ -2670,10 +3353,10 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, * it may be E-Switch master device and representors. * We have to perform identification trough the ports. */ - assert(nl_rdma >= 0); - assert(ns == 0); - assert(nd == 1); - assert(np); + MLX5_ASSERT(nl_rdma >= 0); + MLX5_ASSERT(ns == 0); + MLX5_ASSERT(nd == 1); + MLX5_ASSERT(np); for (i = 1; i <= np; ++i) { list[ns].max_port = np; list[ns].ibv_port = i; @@ -2848,7 +3531,7 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, goto exit; } } - assert(ns); + MLX5_ASSERT(ns); /* * Sort list to probe devices in natural order for users convenience * (i.e. master first, then representors from lowest to highest ID). @@ -2858,6 +3541,7 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, dev_config = (struct mlx5_dev_config){ .hw_padding = 0, .mps = MLX5_ARG_UNSET, + .dbnc = MLX5_ARG_UNSET, .rx_vec_en = 1, .txq_inline_max = MLX5_ARG_UNSET, .txq_inline_min = MLX5_ARG_UNSET, @@ -2867,11 +3551,14 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, .mr_ext_memseg_en = 1, .mprq = { .enabled = 0, /* Disabled by default. */ - .stride_num_n = MLX5_MPRQ_STRIDE_NUM_N, + .stride_num_n = 0, + .stride_size_n = 0, .max_memcpy_len = MLX5_MPRQ_MEMCPY_DEFAULT_LEN, .min_rxqs_num = MLX5_MPRQ_MIN_RXQS, }, .dv_esw_en = 1, + .dv_flow_en = 1, + .log_hp_size = MLX5_ARG_UNSET, }; /* Device specific configuration. */ switch (pci_dev->id.device_id) { @@ -2881,6 +3568,7 @@ mlx5_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, case PCI_DEVICE_ID_MELLANOX_CONNECTX5EXVF: case PCI_DEVICE_ID_MELLANOX_CONNECTX5BFVF: case PCI_DEVICE_ID_MELLANOX_CONNECTX6VF: + case PCI_DEVICE_ID_MELLANOX_CONNECTX6DXVF: dev_config.vf = 1; break; default: @@ -2940,7 +3628,7 @@ exit: close(nl_route); if (list) rte_free(list); - assert(ibv_list); + MLX5_ASSERT(ibv_list); mlx5_glue->free_device_list(ibv_list); return ret; } @@ -3050,6 +3738,18 @@ static const struct rte_pci_id mlx5_pci_id_map[] = { RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, PCI_DEVICE_ID_MELLANOX_CONNECTX6VF) }, + { + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX6DX) + }, + { + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX6DXVF) + }, + { + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX6DXBF) + }, { .vendor_id = 0 } @@ -3068,138 +3768,6 @@ static struct rte_pci_driver mlx5_driver = { RTE_PCI_DRV_PROBE_AGAIN, }; -#ifdef RTE_IBVERBS_LINK_DLOPEN - -/** - * Suffix RTE_EAL_PMD_PATH with "-glue". - * - * This function performs a sanity check on RTE_EAL_PMD_PATH before - * suffixing its last component. - * - * @param buf[out] - * Output buffer, should be large enough otherwise NULL is returned. - * @param size - * Size of @p out. - * - * @return - * Pointer to @p buf or @p NULL in case suffix cannot be appended. - */ -static char * -mlx5_glue_path(char *buf, size_t size) -{ - static const char *const bad[] = { "/", ".", "..", NULL }; - const char *path = RTE_EAL_PMD_PATH; - size_t len = strlen(path); - size_t off; - int i; - - while (len && path[len - 1] == '/') - --len; - for (off = len; off && path[off - 1] != '/'; --off) - ; - for (i = 0; bad[i]; ++i) - if (!strncmp(path + off, bad[i], (int)(len - off))) - goto error; - i = snprintf(buf, size, "%.*s-glue", (int)len, path); - if (i == -1 || (size_t)i >= size) - goto error; - return buf; -error: - DRV_LOG(ERR, - "unable to append \"-glue\" to last component of" - " RTE_EAL_PMD_PATH (\"" RTE_EAL_PMD_PATH "\")," - " please re-configure DPDK"); - return NULL; -} - -/** - * Initialization routine for run-time dependency on rdma-core. - */ -static int -mlx5_glue_init(void) -{ - char glue_path[sizeof(RTE_EAL_PMD_PATH) - 1 + sizeof("-glue")]; - const char *path[] = { - /* - * A basic security check is necessary before trusting - * MLX5_GLUE_PATH, which may override RTE_EAL_PMD_PATH. - */ - (geteuid() == getuid() && getegid() == getgid() ? - getenv("MLX5_GLUE_PATH") : NULL), - /* - * When RTE_EAL_PMD_PATH is set, use its glue-suffixed - * variant, otherwise let dlopen() look up libraries on its - * own. - */ - (*RTE_EAL_PMD_PATH ? - mlx5_glue_path(glue_path, sizeof(glue_path)) : ""), - }; - unsigned int i = 0; - void *handle = NULL; - void **sym; - const char *dlmsg; - - while (!handle && i != RTE_DIM(path)) { - const char *end; - size_t len; - int ret; - - if (!path[i]) { - ++i; - continue; - } - end = strpbrk(path[i], ":;"); - if (!end) - end = path[i] + strlen(path[i]); - len = end - path[i]; - ret = 0; - do { - char name[ret + 1]; - - ret = snprintf(name, sizeof(name), "%.*s%s" MLX5_GLUE, - (int)len, path[i], - (!len || *(end - 1) == '/') ? "" : "/"); - if (ret == -1) - break; - if (sizeof(name) != (size_t)ret + 1) - continue; - DRV_LOG(DEBUG, "looking for rdma-core glue as \"%s\"", - name); - handle = dlopen(name, RTLD_LAZY); - break; - } while (1); - path[i] = end + 1; - if (!*end) - ++i; - } - if (!handle) { - rte_errno = EINVAL; - dlmsg = dlerror(); - if (dlmsg) - DRV_LOG(WARNING, "cannot load glue library: %s", dlmsg); - goto glue_error; - } - sym = dlsym(handle, "mlx5_glue"); - if (!sym || !*sym) { - rte_errno = EINVAL; - dlmsg = dlerror(); - if (dlmsg) - DRV_LOG(ERR, "cannot resolve glue symbol: %s", dlmsg); - goto glue_error; - } - mlx5_glue = *sym; - return 0; -glue_error: - if (handle) - dlclose(handle); - DRV_LOG(WARNING, - "cannot initialize PMD due to missing run-time dependency on" - " rdma-core libraries (libibverbs, libmlx5)"); - return -rte_errno; -} - -#endif - /** * Driver initialization routine. */ @@ -3214,43 +3782,8 @@ RTE_INIT(rte_mlx5_pmd_init) mlx5_set_ptype_table(); mlx5_set_cksum_table(); mlx5_set_swp_types_table(); - /* - * RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use - * huge pages. Calling ibv_fork_init() during init allows - * applications to use fork() safely for purposes other than - * using this PMD, which is not supported in forked processes. - */ - setenv("RDMAV_HUGEPAGES_SAFE", "1", 1); - /* Match the size of Rx completion entry to the size of a cacheline. */ - if (RTE_CACHE_LINE_SIZE == 128) - setenv("MLX5_CQE_SIZE", "128", 0); - /* - * MLX5_DEVICE_FATAL_CLEANUP tells ibv_destroy functions to - * cleanup all the Verbs resources even when the device was removed. - */ - setenv("MLX5_DEVICE_FATAL_CLEANUP", "1", 1); -#ifdef RTE_IBVERBS_LINK_DLOPEN - if (mlx5_glue_init()) - return; - assert(mlx5_glue); -#endif -#ifndef NDEBUG - /* Glue structure must not contain any NULL pointers. */ - { - unsigned int i; - - for (i = 0; i != sizeof(*mlx5_glue) / sizeof(void *); ++i) - assert(((const void *const *)mlx5_glue)[i]); - } -#endif - if (strcmp(mlx5_glue->version, MLX5_GLUE_VERSION)) { - DRV_LOG(ERR, - "rdma-core glue \"%s\" mismatch: \"%s\" is required", - mlx5_glue->version, MLX5_GLUE_VERSION); - return; - } - mlx5_glue->fork_init(); - rte_pci_register(&mlx5_driver); + if (mlx5_glue) + rte_pci_register(&mlx5_driver); } RTE_PMD_EXPORT_NAME(net_mlx5, __COUNTER__);