X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fsfc%2Fsfc_ethdev.c;h=dd9897d7018e43d2b35bd3799f67afa693f112af;hb=538da7a1cad25fbdffe298c8ca76fc4dbd262d1b;hp=f16d5208126006aca7cbdf4033ae5f5909ccc86b;hpb=a6fae8f906ef82ca3caf49344efa3132b0a02b4c;p=dpdk.git diff --git a/drivers/net/sfc/sfc_ethdev.c b/drivers/net/sfc/sfc_ethdev.c index f16d520812..dd9897d701 100644 --- a/drivers/net/sfc/sfc_ethdev.c +++ b/drivers/net/sfc/sfc_ethdev.c @@ -13,6 +13,8 @@ #include #include #include +#include +#include #include "efx.h" @@ -35,7 +37,7 @@ static struct sfc_dp_list sfc_dp_head = static int sfc_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); efx_nic_fw_info_t enfi; int ret; int rc; @@ -83,23 +85,33 @@ sfc_fw_version_get(struct rte_eth_dev *dev, char *fw_version, size_t fw_size) static void sfc_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) { - struct sfc_adapter *sa = dev->data->dev_private; - const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic); + const struct sfc_adapter_priv *sap = sfc_adapter_priv_by_eth_dev(dev); + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); + struct sfc_rss *rss = &sas->rss; uint64_t txq_offloads_def = 0; sfc_log_init(sa, "entry"); - dev_info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); + dev_info->min_mtu = ETHER_MIN_MTU; + dev_info->max_mtu = EFX_MAC_SDU_MAX; + dev_info->max_rx_pktlen = EFX_MAC_PDU_MAX; /* Autonegotiation may be disabled */ dev_info->speed_capa = ETH_LINK_SPEED_FIXED; - if (sa->port.phy_adv_cap_mask & EFX_PHY_CAP_1000FDX) + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_1000FDX)) dev_info->speed_capa |= ETH_LINK_SPEED_1G; - if (sa->port.phy_adv_cap_mask & EFX_PHY_CAP_10000FDX) + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_10000FDX)) dev_info->speed_capa |= ETH_LINK_SPEED_10G; - if (sa->port.phy_adv_cap_mask & EFX_PHY_CAP_40000FDX) + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_25000FDX)) + dev_info->speed_capa |= ETH_LINK_SPEED_25G; + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_40000FDX)) dev_info->speed_capa |= ETH_LINK_SPEED_40G; + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_50000FDX)) + dev_info->speed_capa |= ETH_LINK_SPEED_50G; + if (sa->port.phy_adv_cap_mask & (1u << EFX_PHY_CAP_100000FDX)) + dev_info->speed_capa |= ETH_LINK_SPEED_100G; dev_info->max_rx_queues = sa->rxq_max; dev_info->max_tx_queues = sa->txq_max; @@ -132,66 +144,57 @@ sfc_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) dev_info->default_txconf.offloads |= txq_offloads_def; - dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP; - if ((~sa->dp_tx->features & SFC_DP_TX_FEAT_VLAN_INSERT) || - !encp->enc_hw_tx_insert_vlan_enabled) - dev_info->default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOVLANOFFL; - - if (~sa->dp_tx->features & SFC_DP_TX_FEAT_MULTI_SEG) - dev_info->default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOMULTSEGS; + if (rss->context_type != EFX_RX_SCALE_UNAVAILABLE) { + uint64_t rte_hf = 0; + unsigned int i; - if (~sa->dp_tx->features & SFC_DP_TX_FEAT_MULTI_POOL) - dev_info->default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOMULTMEMP; + for (i = 0; i < rss->hf_map_nb_entries; ++i) + rte_hf |= rss->hf_map[i].rte; - if (~sa->dp_tx->features & SFC_DP_TX_FEAT_REFCNT) - dev_info->default_txconf.txq_flags |= ETH_TXQ_FLAGS_NOREFCOUNT; - -#if EFSYS_OPT_RX_SCALE - if (sa->rss_support != EFX_RX_SCALE_UNAVAILABLE) { dev_info->reta_size = EFX_RSS_TBL_SIZE; dev_info->hash_key_size = EFX_RSS_KEY_SIZE; - dev_info->flow_type_rss_offloads = SFC_RSS_OFFLOADS; + dev_info->flow_type_rss_offloads = rte_hf; } -#endif /* Initialize to hardware limits */ - dev_info->rx_desc_lim.nb_max = EFX_RXQ_MAXNDESCS; - dev_info->rx_desc_lim.nb_min = EFX_RXQ_MINNDESCS; + dev_info->rx_desc_lim.nb_max = sa->rxq_max_entries; + dev_info->rx_desc_lim.nb_min = sa->rxq_min_entries; /* The RXQ hardware requires that the descriptor count is a power * of 2, but rx_desc_lim cannot properly describe that constraint. */ - dev_info->rx_desc_lim.nb_align = EFX_RXQ_MINNDESCS; + dev_info->rx_desc_lim.nb_align = sa->rxq_min_entries; /* Initialize to hardware limits */ dev_info->tx_desc_lim.nb_max = sa->txq_max_entries; - dev_info->tx_desc_lim.nb_min = EFX_TXQ_MINNDESCS; + dev_info->tx_desc_lim.nb_min = sa->txq_min_entries; /* * The TXQ hardware requires that the descriptor count is a power * of 2, but tx_desc_lim cannot properly describe that constraint */ - dev_info->tx_desc_lim.nb_align = EFX_TXQ_MINNDESCS; + dev_info->tx_desc_lim.nb_align = sa->txq_min_entries; + + if (sap->dp_rx->get_dev_info != NULL) + sap->dp_rx->get_dev_info(dev_info); + if (sap->dp_tx->get_dev_info != NULL) + sap->dp_tx->get_dev_info(dev_info); - if (sa->dp_rx->get_dev_info != NULL) - sa->dp_rx->get_dev_info(dev_info); - if (sa->dp_tx->get_dev_info != NULL) - sa->dp_tx->get_dev_info(dev_info); + dev_info->dev_capa = RTE_ETH_DEV_CAPA_RUNTIME_RX_QUEUE_SETUP | + RTE_ETH_DEV_CAPA_RUNTIME_TX_QUEUE_SETUP; } static const uint32_t * sfc_dev_supported_ptypes_get(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; - const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic); - uint32_t tunnel_encaps = encp->enc_tunnel_encapsulations_supported; + const struct sfc_adapter_priv *sap = sfc_adapter_priv_by_eth_dev(dev); - return sa->dp_rx->supported_ptypes_get(tunnel_encaps); + return sap->dp_rx->supported_ptypes_get(sap->shared->tunnel_encaps); } static int sfc_dev_configure(struct rte_eth_dev *dev) { struct rte_eth_dev_data *dev_data = dev->data; - struct sfc_adapter *sa = dev_data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "entry n_rxq=%u n_txq=%u", @@ -220,7 +223,7 @@ sfc_dev_configure(struct rte_eth_dev *dev) static int sfc_dev_start(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "entry"); @@ -237,7 +240,7 @@ sfc_dev_start(struct rte_eth_dev *dev) static int sfc_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct rte_eth_link current_link; int ret; @@ -268,7 +271,7 @@ sfc_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete) static void sfc_dev_stop(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "entry"); @@ -282,7 +285,7 @@ sfc_dev_stop(struct rte_eth_dev *dev) static int sfc_dev_set_link_up(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "entry"); @@ -298,7 +301,7 @@ sfc_dev_set_link_up(struct rte_eth_dev *dev) static int sfc_dev_set_link_down(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "entry"); @@ -312,7 +315,7 @@ sfc_dev_set_link_down(struct rte_eth_dev *dev) static void sfc_dev_close(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "entry"); @@ -343,7 +346,7 @@ sfc_dev_filter_set(struct rte_eth_dev *dev, enum sfc_dev_filter_mode mode, { struct sfc_port *port; boolean_t *toggle; - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); boolean_t allmulti = (mode == SFC_DEV_FILTER_MODE_ALLMULTI); const char *desc = (allmulti) ? "all-multi" : "promiscuous"; @@ -355,7 +358,7 @@ sfc_dev_filter_set(struct rte_eth_dev *dev, enum sfc_dev_filter_mode mode, if (*toggle != enabled) { *toggle = enabled; - if (port->isolated) { + if (sfc_sa2shared(sa)->isolated) { sfc_warn(sa, "isolated mode is active on the port"); sfc_warn(sa, "the change is to be applied on the next " "start provided that isolated mode is " @@ -401,7 +404,8 @@ sfc_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id, const struct rte_eth_rxconf *rx_conf, struct rte_mempool *mb_pool) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "RxQ=%u nb_rx_desc=%u socket_id=%u", @@ -414,7 +418,7 @@ sfc_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id, if (rc != 0) goto fail_rx_qinit; - dev->data->rx_queues[rx_queue_id] = sa->rxq_info[rx_queue_id].rxq->dp; + dev->data->rx_queues[rx_queue_id] = sas->rxq_info[rx_queue_id].dp; sfc_adapter_unlock(sa); @@ -441,12 +445,10 @@ sfc_rx_queue_release(void *queue) sa = rxq->evq->sa; sfc_adapter_lock(sa); - sw_index = sfc_rxq_sw_index(rxq); + sw_index = dp_rxq->dpq.queue_id; sfc_log_init(sa, "RxQ=%u", sw_index); - sa->eth_dev->data->rx_queues[sw_index] = NULL; - sfc_rx_qfini(sa, sw_index); sfc_adapter_unlock(sa); @@ -457,7 +459,8 @@ sfc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id, uint16_t nb_tx_desc, unsigned int socket_id, const struct rte_eth_txconf *tx_conf) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "TxQ = %u, nb_tx_desc = %u, socket_id = %u", @@ -469,7 +472,7 @@ sfc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id, if (rc != 0) goto fail_tx_qinit; - dev->data->tx_queues[tx_queue_id] = sa->txq_info[tx_queue_id].txq->dp; + dev->data->tx_queues[tx_queue_id] = sas->txq_info[tx_queue_id].dp; sfc_adapter_unlock(sa); return 0; @@ -492,7 +495,7 @@ sfc_tx_queue_release(void *queue) return; txq = sfc_txq_by_dp_txq(dp_txq); - sw_index = sfc_txq_sw_index(txq); + sw_index = dp_txq->dpq.queue_id; SFC_ASSERT(txq->evq != NULL); sa = txq->evq->sa; @@ -501,18 +504,38 @@ sfc_tx_queue_release(void *queue) sfc_adapter_lock(sa); - SFC_ASSERT(sw_index < sa->eth_dev->data->nb_tx_queues); - sa->eth_dev->data->tx_queues[sw_index] = NULL; - sfc_tx_qfini(sa, sw_index); sfc_adapter_unlock(sa); } +/* + * Some statistics are computed as A - B where A and B each increase + * monotonically with some hardware counter(s) and the counters are read + * asynchronously. + * + * If packet X is counted in A, but not counted in B yet, computed value is + * greater than real. + * + * If packet X is not counted in A at the moment of reading the counter, + * but counted in B at the moment of reading the counter, computed value + * is less than real. + * + * However, counter which grows backward is worse evil than slightly wrong + * value. So, let's try to guarantee that it never happens except may be + * the case when the MAC stats are zeroed as a result of a NIC reset. + */ +static void +sfc_update_diff_stat(uint64_t *stat, uint64_t newval) +{ + if ((int64_t)(newval - *stat) > 0 || newval == 0) + *stat = newval; +} + static int sfc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; uint64_t *mac_stats; int ret; @@ -543,11 +566,9 @@ sfc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) mac_stats[EFX_MAC_VADAPTER_TX_UNICAST_BYTES] + mac_stats[EFX_MAC_VADAPTER_TX_MULTICAST_BYTES] + mac_stats[EFX_MAC_VADAPTER_TX_BROADCAST_BYTES]; - stats->imissed = mac_stats[EFX_MAC_VADAPTER_RX_OVERFLOW]; - stats->ierrors = mac_stats[EFX_MAC_VADAPTER_RX_BAD_PACKETS]; + stats->imissed = mac_stats[EFX_MAC_VADAPTER_RX_BAD_PACKETS]; stats->oerrors = mac_stats[EFX_MAC_VADAPTER_TX_BAD_PACKETS]; } else { - stats->ipackets = mac_stats[EFX_MAC_RX_PKTS]; stats->opackets = mac_stats[EFX_MAC_TX_PKTS]; stats->ibytes = mac_stats[EFX_MAC_RX_OCTETS]; stats->obytes = mac_stats[EFX_MAC_TX_OCTETS]; @@ -573,6 +594,13 @@ sfc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) mac_stats[EFX_MAC_RX_ALIGN_ERRORS] + mac_stats[EFX_MAC_RX_JABBER_PKTS]; /* no oerrors counters supported on EF10 */ + + /* Exclude missed, errors and pauses from Rx packets */ + sfc_update_diff_stat(&port->ipackets, + mac_stats[EFX_MAC_RX_PKTS] - + mac_stats[EFX_MAC_RX_PAUSE_PKTS] - + stats->imissed - stats->ierrors); + stats->ipackets = port->ipackets; } unlock: @@ -584,7 +612,7 @@ unlock: static void sfc_stats_reset(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; int rc; @@ -606,7 +634,7 @@ static int sfc_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats, unsigned int xstats_count) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; uint64_t *mac_stats; int rc; @@ -645,7 +673,7 @@ sfc_xstats_get_names(struct rte_eth_dev *dev, struct rte_eth_xstat_name *xstats_names, unsigned int xstats_count) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; unsigned int i; unsigned int nstats = 0; @@ -653,7 +681,7 @@ sfc_xstats_get_names(struct rte_eth_dev *dev, for (i = 0; i < EFX_MAC_NSTATS; ++i) { if (EFX_MAC_STAT_SUPPORTED(port->mac_stats_mask, i)) { if (xstats_names != NULL && nstats < xstats_count) - strncpy(xstats_names[nstats].name, + strlcpy(xstats_names[nstats].name, efx_mac_stat_name(sa->nic, i), sizeof(xstats_names[0].name)); nstats++; @@ -667,7 +695,7 @@ static int sfc_xstats_get_by_id(struct rte_eth_dev *dev, const uint64_t *ids, uint64_t *values, unsigned int n) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; uint64_t *mac_stats; unsigned int nb_supported = 0; @@ -714,7 +742,7 @@ sfc_xstats_get_names_by_id(struct rte_eth_dev *dev, struct rte_eth_xstat_name *xstats_names, const uint64_t *ids, unsigned int size) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; unsigned int nb_supported = 0; unsigned int nb_written = 0; @@ -731,9 +759,8 @@ sfc_xstats_get_names_by_id(struct rte_eth_dev *dev, if ((ids == NULL) || (ids[nb_written] == nb_supported)) { char *name = xstats_names[nb_written++].name; - strncpy(name, efx_mac_stat_name(sa->nic, i), + strlcpy(name, efx_mac_stat_name(sa->nic, i), sizeof(xstats_names[0].name)); - name[sizeof(xstats_names[0].name) - 1] = '\0'; } ++nb_supported; @@ -745,7 +772,7 @@ sfc_xstats_get_names_by_id(struct rte_eth_dev *dev, static int sfc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); unsigned int wanted_fc, link_fc; memset(fc_conf, 0, sizeof(*fc_conf)); @@ -785,7 +812,7 @@ sfc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) static int sfc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; unsigned int fcntl; int rc; @@ -838,10 +865,38 @@ fail_inval: return -rc; } +static int +sfc_check_scatter_on_all_rx_queues(struct sfc_adapter *sa, size_t pdu) +{ + struct sfc_adapter_shared * const sas = sfc_sa2shared(sa); + const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic); + boolean_t scatter_enabled; + const char *error; + unsigned int i; + + for (i = 0; i < sas->rxq_count; i++) { + if ((sas->rxq_info[i].state & SFC_RXQ_INITIALIZED) == 0) + continue; + + scatter_enabled = (sas->rxq_info[i].type_flags & + EFX_RXQ_FLAG_SCATTER); + + if (!sfc_rx_check_scatter(pdu, sa->rxq_ctrl[i].buf_size, + encp->enc_rx_prefix_size, + scatter_enabled, &error)) { + sfc_err(sa, "MTU check for RxQ %u failed: %s", i, + error); + return EINVAL; + } + } + + return 0; +} + static int sfc_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); size_t pdu = EFX_MAC_PDU(mtu); size_t old_pdu; int rc; @@ -864,6 +919,10 @@ sfc_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) sfc_adapter_lock(sa); + rc = sfc_check_scatter_on_all_rx_queues(sa, pdu); + if (rc != 0) + goto fail_check_scatter; + if (pdu != sa->port.pdu) { if (sa->state == SFC_ADAPTER_STARTED) { sfc_stop(sa); @@ -879,14 +938,12 @@ sfc_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) } /* - * The driver does not use it, but other PMDs update jumbo_frame + * The driver does not use it, but other PMDs update jumbo frame * flag and max_rx_pkt_len when MTU is set. */ if (mtu > ETHER_MAX_LEN) { struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode; - rxmode->offloads |= DEV_RX_OFFLOAD_JUMBO_FRAME; - rxmode->jumbo_frame = 1; } dev->data->dev_conf.rxmode.max_rx_pkt_len = sa->port.pdu; @@ -902,6 +959,8 @@ fail_start: sfc_err(sa, "cannot start with neither new (%u) nor old (%u) " "PDU max size - port is stopped", (unsigned int)pdu, (unsigned int)old_pdu); + +fail_check_scatter: sfc_adapter_unlock(sa); fail_inval: @@ -909,13 +968,14 @@ fail_inval: SFC_ASSERT(rc > 0); return -rc; } -static void -sfc_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr) +static int +sfc_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); const efx_nic_cfg_t *encp = efx_nic_cfg_get(sa->nic); struct sfc_port *port = &sa->port; - int rc; + struct rte_ether_addr *old_addr = &dev->data->mac_addrs[0]; + int rc = 0; sfc_adapter_lock(sa); @@ -923,11 +983,18 @@ sfc_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr) * Copy the address to the device private data so that * it could be recalled in the case of adapter restart. */ - ether_addr_copy(mac_addr, &port->default_mac_addr); + rte_ether_addr_copy(mac_addr, &port->default_mac_addr); - if (port->isolated) { - sfc_err(sa, "isolated mode is active on the port"); - sfc_err(sa, "will not set MAC address"); + /* + * Neither of the two following checks can return + * an error. The new MAC address is preserved in + * the device private data and can be activated + * on the next port start if the user prevents + * isolated mode from being enabled. + */ + if (sfc_sa2shared(sa)->isolated) { + sfc_warn(sa, "isolated mode is active on the port"); + sfc_warn(sa, "will not set MAC address"); goto unlock; } @@ -951,8 +1018,12 @@ sfc_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr) * we also need to update unicast filters */ rc = sfc_set_rx_mode(sa); - if (rc != 0) + if (rc != 0) { sfc_err(sa, "cannot set filter (rc = %u)", rc); + /* Rollback the old address */ + (void)efx_mac_addr_set(sa->nic, old_addr->addr_bytes); + (void)sfc_set_rx_mode(sa); + } } else { sfc_warn(sa, "cannot set MAC address with filters installed"); sfc_warn(sa, "adapter will be restarted to pick the new MAC"); @@ -971,28 +1042,27 @@ sfc_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr) } unlock: - /* - * In the case of failure sa->port->default_mac_addr does not - * need rollback since no error code is returned, and the upper - * API will anyway update the external MAC address storage. - * To be consistent with that new value it is better to keep - * the device private value the same. - */ + if (rc != 0) + rte_ether_addr_copy(old_addr, &port->default_mac_addr); + sfc_adapter_unlock(sa); + + SFC_ASSERT(rc >= 0); + return -rc; } static int -sfc_set_mc_addr_list(struct rte_eth_dev *dev, struct ether_addr *mc_addr_set, - uint32_t nb_mc_addr) +sfc_set_mc_addr_list(struct rte_eth_dev *dev, + struct rte_ether_addr *mc_addr_set, uint32_t nb_mc_addr) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); struct sfc_port *port = &sa->port; uint8_t *mc_addrs = port->mcast_addrs; int rc; unsigned int i; - if (port->isolated) { + if (sfc_sa2shared(sa)->isolated) { sfc_err(sa, "isolated mode is active on the port"); sfc_err(sa, "will not set multicast address list"); return -ENOTSUP; @@ -1023,7 +1093,7 @@ sfc_set_mc_addr_list(struct rte_eth_dev *dev, struct ether_addr *mc_addr_set, if (rc != 0) sfc_err(sa, "cannot set multicast address list (rc = %u)", rc); - SFC_ASSERT(rc > 0); + SFC_ASSERT(rc >= 0); return -rc; } @@ -1035,32 +1105,23 @@ static void sfc_rx_queue_info_get(struct rte_eth_dev *dev, uint16_t rx_queue_id, struct rte_eth_rxq_info *qinfo) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); struct sfc_rxq_info *rxq_info; - struct sfc_rxq *rxq; - - sfc_adapter_lock(sa); - SFC_ASSERT(rx_queue_id < sa->rxq_count); + SFC_ASSERT(rx_queue_id < sas->rxq_count); - rxq_info = &sa->rxq_info[rx_queue_id]; - rxq = rxq_info->rxq; - SFC_ASSERT(rxq != NULL); + rxq_info = &sas->rxq_info[rx_queue_id]; - qinfo->mp = rxq->refill_mb_pool; - qinfo->conf.rx_free_thresh = rxq->refill_threshold; + qinfo->mp = rxq_info->refill_mb_pool; + qinfo->conf.rx_free_thresh = rxq_info->refill_threshold; qinfo->conf.rx_drop_en = 1; qinfo->conf.rx_deferred_start = rxq_info->deferred_start; - qinfo->conf.offloads = DEV_RX_OFFLOAD_IPV4_CKSUM | - DEV_RX_OFFLOAD_UDP_CKSUM | - DEV_RX_OFFLOAD_TCP_CKSUM; + qinfo->conf.offloads = dev->data->dev_conf.rxmode.offloads; if (rxq_info->type_flags & EFX_RXQ_FLAG_SCATTER) { qinfo->conf.offloads |= DEV_RX_OFFLOAD_SCATTER; qinfo->scattered_rx = 1; } qinfo->nb_desc = rxq_info->entries; - - sfc_adapter_unlock(sa); } /* @@ -1071,67 +1132,91 @@ static void sfc_tx_queue_info_get(struct rte_eth_dev *dev, uint16_t tx_queue_id, struct rte_eth_txq_info *qinfo) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); struct sfc_txq_info *txq_info; - sfc_adapter_lock(sa); - - SFC_ASSERT(tx_queue_id < sa->txq_count); + SFC_ASSERT(tx_queue_id < sas->txq_count); - txq_info = &sa->txq_info[tx_queue_id]; - SFC_ASSERT(txq_info->txq != NULL); + txq_info = &sas->txq_info[tx_queue_id]; memset(qinfo, 0, sizeof(*qinfo)); - qinfo->conf.txq_flags = txq_info->txq->flags; - qinfo->conf.offloads = txq_info->txq->offloads; - qinfo->conf.tx_free_thresh = txq_info->txq->free_thresh; + qinfo->conf.offloads = txq_info->offloads; + qinfo->conf.tx_free_thresh = txq_info->free_thresh; qinfo->conf.tx_deferred_start = txq_info->deferred_start; qinfo->nb_desc = txq_info->entries; - - sfc_adapter_unlock(sa); } +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static uint32_t sfc_rx_queue_count(struct rte_eth_dev *dev, uint16_t rx_queue_id) { - struct sfc_adapter *sa = dev->data->dev_private; + const struct sfc_adapter_priv *sap = sfc_adapter_priv_by_eth_dev(dev); + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_rxq_info *rxq_info; - sfc_log_init(sa, "RxQ=%u", rx_queue_id); + SFC_ASSERT(rx_queue_id < sas->rxq_count); + rxq_info = &sas->rxq_info[rx_queue_id]; - return sfc_rx_qdesc_npending(sa, rx_queue_id); + if ((rxq_info->state & SFC_RXQ_STARTED) == 0) + return 0; + + return sap->dp_rx->qdesc_npending(rxq_info->dp); } +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static int sfc_rx_descriptor_done(void *queue, uint16_t offset) { struct sfc_dp_rxq *dp_rxq = queue; + const struct sfc_dp_rx *dp_rx; + + dp_rx = sfc_dp_rx_by_dp_rxq(dp_rxq); - return sfc_rx_qdesc_done(dp_rxq, offset); + return offset < dp_rx->qdesc_npending(dp_rxq); } +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static int sfc_rx_descriptor_status(void *queue, uint16_t offset) { struct sfc_dp_rxq *dp_rxq = queue; - struct sfc_rxq *rxq = sfc_rxq_by_dp_rxq(dp_rxq); + const struct sfc_dp_rx *dp_rx; + + dp_rx = sfc_dp_rx_by_dp_rxq(dp_rxq); - return rxq->evq->sa->dp_rx->qdesc_status(dp_rxq, offset); + return dp_rx->qdesc_status(dp_rxq, offset); } +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static int sfc_tx_descriptor_status(void *queue, uint16_t offset) { struct sfc_dp_txq *dp_txq = queue; - struct sfc_txq *txq = sfc_txq_by_dp_txq(dp_txq); + const struct sfc_dp_tx *dp_tx; - return txq->evq->sa->dp_tx->qdesc_status(dp_txq, offset); + dp_tx = sfc_dp_tx_by_dp_txq(dp_txq); + + return dp_tx->qdesc_status(dp_txq, offset); } static int sfc_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "RxQ=%u", rx_queue_id); @@ -1142,17 +1227,21 @@ sfc_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id) if (sa->state != SFC_ADAPTER_STARTED) goto fail_not_started; + if (sas->rxq_info[rx_queue_id].state != SFC_RXQ_INITIALIZED) + goto fail_not_setup; + rc = sfc_rx_qstart(sa, rx_queue_id); if (rc != 0) goto fail_rx_qstart; - sa->rxq_info[rx_queue_id].deferred_started = B_TRUE; + sas->rxq_info[rx_queue_id].deferred_started = B_TRUE; sfc_adapter_unlock(sa); return 0; fail_rx_qstart: +fail_not_setup: fail_not_started: sfc_adapter_unlock(sa); SFC_ASSERT(rc > 0); @@ -1162,14 +1251,15 @@ fail_not_started: static int sfc_rx_queue_stop(struct rte_eth_dev *dev, uint16_t rx_queue_id) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "RxQ=%u", rx_queue_id); sfc_adapter_lock(sa); sfc_rx_qstop(sa, rx_queue_id); - sa->rxq_info[rx_queue_id].deferred_started = B_FALSE; + sas->rxq_info[rx_queue_id].deferred_started = B_FALSE; sfc_adapter_unlock(sa); @@ -1179,7 +1269,8 @@ sfc_rx_queue_stop(struct rte_eth_dev *dev, uint16_t rx_queue_id) static int sfc_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc; sfc_log_init(sa, "TxQ = %u", tx_queue_id); @@ -1190,17 +1281,21 @@ sfc_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id) if (sa->state != SFC_ADAPTER_STARTED) goto fail_not_started; + if (sas->txq_info[tx_queue_id].state != SFC_TXQ_INITIALIZED) + goto fail_not_setup; + rc = sfc_tx_qstart(sa, tx_queue_id); if (rc != 0) goto fail_tx_qstart; - sa->txq_info[tx_queue_id].deferred_started = B_TRUE; + sas->txq_info[tx_queue_id].deferred_started = B_TRUE; sfc_adapter_unlock(sa); return 0; fail_tx_qstart: +fail_not_setup: fail_not_started: sfc_adapter_unlock(sa); SFC_ASSERT(rc > 0); @@ -1210,7 +1305,8 @@ fail_not_started: static int sfc_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "TxQ = %u", tx_queue_id); @@ -1218,7 +1314,7 @@ sfc_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id) sfc_tx_qstop(sa, tx_queue_id); - sa->txq_info[tx_queue_id].deferred_started = B_FALSE; + sas->txq_info[tx_queue_id].deferred_started = B_FALSE; sfc_adapter_unlock(sa); return 0; @@ -1247,7 +1343,7 @@ sfc_dev_udp_tunnel_op(struct rte_eth_dev *dev, struct rte_eth_udp_tunnel *tunnel_udp, enum sfc_udp_tunnel_op_e op) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); efx_tunnel_protocol_t tunnel_proto; int rc; @@ -1341,34 +1437,30 @@ sfc_dev_udp_tunnel_port_del(struct rte_eth_dev *dev, return sfc_dev_udp_tunnel_op(dev, tunnel_udp, SFC_UDP_TUNNEL_DEL_PORT); } -#if EFSYS_OPT_RX_SCALE +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static int sfc_dev_rss_hash_conf_get(struct rte_eth_dev *dev, struct rte_eth_rss_conf *rss_conf) { - struct sfc_adapter *sa = dev->data->dev_private; - struct sfc_port *port = &sa->port; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_rss *rss = &sas->rss; - if ((sa->rss_support != EFX_RX_SCALE_EXCLUSIVE) || port->isolated) + if (rss->context_type != EFX_RX_SCALE_EXCLUSIVE) return -ENOTSUP; - if (sa->rss_channels == 0) - return -EINVAL; - - sfc_adapter_lock(sa); - /* * Mapping of hash configuration between RTE and EFX is not one-to-one, * hence, conversion is done here to derive a correct set of ETH_RSS * flags which corresponds to the active EFX configuration stored * locally in 'sfc_adapter' and kept up-to-date */ - rss_conf->rss_hf = sfc_efx_to_rte_hash_type(sa->rss_hash_types); + rss_conf->rss_hf = sfc_rx_hf_efx_to_rte(rss, rss->hash_types); rss_conf->rss_key_len = EFX_RSS_KEY_SIZE; if (rss_conf->rss_key != NULL) - rte_memcpy(rss_conf->rss_key, sa->rss_key, EFX_RSS_KEY_SIZE); - - sfc_adapter_unlock(sa); + rte_memcpy(rss_conf->rss_key, rss->key, EFX_RSS_KEY_SIZE); return 0; } @@ -1377,43 +1469,39 @@ static int sfc_dev_rss_hash_update(struct rte_eth_dev *dev, struct rte_eth_rss_conf *rss_conf) { - struct sfc_adapter *sa = dev->data->dev_private; - struct sfc_port *port = &sa->port; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); + struct sfc_rss *rss = &sfc_sa2shared(sa)->rss; unsigned int efx_hash_types; int rc = 0; - if (port->isolated) + if (sfc_sa2shared(sa)->isolated) return -ENOTSUP; - if (sa->rss_support != EFX_RX_SCALE_EXCLUSIVE) { + if (rss->context_type != EFX_RX_SCALE_EXCLUSIVE) { sfc_err(sa, "RSS is not available"); return -ENOTSUP; } - if (sa->rss_channels == 0) { + if (rss->channels == 0) { sfc_err(sa, "RSS is not configured"); return -EINVAL; } if ((rss_conf->rss_key != NULL) && - (rss_conf->rss_key_len != sizeof(sa->rss_key))) { + (rss_conf->rss_key_len != sizeof(rss->key))) { sfc_err(sa, "RSS key size is wrong (should be %lu)", - sizeof(sa->rss_key)); - return -EINVAL; - } - - if ((rss_conf->rss_hf & ~SFC_RSS_OFFLOADS) != 0) { - sfc_err(sa, "unsupported hash functions requested"); + sizeof(rss->key)); return -EINVAL; } sfc_adapter_lock(sa); - efx_hash_types = sfc_rte_to_efx_hash_type(rss_conf->rss_hf); + rc = sfc_rx_hf_rte_to_efx(sa, rss_conf->rss_hf, &efx_hash_types); + if (rc != 0) + goto fail_rx_hf_rte_to_efx; rc = efx_rx_scale_mode_set(sa->nic, EFX_RSS_CONTEXT_DEFAULT, - EFX_RX_HASHALG_TOEPLITZ, - efx_hash_types, B_TRUE); + rss->hash_alg, efx_hash_types, B_TRUE); if (rc != 0) goto fail_scale_mode_set; @@ -1422,15 +1510,15 @@ sfc_dev_rss_hash_update(struct rte_eth_dev *dev, rc = efx_rx_scale_key_set(sa->nic, EFX_RSS_CONTEXT_DEFAULT, rss_conf->rss_key, - sizeof(sa->rss_key)); + sizeof(rss->key)); if (rc != 0) goto fail_scale_key_set; } - rte_memcpy(sa->rss_key, rss_conf->rss_key, sizeof(sa->rss_key)); + rte_memcpy(rss->key, rss_conf->rss_key, sizeof(rss->key)); } - sa->rss_hash_types = efx_hash_types; + rss->hash_types = efx_hash_types; sfc_adapter_unlock(sa); @@ -1439,44 +1527,45 @@ sfc_dev_rss_hash_update(struct rte_eth_dev *dev, fail_scale_key_set: if (efx_rx_scale_mode_set(sa->nic, EFX_RSS_CONTEXT_DEFAULT, EFX_RX_HASHALG_TOEPLITZ, - sa->rss_hash_types, B_TRUE) != 0) + rss->hash_types, B_TRUE) != 0) sfc_err(sa, "failed to restore RSS mode"); fail_scale_mode_set: +fail_rx_hf_rte_to_efx: sfc_adapter_unlock(sa); return -rc; } +/* + * The function is used by the secondary process as well. It must not + * use any process-local pointers from the adapter data. + */ static int sfc_dev_rss_reta_query(struct rte_eth_dev *dev, struct rte_eth_rss_reta_entry64 *reta_conf, uint16_t reta_size) { - struct sfc_adapter *sa = dev->data->dev_private; - struct sfc_port *port = &sa->port; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_rss *rss = &sas->rss; int entry; - if ((sa->rss_support != EFX_RX_SCALE_EXCLUSIVE) || port->isolated) + if (rss->context_type != EFX_RX_SCALE_EXCLUSIVE || sas->isolated) return -ENOTSUP; - if (sa->rss_channels == 0) + if (rss->channels == 0) return -EINVAL; if (reta_size != EFX_RSS_TBL_SIZE) return -EINVAL; - sfc_adapter_lock(sa); - for (entry = 0; entry < reta_size; entry++) { int grp = entry / RTE_RETA_GROUP_SIZE; int grp_idx = entry % RTE_RETA_GROUP_SIZE; if ((reta_conf[grp].mask >> grp_idx) & 1) - reta_conf[grp].reta[grp_idx] = sa->rss_tbl[entry]; + reta_conf[grp].reta[grp_idx] = rss->tbl[entry]; } - sfc_adapter_unlock(sa); - return 0; } @@ -1485,22 +1574,22 @@ sfc_dev_rss_reta_update(struct rte_eth_dev *dev, struct rte_eth_rss_reta_entry64 *reta_conf, uint16_t reta_size) { - struct sfc_adapter *sa = dev->data->dev_private; - struct sfc_port *port = &sa->port; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); + struct sfc_rss *rss = &sfc_sa2shared(sa)->rss; unsigned int *rss_tbl_new; uint16_t entry; int rc = 0; - if (port->isolated) + if (sfc_sa2shared(sa)->isolated) return -ENOTSUP; - if (sa->rss_support != EFX_RX_SCALE_EXCLUSIVE) { + if (rss->context_type != EFX_RX_SCALE_EXCLUSIVE) { sfc_err(sa, "RSS is not available"); return -ENOTSUP; } - if (sa->rss_channels == 0) { + if (rss->channels == 0) { sfc_err(sa, "RSS is not configured"); return -EINVAL; } @@ -1511,13 +1600,13 @@ sfc_dev_rss_reta_update(struct rte_eth_dev *dev, return -EINVAL; } - rss_tbl_new = rte_zmalloc("rss_tbl_new", sizeof(sa->rss_tbl), 0); + rss_tbl_new = rte_zmalloc("rss_tbl_new", sizeof(rss->tbl), 0); if (rss_tbl_new == NULL) return -ENOMEM; sfc_adapter_lock(sa); - rte_memcpy(rss_tbl_new, sa->rss_tbl, sizeof(sa->rss_tbl)); + rte_memcpy(rss_tbl_new, rss->tbl, sizeof(rss->tbl)); for (entry = 0; entry < reta_size; entry++) { int grp_idx = entry % RTE_RETA_GROUP_SIZE; @@ -1526,7 +1615,7 @@ sfc_dev_rss_reta_update(struct rte_eth_dev *dev, grp = &reta_conf[entry / RTE_RETA_GROUP_SIZE]; if (grp->mask & (1ull << grp_idx)) { - if (grp->reta[grp_idx] >= sa->rss_channels) { + if (grp->reta[grp_idx] >= rss->channels) { rc = EINVAL; goto bad_reta_entry; } @@ -1541,7 +1630,7 @@ sfc_dev_rss_reta_update(struct rte_eth_dev *dev, goto fail_scale_tbl_set; } - rte_memcpy(sa->rss_tbl, rss_tbl_new, sizeof(sa->rss_tbl)); + rte_memcpy(rss->tbl, rss_tbl_new, sizeof(rss->tbl)); fail_scale_tbl_set: bad_reta_entry: @@ -1552,14 +1641,13 @@ bad_reta_entry: SFC_ASSERT(rc >= 0); return -rc; } -#endif static int sfc_dev_filter_ctrl(struct rte_eth_dev *dev, enum rte_filter_type filter_type, enum rte_filter_op filter_op, void *arg) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); int rc = ENOTSUP; sfc_log_init(sa, "entry"); @@ -1610,6 +1698,21 @@ sfc_dev_filter_ctrl(struct rte_eth_dev *dev, enum rte_filter_type filter_type, return -rc; } +static int +sfc_pool_ops_supported(struct rte_eth_dev *dev, const char *pool) +{ + const struct sfc_adapter_priv *sap = sfc_adapter_priv_by_eth_dev(dev); + + /* + * If Rx datapath does not provide callback to check mempool, + * all pools are supported. + */ + if (sap->dp_rx->pool_ops_supported == NULL) + return 1; + + return sap->dp_rx->pool_ops_supported(pool); +} + static const struct eth_dev_ops sfc_eth_dev_ops = { .dev_configure = sfc_dev_configure, .dev_start = sfc_dev_start, @@ -1647,12 +1750,10 @@ static const struct eth_dev_ops sfc_eth_dev_ops = { .mac_addr_set = sfc_mac_addr_set, .udp_tunnel_port_add = sfc_dev_udp_tunnel_port_add, .udp_tunnel_port_del = sfc_dev_udp_tunnel_port_del, -#if EFSYS_OPT_RX_SCALE .reta_update = sfc_dev_rss_reta_update, .reta_query = sfc_dev_rss_reta_query, .rss_hash_update = sfc_dev_rss_hash_update, .rss_hash_conf_get = sfc_dev_rss_hash_conf_get, -#endif .filter_ctrl = sfc_dev_filter_ctrl, .set_mc_addr_list = sfc_set_mc_addr_list, .rxq_info_get = sfc_rx_queue_info_get, @@ -1660,6 +1761,7 @@ static const struct eth_dev_ops sfc_eth_dev_ops = { .fw_version_get = sfc_fw_version_get, .xstats_get_by_id = sfc_xstats_get_by_id, .xstats_get_names_by_id = sfc_xstats_get_names_by_id, + .pool_ops_supported = sfc_pool_ops_supported, }; /** @@ -1688,7 +1790,11 @@ sfc_strdup(const char *str) static int sfc_eth_dev_set_ops(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + const struct sfc_dp_rx *dp_rx; + const struct sfc_dp_tx *dp_tx; + const efx_nic_cfg_t *encp; unsigned int avail_caps = 0; const char *rx_name = NULL; const char *tx_name = NULL; @@ -1697,25 +1803,30 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) switch (sa->family) { case EFX_FAMILY_HUNTINGTON: case EFX_FAMILY_MEDFORD: + case EFX_FAMILY_MEDFORD2: avail_caps |= SFC_DP_HW_FW_CAP_EF10; break; default: break; } + encp = efx_nic_cfg_get(sa->nic); + if (encp->enc_rx_es_super_buffer_supported) + avail_caps |= SFC_DP_HW_FW_CAP_RX_ES_SUPER_BUFFER; + rc = sfc_kvargs_process(sa, SFC_KVARG_RX_DATAPATH, sfc_kvarg_string_handler, &rx_name); if (rc != 0) goto fail_kvarg_rx_datapath; if (rx_name != NULL) { - sa->dp_rx = sfc_dp_find_rx_by_name(&sfc_dp_head, rx_name); - if (sa->dp_rx == NULL) { + dp_rx = sfc_dp_find_rx_by_name(&sfc_dp_head, rx_name); + if (dp_rx == NULL) { sfc_err(sa, "Rx datapath %s not found", rx_name); rc = ENOENT; goto fail_dp_rx; } - if (!sfc_dp_match_hw_fw_caps(&sa->dp_rx->dp, avail_caps)) { + if (!sfc_dp_match_hw_fw_caps(&dp_rx->dp, avail_caps)) { sfc_err(sa, "Insufficient Hw/FW capabilities to use Rx datapath %s", rx_name); @@ -1723,8 +1834,8 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) goto fail_dp_rx_caps; } } else { - sa->dp_rx = sfc_dp_find_rx_by_caps(&sfc_dp_head, avail_caps); - if (sa->dp_rx == NULL) { + dp_rx = sfc_dp_find_rx_by_caps(&sfc_dp_head, avail_caps); + if (dp_rx == NULL) { sfc_err(sa, "Rx datapath by caps %#x not found", avail_caps); rc = ENOENT; @@ -1732,15 +1843,13 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) } } - sa->dp_rx_name = sfc_strdup(sa->dp_rx->dp.name); - if (sa->dp_rx_name == NULL) { + sas->dp_rx_name = sfc_strdup(dp_rx->dp.name); + if (sas->dp_rx_name == NULL) { rc = ENOMEM; goto fail_dp_rx_name; } - sfc_notice(sa, "use %s Rx datapath", sa->dp_rx_name); - - dev->rx_pkt_burst = sa->dp_rx->pkt_burst; + sfc_notice(sa, "use %s Rx datapath", sas->dp_rx_name); rc = sfc_kvargs_process(sa, SFC_KVARG_TX_DATAPATH, sfc_kvarg_string_handler, &tx_name); @@ -1748,13 +1857,13 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) goto fail_kvarg_tx_datapath; if (tx_name != NULL) { - sa->dp_tx = sfc_dp_find_tx_by_name(&sfc_dp_head, tx_name); - if (sa->dp_tx == NULL) { + dp_tx = sfc_dp_find_tx_by_name(&sfc_dp_head, tx_name); + if (dp_tx == NULL) { sfc_err(sa, "Tx datapath %s not found", tx_name); rc = ENOENT; goto fail_dp_tx; } - if (!sfc_dp_match_hw_fw_caps(&sa->dp_tx->dp, avail_caps)) { + if (!sfc_dp_match_hw_fw_caps(&dp_tx->dp, avail_caps)) { sfc_err(sa, "Insufficient Hw/FW capabilities to use Tx datapath %s", tx_name); @@ -1762,8 +1871,8 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) goto fail_dp_tx_caps; } } else { - sa->dp_tx = sfc_dp_find_tx_by_caps(&sfc_dp_head, avail_caps); - if (sa->dp_tx == NULL) { + dp_tx = sfc_dp_find_tx_by_caps(&sfc_dp_head, avail_caps); + if (dp_tx == NULL) { sfc_err(sa, "Tx datapath by caps %#x not found", avail_caps); rc = ENOENT; @@ -1771,15 +1880,20 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) } } - sa->dp_tx_name = sfc_strdup(sa->dp_tx->dp.name); - if (sa->dp_tx_name == NULL) { + sas->dp_tx_name = sfc_strdup(dp_tx->dp.name); + if (sas->dp_tx_name == NULL) { rc = ENOMEM; goto fail_dp_tx_name; } - sfc_notice(sa, "use %s Tx datapath", sa->dp_tx_name); + sfc_notice(sa, "use %s Tx datapath", sas->dp_tx_name); - dev->tx_pkt_burst = sa->dp_tx->pkt_burst; + sa->priv.dp_rx = dp_rx; + sa->priv.dp_tx = dp_tx; + + dev->rx_pkt_burst = dp_rx->pkt_burst; + dev->tx_pkt_prepare = dp_tx->pkt_prepare; + dev->tx_pkt_burst = dp_tx->pkt_burst; dev->dev_ops = &sfc_eth_dev_ops; @@ -1787,17 +1901,13 @@ sfc_eth_dev_set_ops(struct rte_eth_dev *dev) fail_dp_tx_name: fail_dp_tx_caps: - sa->dp_tx = NULL; - fail_dp_tx: fail_kvarg_tx_datapath: - rte_free(sa->dp_rx_name); - sa->dp_rx_name = NULL; + rte_free(sas->dp_rx_name); + sas->dp_rx_name = NULL; fail_dp_rx_name: fail_dp_rx_caps: - sa->dp_rx = NULL; - fail_dp_rx: fail_kvarg_rx_datapath: return rc; @@ -1806,66 +1916,92 @@ fail_kvarg_rx_datapath: static void sfc_eth_dev_clear_ops(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter *sa = sfc_adapter_by_eth_dev(dev); + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); dev->dev_ops = NULL; + dev->tx_pkt_prepare = NULL; dev->rx_pkt_burst = NULL; dev->tx_pkt_burst = NULL; - rte_free(sa->dp_tx_name); - sa->dp_tx_name = NULL; - sa->dp_tx = NULL; + rte_free(sas->dp_tx_name); + sas->dp_tx_name = NULL; + sa->priv.dp_tx = NULL; - rte_free(sa->dp_rx_name); - sa->dp_rx_name = NULL; - sa->dp_rx = NULL; + rte_free(sas->dp_rx_name); + sas->dp_rx_name = NULL; + sa->priv.dp_rx = NULL; } static const struct eth_dev_ops sfc_eth_dev_secondary_ops = { + .dev_supported_ptypes_get = sfc_dev_supported_ptypes_get, + .rx_queue_count = sfc_rx_queue_count, + .rx_descriptor_done = sfc_rx_descriptor_done, + .rx_descriptor_status = sfc_rx_descriptor_status, + .tx_descriptor_status = sfc_tx_descriptor_status, + .reta_query = sfc_dev_rss_reta_query, + .rss_hash_conf_get = sfc_dev_rss_hash_conf_get, .rxq_info_get = sfc_rx_queue_info_get, .txq_info_get = sfc_tx_queue_info_get, }; static int -sfc_eth_dev_secondary_set_ops(struct rte_eth_dev *dev) +sfc_eth_dev_secondary_init(struct rte_eth_dev *dev, uint32_t logtype_main) { - /* - * Device private data has really many process-local pointers. - * Below code should be extremely careful to use data located - * in shared memory only. - */ - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); + struct sfc_adapter_priv *sap; const struct sfc_dp_rx *dp_rx; const struct sfc_dp_tx *dp_tx; int rc; - dp_rx = sfc_dp_find_rx_by_name(&sfc_dp_head, sa->dp_rx_name); + /* + * Allocate process private data from heap, since it should not + * be located in shared memory allocated using rte_malloc() API. + */ + sap = calloc(1, sizeof(*sap)); + if (sap == NULL) { + rc = ENOMEM; + goto fail_alloc_priv; + } + + sap->logtype_main = logtype_main; + + dp_rx = sfc_dp_find_rx_by_name(&sfc_dp_head, sas->dp_rx_name); if (dp_rx == NULL) { - sfc_err(sa, "cannot find %s Rx datapath", sa->dp_tx_name); + SFC_LOG(sas, RTE_LOG_ERR, logtype_main, + "cannot find %s Rx datapath", sas->dp_rx_name); rc = ENOENT; goto fail_dp_rx; } if (~dp_rx->features & SFC_DP_RX_FEAT_MULTI_PROCESS) { - sfc_err(sa, "%s Rx datapath does not support multi-process", - sa->dp_tx_name); + SFC_LOG(sas, RTE_LOG_ERR, logtype_main, + "%s Rx datapath does not support multi-process", + sas->dp_rx_name); rc = EINVAL; goto fail_dp_rx_multi_process; } - dp_tx = sfc_dp_find_tx_by_name(&sfc_dp_head, sa->dp_tx_name); + dp_tx = sfc_dp_find_tx_by_name(&sfc_dp_head, sas->dp_tx_name); if (dp_tx == NULL) { - sfc_err(sa, "cannot find %s Tx datapath", sa->dp_tx_name); + SFC_LOG(sas, RTE_LOG_ERR, logtype_main, + "cannot find %s Tx datapath", sas->dp_tx_name); rc = ENOENT; goto fail_dp_tx; } if (~dp_tx->features & SFC_DP_TX_FEAT_MULTI_PROCESS) { - sfc_err(sa, "%s Tx datapath does not support multi-process", - sa->dp_tx_name); + SFC_LOG(sas, RTE_LOG_ERR, logtype_main, + "%s Tx datapath does not support multi-process", + sas->dp_tx_name); rc = EINVAL; goto fail_dp_tx_multi_process; } + sap->dp_rx = dp_rx; + sap->dp_tx = dp_tx; + + dev->process_private = sap; dev->rx_pkt_burst = dp_rx->pkt_burst; + dev->tx_pkt_prepare = dp_tx->pkt_prepare; dev->tx_pkt_burst = dp_tx->pkt_burst; dev->dev_ops = &sfc_eth_dev_secondary_ops; @@ -1875,13 +2011,19 @@ fail_dp_tx_multi_process: fail_dp_tx: fail_dp_rx_multi_process: fail_dp_rx: + free(sap); + +fail_alloc_priv: return rc; } static void sfc_eth_dev_secondary_clear_ops(struct rte_eth_dev *dev) { + free(dev->process_private); + dev->process_private = NULL; dev->dev_ops = NULL; + dev->tx_pkt_prepare = NULL; dev->tx_pkt_burst = NULL; dev->rx_pkt_burst = NULL; } @@ -1892,6 +2034,7 @@ sfc_register_dp(void) /* Register once */ if (TAILQ_EMPTY(&sfc_dp_head)) { /* Prefer EF10 datapath */ + sfc_dp_register(&sfc_dp_head, &sfc_ef10_essb_rx.dp); sfc_dp_register(&sfc_dp_head, &sfc_ef10_rx.dp); sfc_dp_register(&sfc_dp_head, &sfc_efx_rx.dp); @@ -1904,29 +2047,48 @@ sfc_register_dp(void) static int sfc_eth_dev_init(struct rte_eth_dev *dev) { - struct sfc_adapter *sa = dev->data->dev_private; + struct sfc_adapter_shared *sas = sfc_adapter_shared_by_eth_dev(dev); struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev); + uint32_t logtype_main; + struct sfc_adapter *sa; int rc; const efx_nic_cfg_t *encp; - const struct ether_addr *from; + const struct rte_ether_addr *from; sfc_register_dp(); + logtype_main = sfc_register_logtype(&pci_dev->addr, + SFC_LOGTYPE_MAIN_STR, + RTE_LOG_NOTICE); + if (rte_eal_process_type() != RTE_PROC_PRIMARY) - return -sfc_eth_dev_secondary_set_ops(dev); + return -sfc_eth_dev_secondary_init(dev, logtype_main); /* Required for logging */ - sa->pci_addr = pci_dev->addr; - sa->port_id = dev->data->port_id; + sas->pci_addr = pci_dev->addr; + sas->port_id = dev->data->port_id; + + /* + * Allocate process private data from heap, since it should not + * be located in shared memory allocated using rte_malloc() API. + */ + sa = calloc(1, sizeof(*sa)); + if (sa == NULL) { + rc = ENOMEM; + goto fail_alloc_sa; + } + + dev->process_private = sa; + + /* Required for logging */ + sa->priv.shared = sas; + sa->priv.logtype_main = logtype_main; sa->eth_dev = dev; /* Copy PCI device info to the dev->data */ rte_eth_copy_pci_info(dev, pci_dev); - sa->logtype_main = sfc_register_logtype(sa, SFC_LOGTYPE_MAIN_STR, - RTE_LOG_NOTICE); - rc = sfc_kvargs_parse(sa); if (rc != 0) goto fail_kvargs_parse; @@ -1963,8 +2125,8 @@ sfc_eth_dev_init(struct rte_eth_dev *dev) * The arguments are really reverse order in comparison to * Linux kernel. Copy from NIC config to Ethernet device data. */ - from = (const struct ether_addr *)(encp->enc_mac_addr); - ether_addr_copy(from, &dev->data->mac_addrs[0]); + from = (const struct rte_ether_addr *)(encp->enc_mac_addr); + rte_ether_addr_copy(from, &dev->data->mac_addrs[0]); sfc_adapter_unlock(sa); @@ -1988,6 +2150,10 @@ fail_mac_addrs: fail_kvargs_parse: sfc_log_init(sa, "failed %d", rc); + dev->process_private = NULL; + free(sa); + +fail_alloc_sa: SFC_ASSERT(rc > 0); return -rc; } @@ -2002,7 +2168,7 @@ sfc_eth_dev_uninit(struct rte_eth_dev *dev) return 0; } - sa = dev->data->dev_private; + sa = sfc_adapter_by_eth_dev(dev); sfc_log_init(sa, "entry"); sfc_adapter_lock(sa); @@ -2012,9 +2178,6 @@ sfc_eth_dev_uninit(struct rte_eth_dev *dev) sfc_detach(sa); sfc_unprobe(sa); - rte_free(dev->data->mac_addrs); - dev->data->mac_addrs = NULL; - sfc_kvargs_cleanup(sa); sfc_adapter_unlock(sa); @@ -2024,6 +2187,10 @@ sfc_eth_dev_uninit(struct rte_eth_dev *dev) /* Required for logging, so cleanup last */ sa->eth_dev = NULL; + + dev->process_private = NULL; + free(sa); + return 0; } @@ -2034,6 +2201,8 @@ static const struct rte_pci_id pci_id_sfc_efx_map[] = { { RTE_PCI_DEVICE(EFX_PCI_VENID_SFC, EFX_PCI_DEVID_GREENPORT_VF) }, { RTE_PCI_DEVICE(EFX_PCI_VENID_SFC, EFX_PCI_DEVID_MEDFORD) }, { RTE_PCI_DEVICE(EFX_PCI_VENID_SFC, EFX_PCI_DEVID_MEDFORD_VF) }, + { RTE_PCI_DEVICE(EFX_PCI_VENID_SFC, EFX_PCI_DEVID_MEDFORD2) }, + { RTE_PCI_DEVICE(EFX_PCI_VENID_SFC, EFX_PCI_DEVID_MEDFORD2_VF) }, { .vendor_id = 0 /* sentinel */ } }; @@ -2041,7 +2210,7 @@ static int sfc_eth_dev_pci_probe(struct rte_pci_driver *pci_drv __rte_unused, struct rte_pci_device *pci_dev) { return rte_eth_dev_pci_generic_probe(pci_dev, - sizeof(struct sfc_adapter), sfc_eth_dev_init); + sizeof(struct sfc_adapter_shared), sfc_eth_dev_init); } static int sfc_eth_dev_pci_remove(struct rte_pci_device *pci_dev) @@ -2065,11 +2234,11 @@ RTE_PMD_REGISTER_PARAM_STRING(net_sfc_efx, SFC_KVARG_RX_DATAPATH "=" SFC_KVARG_VALUES_RX_DATAPATH " " SFC_KVARG_TX_DATAPATH "=" SFC_KVARG_VALUES_TX_DATAPATH " " SFC_KVARG_PERF_PROFILE "=" SFC_KVARG_VALUES_PERF_PROFILE " " + SFC_KVARG_FW_VARIANT "=" SFC_KVARG_VALUES_FW_VARIANT " " + SFC_KVARG_RXD_WAIT_TIMEOUT_NS "= " SFC_KVARG_STATS_UPDATE_PERIOD_MS "="); -RTE_INIT(sfc_driver_register_logtype); -static void -sfc_driver_register_logtype(void) +RTE_INIT(sfc_driver_register_logtype) { int ret;