X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx5%2Fmlx5_ethdev.c;h=c0f73e9392d428cabfa622e29af589619b10ae48;hb=188408719888;hp=463b3267284fdc13b56a3dba372321d6c72f98af;hpb=62072098b54ea502764b60f19a626013f838c778;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c index 463b326728..c0f73e9392 100644 --- a/drivers/net/mlx5/mlx5_ethdev.c +++ b/drivers/net/mlx5/mlx5_ethdev.c @@ -44,14 +44,13 @@ #include #include #include -#include #include #include #include /* DPDK headers don't like -pedantic. */ #ifdef PEDANTIC -#pragma GCC diagnostic ignored "-pedantic" +#pragma GCC diagnostic ignored "-Wpedantic" #endif #include #include @@ -59,14 +58,47 @@ #include #include #include +#include #ifdef PEDANTIC -#pragma GCC diagnostic error "-pedantic" +#pragma GCC diagnostic error "-Wpedantic" #endif #include "mlx5.h" #include "mlx5_rxtx.h" #include "mlx5_utils.h" +/** + * Return private structure associated with an Ethernet device. + * + * @param dev + * Pointer to Ethernet device structure. + * + * @return + * Pointer to private structure. + */ +struct priv * +mlx5_get_priv(struct rte_eth_dev *dev) +{ + struct mlx5_secondary_data *sd; + + if (!mlx5_is_secondary()) + return dev->data->dev_private; + sd = &mlx5_secondary_data[dev->data->port_id]; + return sd->data.dev_private; +} + +/** + * Check if running as a secondary process. + * + * @return + * Nonzero if running as a secondary process. + */ +inline int +mlx5_is_secondary(void) +{ + return rte_eal_process_type() != RTE_PROC_PRIMARY; +} + /** * Get interface name from private structure. * @@ -329,6 +361,38 @@ priv_ifreq(const struct priv *priv, int req, struct ifreq *ifr) return ret; } +/** + * Return the number of active VFs for the current device. + * + * @param[in] priv + * Pointer to private structure. + * @param[out] num_vfs + * Number of active VFs. + * + * @return + * 0 on success, -1 on failure and errno is set. + */ +int +priv_get_num_vfs(struct priv *priv, uint16_t *num_vfs) +{ + /* The sysfs entry name depends on the operating system. */ + const char **name = (const char *[]){ + "device/sriov_numvfs", + "device/mlx5_num_vfs", + NULL, + }; + int ret; + + do { + unsigned long ulong_num_vfs; + + ret = priv_get_sysfs_ulong(priv, *name, &ulong_num_vfs); + if (!ret) + *num_vfs = ulong_num_vfs; + } while (*(++name) && ret); + return ret; +} + /** * Get device MTU. * @@ -365,7 +429,15 @@ priv_get_mtu(struct priv *priv, uint16_t *mtu) static int priv_set_mtu(struct priv *priv, uint16_t mtu) { - return priv_set_sysfs_ulong(priv, "mtu", mtu); + uint16_t new_mtu; + + if (priv_set_sysfs_ulong(priv, "mtu", mtu) || + priv_get_mtu(priv, &new_mtu)) + return -1; + if (new_mtu == mtu) + return 0; + errno = EINVAL; + return -1; } /** @@ -389,7 +461,7 @@ priv_set_flags(struct priv *priv, unsigned int keep, unsigned int flags) if (priv_get_sysfs_ulong(priv, "flags", &tmp) == -1) return -1; tmp &= keep; - tmp |= flags; + tmp |= (flags & (~keep)); return priv_set_sysfs_ulong(priv, "flags", tmp); } @@ -414,6 +486,7 @@ dev_configure(struct rte_eth_dev *dev) unsigned int j; unsigned int reta_idx_n; + priv->rss_hf = dev->data->dev_conf.rx_adv_conf.rss_conf.rss_hf; priv->rxqs = (void *)dev->data->rx_queues; priv->txqs = (void *)dev->data->tx_queues; if (txqs_n != priv->txqs_n) { @@ -464,6 +537,9 @@ mlx5_dev_configure(struct rte_eth_dev *dev) struct priv *priv = dev->data->dev_private; int ret; + if (mlx5_is_secondary()) + return -E_RTE_SECONDARY; + priv_lock(priv); ret = dev_configure(dev); assert(ret >= 0); @@ -482,7 +558,7 @@ mlx5_dev_configure(struct rte_eth_dev *dev) void mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) { - struct priv *priv = dev->data->dev_private; + struct priv *priv = mlx5_get_priv(dev); unsigned int max; char ifname[IF_NAMESIZE]; @@ -507,13 +583,15 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) (DEV_RX_OFFLOAD_IPV4_CKSUM | DEV_RX_OFFLOAD_UDP_CKSUM | DEV_RX_OFFLOAD_TCP_CKSUM) : - 0); - info->tx_offload_capa = - (priv->hw_csum ? - (DEV_TX_OFFLOAD_IPV4_CKSUM | - DEV_TX_OFFLOAD_UDP_CKSUM | - DEV_TX_OFFLOAD_TCP_CKSUM) : - 0); + 0) | + (priv->hw_vlan_strip ? DEV_RX_OFFLOAD_VLAN_STRIP : 0); + if (!priv->mps) + info->tx_offload_capa = DEV_TX_OFFLOAD_VLAN_INSERT; + if (priv->hw_csum) + info->tx_offload_capa |= + (DEV_TX_OFFLOAD_IPV4_CKSUM | + DEV_TX_OFFLOAD_UDP_CKSUM | + DEV_TX_OFFLOAD_TCP_CKSUM); if (priv_get_ifname(priv, &ifname) == 0) info->if_index = if_nametoindex(ifname); /* FIXME: RETA update/query API expects the callee to know the size of @@ -522,6 +600,10 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) * size if it is not fixed. * The API should be updated to solve this problem. */ info->reta_size = priv->ind_table_max_size; + info->hash_key_size = ((*priv->rss_conf) ? + (*priv->rss_conf)[0]->rss_key_len : + 0); + info->speed_capa = priv->link_speed_capa; priv_unlock(priv); } @@ -538,14 +620,13 @@ mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev) }; - if (dev->rx_pkt_burst == mlx5_rx_burst || - dev->rx_pkt_burst == mlx5_rx_burst_sp) + if (dev->rx_pkt_burst == mlx5_rx_burst) return ptypes; return NULL; } /** - * DPDK callback to retrieve physical link information (unlocked version). + * Retrieve physical link information (unlocked version using legacy ioctl). * * @param dev * Pointer to Ethernet device structure. @@ -553,11 +634,11 @@ mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev) * Wait for request completion (ignored). */ static int -mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) +mlx5_link_update_unlocked_gset(struct rte_eth_dev *dev, int wait_to_complete) { - struct priv *priv = dev->data->dev_private; + struct priv *priv = mlx5_get_priv(dev); struct ethtool_cmd edata = { - .cmd = ETHTOOL_GSET + .cmd = ETHTOOL_GSET /* Deprecated since Linux v4.5. */ }; struct ifreq ifr; struct rte_eth_link dev_link; @@ -571,7 +652,7 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) memset(&dev_link, 0, sizeof(dev_link)); dev_link.link_status = ((ifr.ifr_flags & IFF_UP) && (ifr.ifr_flags & IFF_RUNNING)); - ifr.ifr_data = &edata; + ifr.ifr_data = (void *)&edata; if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { WARN("ioctl(SIOCETHTOOL, ETHTOOL_GSET) failed: %s", strerror(errno)); @@ -582,8 +663,23 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) dev_link.link_speed = 0; else dev_link.link_speed = link_speed; + priv->link_speed_capa = 0; + if (edata.supported & SUPPORTED_Autoneg) + priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG; + if (edata.supported & (SUPPORTED_1000baseT_Full | + SUPPORTED_1000baseKX_Full)) + priv->link_speed_capa |= ETH_LINK_SPEED_1G; + if (edata.supported & SUPPORTED_10000baseKR_Full) + priv->link_speed_capa |= ETH_LINK_SPEED_10G; + if (edata.supported & (SUPPORTED_40000baseKR4_Full | + SUPPORTED_40000baseCR4_Full | + SUPPORTED_40000baseSR4_Full | + SUPPORTED_40000baseLR4_Full)) + priv->link_speed_capa |= ETH_LINK_SPEED_40G; dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ? ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX); + dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds & + ETH_LINK_SPEED_FIXED); if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) { /* Link status changed. */ dev->data->dev_link = dev_link; @@ -593,6 +689,123 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) return -1; } +/** + * Retrieve physical link information (unlocked version using new ioctl from + * Linux 4.5). + * + * @param dev + * Pointer to Ethernet device structure. + * @param wait_to_complete + * Wait for request completion (ignored). + */ +static int +mlx5_link_update_unlocked_gs(struct rte_eth_dev *dev, int wait_to_complete) +{ +#ifdef ETHTOOL_GLINKSETTINGS + struct priv *priv = mlx5_get_priv(dev); + struct ethtool_link_settings edata = { + .cmd = ETHTOOL_GLINKSETTINGS, + }; + struct ifreq ifr; + struct rte_eth_link dev_link; + uint64_t sc; + + (void)wait_to_complete; + if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) { + WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno)); + return -1; + } + memset(&dev_link, 0, sizeof(dev_link)); + dev_link.link_status = ((ifr.ifr_flags & IFF_UP) && + (ifr.ifr_flags & IFF_RUNNING)); + ifr.ifr_data = (void *)&edata; + if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { + DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s", + strerror(errno)); + return -1; + } + dev_link.link_speed = edata.speed; + sc = edata.link_mode_masks[0] | + ((uint64_t)edata.link_mode_masks[1] << 32); + priv->link_speed_capa = 0; + /* Link speeds available in kernel v4.5. */ + if (sc & ETHTOOL_LINK_MODE_Autoneg_BIT) + priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG; + if (sc & (ETHTOOL_LINK_MODE_1000baseT_Full_BIT | + ETHTOOL_LINK_MODE_1000baseKX_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_1G; + if (sc & (ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT | + ETHTOOL_LINK_MODE_10000baseKR_Full_BIT | + ETHTOOL_LINK_MODE_10000baseR_FEC_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_10G; + if (sc & (ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT | + ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_20G; + if (sc & (ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT | + ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT | + ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT | + ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_40G; + if (sc & (ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT | + ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT | + ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT | + ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_56G; + /* Link speeds available in kernel v4.6. */ +#ifdef HAVE_ETHTOOL_LINK_MODE_25G + if (sc & (ETHTOOL_LINK_MODE_25000baseCR_Full_BIT | + ETHTOOL_LINK_MODE_25000baseKR_Full_BIT | + ETHTOOL_LINK_MODE_25000baseSR_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_25G; +#endif +#ifdef HAVE_ETHTOOL_LINK_MODE_50G + if (sc & (ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT | + ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_50G; +#endif +#ifdef HAVE_ETHTOOL_LINK_MODE_100G + if (sc & (ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT | + ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT | + ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT | + ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_100G; +#endif + dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ? + ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX); + dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds & + ETH_LINK_SPEED_FIXED); + if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) { + /* Link status changed. */ + dev->data->dev_link = dev_link; + return 0; + } +#else + (void)dev; + (void)wait_to_complete; +#endif + /* Link status is still the same. */ + return -1; +} + +/** + * DPDK callback to retrieve physical link information (unlocked version). + * + * @param dev + * Pointer to Ethernet device structure. + * @param wait_to_complete + * Wait for request completion (ignored). + */ +int +mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) +{ + int ret; + + ret = mlx5_link_update_unlocked_gs(dev, wait_to_complete); + if (ret < 0) + ret = mlx5_link_update_unlocked_gset(dev, wait_to_complete); + return ret; +} + /** * DPDK callback to retrieve physical link information. * @@ -604,7 +817,7 @@ mlx5_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) int mlx5_link_update(struct rte_eth_dev *dev, int wait_to_complete) { - struct priv *priv = dev->data->dev_private; + struct priv *priv = mlx5_get_priv(dev); int ret; priv_lock(priv); @@ -638,6 +851,12 @@ mlx5_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) unsigned int i; uint16_t (*rx_func)(void *, struct rte_mbuf **, uint16_t) = mlx5_rx_burst; + unsigned int max_frame_len; + int rehash; + int restart = priv->started; + + if (mlx5_is_secondary()) + return -E_RTE_SECONDARY; priv_lock(priv); /* Set kernel interface MTU first. */ @@ -648,7 +867,6 @@ mlx5_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) goto out; } else DEBUG("adapter port %u MTU set to %u", priv->port, mtu); - priv->mtu = mtu; /* Temporarily replace RX handler with a fake one, assuming it has not * been copied elsewhere. */ dev->rx_pkt_burst = removed_rx_burst; @@ -656,33 +874,94 @@ mlx5_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) * removed_rx_burst() instead. */ rte_wmb(); usleep(1000); + /* MTU does not include header and CRC. */ + max_frame_len = ETHER_HDR_LEN + mtu + ETHER_CRC_LEN; + /* Check if at least one queue is going to need a SGE update. */ + for (i = 0; i != priv->rxqs_n; ++i) { + struct rxq *rxq = (*priv->rxqs)[i]; + unsigned int mb_len; + unsigned int size = RTE_PKTMBUF_HEADROOM + max_frame_len; + unsigned int sges_n; + + if (rxq == NULL) + continue; + mb_len = rte_pktmbuf_data_room_size(rxq->mp); + assert(mb_len >= RTE_PKTMBUF_HEADROOM); + /* + * Determine the number of SGEs needed for a full packet + * and round it to the next power of two. + */ + sges_n = log2above((size / mb_len) + !!(size % mb_len)); + if (sges_n != rxq->sges_n) + break; + } + /* + * If all queues have the right number of SGEs, a simple rehash + * of their buffers is enough, otherwise SGE information can only + * be updated in a queue by recreating it. All resources that depend + * on queues (flows, indirection tables) must be recreated as well in + * that case. + */ + rehash = (i == priv->rxqs_n); + if (!rehash) { + /* Clean up everything as with mlx5_dev_stop(). */ + priv_special_flow_disable_all(priv); + priv_mac_addrs_disable(priv); + priv_destroy_hash_rxqs(priv); + priv_fdir_disable(priv); + priv_dev_interrupt_handler_uninstall(priv, dev); + } +recover: /* Reconfigure each RX queue. */ for (i = 0; (i != priv->rxqs_n); ++i) { struct rxq *rxq = (*priv->rxqs)[i]; - unsigned int max_frame_len; + struct rxq_ctrl *rxq_ctrl = + container_of(rxq, struct rxq_ctrl, rxq); int sp; + unsigned int mb_len; + unsigned int tmp; if (rxq == NULL) continue; - /* Calculate new maximum frame length according to MTU and - * toggle scattered support (sp) if necessary. */ - max_frame_len = (priv->mtu + ETHER_HDR_LEN + - (ETHER_MAX_VLAN_FRAME_LEN - ETHER_MAX_LEN)); - sp = (max_frame_len > (rxq->mb_len - RTE_PKTMBUF_HEADROOM)); + mb_len = rte_pktmbuf_data_room_size(rxq->mp); + assert(mb_len >= RTE_PKTMBUF_HEADROOM); + /* Toggle scattered support (sp) if necessary. */ + sp = (max_frame_len > (mb_len - RTE_PKTMBUF_HEADROOM)); /* Provide new values to rxq_setup(). */ dev->data->dev_conf.rxmode.jumbo_frame = sp; dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame_len; - ret = rxq_rehash(dev, rxq); - if (ret) { - /* Force SP RX if that queue requires it and abort. */ - if (rxq->sp) - rx_func = mlx5_rx_burst_sp; - break; + if (rehash) + ret = rxq_rehash(dev, rxq_ctrl); + else + ret = rxq_ctrl_setup(dev, rxq_ctrl, 1 << rxq->elts_n, + rxq_ctrl->socket, NULL, rxq->mp); + if (!ret) + continue; + /* Attempt to roll back in case of error. */ + tmp = (mb_len << rxq->sges_n) - RTE_PKTMBUF_HEADROOM; + if (max_frame_len != tmp) { + max_frame_len = tmp; + goto recover; } - /* Scattered burst function takes priority. */ - if (rxq->sp) - rx_func = mlx5_rx_burst_sp; + /* Double fault, disable RX. */ + break; } + /* + * Use a safe RX burst function in case of error, otherwise mimic + * mlx5_dev_start(). + */ + if (ret) { + ERROR("unable to reconfigure RX queues, RX disabled"); + rx_func = removed_rx_burst; + } else if (restart && + !rehash && + !priv_create_hash_rxqs(priv) && + !priv_rehash_flows(priv)) { + if (dev->data->dev_conf.fdir_conf.mode == RTE_FDIR_MODE_NONE) + priv_fdir_enable(priv); + priv_dev_interrupt_handler_install(priv, dev); + } + priv->mtu = mtu; /* Burst functions can now be called again. */ rte_wmb(); dev->rx_pkt_burst = rx_func; @@ -713,7 +992,10 @@ mlx5_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) }; int ret; - ifr.ifr_data = ðpause; + if (mlx5_is_secondary()) + return -E_RTE_SECONDARY; + + ifr.ifr_data = (void *)ðpause; priv_lock(priv); if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { ret = errno; @@ -761,7 +1043,10 @@ mlx5_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) }; int ret; - ifr.ifr_data = ðpause; + if (mlx5_is_secondary()) + return -E_RTE_SECONDARY; + + ifr.ifr_data = (void *)ðpause; ethpause.autoneg = fc_conf->autoneg; if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) || (fc_conf->mode & RTE_FC_RX_PAUSE)) @@ -909,7 +1194,7 @@ mlx5_dev_link_status_handler(void *arg) ret = priv_dev_link_status_handler(priv, dev); priv_unlock(priv); if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL); } /** @@ -932,7 +1217,7 @@ mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) ret = priv_dev_link_status_handler(priv, dev); priv_unlock(priv); if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL); } /** @@ -955,7 +1240,7 @@ priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) rte_eal_alarm_cancel(mlx5_dev_link_status_handler, dev); priv->pending_alarm = 0; priv->intr_handle.fd = 0; - priv->intr_handle.type = 0; + priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN; } /** @@ -991,7 +1276,7 @@ priv_dev_interrupt_handler_install(struct priv *priv, struct rte_eth_dev *dev) /** * Change the link state (UP / DOWN). * - * @param dev + * @param priv * Pointer to Ethernet device structure. * @param up * Nonzero for link up, otherwise link down. @@ -1004,23 +1289,13 @@ priv_set_link(struct priv *priv, int up) { struct rte_eth_dev *dev = priv->dev; int err; - unsigned int i; if (up) { err = priv_set_flags(priv, ~IFF_UP, IFF_UP); if (err) return err; - for (i = 0; i < priv->rxqs_n; i++) - if ((*priv->rxqs)[i]->sp) - break; - /* Check if an sp queue exists. - * Note: Some old frames might be received. - */ - if (i == priv->rxqs_n) - dev->rx_pkt_burst = mlx5_rx_burst; - else - dev->rx_pkt_burst = mlx5_rx_burst_sp; - dev->tx_pkt_burst = mlx5_tx_burst; + priv_select_tx_function(priv); + priv_select_rx_function(priv); } else { err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP); if (err) @@ -1072,3 +1347,195 @@ mlx5_set_link_up(struct rte_eth_dev *dev) priv_unlock(priv); return err; } + +/** + * Configure secondary process queues from a private data pointer (primary + * or secondary) and update burst callbacks. Can take place only once. + * + * All queues must have been previously created by the primary process to + * avoid undefined behavior. + * + * @param priv + * Private data pointer from either primary or secondary process. + * + * @return + * Private data pointer from secondary process, NULL in case of error. + */ +struct priv * +mlx5_secondary_data_setup(struct priv *priv) +{ + unsigned int port_id = 0; + struct mlx5_secondary_data *sd; + void **tx_queues; + void **rx_queues; + unsigned int nb_tx_queues; + unsigned int nb_rx_queues; + unsigned int i; + + /* priv must be valid at this point. */ + assert(priv != NULL); + /* priv->dev must also be valid but may point to local memory from + * another process, possibly with the same address and must not + * be dereferenced yet. */ + assert(priv->dev != NULL); + /* Determine port ID by finding out where priv comes from. */ + while (1) { + sd = &mlx5_secondary_data[port_id]; + rte_spinlock_lock(&sd->lock); + /* Primary process? */ + if (sd->primary_priv == priv) + break; + /* Secondary process? */ + if (sd->data.dev_private == priv) + break; + rte_spinlock_unlock(&sd->lock); + if (++port_id == RTE_DIM(mlx5_secondary_data)) + port_id = 0; + } + /* Switch to secondary private structure. If private data has already + * been updated by another thread, there is nothing else to do. */ + priv = sd->data.dev_private; + if (priv->dev->data == &sd->data) + goto end; + /* Sanity checks. Secondary private structure is supposed to point + * to local eth_dev, itself still pointing to the shared device data + * structure allocated by the primary process. */ + assert(sd->shared_dev_data != &sd->data); + assert(sd->data.nb_tx_queues == 0); + assert(sd->data.tx_queues == NULL); + assert(sd->data.nb_rx_queues == 0); + assert(sd->data.rx_queues == NULL); + assert(priv != sd->primary_priv); + assert(priv->dev->data == sd->shared_dev_data); + assert(priv->txqs_n == 0); + assert(priv->txqs == NULL); + assert(priv->rxqs_n == 0); + assert(priv->rxqs == NULL); + nb_tx_queues = sd->shared_dev_data->nb_tx_queues; + nb_rx_queues = sd->shared_dev_data->nb_rx_queues; + /* Allocate local storage for queues. */ + tx_queues = rte_zmalloc("secondary ethdev->tx_queues", + sizeof(sd->data.tx_queues[0]) * nb_tx_queues, + RTE_CACHE_LINE_SIZE); + rx_queues = rte_zmalloc("secondary ethdev->rx_queues", + sizeof(sd->data.rx_queues[0]) * nb_rx_queues, + RTE_CACHE_LINE_SIZE); + if (tx_queues == NULL || rx_queues == NULL) + goto error; + /* Lock to prevent control operations during setup. */ + priv_lock(priv); + /* TX queues. */ + for (i = 0; i != nb_tx_queues; ++i) { + struct txq *primary_txq = (*sd->primary_priv->txqs)[i]; + struct txq_ctrl *primary_txq_ctrl; + struct txq_ctrl *txq_ctrl; + + if (primary_txq == NULL) + continue; + primary_txq_ctrl = container_of(primary_txq, + struct txq_ctrl, txq); + txq_ctrl = rte_calloc_socket("TXQ", 1, sizeof(*txq_ctrl) + + (1 << primary_txq->elts_n) * + sizeof(struct rte_mbuf *), 0, + primary_txq_ctrl->socket); + if (txq_ctrl != NULL) { + if (txq_ctrl_setup(priv->dev, + txq_ctrl, + 1 << primary_txq->elts_n, + primary_txq_ctrl->socket, + NULL) == 0) { + txq_ctrl->txq.stats.idx = + primary_txq->stats.idx; + tx_queues[i] = &txq_ctrl->txq; + continue; + } + rte_free(txq_ctrl); + } + while (i) { + txq_ctrl = tx_queues[--i]; + txq_cleanup(txq_ctrl); + rte_free(txq_ctrl); + } + goto error; + } + /* RX queues. */ + for (i = 0; i != nb_rx_queues; ++i) { + struct rxq_ctrl *primary_rxq = + container_of((*sd->primary_priv->rxqs)[i], + struct rxq_ctrl, rxq); + + if (primary_rxq == NULL) + continue; + /* Not supported yet. */ + rx_queues[i] = NULL; + } + /* Update everything. */ + priv->txqs = (void *)tx_queues; + priv->txqs_n = nb_tx_queues; + priv->rxqs = (void *)rx_queues; + priv->rxqs_n = nb_rx_queues; + sd->data.rx_queues = rx_queues; + sd->data.tx_queues = tx_queues; + sd->data.nb_rx_queues = nb_rx_queues; + sd->data.nb_tx_queues = nb_tx_queues; + sd->data.dev_link = sd->shared_dev_data->dev_link; + sd->data.mtu = sd->shared_dev_data->mtu; + memcpy(sd->data.rx_queue_state, sd->shared_dev_data->rx_queue_state, + sizeof(sd->data.rx_queue_state)); + memcpy(sd->data.tx_queue_state, sd->shared_dev_data->tx_queue_state, + sizeof(sd->data.tx_queue_state)); + sd->data.dev_flags = sd->shared_dev_data->dev_flags; + /* Use local data from now on. */ + rte_mb(); + priv->dev->data = &sd->data; + rte_mb(); + priv_select_tx_function(priv); + priv_select_rx_function(priv); + priv_unlock(priv); +end: + /* More sanity checks. */ + assert(priv->dev->data == &sd->data); + rte_spinlock_unlock(&sd->lock); + return priv; +error: + priv_unlock(priv); + rte_free(tx_queues); + rte_free(rx_queues); + rte_spinlock_unlock(&sd->lock); + return NULL; +} + +/** + * Configure the TX function to use. + * + * @param priv + * Pointer to private structure. + */ +void +priv_select_tx_function(struct priv *priv) +{ + priv->dev->tx_pkt_burst = mlx5_tx_burst; + /* Display warning for unsupported configurations. */ + if (priv->sriov && priv->mps) + WARN("multi-packet send WQE cannot be used on a SR-IOV setup"); + /* Select appropriate TX function. */ + if ((priv->sriov == 0) && priv->mps && priv->txq_inline) { + priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw_inline; + DEBUG("selected MPW inline TX function"); + } else if ((priv->sriov == 0) && priv->mps) { + priv->dev->tx_pkt_burst = mlx5_tx_burst_mpw; + DEBUG("selected MPW TX function"); + } +} + +/** + * Configure the RX function to use. + * + * @param priv + * Pointer to private structure. + */ +void +priv_select_rx_function(struct priv *priv) +{ + priv->dev->rx_pkt_burst = mlx5_rx_burst; +}