X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;ds=sidebyside;f=drivers%2Fnet%2Fmlx5%2Fmlx5_ethdev.c;h=a3cef6891d038b128da57be0a3ff18ba88599a62;hb=b113cb5ee6cec14a9892df1143217e358a58f5d6;hp=84e877cd694d2c11064dd98bda3033e8da912289;hpb=198a3c339a8ff9ecc5916566cfd25e98d98ce484;p=dpdk.git diff --git a/drivers/net/mlx5/mlx5_ethdev.c b/drivers/net/mlx5/mlx5_ethdev.c index 84e877cd69..a3cef6891d 100644 --- a/drivers/net/mlx5/mlx5_ethdev.c +++ b/drivers/net/mlx5/mlx5_ethdev.c @@ -31,6 +31,8 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ +#define _GNU_SOURCE + #include #include #include @@ -43,30 +45,106 @@ #include #include #include +#include #include -#include #include #include +#include #include +#include +#include -/* DPDK headers don't like -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-pedantic" -#endif #include #include +#include #include #include #include #include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-pedantic" -#endif +#include #include "mlx5.h" #include "mlx5_rxtx.h" #include "mlx5_utils.h" +/* Add defines in case the running kernel is not the same as user headers. */ +#ifndef ETHTOOL_GLINKSETTINGS +struct ethtool_link_settings { + uint32_t cmd; + uint32_t speed; + uint8_t duplex; + uint8_t port; + uint8_t phy_address; + uint8_t autoneg; + uint8_t mdio_support; + uint8_t eth_to_mdix; + uint8_t eth_tp_mdix_ctrl; + int8_t link_mode_masks_nwords; + uint32_t reserved[8]; + uint32_t link_mode_masks[]; +}; + +#define ETHTOOL_GLINKSETTINGS 0x0000004c +#define ETHTOOL_LINK_MODE_1000baseT_Full_BIT 5 +#define ETHTOOL_LINK_MODE_Autoneg_BIT 6 +#define ETHTOOL_LINK_MODE_1000baseKX_Full_BIT 17 +#define ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT 18 +#define ETHTOOL_LINK_MODE_10000baseKR_Full_BIT 19 +#define ETHTOOL_LINK_MODE_10000baseR_FEC_BIT 20 +#define ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT 21 +#define ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT 22 +#define ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT 23 +#define ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT 24 +#define ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT 25 +#define ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT 26 +#define ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT 27 +#define ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT 28 +#define ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT 29 +#define ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT 30 +#endif +#ifndef HAVE_ETHTOOL_LINK_MODE_25G +#define ETHTOOL_LINK_MODE_25000baseCR_Full_BIT 31 +#define ETHTOOL_LINK_MODE_25000baseKR_Full_BIT 32 +#define ETHTOOL_LINK_MODE_25000baseSR_Full_BIT 33 +#endif +#ifndef HAVE_ETHTOOL_LINK_MODE_50G +#define ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT 34 +#define ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT 35 +#endif +#ifndef HAVE_ETHTOOL_LINK_MODE_100G +#define ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT 36 +#define ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT 37 +#define ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT 38 +#define ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT 39 +#endif + +/** + * 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) +{ + return dev->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_SECONDARY; +} + /** * Get interface name from private structure. * @@ -88,7 +166,7 @@ priv_get_ifname(const struct priv *priv, char (*ifname)[IF_NAMESIZE]) char match[IF_NAMESIZE] = ""; { - MKSTR(path, "%s/device/net", priv->ctx->device->ibdev_path); + MKSTR(path, "%s/device/net", priv->ibdev_path); dir = opendir(path); if (dir == NULL) @@ -106,7 +184,7 @@ priv_get_ifname(const struct priv *priv, char (*ifname)[IF_NAMESIZE]) continue; MKSTR(path, "%s/device/net/%s/%s", - priv->ctx->device->ibdev_path, name, + priv->ibdev_path, name, (dev_type ? "dev_id" : "dev_port")); file = fopen(path, "rb"); @@ -148,6 +226,23 @@ try_dev_id: return 0; } +/** + * Check if the counter is located on ib counters file. + * + * @param[in] cntr + * Counter name. + * + * @return + * 1 if counter is located on ib counters file , 0 otherwise. + */ +int +priv_is_ib_cntr(const char *cntr) +{ + if (!strcmp(cntr, "out_of_buffer")) + return 1; + return 0; +} + /** * Read from sysfs entry. * @@ -175,10 +270,15 @@ priv_sysfs_read(const struct priv *priv, const char *entry, if (priv_get_ifname(priv, &ifname)) return -1; - MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path, - ifname, entry); - - file = fopen(path, "rb"); + if (priv_is_ib_cntr(entry)) { + MKSTR(path, "%s/ports/1/hw_counters/%s", + priv->ibdev_path, entry); + file = fopen(path, "rb"); + } else { + MKSTR(path, "%s/device/net/%s/%s", + priv->ibdev_path, ifname, entry); + file = fopen(path, "rb"); + } if (file == NULL) return -1; ret = fread(buf, 1, size, file); @@ -219,8 +319,7 @@ priv_sysfs_write(const struct priv *priv, const char *entry, if (priv_get_ifname(priv, &ifname)) return -1; - MKSTR(path, "%s/device/net/%s/%s", priv->ctx->device->ibdev_path, - ifname, entry); + MKSTR(path, "%s/device/net/%s/%s", priv->ibdev_path, ifname, entry); file = fopen(path, "wb"); if (file == NULL) @@ -329,6 +428,38 @@ 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. * @@ -351,6 +482,30 @@ priv_get_mtu(struct priv *priv, uint16_t *mtu) return 0; } +/** + * Read device counter from sysfs. + * + * @param priv + * Pointer to private structure. + * @param name + * Counter name. + * @param[out] cntr + * Counter output buffer. + * + * @return + * 0 on success, -1 on failure and errno is set. + */ +int +priv_get_cntr_sysfs(struct priv *priv, const char *name, uint64_t *cntr) +{ + unsigned long ulong_ctr; + + if (priv_get_sysfs_ulong(priv, name, &ulong_ctr) == -1) + return -1; + *cntr = ulong_ctr; + return 0; +} + /** * Set device MTU. * @@ -365,7 +520,15 @@ 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 +552,7 @@ 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); } @@ -410,7 +573,32 @@ dev_configure(struct rte_eth_dev *dev) struct priv *priv = dev->data->dev_private; unsigned int rxqs_n = dev->data->nb_rx_queues; unsigned int txqs_n = dev->data->nb_tx_queues; - + unsigned int i; + unsigned int j; + unsigned int reta_idx_n; + const uint8_t use_app_rss_key = + !!dev->data->dev_conf.rx_adv_conf.rss_conf.rss_key_len; + + if (use_app_rss_key && + (dev->data->dev_conf.rx_adv_conf.rss_conf.rss_key_len != + rss_hash_default_key_len)) { + /* MLX5 RSS only support 40bytes key. */ + return EINVAL; + } + priv->rss_conf.rss_key = + rte_realloc(priv->rss_conf.rss_key, + rss_hash_default_key_len, 0); + if (!priv->rss_conf.rss_key) { + ERROR("cannot allocate RSS hash key memory (%u)", rxqs_n); + return ENOMEM; + } + memcpy(priv->rss_conf.rss_key, + use_app_rss_key ? + dev->data->dev_conf.rx_adv_conf.rss_conf.rss_key : + rss_hash_default_key, + rss_hash_default_key_len); + priv->rss_conf.rss_key_len = rss_hash_default_key_len; + priv->rss_conf.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) { @@ -418,11 +606,31 @@ dev_configure(struct rte_eth_dev *dev) (void *)dev, priv->txqs_n, txqs_n); priv->txqs_n = txqs_n; } + if (rxqs_n > priv->ind_table_max_size) { + ERROR("cannot handle this many RX queues (%u)", rxqs_n); + return EINVAL; + } if (rxqs_n == priv->rxqs_n) return 0; INFO("%p: RX queues number update: %u -> %u", (void *)dev, priv->rxqs_n, rxqs_n); priv->rxqs_n = rxqs_n; + /* If the requested number of RX queues is not a power of two, use the + * maximum indirection table size for better balancing. + * The result is always rounded to the next power of two. */ + reta_idx_n = (1 << log2above((rxqs_n & (rxqs_n - 1)) ? + priv->ind_table_max_size : + rxqs_n)); + if (priv_rss_reta_index_resize(priv, reta_idx_n)) + return ENOMEM; + /* When the number of RX queues is not a power of two, the remaining + * table entries are padded with reused WQs and hashes are not spread + * uniformly. */ + for (i = 0, j = 0; (i != reta_idx_n); ++i) { + (*priv->reta_idx)[i] = j; + if (++j == rxqs_n) + j = 0; + } return 0; } @@ -441,6 +649,9 @@ 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); @@ -459,10 +670,12 @@ 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]; + info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); + priv_lock(priv); /* FIXME: we should ask the device for these values. */ info->min_rx_bufsize = 32; @@ -471,34 +684,74 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) * Since we need one CQ per QP, the limit is the minimum number * between the two values. */ - max = ((priv->device_attr.max_cq > priv->device_attr.max_qp) ? - priv->device_attr.max_qp : priv->device_attr.max_cq); + max = RTE_MIN(priv->device_attr.orig_attr.max_cq, + priv->device_attr.orig_attr.max_qp); /* If max >= 65535 then max = 0, max_rx_queues is uint16_t. */ if (max >= 65535) max = 65535; info->max_rx_queues = max; info->max_tx_queues = max; - /* Last array entry is reserved for broadcast. */ - info->max_mac_addrs = (RTE_DIM(priv->mac) - 1); + info->max_mac_addrs = RTE_DIM(priv->mac); info->rx_offload_capa = (priv->hw_csum ? (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) | + DEV_RX_OFFLOAD_TIMESTAMP; + + 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->tso) + info->tx_offload_capa |= DEV_TX_OFFLOAD_TCP_TSO; + if (priv->tunnel_en) + info->tx_offload_capa |= (DEV_TX_OFFLOAD_OUTER_IPV4_CKSUM | + DEV_TX_OFFLOAD_VXLAN_TNL_TSO | + DEV_TX_OFFLOAD_GRE_TNL_TSO); if (priv_get_ifname(priv, &ifname) == 0) info->if_index = if_nametoindex(ifname); + info->reta_size = priv->reta_idx_n ? + priv->reta_idx_n : priv->ind_table_max_size; + info->hash_key_size = priv->rss_conf.rss_key_len; + info->speed_capa = priv->link_speed_capa; priv_unlock(priv); } +const uint32_t * +mlx5_dev_supported_ptypes_get(struct rte_eth_dev *dev) +{ + static const uint32_t ptypes[] = { + /* refers to rxq_cq_to_pkt_type() */ + RTE_PTYPE_L2_ETHER, + RTE_PTYPE_L3_IPV4_EXT_UNKNOWN, + RTE_PTYPE_L3_IPV6_EXT_UNKNOWN, + RTE_PTYPE_L4_NONFRAG, + RTE_PTYPE_L4_FRAG, + RTE_PTYPE_L4_TCP, + RTE_PTYPE_L4_UDP, + RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN, + RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN, + RTE_PTYPE_INNER_L4_NONFRAG, + RTE_PTYPE_INNER_L4_FRAG, + RTE_PTYPE_INNER_L4_TCP, + RTE_PTYPE_INNER_L4_UDP, + RTE_PTYPE_UNKNOWN + }; + + if (dev->rx_pkt_burst == mlx5_rx_burst || + dev->rx_pkt_burst == mlx5_rx_burst_vec) + return ptypes; + return NULL; +} + /** - * DPDK callback to retrieve physical link information (unlocked version). + * DPDK callback to retrieve physical link information. * * @param dev * Pointer to Ethernet device structure. @@ -506,16 +759,18 @@ mlx5_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) * 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; int link_speed = 0; + /* priv_lock() is not taken to allow concurrent calls. */ + (void)wait_to_complete; if (priv_ifreq(priv, SIOCGIFFLAGS, &ifr)) { WARN("ioctl(SIOCGIFFLAGS) failed: %s", strerror(errno)); @@ -524,7 +779,7 @@ 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)); @@ -535,8 +790,119 @@ 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; + return 0; + } + /* Link status is still the same. */ + return -1; +} + +/** + * Retrieve physical link information (unlocked version using new ioctl). + * + * @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) +{ + struct priv *priv = mlx5_get_priv(dev); + struct ethtool_link_settings gcmd = { .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 *)&gcmd; + if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { + DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s", + strerror(errno)); + return -1; + } + gcmd.link_mode_masks_nwords = -gcmd.link_mode_masks_nwords; + + alignas(struct ethtool_link_settings) + uint8_t data[offsetof(struct ethtool_link_settings, link_mode_masks) + + sizeof(uint32_t) * gcmd.link_mode_masks_nwords * 3]; + struct ethtool_link_settings *ecmd = (void *)data; + + *ecmd = gcmd; + ifr.ifr_data = (void *)ecmd; + if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { + DEBUG("ioctl(SIOCETHTOOL, ETHTOOL_GLINKSETTINGS) failed: %s", + strerror(errno)); + return -1; + } + dev_link.link_speed = ecmd->speed; + sc = ecmd->link_mode_masks[0] | + ((uint64_t)ecmd->link_mode_masks[1] << 32); + priv->link_speed_capa = 0; + if (sc & MLX5_BITSHIFT(ETHTOOL_LINK_MODE_Autoneg_BIT)) + priv->link_speed_capa |= ETH_LINK_SPEED_AUTONEG; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_1000baseT_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_1000baseKX_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_1G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_10000baseKX4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_10000baseKR_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_10000baseR_FEC_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_10G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_20000baseMLD2_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_20000baseKR2_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_20G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_40000baseKR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_40000baseCR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_40000baseSR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_40000baseLR4_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_40G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_56000baseKR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_56000baseCR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_56000baseSR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_56000baseLR4_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_56G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_25000baseCR_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_25000baseKR_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_25000baseSR_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_25G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_50000baseCR2_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_50000baseKR2_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_50G; + if (sc & (MLX5_BITSHIFT(ETHTOOL_LINK_MODE_100000baseKR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_100000baseSR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_100000baseCR4_Full_BIT) | + MLX5_BITSHIFT(ETHTOOL_LINK_MODE_100000baseLR4_ER4_Full_BIT))) + priv->link_speed_capa |= ETH_LINK_SPEED_100G; + dev_link.link_duplex = ((ecmd->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; @@ -557,24 +923,20 @@ 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; - int ret; - - priv_lock(priv); - ret = mlx5_link_update_unlocked(dev, wait_to_complete); - priv_unlock(priv); - return ret; + struct utsname utsname; + int ver[3]; + + if (uname(&utsname) == -1 || + sscanf(utsname.release, "%d.%d.%d", + &ver[0], &ver[1], &ver[2]) != 3 || + KERNEL_VERSION(ver[0], ver[1], ver[2]) < KERNEL_VERSION(4, 9, 0)) + return mlx5_link_update_unlocked_gset(dev, wait_to_complete); + return mlx5_link_update_unlocked_gs(dev, wait_to_complete); } /** * DPDK callback to change the MTU. * - * Setting the MTU affects hardware MRU (packets larger than the MTU cannot be - * received). Use this as a hint to enable/disable scattered packets support - * and improve performance when not needed. - * Since failure is not an option, reconfiguring queues on the fly is not - * recommended. - * * @param dev * Pointer to Ethernet device structure. * @param in_mtu @@ -587,59 +949,33 @@ int mlx5_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) { struct priv *priv = dev->data->dev_private; + uint16_t kern_mtu; int ret = 0; - unsigned int i; - uint16_t (*rx_func)(void *, struct rte_mbuf **, uint16_t) = - mlx5_rx_burst; + + if (mlx5_is_secondary()) + return -E_RTE_SECONDARY; priv_lock(priv); + ret = priv_get_mtu(priv, &kern_mtu); + if (ret) + goto out; /* Set kernel interface MTU first. */ - if (priv_set_mtu(priv, mtu)) { - ret = errno; - WARN("cannot set port %u MTU to %u: %s", priv->port, mtu, - strerror(ret)); + ret = priv_set_mtu(priv, mtu); + if (ret) + goto out; + ret = priv_get_mtu(priv, &kern_mtu); + if (ret) goto out; - } else + if (kern_mtu == mtu) { + priv->mtu = mtu; 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; - /* Make sure everyone has left mlx5_rx_burst() and uses - * removed_rx_burst() instead. */ - rte_wmb(); - usleep(1000); - /* Reconfigure each RX queue. */ - for (i = 0; (i != priv->rxqs_n); ++i) { - struct rxq *rxq = (*priv->rxqs)[i]; - unsigned int max_frame_len; - int sp; - - 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)); - /* 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; - } - /* Scattered burst function takes priority. */ - if (rxq->sp) - rx_func = mlx5_rx_burst_sp; } - /* Burst functions can now be called again. */ - rte_wmb(); - dev->rx_pkt_burst = rx_func; + priv_unlock(priv); + return 0; out: + ret = errno; + WARN("cannot set port %u MTU to %u: %s", priv->port, mtu, + strerror(ret)); priv_unlock(priv); assert(ret >= 0); return -ret; @@ -666,7 +1002,10 @@ 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; @@ -714,7 +1053,10 @@ 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)) @@ -781,7 +1123,7 @@ mlx5_ibv_device_to_pci_addr(const struct ibv_device *device, /* Extract information. */ if (sscanf(line, "PCI_SLOT_NAME=" - "%" SCNx16 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", + "%" SCNx32 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", &pci_addr->domain, &pci_addr->bus, &pci_addr->devid, @@ -795,52 +1137,77 @@ mlx5_ibv_device_to_pci_addr(const struct ibv_device *device, } /** - * Link status handler. + * Update the link status. * * @param priv * Pointer to private structure. - * @param dev - * Pointer to the rte_eth_dev structure. * * @return - * Nonzero if the callback process can be called immediately. + * Zero if the callback process can be called immediately. */ static int -priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev) +priv_link_status_update(struct priv *priv) +{ + struct rte_eth_link *link = &priv->dev->data->dev_link; + + mlx5_link_update(priv->dev, 0); + if (((link->link_speed == 0) && link->link_status) || + ((link->link_speed != 0) && !link->link_status)) { + /* + * Inconsistent status. Event likely occurred before the + * kernel netdevice exposes the new status. + */ + if (!priv->pending_alarm) { + priv->pending_alarm = 1; + rte_eal_alarm_set(MLX5_ALARM_TIMEOUT_US, + mlx5_dev_link_status_handler, + priv->dev); + } + return 1; + } else if (unlikely(priv->pending_alarm)) { + /* Link interrupt occurred while alarm is already scheduled. */ + priv->pending_alarm = 0; + rte_eal_alarm_cancel(mlx5_dev_link_status_handler, priv->dev); + } + return 0; +} + +/** + * Device status handler. + * + * @param priv + * Pointer to private structure. + * @param events + * Pointer to event flags holder. + * + * @return + * Events bitmap of callback process which can be called immediately. + */ +static uint32_t +priv_dev_status_handler(struct priv *priv) { struct ibv_async_event event; - int port_change = 0; - int ret = 0; + uint32_t ret = 0; /* Read all message and acknowledge them. */ for (;;) { if (ibv_get_async_event(priv->ctx, &event)) break; - - if (event.event_type == IBV_EVENT_PORT_ACTIVE || - event.event_type == IBV_EVENT_PORT_ERR) - port_change = 1; + if ((event.event_type == IBV_EVENT_PORT_ACTIVE || + event.event_type == IBV_EVENT_PORT_ERR) && + (priv->dev->data->dev_conf.intr_conf.lsc == 1)) + ret |= (1 << RTE_ETH_EVENT_INTR_LSC); + else if (event.event_type == IBV_EVENT_DEVICE_FATAL && + priv->dev->data->dev_conf.intr_conf.rmv == 1) + ret |= (1 << RTE_ETH_EVENT_INTR_RMV); else DEBUG("event type %d on port %d not handled", event.event_type, event.element.port_num); ibv_ack_async_event(&event); } - - if (port_change ^ priv->pending_alarm) { - struct rte_eth_link *link = &dev->data->dev_link; - - priv->pending_alarm = 0; - mlx5_link_update_unlocked(dev, 0); - if (((link->link_speed == 0) && link->link_status) || - ((link->link_speed != 0) && !link->link_status)) { - /* Inconsistent status, check again later. */ - priv->pending_alarm = 1; - rte_eal_alarm_set(MLX5_ALARM_TIMEOUT_US, - mlx5_dev_link_status_handler, - dev); - } else - ret = 1; - } + if (ret & (1 << RTE_ETH_EVENT_INTR_LSC)) + if (priv_link_status_update(priv)) + ret &= ~(1 << RTE_ETH_EVENT_INTR_LSC); return ret; } @@ -859,10 +1226,12 @@ mlx5_dev_link_status_handler(void *arg) priv_lock(priv); assert(priv->pending_alarm == 1); - ret = priv_dev_link_status_handler(priv, dev); + priv->pending_alarm = 0; + ret = priv_link_status_update(priv); priv_unlock(priv); - if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); + if (!ret) + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, + NULL); } /** @@ -874,18 +1243,38 @@ mlx5_dev_link_status_handler(void *arg) * Callback argument. */ void -mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) +mlx5_dev_interrupt_handler(void *cb_arg) { struct rte_eth_dev *dev = cb_arg; struct priv *priv = dev->data->dev_private; - int ret; + uint32_t events; - (void)intr_handle; priv_lock(priv); - ret = priv_dev_link_status_handler(priv, dev); + events = priv_dev_status_handler(priv); + priv_unlock(priv); + if (events & (1 << RTE_ETH_EVENT_INTR_LSC)) + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, + NULL); + if (events & (1 << RTE_ETH_EVENT_INTR_RMV)) + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_RMV, NULL, + NULL); +} + +/** + * Handle interrupts from the socket. + * + * @param cb_arg + * Callback argument. + */ +static void +mlx5_dev_handler_socket(void *cb_arg) +{ + struct rte_eth_dev *dev = cb_arg; + struct priv *priv = dev->data->dev_private; + + priv_lock(priv); + priv_socket_handle(priv); priv_unlock(priv); - if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); } /** @@ -899,16 +1288,20 @@ mlx5_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) void priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) { - if (!dev->data->dev_conf.intr_conf.lsc) - return; - rte_intr_callback_unregister(&priv->intr_handle, - mlx5_dev_interrupt_handler, - dev); + if (dev->data->dev_conf.intr_conf.lsc || + dev->data->dev_conf.intr_conf.rmv) + rte_intr_callback_unregister(&priv->intr_handle, + mlx5_dev_interrupt_handler, dev); + if (priv->primary_socket) + rte_intr_callback_unregister(&priv->intr_handle_socket, + mlx5_dev_handler_socket, dev); if (priv->pending_alarm) 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; + priv->intr_handle_socket.fd = 0; + priv->intr_handle_socket.type = RTE_INTR_HANDLE_UNKNOWN; } /** @@ -924,19 +1317,160 @@ priv_dev_interrupt_handler_install(struct priv *priv, struct rte_eth_dev *dev) { int rc, flags; - if (!dev->data->dev_conf.intr_conf.lsc) - return; + assert(!mlx5_is_secondary()); assert(priv->ctx->async_fd > 0); flags = fcntl(priv->ctx->async_fd, F_GETFL); rc = fcntl(priv->ctx->async_fd, F_SETFL, flags | O_NONBLOCK); if (rc < 0) { INFO("failed to change file descriptor async event queue"); dev->data->dev_conf.intr_conf.lsc = 0; - } else { + dev->data->dev_conf.intr_conf.rmv = 0; + } + if (dev->data->dev_conf.intr_conf.lsc || + dev->data->dev_conf.intr_conf.rmv) { priv->intr_handle.fd = priv->ctx->async_fd; priv->intr_handle.type = RTE_INTR_HANDLE_EXT; rte_intr_callback_register(&priv->intr_handle, - mlx5_dev_interrupt_handler, - dev); + mlx5_dev_interrupt_handler, dev); + } + + rc = priv_socket_init(priv); + if (!rc && priv->primary_socket) { + priv->intr_handle_socket.fd = priv->primary_socket; + priv->intr_handle_socket.type = RTE_INTR_HANDLE_EXT; + rte_intr_callback_register(&priv->intr_handle_socket, + mlx5_dev_handler_socket, dev); + } +} + +/** + * Change the link state (UP / DOWN). + * + * @param priv + * Pointer to private data structure. + * @param dev + * Pointer to rte_eth_dev structure. + * @param up + * Nonzero for link up, otherwise link down. + * + * @return + * 0 on success, errno value on failure. + */ +static int +priv_dev_set_link(struct priv *priv, struct rte_eth_dev *dev, int up) +{ + int err; + + if (up) { + err = priv_set_flags(priv, ~IFF_UP, IFF_UP); + if (err) + return err; + priv_dev_select_tx_function(priv, dev); + priv_dev_select_rx_function(priv, dev); + } else { + err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP); + if (err) + return err; + dev->rx_pkt_burst = removed_rx_burst; + dev->tx_pkt_burst = removed_tx_burst; + } + return 0; +} + +/** + * DPDK callback to bring the link DOWN. + * + * @param dev + * Pointer to Ethernet device structure. + * + * @return + * 0 on success, errno value on failure. + */ +int +mlx5_set_link_down(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + int err; + + priv_lock(priv); + err = priv_dev_set_link(priv, dev, 0); + priv_unlock(priv); + return err; +} + +/** + * DPDK callback to bring the link UP. + * + * @param dev + * Pointer to Ethernet device structure. + * + * @return + * 0 on success, errno value on failure. + */ +int +mlx5_set_link_up(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + int err; + + priv_lock(priv); + err = priv_dev_set_link(priv, dev, 1); + priv_unlock(priv); + return err; +} + +/** + * Configure the TX function to use. + * + * @param priv + * Pointer to private data structure. + * @param dev + * Pointer to rte_eth_dev structure. + */ +void +priv_dev_select_tx_function(struct priv *priv, struct rte_eth_dev *dev) +{ + assert(priv != NULL); + assert(dev != NULL); + dev->tx_pkt_burst = mlx5_tx_burst; + /* Select appropriate TX function. */ + if (priv->mps == MLX5_MPW_ENHANCED) { + if (priv_check_vec_tx_support(priv) > 0) { + if (priv_check_raw_vec_tx_support(priv) > 0) + dev->tx_pkt_burst = mlx5_tx_burst_raw_vec; + else + dev->tx_pkt_burst = mlx5_tx_burst_vec; + DEBUG("selected Enhanced MPW TX vectorized function"); + } else { + dev->tx_pkt_burst = mlx5_tx_burst_empw; + DEBUG("selected Enhanced MPW TX function"); + } + } else if (priv->mps && priv->txq_inline) { + dev->tx_pkt_burst = mlx5_tx_burst_mpw_inline; + DEBUG("selected MPW inline TX function"); + } else if (priv->mps) { + dev->tx_pkt_burst = mlx5_tx_burst_mpw; + DEBUG("selected MPW TX function"); + } +} + +/** + * Configure the RX function to use. + * + * @param priv + * Pointer to private data structure. + * @param dev + * Pointer to rte_eth_dev structure. + */ +void +priv_dev_select_rx_function(struct priv *priv, struct rte_eth_dev *dev) +{ + assert(priv != NULL); + assert(dev != NULL); + if (priv_check_vec_rx_support(priv) > 0) { + dev->rx_pkt_burst = mlx5_rx_burst_vec; + DEBUG("selected RX vectorized function"); + } else { + dev->rx_pkt_burst = mlx5_rx_burst; } }