X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx4%2Fmlx4.c;h=30c70ee5df46c6111b14d9d20eef090fe6847345;hb=056eaf2e6d55;hp=da61a856345db8c8c809d80e880e29848e39675d;hpb=83e9d9a3f5a42eebd2f30fc76762378b0d10a1fb;p=dpdk.git diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c index da61a85634..30c70ee5df 100644 --- a/drivers/net/mlx4/mlx4.c +++ b/drivers/net/mlx4/mlx4.c @@ -1,8 +1,8 @@ /*- * BSD LICENSE * - * Copyright 2012-2015 6WIND S.A. - * Copyright 2012 Mellanox. + * Copyright 2012 6WIND S.A. + * Copyright 2012 Mellanox * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -31,12 +31,6 @@ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ -/* - * Known limitations: - * - RSS hash key and options cannot be modified. - * - Hardware counters aren't implemented. - */ - /* System headers. */ #include #include @@ -46,9 +40,7 @@ #include #include #include -#include #include -#include #include #include #include @@ -58,22 +50,9 @@ #include #include -/* Verbs header. */ -/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-Wpedantic" -#endif -#include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-Wpedantic" -#endif - -/* DPDK headers don't like -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-Wpedantic" -#endif #include #include +#include #include #include #include @@ -81,35 +60,20 @@ #include #include #include -#include -#include #include #include #include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-Wpedantic" -#endif +#include +#include +#include +#include /* Generated configuration header. */ #include "mlx4_autoconf.h" -/* PMD header. */ +/* PMD headers. */ #include "mlx4.h" - -/* Runtime logging through RTE_LOG() is enabled when not in debugging mode. - * Intermediate LOG_*() macros add the required end-of-line characters. */ -#ifndef NDEBUG -#define INFO(...) DEBUG(__VA_ARGS__) -#define WARN(...) DEBUG(__VA_ARGS__) -#define ERROR(...) DEBUG(__VA_ARGS__) -#else -#define LOG__(level, m, ...) \ - RTE_LOG(level, PMD, MLX4_DRIVER_NAME ": " m "%c", __VA_ARGS__) -#define LOG_(level, ...) LOG__(level, __VA_ARGS__, '\n') -#define INFO(...) LOG_(INFO, __VA_ARGS__) -#define WARN(...) LOG_(WARNING, __VA_ARGS__) -#define ERROR(...) LOG_(ERR, __VA_ARGS__) -#endif +#include "mlx4_flow.h" /* Convenience macros for accessing mbuf fields. */ #define NEXT(m) ((m)->next) @@ -131,202 +95,31 @@ typedef union { #define WR_ID(o) (((wr_id_t *)&(o))->data) -/* Transpose flags. Useful to convert IBV to DPDK flags. */ -#define TRANSPOSE(val, from, to) \ - (((from) >= (to)) ? \ - (((val) & (from)) / ((from) / (to))) : \ - (((val) & (from)) * ((to) / (from)))) - -struct mlx4_rxq_stats { - unsigned int idx; /**< Mapping index. */ -#ifdef MLX4_PMD_SOFT_COUNTERS - uint64_t ipackets; /**< Total of successfully received packets. */ - uint64_t ibytes; /**< Total of successfully received bytes. */ -#endif - uint64_t idropped; /**< Total of packets dropped when RX ring full. */ - uint64_t rx_nombuf; /**< Total of RX mbuf allocation failures. */ -}; - -struct mlx4_txq_stats { - unsigned int idx; /**< Mapping index. */ -#ifdef MLX4_PMD_SOFT_COUNTERS - uint64_t opackets; /**< Total of successfully sent packets. */ - uint64_t obytes; /**< Total of successfully sent bytes. */ -#endif - uint64_t odropped; /**< Total of packets not sent when TX ring full. */ -}; - -/* RX element (scattered packets). */ -struct rxq_elt_sp { - struct ibv_recv_wr wr; /* Work Request. */ - struct ibv_sge sges[MLX4_PMD_SGE_WR_N]; /* Scatter/Gather Elements. */ - struct rte_mbuf *bufs[MLX4_PMD_SGE_WR_N]; /* SGEs buffers. */ -}; - -/* RX element. */ -struct rxq_elt { - struct ibv_recv_wr wr; /* Work Request. */ - struct ibv_sge sge; /* Scatter/Gather Element. */ - /* mbuf pointer is derived from WR_ID(wr.wr_id).offset. */ -}; - -/* RX queue descriptor. */ -struct rxq { - struct priv *priv; /* Back pointer to private data. */ - struct rte_mempool *mp; /* Memory Pool for allocations. */ - struct ibv_mr *mr; /* Memory Region (for mp). */ - struct ibv_cq *cq; /* Completion Queue. */ - struct ibv_qp *qp; /* Queue Pair. */ - struct ibv_exp_qp_burst_family *if_qp; /* QP burst interface. */ - struct ibv_exp_cq_family *if_cq; /* CQ interface. */ - /* - * Each VLAN ID requires a separate flow steering rule. - */ - BITFIELD_DECLARE(mac_configured, uint32_t, MLX4_MAX_MAC_ADDRESSES); - struct ibv_flow *mac_flow[MLX4_MAX_MAC_ADDRESSES][MLX4_MAX_VLAN_IDS]; - struct ibv_flow *promisc_flow; /* Promiscuous flow. */ - struct ibv_flow *allmulti_flow; /* Multicast flow. */ - unsigned int port_id; /* Port ID for incoming packets. */ - unsigned int elts_n; /* (*elts)[] length. */ - unsigned int elts_head; /* Current index in (*elts)[]. */ - union { - struct rxq_elt_sp (*sp)[]; /* Scattered RX elements. */ - struct rxq_elt (*no_sp)[]; /* RX elements. */ - } elts; - unsigned int sp:1; /* Use scattered RX elements. */ - unsigned int csum:1; /* Enable checksum offloading. */ - unsigned int csum_l2tun:1; /* Same for L2 tunnels. */ - struct mlx4_rxq_stats stats; /* RX queue counters. */ - unsigned int socket; /* CPU socket ID for allocations. */ - struct ibv_exp_res_domain *rd; /* Resource Domain. */ -}; - -/* TX element. */ -struct txq_elt { - struct rte_mbuf *buf; -}; - -/* Linear buffer type. It is used when transmitting buffers with too many - * segments that do not fit the hardware queue (see max_send_sge). - * Extra segments are copied (linearized) in such buffers, replacing the - * last SGE during TX. - * The size is arbitrary but large enough to hold a jumbo frame with - * 8 segments considering mbuf.buf_len is about 2048 bytes. */ -typedef uint8_t linear_t[16384]; - -/* TX queue descriptor. */ -struct txq { - struct priv *priv; /* Back pointer to private data. */ +/** Configuration structure for device arguments. */ +struct mlx4_conf { struct { - const struct rte_mempool *mp; /* Cached Memory Pool. */ - struct ibv_mr *mr; /* Memory Region (for mp). */ - uint32_t lkey; /* mr->lkey */ - } mp2mr[MLX4_PMD_TX_MP_CACHE]; /* MP to MR translation table. */ - struct ibv_cq *cq; /* Completion Queue. */ - struct ibv_qp *qp; /* Queue Pair. */ - struct ibv_exp_qp_burst_family *if_qp; /* QP burst interface. */ - struct ibv_exp_cq_family *if_cq; /* CQ interface. */ -#if MLX4_PMD_MAX_INLINE > 0 - uint32_t max_inline; /* Max inline send size <= MLX4_PMD_MAX_INLINE. */ -#endif - unsigned int elts_n; /* (*elts)[] length. */ - struct txq_elt (*elts)[]; /* TX elements. */ - unsigned int elts_head; /* Current index in (*elts)[]. */ - unsigned int elts_tail; /* First element awaiting completion. */ - unsigned int elts_comp; /* Number of completion requests. */ - unsigned int elts_comp_cd; /* Countdown for next completion request. */ - unsigned int elts_comp_cd_init; /* Initial value for countdown. */ - struct mlx4_txq_stats stats; /* TX queue counters. */ - linear_t (*elts_linear)[]; /* Linearized buffers. */ - struct ibv_mr *mr_linear; /* Memory Region for linearized buffers. */ - unsigned int socket; /* CPU socket ID for allocations. */ - struct ibv_exp_res_domain *rd; /* Resource Domain. */ + uint32_t present; /**< Bit-field for existing ports. */ + uint32_t enabled; /**< Bit-field for user-enabled ports. */ + } ports; }; -struct priv { - struct rte_eth_dev *dev; /* Ethernet device. */ - struct ibv_context *ctx; /* Verbs context. */ - struct ibv_device_attr device_attr; /* Device properties. */ - struct ibv_pd *pd; /* Protection Domain. */ - /* - * MAC addresses array and configuration bit-field. - * An extra entry that cannot be modified by the DPDK is reserved - * for broadcast frames (destination MAC address ff:ff:ff:ff:ff:ff). - */ - struct ether_addr mac[MLX4_MAX_MAC_ADDRESSES]; - BITFIELD_DECLARE(mac_configured, uint32_t, MLX4_MAX_MAC_ADDRESSES); - /* VLAN filters. */ - struct { - unsigned int enabled:1; /* If enabled. */ - unsigned int id:12; /* VLAN ID (0-4095). */ - } vlan_filter[MLX4_MAX_VLAN_IDS]; /* VLAN filters table. */ - /* Device properties. */ - uint16_t mtu; /* Configured MTU. */ - uint8_t port; /* Physical port number. */ - unsigned int started:1; /* Device started, flows enabled. */ - unsigned int promisc:1; /* Device in promiscuous mode. */ - unsigned int allmulti:1; /* Device receives all multicast packets. */ - unsigned int hw_qpg:1; /* QP groups are supported. */ - unsigned int hw_tss:1; /* TSS is supported. */ - unsigned int hw_rss:1; /* RSS is supported. */ - unsigned int hw_csum:1; /* Checksum offload is supported. */ - unsigned int hw_csum_l2tun:1; /* Same for L2 tunnels. */ - unsigned int rss:1; /* RSS is enabled. */ - unsigned int vf:1; /* This is a VF device. */ - unsigned int pending_alarm:1; /* An alarm is pending. */ -#ifdef INLINE_RECV - unsigned int inl_recv_size; /* Inline recv size */ -#endif - unsigned int max_rss_tbl_sz; /* Maximum number of RSS queues. */ - /* RX/TX queues. */ - struct rxq rxq_parent; /* Parent queue when RSS is enabled. */ - unsigned int rxqs_n; /* RX queues array size. */ - unsigned int txqs_n; /* TX queues array size. */ - struct rxq *(*rxqs)[]; /* RX queues. */ - struct txq *(*txqs)[]; /* TX queues. */ - struct rte_intr_handle intr_handle; /* Interrupt handler. */ - rte_spinlock_t lock; /* Lock for control functions. */ +/* Available parameters list. */ +const char *pmd_mlx4_init_params[] = { + MLX4_PMD_PORT_KVARG, + NULL, }; -/* Local storage for secondary process data. */ -struct mlx4_secondary_data { - struct rte_eth_dev_data data; /* Local device data. */ - struct priv *primary_priv; /* Private structure from primary. */ - struct rte_eth_dev_data *shared_dev_data; /* Shared device data. */ - rte_spinlock_t lock; /* Port configuration lock. */ -} mlx4_secondary_data[RTE_MAX_ETHPORTS]; +static int +mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx); -/** - * Check if running as a secondary process. - * - * @return - * Nonzero if running as a secondary process. - */ -static inline int -mlx4_is_secondary(void) -{ - return rte_eal_process_type() != RTE_PROC_PRIMARY; -} +static int +mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx); -/** - * Return private structure associated with an Ethernet device. - * - * @param dev - * Pointer to Ethernet device structure. - * - * @return - * Pointer to private structure. - */ -static struct priv * -mlx4_get_priv(struct rte_eth_dev *dev) -{ - struct mlx4_secondary_data *sd; +static int +priv_rx_intr_vec_enable(struct priv *priv); - if (!mlx4_is_secondary()) - return dev->data->dev_private; - sd = &mlx4_secondary_data[dev->data->port_id]; - return sd->data.dev_private; -} +static void +priv_rx_intr_vec_disable(struct priv *priv); /** * Lock private structure to protect it from concurrent access in the @@ -335,8 +128,7 @@ mlx4_get_priv(struct rte_eth_dev *dev) * @param priv * Pointer to private structure. */ -static void -priv_lock(struct priv *priv) +void priv_lock(struct priv *priv) { rte_spinlock_lock(&priv->lock); } @@ -347,8 +139,7 @@ priv_lock(struct priv *priv) * @param priv * Pointer to private structure. */ -static void -priv_unlock(struct priv *priv) +void priv_unlock(struct priv *priv) { rte_spinlock_unlock(&priv->lock); } @@ -704,17 +495,19 @@ txq_cleanup(struct txq *txq); static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, int inactive, const struct rte_eth_rxconf *conf, + unsigned int socket, const struct rte_eth_rxconf *conf, struct rte_mempool *mp); static void rxq_cleanup(struct rxq *rxq); +static void +priv_mac_addr_del(struct priv *priv); + /** * Ethernet device configuration. * * Prepare the driver for a given number of TX and RX queues. - * Allocate parent RSS queue when several RX queues are requested. * * @param dev * Pointer to Ethernet device structure. @@ -728,8 +521,6 @@ 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 tmp; - int ret; priv->rxqs = (void *)dev->data->rx_queues; priv->txqs = (void *)dev->data->tx_queues; @@ -738,60 +529,12 @@ dev_configure(struct rte_eth_dev *dev) (void *)dev, priv->txqs_n, txqs_n); priv->txqs_n = txqs_n; } - if (rxqs_n == priv->rxqs_n) - return 0; - if (!rte_is_power_of_2(rxqs_n)) { - unsigned n_active; - - n_active = rte_align32pow2(rxqs_n + 1) >> 1; - WARN("%p: number of RX queues must be a power" - " of 2: %u queues among %u will be active", - (void *)dev, n_active, rxqs_n); - } - - INFO("%p: RX queues number update: %u -> %u", - (void *)dev, priv->rxqs_n, rxqs_n); - /* If RSS is enabled, disable it first. */ - if (priv->rss) { - unsigned int i; - - /* Only if there are no remaining child RX queues. */ - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - return EINVAL; - rxq_cleanup(&priv->rxq_parent); - priv->rss = 0; - priv->rxqs_n = 0; - } - if (rxqs_n <= 1) { - /* Nothing else to do. */ + if (rxqs_n != priv->rxqs_n) { + INFO("%p: Rx queues number update: %u -> %u", + (void *)dev, priv->rxqs_n, rxqs_n); priv->rxqs_n = rxqs_n; - return 0; } - /* Allocate a new RSS parent queue if supported by hardware. */ - if (!priv->hw_rss) { - ERROR("%p: only a single RX queue can be configured when" - " hardware doesn't support RSS", - (void *)dev); - return EINVAL; - } - /* Fail if hardware doesn't support that many RSS queues. */ - if (rxqs_n >= priv->max_rss_tbl_sz) { - ERROR("%p: only %u RX queues can be configured for RSS", - (void *)dev, priv->max_rss_tbl_sz); - return EINVAL; - } - priv->rss = 1; - tmp = priv->rxqs_n; - priv->rxqs_n = rxqs_n; - ret = rxq_setup(dev, &priv->rxq_parent, 0, 0, 0, NULL, NULL); - if (!ret) - return 0; - /* Failure, rollback. */ - priv->rss = 0; - priv->rxqs_n = tmp; - assert(ret > 0); - return ret; + return 0; } /** @@ -809,8 +552,6 @@ mlx4_dev_configure(struct rte_eth_dev *dev) struct priv *priv = dev->data->dev_private; int ret; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; priv_lock(priv); ret = dev_configure(dev); assert(ret >= 0); @@ -821,157 +562,6 @@ mlx4_dev_configure(struct rte_eth_dev *dev) static uint16_t mlx4_tx_burst(void *, struct rte_mbuf **, uint16_t); static uint16_t removed_rx_burst(void *, struct rte_mbuf **, uint16_t); -/** - * 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. - */ -static struct priv * -mlx4_secondary_data_setup(struct priv *priv) -{ - unsigned int port_id = 0; - struct mlx4_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 = &mlx4_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(mlx4_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 *txq; - - if (primary_txq == NULL) - continue; - txq = rte_calloc_socket("TXQ", 1, sizeof(*txq), 0, - primary_txq->socket); - if (txq != NULL) { - if (txq_setup(priv->dev, - txq, - primary_txq->elts_n * MLX4_PMD_SGE_WR_N, - primary_txq->socket, - NULL) == 0) { - txq->stats.idx = primary_txq->stats.idx; - tx_queues[i] = txq; - continue; - } - rte_free(txq); - } - while (i) { - txq = tx_queues[--i]; - txq_cleanup(txq); - rte_free(txq); - } - goto error; - } - /* RX queues. */ - for (i = 0; i != nb_rx_queues; ++i) { - struct rxq *primary_rxq = (*sd->primary_priv->rxqs)[i]; - - 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->dev->tx_pkt_burst = mlx4_tx_burst; - priv->dev->rx_pkt_burst = removed_rx_burst; - priv_unlock(priv); -end: - /* More sanity checks. */ - assert(priv->dev->tx_pkt_burst == mlx4_tx_burst); - assert(priv->dev->rx_pkt_burst == removed_rx_burst); - 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; -} - /* TX queues handling. */ /** @@ -991,26 +581,13 @@ txq_alloc_elts(struct txq *txq, unsigned int elts_n) unsigned int i; struct txq_elt (*elts)[elts_n] = rte_calloc_socket("TXQ", 1, sizeof(*elts), 0, txq->socket); - linear_t (*elts_linear)[elts_n] = - rte_calloc_socket("TXQ", 1, sizeof(*elts_linear), 0, - txq->socket); - struct ibv_mr *mr_linear = NULL; int ret = 0; - if ((elts == NULL) || (elts_linear == NULL)) { + if (elts == NULL) { ERROR("%p: can't allocate packets array", (void *)txq); ret = ENOMEM; goto error; } - mr_linear = - ibv_reg_mr(txq->priv->pd, elts_linear, sizeof(*elts_linear), - IBV_ACCESS_LOCAL_WRITE); - if (mr_linear == NULL) { - ERROR("%p: unable to configure MR, ibv_reg_mr() failed", - (void *)txq); - ret = EINVAL; - goto error; - } for (i = 0; (i != elts_n); ++i) { struct txq_elt *elt = &(*elts)[i]; @@ -1028,15 +605,9 @@ txq_alloc_elts(struct txq *txq, unsigned int elts_n) ((MLX4_PMD_TX_PER_COMP_REQ < (elts_n / 4)) ? MLX4_PMD_TX_PER_COMP_REQ : (elts_n / 4)); txq->elts_comp_cd = txq->elts_comp_cd_init; - txq->elts_linear = elts_linear; - txq->mr_linear = mr_linear; assert(ret == 0); return 0; error: - if (mr_linear != NULL) - claim_zero(ibv_dereg_mr(mr_linear)); - - rte_free(elts_linear); rte_free(elts); DEBUG("%p: failed, freed everything", (void *)txq); @@ -1057,8 +628,6 @@ txq_free_elts(struct txq *txq) unsigned int elts_head = txq->elts_head; unsigned int elts_tail = txq->elts_tail; struct txq_elt (*elts)[elts_n] = txq->elts; - linear_t (*elts_linear)[elts_n] = txq->elts_linear; - struct ibv_mr *mr_linear = txq->mr_linear; DEBUG("%p: freeing WRs", (void *)txq); txq->elts_n = 0; @@ -1068,12 +637,6 @@ txq_free_elts(struct txq *txq) txq->elts_comp_cd = 0; txq->elts_comp_cd_init = 0; txq->elts = NULL; - txq->elts_linear = NULL; - txq->mr_linear = NULL; - if (mr_linear != NULL) - claim_zero(ibv_dereg_mr(mr_linear)); - - rte_free(elts_linear); if (elts == NULL) return; while (elts_tail != elts_head) { @@ -1178,10 +741,6 @@ txq_complete(struct txq *txq) if (unlikely(elts_comp == 0)) return 0; -#ifdef DEBUG_SEND - DEBUG("%p: processing %u work requests completions", - (void *)txq, elts_comp); -#endif wcs_n = txq->if_cq->poll_cnt(txq->cq, elts_comp); if (unlikely(wcs_n == 0)) return 0; @@ -1269,7 +828,7 @@ static int mlx4_check_mempool(struct rte_mempool *mp, uintptr_t *start, /* For best performance, this function should not be inlined. */ static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, struct rte_mempool *) - __attribute__((noinline)); + __rte_noinline; /** * Register mempool as a memory region. @@ -1449,156 +1008,6 @@ txq_mp2mr_iter(struct rte_mempool *mp, void *arg) txq_mp2mr(txq, mp); } -#if MLX4_PMD_SGE_WR_N > 1 - -/** - * Copy scattered mbuf contents to a single linear buffer. - * - * @param[out] linear - * Linear output buffer. - * @param[in] buf - * Scattered input buffer. - * - * @return - * Number of bytes copied to the output buffer or 0 if not large enough. - */ -static unsigned int -linearize_mbuf(linear_t *linear, struct rte_mbuf *buf) -{ - unsigned int size = 0; - unsigned int offset; - - do { - unsigned int len = DATA_LEN(buf); - - offset = size; - size += len; - if (unlikely(size > sizeof(*linear))) - return 0; - memcpy(&(*linear)[offset], - rte_pktmbuf_mtod(buf, uint8_t *), - len); - buf = NEXT(buf); - } while (buf != NULL); - return size; -} - -/** - * Handle scattered buffers for mlx4_tx_burst(). - * - * @param txq - * TX queue structure. - * @param segs - * Number of segments in buf. - * @param elt - * TX queue element to fill. - * @param[in] buf - * Buffer to process. - * @param elts_head - * Index of the linear buffer to use if necessary (normally txq->elts_head). - * @param[out] sges - * Array filled with SGEs on success. - * - * @return - * A structure containing the processed packet size in bytes and the - * number of SGEs. Both fields are set to (unsigned int)-1 in case of - * failure. - */ -static struct tx_burst_sg_ret { - unsigned int length; - unsigned int num; -} -tx_burst_sg(struct txq *txq, unsigned int segs, struct txq_elt *elt, - struct rte_mbuf *buf, unsigned int elts_head, - struct ibv_sge (*sges)[MLX4_PMD_SGE_WR_N]) -{ - unsigned int sent_size = 0; - unsigned int j; - int linearize = 0; - - /* When there are too many segments, extra segments are - * linearized in the last SGE. */ - if (unlikely(segs > elemof(*sges))) { - segs = (elemof(*sges) - 1); - linearize = 1; - } - /* Update element. */ - elt->buf = buf; - /* Register segments as SGEs. */ - for (j = 0; (j != segs); ++j) { - struct ibv_sge *sge = &(*sges)[j]; - uint32_t lkey; - - /* Retrieve Memory Region key for this memory pool. */ - lkey = txq_mp2mr(txq, txq_mb2mp(buf)); - if (unlikely(lkey == (uint32_t)-1)) { - /* MR does not exist. */ - DEBUG("%p: unable to get MP <-> MR association", - (void *)txq); - /* Clean up TX element. */ - elt->buf = NULL; - goto stop; - } - /* Update SGE. */ - sge->addr = rte_pktmbuf_mtod(buf, uintptr_t); - if (txq->priv->vf) - rte_prefetch0((volatile void *) - (uintptr_t)sge->addr); - sge->length = DATA_LEN(buf); - sge->lkey = lkey; - sent_size += sge->length; - buf = NEXT(buf); - } - /* If buf is not NULL here and is not going to be linearized, - * nb_segs is not valid. */ - assert(j == segs); - assert((buf == NULL) || (linearize)); - /* Linearize extra segments. */ - if (linearize) { - struct ibv_sge *sge = &(*sges)[segs]; - linear_t *linear = &(*txq->elts_linear)[elts_head]; - unsigned int size = linearize_mbuf(linear, buf); - - assert(segs == (elemof(*sges) - 1)); - if (size == 0) { - /* Invalid packet. */ - DEBUG("%p: packet too large to be linearized.", - (void *)txq); - /* Clean up TX element. */ - elt->buf = NULL; - goto stop; - } - /* If MLX4_PMD_SGE_WR_N is 1, free mbuf immediately. */ - if (elemof(*sges) == 1) { - do { - struct rte_mbuf *next = NEXT(buf); - - rte_pktmbuf_free_seg(buf); - buf = next; - } while (buf != NULL); - elt->buf = NULL; - } - /* Update SGE. */ - sge->addr = (uintptr_t)&(*linear)[0]; - sge->length = size; - sge->lkey = txq->mr_linear->lkey; - sent_size += size; - /* Include last segment. */ - segs++; - } - return (struct tx_burst_sg_ret){ - .length = sent_size, - .num = segs, - }; -stop: - return (struct tx_burst_sg_ret){ - .length = -1, - .num = -1, - }; -} - -#endif /* MLX4_PMD_SGE_WR_N > 1 */ - /** * DPDK callback for TX. * @@ -1644,9 +1053,7 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) struct txq_elt *elt_next = &(*txq->elts)[elts_head_next]; struct txq_elt *elt = &(*txq->elts)[elts_head]; unsigned int segs = NB_SEGS(buf); -#ifdef MLX4_PMD_SOFT_COUNTERS unsigned int sent_size = 0; -#endif uint32_t send_flags = 0; /* Clean up old buffer. */ @@ -1671,17 +1078,6 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) ++elts_comp; send_flags |= IBV_EXP_QP_BURST_SIGNALED; } - /* Should we enable HW CKSUM offload */ - if (buf->ol_flags & - (PKT_TX_IP_CKSUM | PKT_TX_TCP_CKSUM | PKT_TX_UDP_CKSUM)) { - send_flags |= IBV_EXP_QP_BURST_IP_CSUM; - /* HW does not support checksum offloads at arbitrary - * offsets but automatically recognizes the packet - * type. For inner L3/L4 checksums, only VXLAN (UDP) - * tunnels are currently supported. */ - if (RTE_ETH_IS_TUNNEL_PKT(buf->packet_type)) - send_flags |= IBV_EXP_QP_BURST_TUNNEL; - } if (likely(segs == 1)) { uintptr_t addr; uint32_t length; @@ -1707,7 +1103,6 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) (uintptr_t)addr); RTE_MBUF_PREFETCH_TO_FREE(elt_next->buf); /* Put packet into send queue. */ -#if MLX4_PMD_MAX_INLINE > 0 if (length <= txq->max_inline) err = txq->if_qp->send_pending_inline (txq->qp, @@ -1715,7 +1110,6 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) length, send_flags); else -#endif err = txq->if_qp->send_pending (txq->qp, addr, @@ -1724,50 +1118,21 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) send_flags); if (unlikely(err)) goto stop; -#ifdef MLX4_PMD_SOFT_COUNTERS sent_size += length; -#endif } else { -#if MLX4_PMD_SGE_WR_N > 1 - struct ibv_sge sges[MLX4_PMD_SGE_WR_N]; - struct tx_burst_sg_ret ret; - - ret = tx_burst_sg(txq, segs, elt, buf, elts_head, - &sges); - if (ret.length == (unsigned int)-1) - goto stop; - RTE_MBUF_PREFETCH_TO_FREE(elt_next->buf); - /* Put SG list into send queue. */ - err = txq->if_qp->send_pending_sg_list - (txq->qp, - sges, - ret.num, - send_flags); - if (unlikely(err)) - goto stop; -#ifdef MLX4_PMD_SOFT_COUNTERS - sent_size += ret.length; -#endif -#else /* MLX4_PMD_SGE_WR_N > 1 */ - DEBUG("%p: TX scattered buffers support not" - " compiled in", (void *)txq); + err = -1; goto stop; -#endif /* MLX4_PMD_SGE_WR_N > 1 */ } elts_head = elts_head_next; -#ifdef MLX4_PMD_SOFT_COUNTERS /* Increment sent bytes counter. */ txq->stats.obytes += sent_size; -#endif } stop: /* Take a shortcut if nothing must be sent. */ if (unlikely(i == 0)) return 0; -#ifdef MLX4_PMD_SOFT_COUNTERS /* Increment sent packets counter. */ txq->stats.opackets += i; -#endif /* Ring QP doorbell. */ err = txq->if_qp->send_flush(txq->qp); if (unlikely(err)) { @@ -1782,46 +1147,6 @@ stop: return i; } -/** - * DPDK callback for TX in secondary processes. - * - * This function configures all queues from primary process information - * if necessary before reverting to the normal TX burst callback. - * - * @param dpdk_txq - * Generic pointer to TX queue structure. - * @param[in] pkts - * Packets to transmit. - * @param pkts_n - * Number of packets in array. - * - * @return - * Number of packets successfully transmitted (<= pkts_n). - */ -static uint16_t -mlx4_tx_burst_secondary_setup(void *dpdk_txq, struct rte_mbuf **pkts, - uint16_t pkts_n) -{ - struct txq *txq = dpdk_txq; - struct priv *priv = mlx4_secondary_data_setup(txq->priv); - struct priv *primary_priv; - unsigned int index; - - if (priv == NULL) - return 0; - primary_priv = - mlx4_secondary_data[priv->dev->data->port_id].primary_priv; - /* Look for queue index in both private structures. */ - for (index = 0; index != priv->txqs_n; ++index) - if (((*primary_priv->txqs)[index] == txq) || - ((*priv->txqs)[index] == txq)) - break; - if (index == priv->txqs_n) - return 0; - txq = (*priv->txqs)[index]; - return priv->dev->tx_pkt_burst(txq, pkts, pkts_n); -} - /** * Configure a TX queue. * @@ -1843,7 +1168,7 @@ static int txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, unsigned int socket, const struct rte_eth_txconf *conf) { - struct priv *priv = mlx4_get_priv(dev); + struct priv *priv = dev->data->dev_private; struct txq tmpl = { .priv = priv, .socket = socket @@ -1861,12 +1186,10 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, (void)conf; /* Thresholds configuration (ignored). */ if (priv == NULL) return EINVAL; - if ((desc == 0) || (desc % MLX4_PMD_SGE_WR_N)) { - ERROR("%p: invalid number of TX descriptors (must be a" - " multiple of %d)", (void *)dev, MLX4_PMD_SGE_WR_N); + if (desc == 0) { + ERROR("%p: invalid number of Tx descriptors", (void *)dev); return EINVAL; } - desc /= MLX4_PMD_SGE_WR_N; /* MRs will be registered in mp2mr[] later. */ attr.rd = (struct ibv_exp_res_domain_init_attr){ .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL | @@ -1907,13 +1230,8 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, priv->device_attr.max_qp_wr : desc), /* Max number of scatter/gather elements in a WR. */ - .max_send_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), -#if MLX4_PMD_MAX_INLINE > 0 + .max_send_sge = 1, .max_inline_data = MLX4_PMD_MAX_INLINE, -#endif }, .qp_type = IBV_QPT_RAW_PACKET, /* Do *NOT* enable this, completions events are managed per @@ -1931,10 +1249,8 @@ txq_setup(struct rte_eth_dev *dev, struct txq *txq, uint16_t desc, (void *)dev, strerror(ret)); goto error; } -#if MLX4_PMD_MAX_INLINE > 0 /* ibv_create_qp() updates this value. */ tmpl.max_inline = attr.init.cap.max_inline_data; -#endif attr.mod = (struct ibv_exp_qp_attr){ /* Move the QP to this state. */ .qp_state = IBV_QPS_INIT, @@ -2039,8 +1355,6 @@ mlx4_tx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, struct txq *txq = (*priv->txqs)[idx]; int ret; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; priv_lock(priv); DEBUG("%p: configuring queue %u for %u descriptors", (void *)dev, idx, desc); @@ -2096,8 +1410,6 @@ mlx4_tx_queue_release(void *dpdk_txq) struct priv *priv; unsigned int i; - if (mlx4_is_secondary()) - return; if (txq == NULL) return; priv = txq->priv; @@ -2117,25 +1429,21 @@ mlx4_tx_queue_release(void *dpdk_txq) /* RX queues handling. */ /** - * Allocate RX queue elements with scattered packets support. + * Allocate RX queue elements. * * @param rxq * Pointer to RX queue structure. * @param elts_n * Number of elements to allocate. - * @param[in] pool - * If not NULL, fetch buffers from this array instead of allocating them - * with rte_pktmbuf_alloc(). * * @return * 0 on success, errno value on failure. */ static int -rxq_alloc_elts_sp(struct rxq *rxq, unsigned int elts_n, - struct rte_mbuf **pool) +rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n) { unsigned int i; - struct rxq_elt_sp (*elts)[elts_n] = + struct rxq_elt (*elts)[elts_n] = rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0, rxq->socket); int ret = 0; @@ -2147,82 +1455,74 @@ rxq_alloc_elts_sp(struct rxq *rxq, unsigned int elts_n, } /* For each WR (packet). */ for (i = 0; (i != elts_n); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; + struct rxq_elt *elt = &(*elts)[i]; struct ibv_recv_wr *wr = &elt->wr; - struct ibv_sge (*sges)[(elemof(elt->sges))] = &elt->sges; + struct ibv_sge *sge = &(*elts)[i].sge; + struct rte_mbuf *buf = rte_pktmbuf_alloc(rxq->mp); - /* These two arrays must have the same size. */ - assert(elemof(elt->sges) == elemof(elt->bufs)); - /* Configure WR. */ - wr->wr_id = i; + if (buf == NULL) { + ERROR("%p: empty mbuf pool", (void *)rxq); + ret = ENOMEM; + goto error; + } + /* Configure WR. Work request ID contains its own index in + * the elts array and the offset between SGE buffer header and + * its data. */ + WR_ID(wr->wr_id).id = i; + WR_ID(wr->wr_id).offset = + (((uintptr_t)buf->buf_addr + RTE_PKTMBUF_HEADROOM) - + (uintptr_t)buf); wr->next = &(*elts)[(i + 1)].wr; - wr->sg_list = &(*sges)[0]; - wr->num_sge = elemof(*sges); - /* For each SGE (segment). */ - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct ibv_sge *sge = &(*sges)[j]; - struct rte_mbuf *buf; - - if (pool != NULL) { - buf = *(pool++); - assert(buf != NULL); - rte_pktmbuf_reset(buf); - } else - buf = rte_pktmbuf_alloc(rxq->mp); - if (buf == NULL) { - assert(pool == NULL); - ERROR("%p: empty mbuf pool", (void *)rxq); - ret = ENOMEM; - goto error; - } - elt->bufs[j] = buf; - /* Headroom is reserved by rte_pktmbuf_alloc(). */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); - /* Buffer is supposed to be empty. */ - assert(rte_pktmbuf_data_len(buf) == 0); - assert(rte_pktmbuf_pkt_len(buf) == 0); - /* sge->addr must be able to store a pointer. */ - assert(sizeof(sge->addr) >= sizeof(uintptr_t)); - if (j == 0) { - /* The first SGE keeps its headroom. */ - sge->addr = rte_pktmbuf_mtod(buf, uintptr_t); - sge->length = (buf->buf_len - - RTE_PKTMBUF_HEADROOM); - } else { - /* Subsequent SGEs lose theirs. */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); - SET_DATA_OFF(buf, 0); - sge->addr = (uintptr_t)buf->buf_addr; - sge->length = buf->buf_len; - } - sge->lkey = rxq->mr->lkey; - /* Redundant check for tailroom. */ - assert(sge->length == rte_pktmbuf_tailroom(buf)); + wr->sg_list = sge; + wr->num_sge = 1; + /* Headroom is reserved by rte_pktmbuf_alloc(). */ + assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); + /* Buffer is supposed to be empty. */ + assert(rte_pktmbuf_data_len(buf) == 0); + assert(rte_pktmbuf_pkt_len(buf) == 0); + /* sge->addr must be able to store a pointer. */ + assert(sizeof(sge->addr) >= sizeof(uintptr_t)); + /* SGE keeps its headroom. */ + sge->addr = (uintptr_t) + ((uint8_t *)buf->buf_addr + RTE_PKTMBUF_HEADROOM); + sge->length = (buf->buf_len - RTE_PKTMBUF_HEADROOM); + sge->lkey = rxq->mr->lkey; + /* Redundant check for tailroom. */ + assert(sge->length == rte_pktmbuf_tailroom(buf)); + /* Make sure elts index and SGE mbuf pointer can be deduced + * from WR ID. */ + if ((WR_ID(wr->wr_id).id != i) || + ((void *)((uintptr_t)sge->addr - + WR_ID(wr->wr_id).offset) != buf)) { + ERROR("%p: cannot store index and offset in WR ID", + (void *)rxq); + sge->addr = 0; + rte_pktmbuf_free(buf); + ret = EOVERFLOW; + goto error; } } /* The last WR pointer must be NULL. */ (*elts)[(i - 1)].wr.next = NULL; - DEBUG("%p: allocated and configured %u WRs (%zu segments)", - (void *)rxq, elts_n, (elts_n * elemof((*elts)[0].sges))); + DEBUG("%p: allocated and configured %u single-segment WRs", + (void *)rxq, elts_n); rxq->elts_n = elts_n; rxq->elts_head = 0; - rxq->elts.sp = elts; + rxq->elts = elts; assert(ret == 0); return 0; error: if (elts != NULL) { - assert(pool == NULL); for (i = 0; (i != elemof(*elts)); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct rte_mbuf *buf = elt->bufs[j]; + struct rxq_elt *elt = &(*elts)[i]; + struct rte_mbuf *buf; - if (buf != NULL) - rte_pktmbuf_free_seg(buf); - } + if (elt->sge.addr == 0) + continue; + assert(WR_ID(elt->wr.wr_id).id == i); + buf = (void *)((uintptr_t)elt->sge.addr - + WR_ID(elt->wr.wr_id).offset); + rte_pktmbuf_free_seg(buf); } rte_free(elts); } @@ -2232,282 +1532,86 @@ error: } /** - * Free RX queue elements with scattered packets support. + * Free RX queue elements. * * @param rxq * Pointer to RX queue structure. */ static void -rxq_free_elts_sp(struct rxq *rxq) +rxq_free_elts(struct rxq *rxq) { unsigned int i; unsigned int elts_n = rxq->elts_n; - struct rxq_elt_sp (*elts)[elts_n] = rxq->elts.sp; + struct rxq_elt (*elts)[elts_n] = rxq->elts; DEBUG("%p: freeing WRs", (void *)rxq); rxq->elts_n = 0; - rxq->elts.sp = NULL; + rxq->elts = NULL; if (elts == NULL) return; for (i = 0; (i != elemof(*elts)); ++i) { - unsigned int j; - struct rxq_elt_sp *elt = &(*elts)[i]; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - struct rte_mbuf *buf = elt->bufs[j]; + struct rxq_elt *elt = &(*elts)[i]; + struct rte_mbuf *buf; - if (buf != NULL) - rte_pktmbuf_free_seg(buf); - } + if (elt->sge.addr == 0) + continue; + assert(WR_ID(elt->wr.wr_id).id == i); + buf = (void *)((uintptr_t)elt->sge.addr - + WR_ID(elt->wr.wr_id).offset); + rte_pktmbuf_free_seg(buf); } rte_free(elts); } /** - * Allocate RX queue elements. - * - * @param rxq - * Pointer to RX queue structure. - * @param elts_n - * Number of elements to allocate. - * @param[in] pool - * If not NULL, fetch buffers from this array instead of allocating them - * with rte_pktmbuf_alloc(). - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_alloc_elts(struct rxq *rxq, unsigned int elts_n, struct rte_mbuf **pool) -{ - unsigned int i; - struct rxq_elt (*elts)[elts_n] = - rte_calloc_socket("RXQ elements", 1, sizeof(*elts), 0, - rxq->socket); - int ret = 0; - - if (elts == NULL) { - ERROR("%p: can't allocate packets array", (void *)rxq); - ret = ENOMEM; - goto error; - } - /* For each WR (packet). */ - for (i = 0; (i != elts_n); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct ibv_recv_wr *wr = &elt->wr; - struct ibv_sge *sge = &(*elts)[i].sge; - struct rte_mbuf *buf; - - if (pool != NULL) { - buf = *(pool++); - assert(buf != NULL); - rte_pktmbuf_reset(buf); - } else - buf = rte_pktmbuf_alloc(rxq->mp); - if (buf == NULL) { - assert(pool == NULL); - ERROR("%p: empty mbuf pool", (void *)rxq); - ret = ENOMEM; - goto error; - } - /* Configure WR. Work request ID contains its own index in - * the elts array and the offset between SGE buffer header and - * its data. */ - WR_ID(wr->wr_id).id = i; - WR_ID(wr->wr_id).offset = - (((uintptr_t)buf->buf_addr + RTE_PKTMBUF_HEADROOM) - - (uintptr_t)buf); - wr->next = &(*elts)[(i + 1)].wr; - wr->sg_list = sge; - wr->num_sge = 1; - /* Headroom is reserved by rte_pktmbuf_alloc(). */ - assert(DATA_OFF(buf) == RTE_PKTMBUF_HEADROOM); - /* Buffer is supposed to be empty. */ - assert(rte_pktmbuf_data_len(buf) == 0); - assert(rte_pktmbuf_pkt_len(buf) == 0); - /* sge->addr must be able to store a pointer. */ - assert(sizeof(sge->addr) >= sizeof(uintptr_t)); - /* SGE keeps its headroom. */ - sge->addr = (uintptr_t) - ((uint8_t *)buf->buf_addr + RTE_PKTMBUF_HEADROOM); - sge->length = (buf->buf_len - RTE_PKTMBUF_HEADROOM); - sge->lkey = rxq->mr->lkey; - /* Redundant check for tailroom. */ - assert(sge->length == rte_pktmbuf_tailroom(buf)); - /* Make sure elts index and SGE mbuf pointer can be deduced - * from WR ID. */ - if ((WR_ID(wr->wr_id).id != i) || - ((void *)((uintptr_t)sge->addr - - WR_ID(wr->wr_id).offset) != buf)) { - ERROR("%p: cannot store index and offset in WR ID", - (void *)rxq); - sge->addr = 0; - rte_pktmbuf_free(buf); - ret = EOVERFLOW; - goto error; - } - } - /* The last WR pointer must be NULL. */ - (*elts)[(i - 1)].wr.next = NULL; - DEBUG("%p: allocated and configured %u single-segment WRs", - (void *)rxq, elts_n); - rxq->elts_n = elts_n; - rxq->elts_head = 0; - rxq->elts.no_sp = elts; - assert(ret == 0); - return 0; -error: - if (elts != NULL) { - assert(pool == NULL); - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf; - - if (elt->sge.addr == 0) - continue; - assert(WR_ID(elt->wr.wr_id).id == i); - buf = (void *)((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - rte_pktmbuf_free_seg(buf); - } - rte_free(elts); - } - DEBUG("%p: failed, freed everything", (void *)rxq); - assert(ret > 0); - return ret; -} - -/** - * Free RX queue elements. - * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_free_elts(struct rxq *rxq) -{ - unsigned int i; - unsigned int elts_n = rxq->elts_n; - struct rxq_elt (*elts)[elts_n] = rxq->elts.no_sp; - - DEBUG("%p: freeing WRs", (void *)rxq); - rxq->elts_n = 0; - rxq->elts.no_sp = NULL; - if (elts == NULL) - return; - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf; - - if (elt->sge.addr == 0) - continue; - assert(WR_ID(elt->wr.wr_id).id == i); - buf = (void *)((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - rte_pktmbuf_free_seg(buf); - } - rte_free(elts); -} - -/** - * Delete flow steering rule. + * Unregister a MAC address. * - * @param rxq - * Pointer to RX queue structure. - * @param mac_index - * MAC address index. - * @param vlan_index - * VLAN index. + * @param priv + * Pointer to private structure. */ static void -rxq_del_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index) +priv_mac_addr_del(struct priv *priv) { #ifndef NDEBUG - struct priv *priv = rxq->priv; - const uint8_t (*mac)[ETHER_ADDR_LEN] = - (const uint8_t (*)[ETHER_ADDR_LEN]) - priv->mac[mac_index].addr_bytes; + uint8_t (*mac)[ETHER_ADDR_LEN] = &priv->mac.addr_bytes; #endif - assert(rxq->mac_flow[mac_index][vlan_index] != NULL); - DEBUG("%p: removing MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u" - " (VLAN ID %" PRIu16 ")", - (void *)rxq, - (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5], - mac_index, priv->vlan_filter[vlan_index].id); - claim_zero(ibv_destroy_flow(rxq->mac_flow[mac_index][vlan_index])); - rxq->mac_flow[mac_index][vlan_index] = NULL; -} - -/** - * Unregister a MAC address from a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - * @param mac_index - * MAC address index. - */ -static void -rxq_mac_addr_del(struct rxq *rxq, unsigned int mac_index) -{ - struct priv *priv = rxq->priv; - unsigned int i; - unsigned int vlans = 0; - assert(mac_index < elemof(priv->mac)); - if (!BITFIELD_ISSET(rxq->mac_configured, mac_index)) + if (!priv->mac_flow) return; - for (i = 0; (i != elemof(priv->vlan_filter)); ++i) { - if (!priv->vlan_filter[i].enabled) - continue; - rxq_del_flow(rxq, mac_index, i); - vlans++; - } - if (!vlans) { - rxq_del_flow(rxq, mac_index, 0); - } - BITFIELD_RESET(rxq->mac_configured, mac_index); + DEBUG("%p: removing MAC address %02x:%02x:%02x:%02x:%02x:%02x", + (void *)priv, + (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5]); + claim_zero(ibv_destroy_flow(priv->mac_flow)); + priv->mac_flow = NULL; } /** - * Unregister all MAC addresses from a RX queue. + * Register a MAC address. * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_mac_addrs_del(struct rxq *rxq) -{ - struct priv *priv = rxq->priv; - unsigned int i; - - for (i = 0; (i != elemof(priv->mac)); ++i) - rxq_mac_addr_del(rxq, i); -} - -static int rxq_promiscuous_enable(struct rxq *); -static void rxq_promiscuous_disable(struct rxq *); - -/** - * Add single flow steering rule. + * The MAC address is registered in queue 0. * - * @param rxq - * Pointer to RX queue structure. - * @param mac_index - * MAC address index to register. - * @param vlan_index - * VLAN index. Use -1 for a flow without VLAN. + * @param priv + * Pointer to private structure. * * @return * 0 on success, errno value on failure. */ static int -rxq_add_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index) +priv_mac_addr_add(struct priv *priv) { + uint8_t (*mac)[ETHER_ADDR_LEN] = &priv->mac.addr_bytes; + struct rxq *rxq; struct ibv_flow *flow; - struct priv *priv = rxq->priv; - const uint8_t (*mac)[ETHER_ADDR_LEN] = - (const uint8_t (*)[ETHER_ADDR_LEN]) - priv->mac[mac_index].addr_bytes; + + /* If device isn't started, this is all we need to do. */ + if (!priv->started) + return 0; + if (priv->isolated) + return 0; + if (*priv->rxqs && (*priv->rxqs)[0]) + rxq = (*priv->rxqs)[0]; + else + return 0; /* Allocate flow specification on the stack. */ struct __attribute__((packed)) { @@ -2517,8 +1621,8 @@ rxq_add_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index) struct ibv_flow_attr *attr = &data.attr; struct ibv_flow_spec_eth *spec = &data.spec; - assert(mac_index < elemof(priv->mac)); - assert((vlan_index < elemof(priv->vlan_filter)) || (vlan_index == -1u)); + if (priv->mac_flow) + priv_mac_addr_del(priv); /* * No padding must be inserted by the compiler between attr and spec. * This layout is expected by libibverbs. @@ -2526,6 +1630,7 @@ rxq_add_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index) assert(((uint8_t *)attr + sizeof(*attr)) == (uint8_t *)spec); *attr = (struct ibv_flow_attr){ .type = IBV_FLOW_ATTR_NORMAL, + .priority = 3, .num_of_specs = 1, .port = priv->port, .flags = 0 @@ -2534,710 +1639,98 @@ rxq_add_flow(struct rxq *rxq, unsigned int mac_index, unsigned int vlan_index) .type = IBV_FLOW_SPEC_ETH, .size = sizeof(*spec), .val = { - .dst_mac = { - (*mac)[0], (*mac)[1], (*mac)[2], - (*mac)[3], (*mac)[4], (*mac)[5] - }, - .vlan_tag = ((vlan_index != -1u) ? - htons(priv->vlan_filter[vlan_index].id) : - 0), - }, - .mask = { - .dst_mac = "\xff\xff\xff\xff\xff\xff", - .vlan_tag = ((vlan_index != -1u) ? htons(0xfff) : 0), - } - }; - DEBUG("%p: adding MAC address %02x:%02x:%02x:%02x:%02x:%02x index %u" - " (VLAN %s %" PRIu16 ")", - (void *)rxq, - (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5], - mac_index, - ((vlan_index != -1u) ? "ID" : "index"), - ((vlan_index != -1u) ? priv->vlan_filter[vlan_index].id : -1u)); - /* Create related flow. */ - errno = 0; - flow = ibv_create_flow(rxq->qp, attr); - if (flow == NULL) { - /* It's not clear whether errno is always set in this case. */ - ERROR("%p: flow configuration failed, errno=%d: %s", - (void *)rxq, errno, - (errno ? strerror(errno) : "Unknown error")); - if (errno) - return errno; - return EINVAL; - } - if (vlan_index == -1u) - vlan_index = 0; - assert(rxq->mac_flow[mac_index][vlan_index] == NULL); - rxq->mac_flow[mac_index][vlan_index] = flow; - return 0; -} - -/** - * Register a MAC address in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - * @param mac_index - * MAC address index to register. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_mac_addr_add(struct rxq *rxq, unsigned int mac_index) -{ - struct priv *priv = rxq->priv; - unsigned int i; - unsigned int vlans = 0; - int ret; - - assert(mac_index < elemof(priv->mac)); - if (BITFIELD_ISSET(rxq->mac_configured, mac_index)) - rxq_mac_addr_del(rxq, mac_index); - /* Fill VLAN specifications. */ - for (i = 0; (i != elemof(priv->vlan_filter)); ++i) { - if (!priv->vlan_filter[i].enabled) - continue; - /* Create related flow. */ - ret = rxq_add_flow(rxq, mac_index, i); - if (!ret) { - vlans++; - continue; - } - /* Failure, rollback. */ - while (i != 0) - if (priv->vlan_filter[--i].enabled) - rxq_del_flow(rxq, mac_index, i); - assert(ret > 0); - return ret; - } - /* In case there is no VLAN filter. */ - if (!vlans) { - ret = rxq_add_flow(rxq, mac_index, -1); - if (ret) - return ret; - } - BITFIELD_SET(rxq->mac_configured, mac_index); - return 0; -} - -/** - * Register all MAC addresses in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_mac_addrs_add(struct rxq *rxq) -{ - struct priv *priv = rxq->priv; - unsigned int i; - int ret; - - for (i = 0; (i != elemof(priv->mac)); ++i) { - if (!BITFIELD_ISSET(priv->mac_configured, i)) - continue; - ret = rxq_mac_addr_add(rxq, i); - if (!ret) - continue; - /* Failure, rollback. */ - while (i != 0) - rxq_mac_addr_del(rxq, --i); - assert(ret > 0); - return ret; - } - return 0; -} - -/** - * Unregister a MAC address. - * - * In RSS mode, the MAC address is unregistered from the parent queue, - * otherwise it is unregistered from each queue directly. - * - * @param priv - * Pointer to private structure. - * @param mac_index - * MAC address index. - */ -static void -priv_mac_addr_del(struct priv *priv, unsigned int mac_index) -{ - unsigned int i; - - assert(mac_index < elemof(priv->mac)); - if (!BITFIELD_ISSET(priv->mac_configured, mac_index)) - return; - if (priv->rss) { - rxq_mac_addr_del(&priv->rxq_parent, mac_index); - goto end; - } - for (i = 0; (i != priv->dev->data->nb_rx_queues); ++i) - rxq_mac_addr_del((*priv->rxqs)[i], mac_index); -end: - BITFIELD_RESET(priv->mac_configured, mac_index); -} - -/** - * Register a MAC address. - * - * In RSS mode, the MAC address is registered in the parent queue, - * otherwise it is registered in each queue directly. - * - * @param priv - * Pointer to private structure. - * @param mac_index - * MAC address index to use. - * @param mac - * MAC address to register. - * - * @return - * 0 on success, errno value on failure. - */ -static int -priv_mac_addr_add(struct priv *priv, unsigned int mac_index, - const uint8_t (*mac)[ETHER_ADDR_LEN]) -{ - unsigned int i; - int ret; - - assert(mac_index < elemof(priv->mac)); - /* First, make sure this address isn't already configured. */ - for (i = 0; (i != elemof(priv->mac)); ++i) { - /* Skip this index, it's going to be reconfigured. */ - if (i == mac_index) - continue; - if (!BITFIELD_ISSET(priv->mac_configured, i)) - continue; - if (memcmp(priv->mac[i].addr_bytes, *mac, sizeof(*mac))) - continue; - /* Address already configured elsewhere, return with error. */ - return EADDRINUSE; - } - if (BITFIELD_ISSET(priv->mac_configured, mac_index)) - priv_mac_addr_del(priv, mac_index); - priv->mac[mac_index] = (struct ether_addr){ - { - (*mac)[0], (*mac)[1], (*mac)[2], - (*mac)[3], (*mac)[4], (*mac)[5] - } - }; - /* If device isn't started, this is all we need to do. */ - if (!priv->started) { -#ifndef NDEBUG - /* Verify that all queues have this index disabled. */ - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - assert(!BITFIELD_ISSET - ((*priv->rxqs)[i]->mac_configured, mac_index)); - } -#endif - goto end; - } - if (priv->rss) { - ret = rxq_mac_addr_add(&priv->rxq_parent, mac_index); - if (ret) - return ret; - goto end; - } - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - ret = rxq_mac_addr_add((*priv->rxqs)[i], mac_index); - if (!ret) - continue; - /* Failure, rollback. */ - while (i != 0) - if ((*priv->rxqs)[(--i)] != NULL) - rxq_mac_addr_del((*priv->rxqs)[i], mac_index); - return ret; - } -end: - BITFIELD_SET(priv->mac_configured, mac_index); - return 0; -} - -/** - * Enable allmulti mode in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_allmulticast_enable(struct rxq *rxq) -{ - struct ibv_flow *flow; - struct ibv_flow_attr attr = { - .type = IBV_FLOW_ATTR_MC_DEFAULT, - .num_of_specs = 0, - .port = rxq->priv->port, - .flags = 0 - }; - - DEBUG("%p: enabling allmulticast mode", (void *)rxq); - if (rxq->allmulti_flow != NULL) - return EBUSY; - errno = 0; - flow = ibv_create_flow(rxq->qp, &attr); - if (flow == NULL) { - /* It's not clear whether errno is always set in this case. */ - ERROR("%p: flow configuration failed, errno=%d: %s", - (void *)rxq, errno, - (errno ? strerror(errno) : "Unknown error")); - if (errno) - return errno; - return EINVAL; - } - rxq->allmulti_flow = flow; - DEBUG("%p: allmulticast mode enabled", (void *)rxq); - return 0; -} - -/** - * Disable allmulti mode in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_allmulticast_disable(struct rxq *rxq) -{ - DEBUG("%p: disabling allmulticast mode", (void *)rxq); - if (rxq->allmulti_flow == NULL) - return; - claim_zero(ibv_destroy_flow(rxq->allmulti_flow)); - rxq->allmulti_flow = NULL; - DEBUG("%p: allmulticast mode disabled", (void *)rxq); -} - -/** - * Enable promiscuous mode in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_promiscuous_enable(struct rxq *rxq) -{ - struct ibv_flow *flow; - struct ibv_flow_attr attr = { - .type = IBV_FLOW_ATTR_ALL_DEFAULT, - .num_of_specs = 0, - .port = rxq->priv->port, - .flags = 0 - }; - - if (rxq->priv->vf) - return 0; - DEBUG("%p: enabling promiscuous mode", (void *)rxq); - if (rxq->promisc_flow != NULL) - return EBUSY; - errno = 0; - flow = ibv_create_flow(rxq->qp, &attr); - if (flow == NULL) { - /* It's not clear whether errno is always set in this case. */ - ERROR("%p: flow configuration failed, errno=%d: %s", - (void *)rxq, errno, - (errno ? strerror(errno) : "Unknown error")); - if (errno) - return errno; - return EINVAL; - } - rxq->promisc_flow = flow; - DEBUG("%p: promiscuous mode enabled", (void *)rxq); - return 0; -} - -/** - * Disable promiscuous mode in a RX queue. - * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_promiscuous_disable(struct rxq *rxq) -{ - if (rxq->priv->vf) - return; - DEBUG("%p: disabling promiscuous mode", (void *)rxq); - if (rxq->promisc_flow == NULL) - return; - claim_zero(ibv_destroy_flow(rxq->promisc_flow)); - rxq->promisc_flow = NULL; - DEBUG("%p: promiscuous mode disabled", (void *)rxq); -} - -/** - * Clean up a RX queue. - * - * Destroy objects, free allocated memory and reset the structure for reuse. - * - * @param rxq - * Pointer to RX queue structure. - */ -static void -rxq_cleanup(struct rxq *rxq) -{ - struct ibv_exp_release_intf_params params; - - DEBUG("cleaning up %p", (void *)rxq); - if (rxq->sp) - rxq_free_elts_sp(rxq); - else - rxq_free_elts(rxq); - if (rxq->if_qp != NULL) { - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - assert(rxq->qp != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(rxq->priv->ctx, - rxq->if_qp, - ¶ms)); - } - if (rxq->if_cq != NULL) { - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - assert(rxq->cq != NULL); - params = (struct ibv_exp_release_intf_params){ - .comp_mask = 0, - }; - claim_zero(ibv_exp_release_intf(rxq->priv->ctx, - rxq->if_cq, - ¶ms)); - } - if (rxq->qp != NULL) { - rxq_promiscuous_disable(rxq); - rxq_allmulticast_disable(rxq); - rxq_mac_addrs_del(rxq); - claim_zero(ibv_destroy_qp(rxq->qp)); - } - if (rxq->cq != NULL) - claim_zero(ibv_destroy_cq(rxq->cq)); - if (rxq->rd != NULL) { - struct ibv_exp_destroy_res_domain_attr attr = { - .comp_mask = 0, - }; - - assert(rxq->priv != NULL); - assert(rxq->priv->ctx != NULL); - claim_zero(ibv_exp_destroy_res_domain(rxq->priv->ctx, - rxq->rd, - &attr)); - } - if (rxq->mr != NULL) - claim_zero(ibv_dereg_mr(rxq->mr)); - memset(rxq, 0, sizeof(*rxq)); -} - -/** - * Translate RX completion flags to packet type. - * - * @param flags - * RX completion flags returned by poll_length_flags(). - * - * @note: fix mlx4_dev_supported_ptypes_get() if any change here. - * - * @return - * Packet type for struct rte_mbuf. - */ -static inline uint32_t -rxq_cq_to_pkt_type(uint32_t flags) -{ - uint32_t pkt_type; - - if (flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) - pkt_type = - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV4_PACKET, RTE_PTYPE_L3_IPV4) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV6_PACKET, RTE_PTYPE_L3_IPV6) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, RTE_PTYPE_INNER_L3_IPV4) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, RTE_PTYPE_INNER_L3_IPV6); - else - pkt_type = - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, RTE_PTYPE_L3_IPV4) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, RTE_PTYPE_L3_IPV6); - return pkt_type; -} - -/** - * Translate RX completion flags to offload flags. - * - * @param[in] rxq - * Pointer to RX queue structure. - * @param flags - * RX completion flags returned by poll_length_flags(). - * - * @return - * Offload flags (ol_flags) for struct rte_mbuf. - */ -static inline uint32_t -rxq_cq_to_ol_flags(const struct rxq *rxq, uint32_t flags) -{ - uint32_t ol_flags = 0; - - if (rxq->csum) - ol_flags |= - TRANSPOSE(flags, - IBV_EXP_CQ_RX_IP_CSUM_OK, - PKT_RX_IP_CKSUM_GOOD) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_GOOD); - if ((flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) && (rxq->csum_l2tun)) - ol_flags |= - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IP_CSUM_OK, - PKT_RX_IP_CKSUM_GOOD) | - TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_GOOD); - return ol_flags; -} - -static uint16_t -mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n); - -/** - * DPDK callback for RX with scattered packets support. - * - * @param dpdk_rxq - * Generic pointer to RX queue structure. - * @param[out] pkts - * Array to store received packets. - * @param pkts_n - * Maximum number of packets in array. - * - * @return - * Number of packets successfully received (<= pkts_n). - */ -static uint16_t -mlx4_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) -{ - struct rxq *rxq = (struct rxq *)dpdk_rxq; - struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp; - const unsigned int elts_n = rxq->elts_n; - unsigned int elts_head = rxq->elts_head; - struct ibv_recv_wr head; - struct ibv_recv_wr **next = &head.next; - struct ibv_recv_wr *bad_wr; - unsigned int i; - unsigned int pkts_ret = 0; - int ret; - - if (unlikely(!rxq->sp)) - return mlx4_rx_burst(dpdk_rxq, pkts, pkts_n); - if (unlikely(elts == NULL)) /* See RTE_DEV_CMD_SET_MTU. */ - return 0; - for (i = 0; (i != pkts_n); ++i) { - struct rxq_elt_sp *elt = &(*elts)[elts_head]; - struct ibv_recv_wr *wr = &elt->wr; - uint64_t wr_id = wr->wr_id; - unsigned int len; - unsigned int pkt_buf_len; - struct rte_mbuf *pkt_buf = NULL; /* Buffer returned in pkts. */ - struct rte_mbuf **pkt_buf_next = &pkt_buf; - unsigned int seg_headroom = RTE_PKTMBUF_HEADROOM; - unsigned int j = 0; - uint32_t flags; - - /* Sanity checks. */ -#ifdef NDEBUG - (void)wr_id; -#endif - assert(wr_id < rxq->elts_n); - assert(wr->sg_list == elt->sges); - assert(wr->num_sge == elemof(elt->sges)); - assert(elts_head < rxq->elts_n); - assert(rxq->elts_head < rxq->elts_n); - ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL, - &flags); - if (unlikely(ret < 0)) { - struct ibv_wc wc; - int wcs_n; - - DEBUG("rxq=%p, poll_length() failed (ret=%d)", - (void *)rxq, ret); - /* ibv_poll_cq() must be used in case of failure. */ - wcs_n = ibv_poll_cq(rxq->cq, 1, &wc); - if (unlikely(wcs_n == 0)) - break; - if (unlikely(wcs_n < 0)) { - DEBUG("rxq=%p, ibv_poll_cq() failed (wcs_n=%d)", - (void *)rxq, wcs_n); - break; - } - assert(wcs_n == 1); - if (unlikely(wc.status != IBV_WC_SUCCESS)) { - /* Whatever, just repost the offending WR. */ - DEBUG("rxq=%p, wr_id=%" PRIu64 ": bad work" - " completion status (%d): %s", - (void *)rxq, wc.wr_id, wc.status, - ibv_wc_status_str(wc.status)); -#ifdef MLX4_PMD_SOFT_COUNTERS - /* Increment dropped packets counter. */ - ++rxq->stats.idropped; -#endif - /* Link completed WRs together for repost. */ - *next = wr; - next = &wr->next; - goto repost; - } - ret = wc.byte_len; + .dst_mac = { + (*mac)[0], (*mac)[1], (*mac)[2], + (*mac)[3], (*mac)[4], (*mac)[5] + }, + }, + .mask = { + .dst_mac = "\xff\xff\xff\xff\xff\xff", } - if (ret == 0) - break; - len = ret; - pkt_buf_len = len; - /* Link completed WRs together for repost. */ - *next = wr; - next = &wr->next; - /* - * Replace spent segments with new ones, concatenate and - * return them as pkt_buf. - */ - while (1) { - struct ibv_sge *sge = &elt->sges[j]; - struct rte_mbuf *seg = elt->bufs[j]; - struct rte_mbuf *rep; - unsigned int seg_tailroom; + }; + DEBUG("%p: adding MAC address %02x:%02x:%02x:%02x:%02x:%02x", + (void *)priv, + (*mac)[0], (*mac)[1], (*mac)[2], (*mac)[3], (*mac)[4], (*mac)[5]); + /* Create related flow. */ + errno = 0; + flow = ibv_create_flow(rxq->qp, attr); + if (flow == NULL) { + /* It's not clear whether errno is always set in this case. */ + ERROR("%p: flow configuration failed, errno=%d: %s", + (void *)rxq, errno, + (errno ? strerror(errno) : "Unknown error")); + if (errno) + return errno; + return EINVAL; + } + assert(priv->mac_flow == NULL); + priv->mac_flow = flow; + return 0; +} - /* - * Fetch initial bytes of packet descriptor into a - * cacheline while allocating rep. - */ - rte_prefetch0(seg); - rep = rte_mbuf_raw_alloc(rxq->mp); - if (unlikely(rep == NULL)) { - /* - * Unable to allocate a replacement mbuf, - * repost WR. - */ - DEBUG("rxq=%p, wr_id=%" PRIu64 ":" - " can't allocate a new mbuf", - (void *)rxq, wr_id); - if (pkt_buf != NULL) { - *pkt_buf_next = NULL; - rte_pktmbuf_free(pkt_buf); - } - /* Increase out of memory counters. */ - ++rxq->stats.rx_nombuf; - ++rxq->priv->dev->data->rx_mbuf_alloc_failed; - goto repost; - } -#ifndef NDEBUG - /* Poison user-modifiable fields in rep. */ - NEXT(rep) = (void *)((uintptr_t)-1); - SET_DATA_OFF(rep, 0xdead); - DATA_LEN(rep) = 0xd00d; - PKT_LEN(rep) = 0xdeadd00d; - NB_SEGS(rep) = 0x2a; - PORT(rep) = 0x2a; - rep->ol_flags = -1; -#endif - assert(rep->buf_len == seg->buf_len); - /* Reconfigure sge to use rep instead of seg. */ - assert(sge->lkey == rxq->mr->lkey); - sge->addr = ((uintptr_t)rep->buf_addr + seg_headroom); - elt->bufs[j] = rep; - ++j; - /* Update pkt_buf if it's the first segment, or link - * seg to the previous one and update pkt_buf_next. */ - *pkt_buf_next = seg; - pkt_buf_next = &NEXT(seg); - /* Update seg information. */ - seg_tailroom = (seg->buf_len - seg_headroom); - assert(sge->length == seg_tailroom); - SET_DATA_OFF(seg, seg_headroom); - if (likely(len <= seg_tailroom)) { - /* Last segment. */ - DATA_LEN(seg) = len; - PKT_LEN(seg) = len; - /* Sanity check. */ - assert(rte_pktmbuf_headroom(seg) == - seg_headroom); - assert(rte_pktmbuf_tailroom(seg) == - (seg_tailroom - len)); - break; - } - DATA_LEN(seg) = seg_tailroom; - PKT_LEN(seg) = seg_tailroom; - /* Sanity check. */ - assert(rte_pktmbuf_headroom(seg) == seg_headroom); - assert(rte_pktmbuf_tailroom(seg) == 0); - /* Fix len and clear headroom for next segments. */ - len -= seg_tailroom; - seg_headroom = 0; - } - /* Update head and tail segments. */ - *pkt_buf_next = NULL; - assert(pkt_buf != NULL); - assert(j != 0); - NB_SEGS(pkt_buf) = j; - PORT(pkt_buf) = rxq->port_id; - PKT_LEN(pkt_buf) = pkt_buf_len; - pkt_buf->packet_type = rxq_cq_to_pkt_type(flags); - pkt_buf->ol_flags = rxq_cq_to_ol_flags(rxq, flags); +/** + * Clean up a RX queue. + * + * Destroy objects, free allocated memory and reset the structure for reuse. + * + * @param rxq + * Pointer to RX queue structure. + */ +static void +rxq_cleanup(struct rxq *rxq) +{ + struct ibv_exp_release_intf_params params; - /* Return packet. */ - *(pkts++) = pkt_buf; - ++pkts_ret; -#ifdef MLX4_PMD_SOFT_COUNTERS - /* Increase bytes counter. */ - rxq->stats.ibytes += pkt_buf_len; -#endif -repost: - if (++elts_head >= elts_n) - elts_head = 0; - continue; + DEBUG("cleaning up %p", (void *)rxq); + rxq_free_elts(rxq); + if (rxq->if_qp != NULL) { + assert(rxq->priv != NULL); + assert(rxq->priv->ctx != NULL); + assert(rxq->qp != NULL); + params = (struct ibv_exp_release_intf_params){ + .comp_mask = 0, + }; + claim_zero(ibv_exp_release_intf(rxq->priv->ctx, + rxq->if_qp, + ¶ms)); } - if (unlikely(i == 0)) - return 0; - *next = NULL; - /* Repost WRs. */ -#ifdef DEBUG_RECV - DEBUG("%p: reposting %d WRs", (void *)rxq, i); -#endif - ret = ibv_post_recv(rxq->qp, head.next, &bad_wr); - if (unlikely(ret)) { - /* Inability to repost WRs is fatal. */ - DEBUG("%p: ibv_post_recv(): failed for WR %p: %s", - (void *)rxq->priv, - (void *)bad_wr, - strerror(ret)); - abort(); + if (rxq->if_cq != NULL) { + assert(rxq->priv != NULL); + assert(rxq->priv->ctx != NULL); + assert(rxq->cq != NULL); + params = (struct ibv_exp_release_intf_params){ + .comp_mask = 0, + }; + claim_zero(ibv_exp_release_intf(rxq->priv->ctx, + rxq->if_cq, + ¶ms)); } - rxq->elts_head = elts_head; -#ifdef MLX4_PMD_SOFT_COUNTERS - /* Increase packets counter. */ - rxq->stats.ipackets += pkts_ret; -#endif - return pkts_ret; + if (rxq->qp != NULL) + claim_zero(ibv_destroy_qp(rxq->qp)); + if (rxq->cq != NULL) + claim_zero(ibv_destroy_cq(rxq->cq)); + if (rxq->channel != NULL) + claim_zero(ibv_destroy_comp_channel(rxq->channel)); + if (rxq->rd != NULL) { + struct ibv_exp_destroy_res_domain_attr attr = { + .comp_mask = 0, + }; + + assert(rxq->priv != NULL); + assert(rxq->priv->ctx != NULL); + claim_zero(ibv_exp_destroy_res_domain(rxq->priv->ctx, + rxq->rd, + &attr)); + } + if (rxq->mr != NULL) + claim_zero(ibv_dereg_mr(rxq->mr)); + memset(rxq, 0, sizeof(*rxq)); } /** * DPDK callback for RX. * - * The following function is the same as mlx4_rx_burst_sp(), except it doesn't - * manage scattered packets. Improves performance when MRU is lower than the - * size of the first segment. + * The following function doesn't manage scattered packets. * * @param dpdk_rxq * Generic pointer to RX queue structure. @@ -3253,7 +1746,7 @@ static uint16_t mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) { struct rxq *rxq = (struct rxq *)dpdk_rxq; - struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp; + struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts; const unsigned int elts_n = rxq->elts_n; unsigned int elts_head = rxq->elts_head; struct ibv_sge sges[pkts_n]; @@ -3261,8 +1754,6 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) unsigned int pkts_ret = 0; int ret; - if (unlikely(rxq->sp)) - return mlx4_rx_burst_sp(dpdk_rxq, pkts, pkts_n); for (i = 0; (i != pkts_n); ++i) { struct rxq_elt *elt = &(*elts)[elts_head]; struct ibv_recv_wr *wr = &elt->wr; @@ -3309,10 +1800,8 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) " completion status (%d): %s", (void *)rxq, wc.wr_id, wc.status, ibv_wc_status_str(wc.status)); -#ifdef MLX4_PMD_SOFT_COUNTERS /* Increment dropped packets counter. */ ++rxq->stats.idropped; -#endif /* Add SGE to array for repost. */ sges[i] = elt->sge; goto repost; @@ -3334,6 +1823,8 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) /* Increase out of memory counters. */ ++rxq->stats.rx_nombuf; ++rxq->priv->dev->data->rx_mbuf_alloc_failed; + /* Add SGE to array for repost. */ + sges[i] = elt->sge; goto repost; } @@ -3355,16 +1846,14 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) NEXT(seg) = NULL; PKT_LEN(seg) = len; DATA_LEN(seg) = len; - seg->packet_type = rxq_cq_to_pkt_type(flags); - seg->ol_flags = rxq_cq_to_ol_flags(rxq, flags); + seg->packet_type = 0; + seg->ol_flags = 0; /* Return packet. */ *(pkts++) = seg; ++pkts_ret; -#ifdef MLX4_PMD_SOFT_COUNTERS /* Increase bytes counter. */ rxq->stats.ibytes += len; -#endif repost: if (++elts_head >= elts_n) elts_head = 0; @@ -3373,9 +1862,6 @@ repost: if (unlikely(i == 0)) return 0; /* Repost WRs. */ -#ifdef DEBUG_RECV - DEBUG("%p: reposting %u WRs", (void *)rxq, i); -#endif ret = rxq->if_qp->recv_burst(rxq->qp, sges, i); if (unlikely(ret)) { /* Inability to repost WRs is fatal. */ @@ -3385,53 +1871,11 @@ repost: abort(); } rxq->elts_head = elts_head; -#ifdef MLX4_PMD_SOFT_COUNTERS /* Increase packets counter. */ rxq->stats.ipackets += pkts_ret; -#endif return pkts_ret; } -/** - * DPDK callback for RX in secondary processes. - * - * This function configures all queues from primary process information - * if necessary before reverting to the normal RX burst callback. - * - * @param dpdk_rxq - * Generic pointer to RX queue structure. - * @param[out] pkts - * Array to store received packets. - * @param pkts_n - * Maximum number of packets in array. - * - * @return - * Number of packets successfully received (<= pkts_n). - */ -static uint16_t -mlx4_rx_burst_secondary_setup(void *dpdk_rxq, struct rte_mbuf **pkts, - uint16_t pkts_n) -{ - struct rxq *rxq = dpdk_rxq; - struct priv *priv = mlx4_secondary_data_setup(rxq->priv); - struct priv *primary_priv; - unsigned int index; - - if (priv == NULL) - return 0; - primary_priv = - mlx4_secondary_data[priv->dev->data->port_id].primary_priv; - /* Look for queue index in both private structures. */ - for (index = 0; index != priv->rxqs_n; ++index) - if (((*primary_priv->rxqs)[index] == rxq) || - ((*priv->rxqs)[index] == rxq)) - break; - if (index == priv->rxqs_n) - return 0; - rxq = (*priv->rxqs)[index]; - return priv->dev->rx_pkt_burst(rxq, pkts, pkts_n); -} - /** * Allocate a Queue Pair. * Optionally setup inline receive if supported. @@ -3461,10 +1905,7 @@ rxq_setup_qp(struct priv *priv, struct ibv_cq *cq, uint16_t desc, priv->device_attr.max_qp_wr : desc), /* Max number of scatter/gather elements in a WR. */ - .max_recv_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), + .max_recv_sge = 1, }, .qp_type = IBV_QPT_RAW_PACKET, .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD | @@ -3473,283 +1914,9 @@ rxq_setup_qp(struct priv *priv, struct ibv_cq *cq, uint16_t desc, .res_domain = rd, }; -#ifdef INLINE_RECV - attr.max_inl_recv = priv->inl_recv_size; - attr.comp_mask |= IBV_EXP_QP_INIT_ATTR_INL_RECV; -#endif - return ibv_exp_create_qp(priv->ctx, &attr); -} - -#ifdef RSS_SUPPORT - -/** - * Allocate a RSS Queue Pair. - * Optionally setup inline receive if supported. - * - * @param priv - * Pointer to private structure. - * @param cq - * Completion queue to associate with QP. - * @param desc - * Number of descriptors in QP (hint only). - * @param parent - * If nonzero, create a parent QP, otherwise a child. - * - * @return - * QP pointer or NULL in case of error. - */ -static struct ibv_qp * -rxq_setup_qp_rss(struct priv *priv, struct ibv_cq *cq, uint16_t desc, - int parent, struct ibv_exp_res_domain *rd) -{ - struct ibv_exp_qp_init_attr attr = { - /* CQ to be associated with the send queue. */ - .send_cq = cq, - /* CQ to be associated with the receive queue. */ - .recv_cq = cq, - .cap = { - /* Max number of outstanding WRs. */ - .max_recv_wr = ((priv->device_attr.max_qp_wr < desc) ? - priv->device_attr.max_qp_wr : - desc), - /* Max number of scatter/gather elements in a WR. */ - .max_recv_sge = ((priv->device_attr.max_sge < - MLX4_PMD_SGE_WR_N) ? - priv->device_attr.max_sge : - MLX4_PMD_SGE_WR_N), - }, - .qp_type = IBV_QPT_RAW_PACKET, - .comp_mask = (IBV_EXP_QP_INIT_ATTR_PD | - IBV_EXP_QP_INIT_ATTR_RES_DOMAIN | - IBV_EXP_QP_INIT_ATTR_QPG), - .pd = priv->pd, - .res_domain = rd, - }; - -#ifdef INLINE_RECV - attr.max_inl_recv = priv->inl_recv_size, - attr.comp_mask |= IBV_EXP_QP_INIT_ATTR_INL_RECV; -#endif - if (parent) { - attr.qpg.qpg_type = IBV_EXP_QPG_PARENT; - /* TSS isn't necessary. */ - attr.qpg.parent_attrib.tss_child_count = 0; - attr.qpg.parent_attrib.rss_child_count = - rte_align32pow2(priv->rxqs_n + 1) >> 1; - DEBUG("initializing parent RSS queue"); - } else { - attr.qpg.qpg_type = IBV_EXP_QPG_CHILD_RX; - attr.qpg.qpg_parent = priv->rxq_parent.qp; - DEBUG("initializing child RSS queue"); - } return ibv_exp_create_qp(priv->ctx, &attr); } -#endif /* RSS_SUPPORT */ - -/** - * Reconfigure a RX queue with new parameters. - * - * rxq_rehash() does not allocate mbufs, which, if not done from the right - * thread (such as a control thread), may corrupt the pool. - * In case of failure, the queue is left untouched. - * - * @param dev - * Pointer to Ethernet device structure. - * @param rxq - * RX queue pointer. - * - * @return - * 0 on success, errno value on failure. - */ -static int -rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq) -{ - struct priv *priv = rxq->priv; - struct rxq tmpl = *rxq; - unsigned int mbuf_n; - unsigned int desc_n; - struct rte_mbuf **pool; - unsigned int i, k; - struct ibv_exp_qp_attr mod; - struct ibv_recv_wr *bad_wr; - unsigned int mb_len; - int err; - int parent = (rxq == &priv->rxq_parent); - - if (parent) { - ERROR("%p: cannot rehash parent queue %p", - (void *)dev, (void *)rxq); - return EINVAL; - } - mb_len = rte_pktmbuf_data_room_size(rxq->mp); - DEBUG("%p: rehashing queue %p", (void *)dev, (void *)rxq); - /* Number of descriptors and mbufs currently allocated. */ - desc_n = (tmpl.elts_n * (tmpl.sp ? MLX4_PMD_SGE_WR_N : 1)); - mbuf_n = desc_n; - /* Toggle RX checksum offload if hardware supports it. */ - if (priv->hw_csum) { - tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - rxq->csum = tmpl.csum; - } - if (priv->hw_csum_l2tun) { - tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - rxq->csum_l2tun = tmpl.csum_l2tun; - } - /* Enable scattered packets support for this queue if necessary. */ - assert(mb_len >= RTE_PKTMBUF_HEADROOM); - if ((dev->data->dev_conf.rxmode.jumbo_frame) && - (dev->data->dev_conf.rxmode.max_rx_pkt_len > - (mb_len - RTE_PKTMBUF_HEADROOM))) { - tmpl.sp = 1; - desc_n /= MLX4_PMD_SGE_WR_N; - } else - tmpl.sp = 0; - DEBUG("%p: %s scattered packets support (%u WRs)", - (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc_n); - /* If scatter mode is the same as before, nothing to do. */ - if (tmpl.sp == rxq->sp) { - DEBUG("%p: nothing to do", (void *)dev); - return 0; - } - /* Remove attached flows if RSS is disabled (no parent queue). */ - if (!priv->rss) { - rxq_allmulticast_disable(&tmpl); - rxq_promiscuous_disable(&tmpl); - rxq_mac_addrs_del(&tmpl); - /* Update original queue in case of failure. */ - rxq->allmulti_flow = tmpl.allmulti_flow; - rxq->promisc_flow = tmpl.promisc_flow; - memcpy(rxq->mac_configured, tmpl.mac_configured, - sizeof(rxq->mac_configured)); - memcpy(rxq->mac_flow, tmpl.mac_flow, sizeof(rxq->mac_flow)); - } - /* From now on, any failure will render the queue unusable. - * Reinitialize QP. */ - mod = (struct ibv_exp_qp_attr){ .qp_state = IBV_QPS_RESET }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, IBV_EXP_QP_STATE); - if (err) { - ERROR("%p: cannot reset QP: %s", (void *)dev, strerror(err)); - assert(err > 0); - return err; - } - err = ibv_resize_cq(tmpl.cq, desc_n); - if (err) { - ERROR("%p: cannot resize CQ: %s", (void *)dev, strerror(err)); - assert(err > 0); - return err; - } - mod = (struct ibv_exp_qp_attr){ - /* Move the QP to this state. */ - .qp_state = IBV_QPS_INIT, - /* Primary port number. */ - .port_num = priv->port - }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, - (IBV_EXP_QP_STATE | -#ifdef RSS_SUPPORT - (parent ? IBV_EXP_QP_GROUP_RSS : 0) | -#endif /* RSS_SUPPORT */ - IBV_EXP_QP_PORT)); - if (err) { - ERROR("%p: QP state to IBV_QPS_INIT failed: %s", - (void *)dev, strerror(err)); - assert(err > 0); - return err; - }; - /* Reconfigure flows. Do not care for errors. */ - if (!priv->rss) { - rxq_mac_addrs_add(&tmpl); - if (priv->promisc) - rxq_promiscuous_enable(&tmpl); - if (priv->allmulti) - rxq_allmulticast_enable(&tmpl); - /* Update original queue in case of failure. */ - rxq->allmulti_flow = tmpl.allmulti_flow; - rxq->promisc_flow = tmpl.promisc_flow; - memcpy(rxq->mac_configured, tmpl.mac_configured, - sizeof(rxq->mac_configured)); - memcpy(rxq->mac_flow, tmpl.mac_flow, sizeof(rxq->mac_flow)); - } - /* Allocate pool. */ - pool = rte_malloc(__func__, (mbuf_n * sizeof(*pool)), 0); - if (pool == NULL) { - ERROR("%p: cannot allocate memory", (void *)dev); - return ENOBUFS; - } - /* Snatch mbufs from original queue. */ - k = 0; - if (rxq->sp) { - struct rxq_elt_sp (*elts)[rxq->elts_n] = rxq->elts.sp; - - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt_sp *elt = &(*elts)[i]; - unsigned int j; - - for (j = 0; (j != elemof(elt->bufs)); ++j) { - assert(elt->bufs[j] != NULL); - pool[k++] = elt->bufs[j]; - } - } - } else { - struct rxq_elt (*elts)[rxq->elts_n] = rxq->elts.no_sp; - - for (i = 0; (i != elemof(*elts)); ++i) { - struct rxq_elt *elt = &(*elts)[i]; - struct rte_mbuf *buf = (void *) - ((uintptr_t)elt->sge.addr - - WR_ID(elt->wr.wr_id).offset); - - assert(WR_ID(elt->wr.wr_id).id == i); - pool[k++] = buf; - } - } - assert(k == mbuf_n); - tmpl.elts_n = 0; - tmpl.elts.sp = NULL; - assert((void *)&tmpl.elts.sp == (void *)&tmpl.elts.no_sp); - err = ((tmpl.sp) ? - rxq_alloc_elts_sp(&tmpl, desc_n, pool) : - rxq_alloc_elts(&tmpl, desc_n, pool)); - if (err) { - ERROR("%p: cannot reallocate WRs, aborting", (void *)dev); - rte_free(pool); - assert(err > 0); - return err; - } - assert(tmpl.elts_n == desc_n); - assert(tmpl.elts.sp != NULL); - rte_free(pool); - /* Clean up original data. */ - rxq->elts_n = 0; - rte_free(rxq->elts.sp); - rxq->elts.sp = NULL; - /* Post WRs. */ - err = ibv_post_recv(tmpl.qp, - (tmpl.sp ? - &(*tmpl.elts.sp)[0].wr : - &(*tmpl.elts.no_sp)[0].wr), - &bad_wr); - if (err) { - ERROR("%p: ibv_post_recv() failed for WR %p: %s", - (void *)dev, - (void *)bad_wr, - strerror(err)); - goto skip_rtr; - } - mod = (struct ibv_exp_qp_attr){ - .qp_state = IBV_QPS_RTR - }; - err = ibv_exp_modify_qp(tmpl.qp, &mod, IBV_EXP_QP_STATE); - if (err) - ERROR("%p: QP state to IBV_QPS_RTR failed: %s", - (void *)dev, strerror(err)); -skip_rtr: - *rxq = tmpl; - assert(err >= 0); - return err; -} - /** * Configure a RX queue. * @@ -3761,9 +1928,6 @@ skip_rtr: * Number of descriptors to configure in queue. * @param socket * NUMA socket on which memory must be allocated. - * @param inactive - * If true, the queue is disabled because its index is higher or - * equal to the real number of queues, which must be a power of 2. * @param[in] conf * Thresholds parameters. * @param mp @@ -3774,7 +1938,7 @@ skip_rtr: */ static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, int inactive, const struct rte_eth_rxconf *conf, + unsigned int socket, const struct rte_eth_rxconf *conf, struct rte_mempool *mp) { struct priv *priv = dev->data->dev_private; @@ -3793,41 +1957,30 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, struct ibv_recv_wr *bad_wr; unsigned int mb_len; int ret = 0; - int parent = (rxq == &priv->rxq_parent); (void)conf; /* Thresholds configuration (ignored). */ - /* - * If this is a parent queue, hardware must support RSS and - * RSS must be enabled. - */ - assert((!parent) || ((priv->hw_rss) && (priv->rss))); - if (parent) { - /* Even if unused, ibv_create_cq() requires at least one - * descriptor. */ - desc = 1; - goto skip_mr; - } mb_len = rte_pktmbuf_data_room_size(mp); - if ((desc == 0) || (desc % MLX4_PMD_SGE_WR_N)) { - ERROR("%p: invalid number of RX descriptors (must be a" - " multiple of %d)", (void *)dev, MLX4_PMD_SGE_WR_N); + if (desc == 0) { + ERROR("%p: invalid number of Rx descriptors", (void *)dev); return EINVAL; } - /* Toggle RX checksum offload if hardware supports it. */ - if (priv->hw_csum) - tmpl.csum = !!dev->data->dev_conf.rxmode.hw_ip_checksum; - if (priv->hw_csum_l2tun) - tmpl.csum_l2tun = !!dev->data->dev_conf.rxmode.hw_ip_checksum; /* Enable scattered packets support for this queue if necessary. */ assert(mb_len >= RTE_PKTMBUF_HEADROOM); - if ((dev->data->dev_conf.rxmode.jumbo_frame) && - (dev->data->dev_conf.rxmode.max_rx_pkt_len > - (mb_len - RTE_PKTMBUF_HEADROOM))) { - tmpl.sp = 1; - desc /= MLX4_PMD_SGE_WR_N; + if (dev->data->dev_conf.rxmode.max_rx_pkt_len <= + (mb_len - RTE_PKTMBUF_HEADROOM)) { + ; + } else if (dev->data->dev_conf.rxmode.enable_scatter) { + WARN("%p: scattered mode has been requested but is" + " not supported, this may lead to packet loss", + (void *)dev); + } else { + WARN("%p: the requested maximum Rx packet size (%u) is" + " larger than a single mbuf (%u) and scattered" + " mode has not been requested", + (void *)dev, + dev->data->dev_conf.rxmode.max_rx_pkt_len, + mb_len - RTE_PKTMBUF_HEADROOM); } - DEBUG("%p: %s scattered packets support (%u WRs)", - (void *)dev, (tmpl.sp ? "enabling" : "disabling"), desc); /* Use the entire RX mempool as the memory region. */ tmpl.mr = mlx4_mp2mr(priv->pd, mp); if (tmpl.mr == NULL) { @@ -3836,7 +1989,6 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, (void *)dev, strerror(ret)); goto error; } -skip_mr: attr.rd = (struct ibv_exp_res_domain_init_attr){ .comp_mask = (IBV_EXP_RES_DOMAIN_THREAD_MODEL | IBV_EXP_RES_DOMAIN_MSG_MODEL), @@ -3850,11 +2002,22 @@ skip_mr: (void *)dev, strerror(ret)); goto error; } + if (dev->data->dev_conf.intr_conf.rxq) { + tmpl.channel = ibv_create_comp_channel(priv->ctx); + if (tmpl.channel == NULL) { + ret = ENOMEM; + ERROR("%p: Rx interrupt completion channel creation" + " failure: %s", + (void *)dev, strerror(ret)); + goto error; + } + } attr.cq = (struct ibv_exp_cq_init_attr){ .comp_mask = IBV_EXP_CQ_INIT_ATTR_RES_DOMAIN, .res_domain = tmpl.rd, }; - tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, NULL, 0, &attr.cq); + tmpl.cq = ibv_exp_create_cq(priv->ctx, desc, NULL, tmpl.channel, 0, + &attr.cq); if (tmpl.cq == NULL) { ret = ENOMEM; ERROR("%p: CQ creation failure: %s", @@ -3865,13 +2028,7 @@ skip_mr: priv->device_attr.max_qp_wr); DEBUG("priv->device_attr.max_sge is %d", priv->device_attr.max_sge); -#ifdef RSS_SUPPORT - if (priv->rss && !inactive) - tmpl.qp = rxq_setup_qp_rss(priv, tmpl.cq, desc, parent, - tmpl.rd); - else -#endif /* RSS_SUPPORT */ - tmpl.qp = rxq_setup_qp(priv, tmpl.cq, desc, tmpl.rd); + tmpl.qp = rxq_setup_qp(priv, tmpl.cq, desc, tmpl.rd); if (tmpl.qp == NULL) { ret = (errno ? errno : EINVAL); ERROR("%p: QP creation failure: %s", @@ -3885,42 +2042,20 @@ skip_mr: .port_num = priv->port }; ret = ibv_exp_modify_qp(tmpl.qp, &mod, - (IBV_EXP_QP_STATE | -#ifdef RSS_SUPPORT - (parent ? IBV_EXP_QP_GROUP_RSS : 0) | -#endif /* RSS_SUPPORT */ - IBV_EXP_QP_PORT)); + IBV_EXP_QP_STATE | + IBV_EXP_QP_PORT); if (ret) { ERROR("%p: QP state to IBV_QPS_INIT failed: %s", (void *)dev, strerror(ret)); goto error; } - if ((parent) || (!priv->rss)) { - /* Configure MAC and broadcast addresses. */ - ret = rxq_mac_addrs_add(&tmpl); - if (ret) { - ERROR("%p: QP flow attachment failed: %s", - (void *)dev, strerror(ret)); - goto error; - } - } - /* Allocate descriptors for RX queues, except for the RSS parent. */ - if (parent) - goto skip_alloc; - if (tmpl.sp) - ret = rxq_alloc_elts_sp(&tmpl, desc, NULL); - else - ret = rxq_alloc_elts(&tmpl, desc, NULL); + ret = rxq_alloc_elts(&tmpl, desc); if (ret) { ERROR("%p: RXQ allocation failed: %s", (void *)dev, strerror(ret)); goto error; } - ret = ibv_post_recv(tmpl.qp, - (tmpl.sp ? - &(*tmpl.elts.sp)[0].wr : - &(*tmpl.elts.no_sp)[0].wr), - &bad_wr); + ret = ibv_post_recv(tmpl.qp, &(*tmpl.elts)[0].wr, &bad_wr); if (ret) { ERROR("%p: ibv_post_recv() failed for WR %p: %s", (void *)dev, @@ -3928,7 +2063,6 @@ skip_mr: strerror(ret)); goto error; } -skip_alloc: mod = (struct ibv_exp_qp_attr){ .qp_state = IBV_QPS_RTR }; @@ -4002,11 +2136,8 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, { struct priv *priv = dev->data->dev_private; struct rxq *rxq = (*priv->rxqs)[idx]; - int inactive = 0; int ret; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; priv_lock(priv); DEBUG("%p: configuring queue %u for %u descriptors", (void *)dev, idx, desc); @@ -4024,6 +2155,8 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, return -EEXIST; } (*priv->rxqs)[idx] = NULL; + if (idx == 0) + priv_mac_addr_del(priv); rxq_cleanup(rxq); } else { rxq = rte_calloc_socket("RXQ", 1, sizeof(*rxq), 0, socket); @@ -4034,9 +2167,7 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, return -ENOMEM; } } - if (idx >= rte_align32pow2(priv->rxqs_n + 1) >> 1) - inactive = 1; - ret = rxq_setup(dev, rxq, desc, socket, inactive, conf, mp); + ret = rxq_setup(dev, rxq, desc, socket, conf, mp); if (ret) rte_free(rxq); else { @@ -4045,10 +2176,7 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, (void *)dev, (void *)rxq); (*priv->rxqs)[idx] = rxq; /* Update receive callback. */ - if (rxq->sp) - dev->rx_pkt_burst = mlx4_rx_burst_sp; - else - dev->rx_pkt_burst = mlx4_rx_burst; + dev->rx_pkt_burst = mlx4_rx_burst; } priv_unlock(priv); return -ret; @@ -4067,18 +2195,17 @@ mlx4_rx_queue_release(void *dpdk_rxq) struct priv *priv; unsigned int i; - if (mlx4_is_secondary()) - return; if (rxq == NULL) return; priv = rxq->priv; priv_lock(priv); - assert(rxq != &priv->rxq_parent); for (i = 0; (i != priv->rxqs_n); ++i) if ((*priv->rxqs)[i] == rxq) { DEBUG("%p: removing RX queue %p from list", (void *)priv->dev, (void *)rxq); (*priv->rxqs)[i] = NULL; + if (i == 0) + priv_mac_addr_del(priv); break; } rxq_cleanup(rxq); @@ -4086,9 +2213,15 @@ mlx4_rx_queue_release(void *dpdk_rxq) priv_unlock(priv); } -static void +static int priv_dev_interrupt_handler_install(struct priv *, struct rte_eth_dev *); +static int +priv_dev_removal_interrupt_handler_install(struct priv *, struct rte_eth_dev *); + +static int +priv_dev_link_interrupt_handler_install(struct priv *, struct rte_eth_dev *); + /** * DPDK callback to start the device. * @@ -4104,12 +2237,8 @@ static int mlx4_dev_start(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; - unsigned int i = 0; - unsigned int r; - struct rxq *rxq; + int ret; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; priv_lock(priv); if (priv->started) { priv_unlock(priv); @@ -4117,45 +2246,41 @@ mlx4_dev_start(struct rte_eth_dev *dev) } DEBUG("%p: attaching configured flows to all RX queues", (void *)dev); priv->started = 1; - if (priv->rss) { - rxq = &priv->rxq_parent; - r = 1; - } else { - rxq = (*priv->rxqs)[0]; - r = priv->rxqs_n; + ret = priv_mac_addr_add(priv); + if (ret) + goto err; + ret = priv_dev_link_interrupt_handler_install(priv, dev); + if (ret) { + ERROR("%p: LSC handler install failed", + (void *)dev); + goto err; + } + ret = priv_dev_removal_interrupt_handler_install(priv, dev); + if (ret) { + ERROR("%p: RMV handler install failed", + (void *)dev); + goto err; + } + ret = priv_rx_intr_vec_enable(priv); + if (ret) { + ERROR("%p: Rx interrupt vector creation failed", + (void *)dev); + goto err; + } + ret = mlx4_priv_flow_start(priv); + if (ret) { + ERROR("%p: flow start failed: %s", + (void *)dev, strerror(ret)); + goto err; } - /* Iterate only once when RSS is enabled. */ - do { - int ret; - - /* Ignore nonexistent RX queues. */ - if (rxq == NULL) - continue; - ret = rxq_mac_addrs_add(rxq); - if (!ret && priv->promisc) - ret = rxq_promiscuous_enable(rxq); - if (!ret && priv->allmulti) - ret = rxq_allmulticast_enable(rxq); - if (!ret) - continue; - WARN("%p: QP flow attachment failed: %s", - (void *)dev, strerror(ret)); - /* Rollback. */ - while (i != 0) { - rxq = (*priv->rxqs)[--i]; - if (rxq != NULL) { - rxq_allmulticast_disable(rxq); - rxq_promiscuous_disable(rxq); - rxq_mac_addrs_del(rxq); - } - } - priv->started = 0; - priv_unlock(priv); - return -ret; - } while ((--r) && ((rxq = (*priv->rxqs)[++i]), i)); - priv_dev_interrupt_handler_install(priv, dev); priv_unlock(priv); return 0; +err: + /* Rollback. */ + priv_mac_addr_del(priv); + priv->started = 0; + priv_unlock(priv); + return -ret; } /** @@ -4170,12 +2295,7 @@ static void mlx4_dev_stop(struct rte_eth_dev *dev) { struct priv *priv = dev->data->dev_private; - unsigned int i = 0; - unsigned int r; - struct rxq *rxq; - if (mlx4_is_secondary()) - return; priv_lock(priv); if (!priv->started) { priv_unlock(priv); @@ -4183,22 +2303,8 @@ mlx4_dev_stop(struct rte_eth_dev *dev) } DEBUG("%p: detaching flows from all RX queues", (void *)dev); priv->started = 0; - if (priv->rss) { - rxq = &priv->rxq_parent; - r = 1; - } else { - rxq = (*priv->rxqs)[0]; - r = priv->rxqs_n; - } - /* Iterate only once when RSS is enabled. */ - do { - /* Ignore nonexistent RX queues. */ - if (rxq == NULL) - continue; - rxq_allmulticast_disable(rxq); - rxq_promiscuous_disable(rxq); - rxq_mac_addrs_del(rxq); - } while ((--r) && ((rxq = (*priv->rxqs)[++i]), i)); + mlx4_priv_flow_stop(priv); + priv_mac_addr_del(priv); priv_unlock(priv); } @@ -4252,9 +2358,16 @@ removed_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) return 0; } -static void +static int priv_dev_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *); +static int +priv_dev_removal_interrupt_handler_uninstall(struct priv *, + struct rte_eth_dev *); + +static int +priv_dev_link_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *); + /** * DPDK callback to close the device. * @@ -4266,7 +2379,7 @@ priv_dev_interrupt_handler_uninstall(struct priv *, struct rte_eth_dev *); static void mlx4_dev_close(struct rte_eth_dev *dev) { - struct priv *priv = mlx4_get_priv(dev); + struct priv *priv = dev->data->dev_private; void *tmp; unsigned int i; @@ -4276,6 +2389,7 @@ mlx4_dev_close(struct rte_eth_dev *dev) DEBUG("%p: closing device \"%s\"", (void *)dev, ((priv->ctx != NULL) ? priv->ctx->device->name : "")); + priv_mac_addr_del(priv); /* Prevent crashes when queues are still in use. This is unfortunately * still required for DPDK 1.3 because some programs (such as testpmd) * never release them before closing the device. */ @@ -4309,15 +2423,15 @@ mlx4_dev_close(struct rte_eth_dev *dev) priv->txqs_n = 0; priv->txqs = NULL; } - if (priv->rss) - rxq_cleanup(&priv->rxq_parent); if (priv->pd != NULL) { assert(priv->ctx != NULL); claim_zero(ibv_dealloc_pd(priv->pd)); claim_zero(ibv_close_device(priv->ctx)); } else assert(priv->ctx == NULL); - priv_dev_interrupt_handler_uninstall(priv, dev); + priv_dev_removal_interrupt_handler_uninstall(priv, dev); + priv_dev_link_interrupt_handler_uninstall(priv, dev); + priv_rx_intr_vec_disable(priv); priv_unlock(priv); memset(priv, 0, sizeof(*priv)); } @@ -4338,23 +2452,12 @@ 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 = mlx4_rx_burst; - else - dev->rx_pkt_burst = mlx4_rx_burst_sp; - dev->tx_pkt_burst = mlx4_tx_burst; + dev->rx_pkt_burst = mlx4_rx_burst; } else { err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP); if (err) @@ -4417,10 +2520,12 @@ mlx4_set_link_up(struct rte_eth_dev *dev) static void mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) { - struct priv *priv = mlx4_get_priv(dev); + struct priv *priv = dev->data->dev_private; unsigned int max; char ifname[IF_NAMESIZE]; + info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); + if (priv == NULL) return; priv_lock(priv); @@ -4439,19 +2544,9 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) info->max_rx_queues = max; info->max_tx_queues = max; /* Last array entry is reserved for broadcast. */ - info->max_mac_addrs = (elemof(priv->mac) - 1); - 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); + info->max_mac_addrs = 1; + info->rx_offload_capa = 0; + info->tx_offload_capa = 0; if (priv_get_ifname(priv, &ifname) == 0) info->if_index = if_nametoindex(ifname); info->speed_capa = @@ -4463,24 +2558,6 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) priv_unlock(priv); } -static const uint32_t * -mlx4_dev_supported_ptypes_get(struct rte_eth_dev *dev) -{ - static const uint32_t ptypes[] = { - /* refers to rxq_cq_to_pkt_type() */ - RTE_PTYPE_L3_IPV4, - RTE_PTYPE_L3_IPV6, - RTE_PTYPE_INNER_L3_IPV4, - RTE_PTYPE_INNER_L3_IPV6, - RTE_PTYPE_UNKNOWN - }; - - if (dev->rx_pkt_burst == mlx4_rx_burst || - dev->rx_pkt_burst == mlx4_rx_burst_sp) - return ptypes; - return NULL; -} - /** * DPDK callback to get device statistics. * @@ -4492,7 +2569,7 @@ mlx4_dev_supported_ptypes_get(struct rte_eth_dev *dev) static void mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) { - struct priv *priv = mlx4_get_priv(dev); + struct priv *priv = dev->data->dev_private; struct rte_eth_stats tmp = {0}; unsigned int i; unsigned int idx; @@ -4508,17 +2585,13 @@ mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) continue; idx = rxq->stats.idx; if (idx < RTE_ETHDEV_QUEUE_STAT_CNTRS) { -#ifdef MLX4_PMD_SOFT_COUNTERS tmp.q_ipackets[idx] += rxq->stats.ipackets; tmp.q_ibytes[idx] += rxq->stats.ibytes; -#endif tmp.q_errors[idx] += (rxq->stats.idropped + rxq->stats.rx_nombuf); } -#ifdef MLX4_PMD_SOFT_COUNTERS tmp.ipackets += rxq->stats.ipackets; tmp.ibytes += rxq->stats.ibytes; -#endif tmp.ierrors += rxq->stats.idropped; tmp.rx_nombuf += rxq->stats.rx_nombuf; } @@ -4529,21 +2602,14 @@ mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) continue; idx = txq->stats.idx; if (idx < RTE_ETHDEV_QUEUE_STAT_CNTRS) { -#ifdef MLX4_PMD_SOFT_COUNTERS tmp.q_opackets[idx] += txq->stats.opackets; tmp.q_obytes[idx] += txq->stats.obytes; -#endif tmp.q_errors[idx] += txq->stats.odropped; } -#ifdef MLX4_PMD_SOFT_COUNTERS tmp.opackets += txq->stats.opackets; tmp.obytes += txq->stats.obytes; -#endif tmp.oerrors += txq->stats.odropped; } -#ifndef MLX4_PMD_SOFT_COUNTERS - /* FIXME: retrieve and add hardware counters. */ -#endif *stats = tmp; priv_unlock(priv); } @@ -4557,7 +2623,7 @@ mlx4_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats) static void mlx4_stats_reset(struct rte_eth_dev *dev) { - struct priv *priv = mlx4_get_priv(dev); + struct priv *priv = dev->data->dev_private; unsigned int i; unsigned int idx; @@ -4578,250 +2644,11 @@ mlx4_stats_reset(struct rte_eth_dev *dev) (*priv->txqs)[i]->stats = (struct mlx4_txq_stats){ .idx = idx }; } -#ifndef MLX4_PMD_SOFT_COUNTERS - /* FIXME: reset hardware counters. */ -#endif - priv_unlock(priv); -} - -/** - * DPDK callback to remove a MAC address. - * - * @param dev - * Pointer to Ethernet device structure. - * @param index - * MAC address index. - */ -static void -mlx4_mac_addr_remove(struct rte_eth_dev *dev, uint32_t index) -{ - struct priv *priv = dev->data->dev_private; - - if (mlx4_is_secondary()) - return; - priv_lock(priv); - DEBUG("%p: removing MAC address from index %" PRIu32, - (void *)dev, index); - /* Last array entry is reserved for broadcast. */ - if (index >= (elemof(priv->mac) - 1)) - goto end; - priv_mac_addr_del(priv, index); -end: - priv_unlock(priv); -} - -/** - * DPDK callback to add a MAC address. - * - * @param dev - * Pointer to Ethernet device structure. - * @param mac_addr - * MAC address to register. - * @param index - * MAC address index. - * @param vmdq - * VMDq pool index to associate address with (ignored). - */ -static void -mlx4_mac_addr_add(struct rte_eth_dev *dev, struct ether_addr *mac_addr, - uint32_t index, uint32_t vmdq) -{ - struct priv *priv = dev->data->dev_private; - - if (mlx4_is_secondary()) - return; - (void)vmdq; - priv_lock(priv); - DEBUG("%p: adding MAC address at index %" PRIu32, - (void *)dev, index); - /* Last array entry is reserved for broadcast. */ - if (index >= (elemof(priv->mac) - 1)) - goto end; - priv_mac_addr_add(priv, index, - (const uint8_t (*)[ETHER_ADDR_LEN]) - mac_addr->addr_bytes); -end: - priv_unlock(priv); -} - -/** - * DPDK callback to set the primary MAC address. - * - * @param dev - * Pointer to Ethernet device structure. - * @param mac_addr - * MAC address to register. - */ -static void -mlx4_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr) -{ - DEBUG("%p: setting primary MAC address", (void *)dev); - mlx4_mac_addr_remove(dev, 0); - mlx4_mac_addr_add(dev, mac_addr, 0, 0); -} - -/** - * DPDK callback to enable promiscuous mode. - * - * @param dev - * Pointer to Ethernet device structure. - */ -static void -mlx4_promiscuous_enable(struct rte_eth_dev *dev) -{ - struct priv *priv = dev->data->dev_private; - unsigned int i; - int ret; - - if (mlx4_is_secondary()) - return; - priv_lock(priv); - if (priv->promisc) { - priv_unlock(priv); - return; - } - /* If device isn't started, this is all we need to do. */ - if (!priv->started) - goto end; - if (priv->rss) { - ret = rxq_promiscuous_enable(&priv->rxq_parent); - if (ret) { - priv_unlock(priv); - return; - } - goto end; - } - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - ret = rxq_promiscuous_enable((*priv->rxqs)[i]); - if (!ret) - continue; - /* Failure, rollback. */ - while (i != 0) - if ((*priv->rxqs)[--i] != NULL) - rxq_promiscuous_disable((*priv->rxqs)[i]); - priv_unlock(priv); - return; - } -end: - priv->promisc = 1; - priv_unlock(priv); -} - -/** - * DPDK callback to disable promiscuous mode. - * - * @param dev - * Pointer to Ethernet device structure. - */ -static void -mlx4_promiscuous_disable(struct rte_eth_dev *dev) -{ - struct priv *priv = dev->data->dev_private; - unsigned int i; - - if (mlx4_is_secondary()) - return; - priv_lock(priv); - if (!priv->promisc) { - priv_unlock(priv); - return; - } - if (priv->rss) { - rxq_promiscuous_disable(&priv->rxq_parent); - goto end; - } - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - rxq_promiscuous_disable((*priv->rxqs)[i]); -end: - priv->promisc = 0; - priv_unlock(priv); -} - -/** - * DPDK callback to enable allmulti mode. - * - * @param dev - * Pointer to Ethernet device structure. - */ -static void -mlx4_allmulticast_enable(struct rte_eth_dev *dev) -{ - struct priv *priv = dev->data->dev_private; - unsigned int i; - int ret; - - if (mlx4_is_secondary()) - return; - priv_lock(priv); - if (priv->allmulti) { - priv_unlock(priv); - return; - } - /* If device isn't started, this is all we need to do. */ - if (!priv->started) - goto end; - if (priv->rss) { - ret = rxq_allmulticast_enable(&priv->rxq_parent); - if (ret) { - priv_unlock(priv); - return; - } - goto end; - } - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - ret = rxq_allmulticast_enable((*priv->rxqs)[i]); - if (!ret) - continue; - /* Failure, rollback. */ - while (i != 0) - if ((*priv->rxqs)[--i] != NULL) - rxq_allmulticast_disable((*priv->rxqs)[i]); - priv_unlock(priv); - return; - } -end: - priv->allmulti = 1; - priv_unlock(priv); -} - -/** - * DPDK callback to disable allmulti mode. - * - * @param dev - * Pointer to Ethernet device structure. - */ -static void -mlx4_allmulticast_disable(struct rte_eth_dev *dev) -{ - struct priv *priv = dev->data->dev_private; - unsigned int i; - - if (mlx4_is_secondary()) - return; - priv_lock(priv); - if (!priv->allmulti) { - priv_unlock(priv); - return; - } - if (priv->rss) { - rxq_allmulticast_disable(&priv->rxq_parent); - goto end; - } - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - rxq_allmulticast_disable((*priv->rxqs)[i]); -end: - priv->allmulti = 0; priv_unlock(priv); } /** - * DPDK callback to retrieve physical link information (unlocked version). + * DPDK callback to retrieve physical link information. * * @param dev * Pointer to Ethernet device structure. @@ -4829,9 +2656,9 @@ end: * Wait for request completion (ignored). */ static int -mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) +mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) { - struct priv *priv = mlx4_get_priv(dev); + const struct priv *priv = dev->data->dev_private; struct ethtool_cmd edata = { .cmd = ETHTOOL_GSET }; @@ -4839,6 +2666,8 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) struct rte_eth_link dev_link; int link_speed = 0; + /* priv_lock() is not taken to allow concurrent calls. */ + if (priv == NULL) return -EINVAL; (void)wait_to_complete; @@ -4873,37 +2702,9 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) return -1; } -/** - * DPDK callback to retrieve physical link information. - * - * @param dev - * Pointer to Ethernet device structure. - * @param wait_to_complete - * Wait for request completion (ignored). - */ -static int -mlx4_link_update(struct rte_eth_dev *dev, int wait_to_complete) -{ - struct priv *priv = mlx4_get_priv(dev); - int ret; - - if (priv == NULL) - return -EINVAL; - priv_lock(priv); - ret = mlx4_link_update_unlocked(dev, wait_to_complete); - priv_unlock(priv); - return ret; -} - /** * 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 @@ -4917,71 +2718,17 @@ mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) { struct priv *priv = dev->data->dev_private; int ret = 0; - unsigned int i; - uint16_t (*rx_func)(void *, struct rte_mbuf **, uint16_t) = - mlx4_rx_burst; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; priv_lock(priv); /* 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)); - 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; - /* Make sure everyone has left mlx4_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 mb_len; - 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)); - mb_len = rte_pktmbuf_data_room_size(rxq->mp); - assert(mb_len >= RTE_PKTMBUF_HEADROOM); - 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 = mlx4_rx_burst_sp; - break; - } - /* Reenable non-RSS queue attributes. No need to check - * for errors at this stage. */ - if (!priv->rss) { - rxq_mac_addrs_add(rxq); - if (priv->promisc) - rxq_promiscuous_enable(rxq); - if (priv->allmulti) - rxq_allmulticast_enable(rxq); - } - /* Scattered burst function takes priority. */ - if (rxq->sp) - rx_func = mlx4_rx_burst_sp; - } - /* Burst functions can now be called again. */ - rte_wmb(); - dev->rx_pkt_burst = rx_func; + goto out; + } else + DEBUG("adapter port %u MTU set to %u", priv->port, mtu); + priv->mtu = mtu; out: priv_unlock(priv); assert(ret >= 0); @@ -5009,8 +2756,6 @@ mlx4_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) }; int ret; - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; ifr.ifr_data = (void *)ðpause; priv_lock(priv); if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { @@ -5059,8 +2804,6 @@ mlx4_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) }; int ret; - if (mlx4_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) || @@ -5091,121 +2834,49 @@ out: return -ret; } -/** - * Configure a VLAN filter. - * - * @param dev - * Pointer to Ethernet device structure. - * @param vlan_id - * VLAN ID to filter. - * @param on - * Toggle filter. - * - * @return - * 0 on success, errno value on failure. - */ -static int -vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on) -{ - struct priv *priv = dev->data->dev_private; - unsigned int i; - unsigned int j = -1; - - DEBUG("%p: %s VLAN filter ID %" PRIu16, - (void *)dev, (on ? "enable" : "disable"), vlan_id); - for (i = 0; (i != elemof(priv->vlan_filter)); ++i) { - if (!priv->vlan_filter[i].enabled) { - /* Unused index, remember it. */ - j = i; - continue; - } - if (priv->vlan_filter[i].id != vlan_id) - continue; - /* This VLAN ID is already known, use its index. */ - j = i; - break; - } - /* Check if there's room for another VLAN filter. */ - if (j == (unsigned int)-1) - return ENOMEM; - /* - * VLAN filters apply to all configured MAC addresses, flow - * specifications must be reconfigured accordingly. - */ - priv->vlan_filter[j].id = vlan_id; - if ((on) && (!priv->vlan_filter[j].enabled)) { - /* - * Filter is disabled, enable it. - * Rehashing flows in all RX queues is necessary. - */ - if (priv->rss) - rxq_mac_addrs_del(&priv->rxq_parent); - else - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - rxq_mac_addrs_del((*priv->rxqs)[i]); - priv->vlan_filter[j].enabled = 1; - if (priv->started) { - if (priv->rss) - rxq_mac_addrs_add(&priv->rxq_parent); - else - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - rxq_mac_addrs_add((*priv->rxqs)[i]); - } - } - } else if ((!on) && (priv->vlan_filter[j].enabled)) { - /* - * Filter is enabled, disable it. - * Rehashing flows in all RX queues is necessary. - */ - if (priv->rss) - rxq_mac_addrs_del(&priv->rxq_parent); - else - for (i = 0; (i != priv->rxqs_n); ++i) - if ((*priv->rxqs)[i] != NULL) - rxq_mac_addrs_del((*priv->rxqs)[i]); - priv->vlan_filter[j].enabled = 0; - if (priv->started) { - if (priv->rss) - rxq_mac_addrs_add(&priv->rxq_parent); - else - for (i = 0; (i != priv->rxqs_n); ++i) { - if ((*priv->rxqs)[i] == NULL) - continue; - rxq_mac_addrs_add((*priv->rxqs)[i]); - } - } - } - return 0; -} +const struct rte_flow_ops mlx4_flow_ops = { + .validate = mlx4_flow_validate, + .create = mlx4_flow_create, + .destroy = mlx4_flow_destroy, + .flush = mlx4_flow_flush, + .query = NULL, + .isolate = mlx4_flow_isolate, +}; /** - * DPDK callback to configure a VLAN filter. + * Manage filter operations. * * @param dev * Pointer to Ethernet device structure. - * @param vlan_id - * VLAN ID to filter. - * @param on - * Toggle filter. + * @param filter_type + * Filter type. + * @param filter_op + * Operation to perform. + * @param arg + * Pointer to operation-specific structure. * * @return * 0 on success, negative errno value on failure. */ static int -mlx4_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on) -{ - struct priv *priv = dev->data->dev_private; - int ret; - - if (mlx4_is_secondary()) - return -E_RTE_SECONDARY; - priv_lock(priv); - ret = vlan_filter_set(dev, vlan_id, on); - priv_unlock(priv); - assert(ret >= 0); +mlx4_dev_filter_ctrl(struct rte_eth_dev *dev, + enum rte_filter_type filter_type, + enum rte_filter_op filter_op, + void *arg) +{ + int ret = EINVAL; + + switch (filter_type) { + case RTE_ETH_FILTER_GENERIC: + if (filter_op != RTE_ETH_FILTER_GET) + return -EINVAL; + *(const void **)arg = &mlx4_flow_ops; + return 0; + default: + ERROR("%p: filter type (%d) not supported", + (void *)dev, filter_type); + break; + } return -ret; } @@ -5216,33 +2887,20 @@ static const struct eth_dev_ops mlx4_dev_ops = { .dev_set_link_down = mlx4_set_link_down, .dev_set_link_up = mlx4_set_link_up, .dev_close = mlx4_dev_close, - .promiscuous_enable = mlx4_promiscuous_enable, - .promiscuous_disable = mlx4_promiscuous_disable, - .allmulticast_enable = mlx4_allmulticast_enable, - .allmulticast_disable = mlx4_allmulticast_disable, .link_update = mlx4_link_update, .stats_get = mlx4_stats_get, .stats_reset = mlx4_stats_reset, - .queue_stats_mapping_set = NULL, .dev_infos_get = mlx4_dev_infos_get, - .dev_supported_ptypes_get = mlx4_dev_supported_ptypes_get, - .vlan_filter_set = mlx4_vlan_filter_set, - .vlan_tpid_set = NULL, - .vlan_strip_queue_set = NULL, - .vlan_offload_set = NULL, .rx_queue_setup = mlx4_rx_queue_setup, .tx_queue_setup = mlx4_tx_queue_setup, .rx_queue_release = mlx4_rx_queue_release, .tx_queue_release = mlx4_tx_queue_release, - .dev_led_on = NULL, - .dev_led_off = NULL, .flow_ctrl_get = mlx4_dev_get_flow_ctrl, .flow_ctrl_set = mlx4_dev_set_flow_ctrl, - .priority_flow_ctrl_set = NULL, - .mac_addr_remove = mlx4_mac_addr_remove, - .mac_addr_add = mlx4_mac_addr_add, - .mac_addr_set = mlx4_mac_addr_set, .mtu_set = mlx4_dev_set_mtu, + .filter_ctrl = mlx4_dev_filter_ctrl, + .rx_queue_intr_enable = mlx4_rx_intr_enable, + .rx_queue_intr_disable = mlx4_rx_intr_disable, }; /** @@ -5254,275 +2912,615 @@ static const struct eth_dev_ops mlx4_dev_ops = { * PCI bus address output buffer. * * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, -1 on failure and errno is set. + */ +static int +mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, + struct rte_pci_addr *pci_addr) +{ + FILE *file; + char line[32]; + MKSTR(path, "%s/device/uevent", device->ibdev_path); + + file = fopen(path, "rb"); + if (file == NULL) + return -1; + while (fgets(line, sizeof(line), file) == line) { + size_t len = strlen(line); + int ret; + + /* Truncate long lines. */ + if (len == (sizeof(line) - 1)) + while (line[(len - 1)] != '\n') { + ret = fgetc(file); + if (ret == EOF) + break; + line[(len - 1)] = ret; + } + /* Extract information. */ + if (sscanf(line, + "PCI_SLOT_NAME=" + "%" SCNx32 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", + &pci_addr->domain, + &pci_addr->bus, + &pci_addr->devid, + &pci_addr->function) == 4) { + ret = 0; + break; + } + } + fclose(file); + return 0; +} + +/** + * Get MAC address by querying netdevice. + * + * @param[in] priv + * struct priv for the requested device. + * @param[out] mac + * MAC address output buffer. + * + * @return + * 0 on success, -1 on failure and errno is set. + */ +static int +priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN]) +{ + struct ifreq request; + + if (priv_ifreq(priv, SIOCGIFHWADDR, &request)) + return -1; + memcpy(mac, request.ifr_hwaddr.sa_data, ETHER_ADDR_LEN); + return 0; +} + +static void +mlx4_dev_link_status_handler(void *); +static void +mlx4_dev_interrupt_handler(void *); + +/** + * Link/device status handler. + * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. + * @param events + * Pointer to event flags holder. + * + * @return + * Number of events + */ +static int +priv_dev_status_handler(struct priv *priv, struct rte_eth_dev *dev, + uint32_t *events) +{ + struct ibv_async_event event; + int port_change = 0; + struct rte_eth_link *link = &dev->data->dev_link; + int ret = 0; + + *events = 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) && + (priv->intr_conf.lsc == 1)) { + port_change = 1; + ret++; + } else if (event.event_type == IBV_EVENT_DEVICE_FATAL && + priv->intr_conf.rmv == 1) { + *events |= (1 << RTE_ETH_EVENT_INTR_RMV); + ret++; + } 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) + return ret; + mlx4_link_update(dev, 0); + if (((link->link_speed == 0) && link->link_status) || + ((link->link_speed != 0) && !link->link_status)) { + if (!priv->pending_alarm) { + /* Inconsistent status, check again later. */ + priv->pending_alarm = 1; + rte_eal_alarm_set(MLX4_ALARM_TIMEOUT_US, + mlx4_dev_link_status_handler, + dev); + } + } else { + *events |= (1 << RTE_ETH_EVENT_INTR_LSC); + } + return ret; +} + +/** + * Handle delayed link status event. + * + * @param arg + * Registered argument. + */ +static void +mlx4_dev_link_status_handler(void *arg) +{ + struct rte_eth_dev *dev = arg; + struct priv *priv = dev->data->dev_private; + uint32_t events; + int ret; + + priv_lock(priv); + assert(priv->pending_alarm == 1); + priv->pending_alarm = 0; + ret = priv_dev_status_handler(priv, dev, &events); + priv_unlock(priv); + if (ret > 0 && events & (1 << RTE_ETH_EVENT_INTR_LSC)) + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, + NULL); +} + +/** + * Handle interrupts from the NIC. + * + * @param[in] intr_handle + * Interrupt handler. + * @param cb_arg + * Callback argument. + */ +static void +mlx4_dev_interrupt_handler(void *cb_arg) +{ + struct rte_eth_dev *dev = cb_arg; + struct priv *priv = dev->data->dev_private; + int ret; + uint32_t ev; + int i; + + priv_lock(priv); + ret = priv_dev_status_handler(priv, dev, &ev); + priv_unlock(priv); + if (ret > 0) { + for (i = RTE_ETH_EVENT_UNKNOWN; + i < RTE_ETH_EVENT_MAX; + i++) { + if (ev & (1 << i)) { + ev &= ~(1 << i); + _rte_eth_dev_callback_process(dev, i, NULL, + NULL); + ret--; + } + } + if (ret) + WARN("%d event%s not processed", ret, + (ret > 1 ? "s were" : " was")); + } +} + +/** + * Uninstall interrupt handler. + * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. + * @return + * 0 on success, negative errno value on failure. + */ +static int +priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) +{ + int ret; + + if (priv->intr_conf.lsc || + priv->intr_conf.rmv) + return 0; + ret = rte_intr_callback_unregister(&priv->intr_handle, + mlx4_dev_interrupt_handler, + dev); + if (ret < 0) { + ERROR("rte_intr_callback_unregister failed with %d" + "%s%s%s", ret, + (errno ? " (errno: " : ""), + (errno ? strerror(errno) : ""), + (errno ? ")" : "")); + } + priv->intr_handle.fd = 0; + priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN; + return ret; +} + +/** + * Install interrupt handler. + * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. + * @return + * 0 on success, negative errno value on failure. + */ +static int +priv_dev_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) +{ + int flags; + int rc; + + /* Check whether the interrupt handler has already been installed + * for either type of interrupt + */ + if (priv->intr_conf.lsc && + priv->intr_conf.rmv && + priv->intr_handle.fd) + return 0; + 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; + dev->data->dev_conf.intr_conf.rmv = 0; + return -errno; + } else { + priv->intr_handle.fd = priv->ctx->async_fd; + priv->intr_handle.type = RTE_INTR_HANDLE_EXT; + rc = rte_intr_callback_register(&priv->intr_handle, + mlx4_dev_interrupt_handler, + dev); + if (rc) { + ERROR("rte_intr_callback_register failed " + " (errno: %s)", strerror(errno)); + return rc; + } + } + return 0; +} + +/** + * Uninstall interrupt handler. + * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. + * @return + * 0 on success, negative value on error. */ static int -mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, - struct rte_pci_addr *pci_addr) +priv_dev_removal_interrupt_handler_uninstall(struct priv *priv, + struct rte_eth_dev *dev) { - FILE *file; - char line[32]; - MKSTR(path, "%s/device/uevent", device->ibdev_path); - - file = fopen(path, "rb"); - if (file == NULL) - return -1; - while (fgets(line, sizeof(line), file) == line) { - size_t len = strlen(line); - int ret; - - /* Truncate long lines. */ - if (len == (sizeof(line) - 1)) - while (line[(len - 1)] != '\n') { - ret = fgetc(file); - if (ret == EOF) - break; - line[(len - 1)] = ret; - } - /* Extract information. */ - if (sscanf(line, - "PCI_SLOT_NAME=" - "%" SCNx16 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", - &pci_addr->domain, - &pci_addr->bus, - &pci_addr->devid, - &pci_addr->function) == 4) { - ret = 0; - break; - } + if (dev->data->dev_conf.intr_conf.rmv) { + priv->intr_conf.rmv = 0; + return priv_dev_interrupt_handler_uninstall(priv, dev); } - fclose(file); return 0; } /** - * Get MAC address by querying netdevice. - * - * @param[in] priv - * struct priv for the requested device. - * @param[out] mac - * MAC address output buffer. + * Uninstall interrupt handler. * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. * @return - * 0 on success, -1 on failure and errno is set. + * 0 on success, negative value on error, */ static int -priv_get_mac(struct priv *priv, uint8_t (*mac)[ETHER_ADDR_LEN]) +priv_dev_link_interrupt_handler_uninstall(struct priv *priv, + struct rte_eth_dev *dev) { - struct ifreq request; + int ret = 0; - if (priv_ifreq(priv, SIOCGIFHWADDR, &request)) - return -1; - memcpy(mac, request.ifr_hwaddr.sa_data, ETHER_ADDR_LEN); + if (dev->data->dev_conf.intr_conf.lsc) { + priv->intr_conf.lsc = 0; + ret = priv_dev_interrupt_handler_uninstall(priv, dev); + if (ret) + return ret; + } + if (priv->pending_alarm) + if (rte_eal_alarm_cancel(mlx4_dev_link_status_handler, + dev)) { + ERROR("rte_eal_alarm_cancel failed " + " (errno: %s)", strerror(rte_errno)); + return -rte_errno; + } + priv->pending_alarm = 0; return 0; } -/* Support up to 32 adapters. */ -static struct { - struct rte_pci_addr pci_addr; /* associated PCI address */ - uint32_t ports; /* physical ports bitfield. */ -} mlx4_dev[32]; - /** - * Get device index in mlx4_dev[] from PCI bus address. - * - * @param[in] pci_addr - * PCI bus address to look for. + * Install link interrupt handler. * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. * @return - * mlx4_dev[] index on success, -1 on failure. + * 0 on success, negative value on error. */ static int -mlx4_dev_idx(struct rte_pci_addr *pci_addr) +priv_dev_link_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) { - unsigned int i; - int ret = -1; + int ret; - assert(pci_addr != NULL); - for (i = 0; (i != elemof(mlx4_dev)); ++i) { - if ((mlx4_dev[i].pci_addr.domain == pci_addr->domain) && - (mlx4_dev[i].pci_addr.bus == pci_addr->bus) && - (mlx4_dev[i].pci_addr.devid == pci_addr->devid) && - (mlx4_dev[i].pci_addr.function == pci_addr->function)) - return i; - if ((mlx4_dev[i].ports == 0) && (ret == -1)) - ret = i; + if (dev->data->dev_conf.intr_conf.lsc) { + ret = priv_dev_interrupt_handler_install(priv, dev); + if (ret) + return ret; + priv->intr_conf.lsc = 1; } - return ret; + return 0; } /** - * Retrieve integer value from environment variable. - * - * @param[in] name - * Environment variable name. + * Install removal interrupt handler. * + * @param priv + * Pointer to private structure. + * @param dev + * Pointer to the rte_eth_dev structure. * @return - * Integer value, 0 if the variable is not set. + * 0 on success, negative value on error. */ static int -mlx4_getenv_int(const char *name) +priv_dev_removal_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) { - const char *val = getenv(name); + int ret; - if (val == NULL) - return 0; - return atoi(val); + if (dev->data->dev_conf.intr_conf.rmv) { + ret = priv_dev_interrupt_handler_install(priv, dev); + if (ret) + return ret; + priv->intr_conf.rmv = 1; + } + return 0; } -static void -mlx4_dev_link_status_handler(void *); -static void -mlx4_dev_interrupt_handler(struct rte_intr_handle *, void *); - /** - * Link status handler. + * Allocate queue vector and fill epoll fd list for Rx interrupts. * * @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. + * 0 on success, negative on failure. */ static int -priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev) +priv_rx_intr_vec_enable(struct priv *priv) { - struct ibv_async_event event; - int port_change = 0; - int ret = 0; - - /* Read all message and acknowledge them. */ - for (;;) { - if (ibv_get_async_event(priv->ctx, &event)) - break; + unsigned int i; + unsigned int rxqs_n = priv->rxqs_n; + unsigned int n = RTE_MIN(rxqs_n, (uint32_t)RTE_MAX_RXTX_INTR_VEC_ID); + unsigned int count = 0; + struct rte_intr_handle *intr_handle = priv->dev->intr_handle; - if (event.event_type == IBV_EVENT_PORT_ACTIVE || - event.event_type == IBV_EVENT_PORT_ERR) - port_change = 1; - else - DEBUG("event type %d on port %d not handled", - event.event_type, event.element.port_num); - ibv_ack_async_event(&event); + if (!priv->dev->data->dev_conf.intr_conf.rxq) + return 0; + priv_rx_intr_vec_disable(priv); + intr_handle->intr_vec = malloc(sizeof(intr_handle->intr_vec[rxqs_n])); + if (intr_handle->intr_vec == NULL) { + ERROR("failed to allocate memory for interrupt vector," + " Rx interrupts will not be supported"); + return -ENOMEM; } - - if (port_change ^ priv->pending_alarm) { - struct rte_eth_link *link = &dev->data->dev_link; - - priv->pending_alarm = 0; - mlx4_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(MLX4_ALARM_TIMEOUT_US, - mlx4_dev_link_status_handler, - dev); - } else - ret = 1; + intr_handle->type = RTE_INTR_HANDLE_EXT; + for (i = 0; i != n; ++i) { + struct rxq *rxq = (*priv->rxqs)[i]; + int fd; + int flags; + int rc; + + /* Skip queues that cannot request interrupts. */ + if (!rxq || !rxq->channel) { + /* Use invalid intr_vec[] index to disable entry. */ + intr_handle->intr_vec[i] = + RTE_INTR_VEC_RXTX_OFFSET + + RTE_MAX_RXTX_INTR_VEC_ID; + continue; + } + if (count >= RTE_MAX_RXTX_INTR_VEC_ID) { + ERROR("too many Rx queues for interrupt vector size" + " (%d), Rx interrupts cannot be enabled", + RTE_MAX_RXTX_INTR_VEC_ID); + priv_rx_intr_vec_disable(priv); + return -1; + } + fd = rxq->channel->fd; + flags = fcntl(fd, F_GETFL); + rc = fcntl(fd, F_SETFL, flags | O_NONBLOCK); + if (rc < 0) { + ERROR("failed to make Rx interrupt file descriptor" + " %d non-blocking for queue index %d", fd, i); + priv_rx_intr_vec_disable(priv); + return rc; + } + intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + count; + intr_handle->efds[count] = fd; + count++; } - return ret; + if (!count) + priv_rx_intr_vec_disable(priv); + else + intr_handle->nb_efd = count; + return 0; } /** - * Handle delayed link status event. + * Clean up Rx interrupts handler. * - * @param arg - * Registered argument. + * @param priv + * Pointer to private structure. */ static void -mlx4_dev_link_status_handler(void *arg) +priv_rx_intr_vec_disable(struct priv *priv) +{ + struct rte_intr_handle *intr_handle = priv->dev->intr_handle; + + rte_intr_free_epoll_fd(intr_handle); + free(intr_handle->intr_vec); + intr_handle->nb_efd = 0; + intr_handle->intr_vec = NULL; +} + +/** + * DPDK callback for Rx queue interrupt enable. + * + * @param dev + * Pointer to Ethernet device structure. + * @param idx + * Rx queue index. + * + * @return + * 0 on success, negative on failure. + */ +static int +mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx) { - struct rte_eth_dev *dev = arg; struct priv *priv = dev->data->dev_private; + struct rxq *rxq = (*priv->rxqs)[idx]; int ret; - priv_lock(priv); - assert(priv->pending_alarm == 1); - ret = priv_dev_link_status_handler(priv, dev); - priv_unlock(priv); + if (!rxq || !rxq->channel) + ret = EINVAL; + else + ret = ibv_req_notify_cq(rxq->cq, 0); if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL); + WARN("unable to arm interrupt on rx queue %d", idx); + return -ret; } /** - * Handle interrupts from the NIC. + * DPDK callback for Rx queue interrupt disable. * - * @param[in] intr_handle - * Interrupt handler. - * @param cb_arg - * Callback argument. + * @param dev + * Pointer to Ethernet device structure. + * @param idx + * Rx queue index. + * + * @return + * 0 on success, negative on failure. */ -static void -mlx4_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) +static int +mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx) { - struct rte_eth_dev *dev = cb_arg; struct priv *priv = dev->data->dev_private; + struct rxq *rxq = (*priv->rxqs)[idx]; + struct ibv_cq *ev_cq; + void *ev_ctx; int ret; - (void)intr_handle; - priv_lock(priv); - ret = priv_dev_link_status_handler(priv, dev); - priv_unlock(priv); + if (!rxq || !rxq->channel) { + ret = EINVAL; + } else { + ret = ibv_get_cq_event(rxq->cq->channel, &ev_cq, &ev_ctx); + if (ret || ev_cq != rxq->cq) + ret = EINVAL; + } if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL); + WARN("unable to disable interrupt on rx queue %d", + idx); + else + ibv_ack_cq_events(rxq->cq, 1); + return -ret; } /** - * Uninstall interrupt handler. + * Verify and store value for device argument. * - * @param priv - * Pointer to private structure. - * @param dev - * Pointer to the rte_eth_dev structure. + * @param[in] key + * Key argument to verify. + * @param[in] val + * Value associated with key. + * @param[in, out] conf + * Shared configuration data. + * + * @return + * 0 on success, negative errno value on failure. */ -static void -priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) +static int +mlx4_arg_parse(const char *key, const char *val, struct mlx4_conf *conf) { - if (!dev->data->dev_conf.intr_conf.lsc) - return; - rte_intr_callback_unregister(&priv->intr_handle, - mlx4_dev_interrupt_handler, - dev); - if (priv->pending_alarm) - rte_eal_alarm_cancel(mlx4_dev_link_status_handler, dev); - priv->pending_alarm = 0; - priv->intr_handle.fd = 0; - priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN; + unsigned long tmp; + + errno = 0; + tmp = strtoul(val, NULL, 0); + if (errno) { + WARN("%s: \"%s\" is not a valid integer", key, val); + return -errno; + } + if (strcmp(MLX4_PMD_PORT_KVARG, key) == 0) { + uint32_t ports = rte_log2_u32(conf->ports.present); + + if (tmp >= ports) { + ERROR("port index %lu outside range [0,%" PRIu32 ")", + tmp, ports); + return -EINVAL; + } + if (!(conf->ports.present & (1 << tmp))) { + ERROR("invalid port index %lu", tmp); + return -EINVAL; + } + conf->ports.enabled |= 1 << tmp; + } else { + WARN("%s: unknown parameter", key); + return -EINVAL; + } + return 0; } /** - * Install interrupt handler. + * Parse device parameters. * - * @param priv - * Pointer to private structure. - * @param dev - * Pointer to the rte_eth_dev structure. + * @param devargs + * Device arguments structure. + * + * @return + * 0 on success, negative errno value on failure. */ -static void -priv_dev_interrupt_handler_install(struct priv *priv, struct rte_eth_dev *dev) +static int +mlx4_args(struct rte_devargs *devargs, struct mlx4_conf *conf) { - int rc, flags; + struct rte_kvargs *kvlist; + unsigned int arg_count; + int ret = 0; + int i; - if (!dev->data->dev_conf.intr_conf.lsc) - return; - 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 { - priv->intr_handle.fd = priv->ctx->async_fd; - priv->intr_handle.type = RTE_INTR_HANDLE_EXT; - rte_intr_callback_register(&priv->intr_handle, - mlx4_dev_interrupt_handler, - dev); + if (devargs == NULL) + return 0; + kvlist = rte_kvargs_parse(devargs->args, pmd_mlx4_init_params); + if (kvlist == NULL) { + ERROR("failed to parse kvargs"); + return -EINVAL; + } + /* Process parameters. */ + for (i = 0; pmd_mlx4_init_params[i]; ++i) { + arg_count = rte_kvargs_count(kvlist, MLX4_PMD_PORT_KVARG); + while (arg_count-- > 0) { + ret = rte_kvargs_process(kvlist, + MLX4_PMD_PORT_KVARG, + (int (*)(const char *, + const char *, + void *)) + mlx4_arg_parse, + conf); + if (ret != 0) + goto free_kvlist; + } } +free_kvlist: + rte_kvargs_free(kvlist); + return ret; } -static struct eth_driver mlx4_driver; +static struct rte_pci_driver mlx4_driver; /** * DPDK callback to register a PCI device. @@ -5546,29 +3544,20 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) int err = 0; struct ibv_context *attr_ctx = NULL; struct ibv_device_attr device_attr; + struct mlx4_conf conf = { + .ports.present = 0, + }; unsigned int vf; - int idx; int i; (void)pci_drv; - assert(pci_drv == &mlx4_driver.pci_drv); - /* Get mlx4_dev[] index. */ - idx = mlx4_dev_idx(&pci_dev->addr); - if (idx == -1) { - ERROR("this driver cannot support any more adapters"); - return -ENOMEM; - } - DEBUG("using driver device index %d", idx); + assert(pci_drv == &mlx4_driver); - /* Save PCI address. */ - mlx4_dev[idx].pci_addr = pci_dev->addr; list = ibv_get_device_list(&i); if (list == NULL) { assert(errno); - if (errno == ENOSYS) { - WARN("cannot list devices, is ib_uverbs loaded?"); - return 0; - } + if (errno == ENOSYS) + ERROR("cannot list devices, is ib_uverbs loaded?"); return -errno; } assert(i >= 0); @@ -5600,11 +3589,11 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) ibv_free_device_list(list); switch (err) { case 0: - WARN("cannot access device, is mlx4_ib loaded?"); - return 0; + ERROR("cannot access device, is mlx4_ib loaded?"); + return -ENODEV; case EINVAL: - WARN("cannot use device, are drivers up to date?"); - return 0; + ERROR("cannot use device, are drivers up to date?"); + return -EINVAL; } assert(err > 0); return -err; @@ -5612,46 +3601,54 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) ibv_dev = list[i]; DEBUG("device opened"); - if (ibv_query_device(attr_ctx, &device_attr)) + if (ibv_query_device(attr_ctx, &device_attr)) { + err = ENODEV; goto error; + } INFO("%u port(s) detected", device_attr.phys_port_cnt); + conf.ports.present |= (UINT64_C(1) << device_attr.phys_port_cnt) - 1; + if (mlx4_args(pci_dev->device.devargs, &conf)) { + ERROR("failed to process device arguments"); + err = EINVAL; + goto error; + } + /* Use all ports when none are defined */ + if (!conf.ports.enabled) + conf.ports.enabled = conf.ports.present; for (i = 0; i < device_attr.phys_port_cnt; i++) { uint32_t port = i + 1; /* ports are indexed from one */ - uint32_t test = (1 << i); struct ibv_context *ctx = NULL; struct ibv_port_attr port_attr; struct ibv_pd *pd = NULL; struct priv *priv = NULL; struct rte_eth_dev *eth_dev = NULL; -#ifdef HAVE_EXP_QUERY_DEVICE - struct ibv_exp_device_attr exp_device_attr; -#endif /* HAVE_EXP_QUERY_DEVICE */ struct ether_addr mac; -#ifdef HAVE_EXP_QUERY_DEVICE - exp_device_attr.comp_mask = IBV_EXP_DEVICE_ATTR_EXP_CAP_FLAGS; -#ifdef RSS_SUPPORT - exp_device_attr.comp_mask |= IBV_EXP_DEVICE_ATTR_RSS_TBL_SZ; -#endif /* RSS_SUPPORT */ -#endif /* HAVE_EXP_QUERY_DEVICE */ + /* If port is not enabled, skip. */ + if (!(conf.ports.enabled & (1 << i))) + continue; - DEBUG("using port %u (%08" PRIx32 ")", port, test); + DEBUG("using port %u", port); ctx = ibv_open_device(ibv_dev); - if (ctx == NULL) + if (ctx == NULL) { + err = ENODEV; goto port_error; + } /* Check port status. */ err = ibv_query_port(ctx, port, &port_attr); if (err) { ERROR("port query failed: %s", strerror(err)); + err = ENODEV; goto port_error; } if (port_attr.link_layer != IBV_LINK_LAYER_ETHERNET) { ERROR("port %d is not configured in Ethernet mode", port); + err = EINVAL; goto port_error; } @@ -5668,8 +3665,6 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) goto port_error; } - mlx4_dev[idx].ports |= test; - /* from rte_ethdev.c */ priv = rte_zmalloc("ethdev private structure", sizeof(*priv), @@ -5685,84 +3680,13 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) priv->port = port; priv->pd = pd; priv->mtu = ETHER_MTU; -#ifdef HAVE_EXP_QUERY_DEVICE - if (ibv_exp_query_device(ctx, &exp_device_attr)) { - ERROR("ibv_exp_query_device() failed"); - goto port_error; - } -#ifdef RSS_SUPPORT - if ((exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_QPG) && - (exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_UD_RSS) && - (exp_device_attr.comp_mask & - IBV_EXP_DEVICE_ATTR_RSS_TBL_SZ) && - (exp_device_attr.max_rss_tbl_sz > 0)) { - priv->hw_qpg = 1; - priv->hw_rss = 1; - priv->max_rss_tbl_sz = exp_device_attr.max_rss_tbl_sz; - } else { - priv->hw_qpg = 0; - priv->hw_rss = 0; - priv->max_rss_tbl_sz = 0; - } - priv->hw_tss = !!(exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_UD_TSS); - DEBUG("device flags: %s%s%s", - (priv->hw_qpg ? "IBV_DEVICE_QPG " : ""), - (priv->hw_tss ? "IBV_DEVICE_TSS " : ""), - (priv->hw_rss ? "IBV_DEVICE_RSS " : "")); - if (priv->hw_rss) - DEBUG("maximum RSS indirection table size: %u", - exp_device_attr.max_rss_tbl_sz); -#endif /* RSS_SUPPORT */ - - priv->hw_csum = - ((exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_RX_CSUM_TCP_UDP_PKT) && - (exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_RX_CSUM_IP_PKT)); - DEBUG("checksum offloading is %ssupported", - (priv->hw_csum ? "" : "not ")); - - priv->hw_csum_l2tun = !!(exp_device_attr.exp_device_cap_flags & - IBV_EXP_DEVICE_VXLAN_SUPPORT); - DEBUG("L2 tunnel checksum offloads are %ssupported", - (priv->hw_csum_l2tun ? "" : "not ")); - -#ifdef INLINE_RECV - priv->inl_recv_size = mlx4_getenv_int("MLX4_INLINE_RECV_SIZE"); - - if (priv->inl_recv_size) { - exp_device_attr.comp_mask = - IBV_EXP_DEVICE_ATTR_INLINE_RECV_SZ; - if (ibv_exp_query_device(ctx, &exp_device_attr)) { - INFO("Couldn't query device for inline-receive" - " capabilities."); - priv->inl_recv_size = 0; - } else { - if ((unsigned)exp_device_attr.inline_recv_sz < - priv->inl_recv_size) { - INFO("Max inline-receive (%d) <" - " requested inline-receive (%u)", - exp_device_attr.inline_recv_sz, - priv->inl_recv_size); - priv->inl_recv_size = - exp_device_attr.inline_recv_sz; - } - } - INFO("Set inline receive size to %u", - priv->inl_recv_size); - } -#endif /* INLINE_RECV */ -#endif /* HAVE_EXP_QUERY_DEVICE */ - (void)mlx4_getenv_int; priv->vf = vf; /* Configure the first MAC address by default. */ if (priv_get_mac(priv, &mac.addr_bytes)) { ERROR("cannot get MAC address, is mlx4_en loaded?" " (errno: %s)", strerror(errno)); + err = ENODEV; goto port_error; } INFO("port %u MAC address is %02x:%02x:%02x:%02x:%02x:%02x", @@ -5770,13 +3694,10 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) mac.addr_bytes[0], mac.addr_bytes[1], mac.addr_bytes[2], mac.addr_bytes[3], mac.addr_bytes[4], mac.addr_bytes[5]); - /* Register MAC and broadcast addresses. */ - claim_zero(priv_mac_addr_add(priv, 0, - (const uint8_t (*)[ETHER_ADDR_LEN]) - mac.addr_bytes)); - claim_zero(priv_mac_addr_add(priv, (elemof(priv->mac) - 1), - &(const uint8_t [ETHER_ADDR_LEN]) - { "\xff\xff\xff\xff\xff\xff" })); + /* Register MAC address. */ + priv->mac = mac; + if (priv_mac_addr_add(priv)) + goto port_error; #ifndef NDEBUG { char ifname[IF_NAMESIZE]; @@ -5806,51 +3727,33 @@ mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) goto port_error; } - /* Secondary processes have to use local storage for their - * private data as well as a copy of eth_dev->data, but this - * pointer must not be modified before burst functions are - * actually called. */ - if (mlx4_is_secondary()) { - struct mlx4_secondary_data *sd = - &mlx4_secondary_data[eth_dev->data->port_id]; - - sd->primary_priv = eth_dev->data->dev_private; - if (sd->primary_priv == NULL) { - ERROR("no private data for port %u", - eth_dev->data->port_id); - err = EINVAL; - goto port_error; - } - sd->shared_dev_data = eth_dev->data; - rte_spinlock_init(&sd->lock); - memcpy(sd->data.name, sd->shared_dev_data->name, - sizeof(sd->data.name)); - sd->data.dev_private = priv; - sd->data.rx_mbuf_alloc_failed = 0; - sd->data.mtu = ETHER_MTU; - sd->data.port_id = sd->shared_dev_data->port_id; - sd->data.mac_addrs = priv->mac; - eth_dev->tx_pkt_burst = mlx4_tx_burst_secondary_setup; - eth_dev->rx_pkt_burst = mlx4_rx_burst_secondary_setup; - } else { - eth_dev->data->dev_private = priv; - eth_dev->data->rx_mbuf_alloc_failed = 0; - eth_dev->data->mtu = ETHER_MTU; - eth_dev->data->mac_addrs = priv->mac; - } - eth_dev->pci_dev = pci_dev; + eth_dev->data->dev_private = priv; + eth_dev->data->mac_addrs = &priv->mac; + eth_dev->device = &pci_dev->device; rte_eth_copy_pci_info(eth_dev, pci_dev); - eth_dev->driver = &mlx4_driver; + eth_dev->device->driver = &mlx4_driver.driver; + + /* + * Copy and override interrupt handle to prevent it from + * being shared between all ethdev instances of a given PCI + * device. This is required to properly handle Rx interrupts + * on all ports. + */ + priv->intr_handle_dev = *eth_dev->intr_handle; + eth_dev->intr_handle = &priv->intr_handle_dev; priv->dev = eth_dev; eth_dev->dev_ops = &mlx4_dev_ops; - TAILQ_INIT(ð_dev->link_intr_cbs); + eth_dev->data->dev_flags |= RTE_ETH_DEV_DETACHABLE; /* Bring Ethernet device up. */ DEBUG("forcing Ethernet interface up"); priv_set_flags(priv, ~IFF_UP, IFF_UP); + /* Update link status once if waiting for LSC. */ + if (eth_dev->data->dev_flags & RTE_ETH_DEV_INTR_LSC) + mlx4_link_update(eth_dev, 0); continue; port_error: @@ -5863,6 +3766,8 @@ port_error: rte_eth_dev_release_port(eth_dev); break; } + if (i == device_attr.phys_port_cnt) + return 0; /* * XXX if something went wrong in the loop above, there is a resource @@ -5871,12 +3776,6 @@ port_error: * way to enumerate the registered ethdevs to free the previous ones. */ - /* no port found, complain */ - if (!mlx4_dev[idx].ports) { - err = ENODEV; - goto error; - } - error: if (attr_ctx) claim_zero(ibv_close_device(attr_ctx)); @@ -5904,16 +3803,14 @@ static const struct rte_pci_id mlx4_pci_id_map[] = { } }; -static struct eth_driver mlx4_driver = { - .pci_drv = { - .driver = { - .name = MLX4_DRIVER_NAME - }, - .id_table = mlx4_pci_id_map, - .probe = mlx4_pci_probe, - .drv_flags = RTE_PCI_DRV_INTR_LSC, +static struct rte_pci_driver mlx4_driver = { + .driver = { + .name = MLX4_DRIVER_NAME }, - .dev_private_size = sizeof(struct priv) + .id_table = mlx4_pci_id_map, + .probe = mlx4_pci_probe, + .drv_flags = RTE_PCI_DRV_INTR_LSC | + RTE_PCI_DRV_INTR_RMV, }; /** @@ -5932,8 +3829,10 @@ rte_mlx4_pmd_init(void) */ setenv("RDMAV_HUGEPAGES_SAFE", "1", 1); ibv_fork_init(); - rte_eal_pci_register(&mlx4_driver.pci_drv); + rte_pci_register(&mlx4_driver); } RTE_PMD_EXPORT_NAME(net_mlx4, __COUNTER__); RTE_PMD_REGISTER_PCI_TABLE(net_mlx4, mlx4_pci_id_map); +RTE_PMD_REGISTER_KMOD_DEP(net_mlx4, + "* ib_uverbs & mlx4_en & mlx4_core & mlx4_ib");