X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fmlx4%2Fmlx4.c;h=ae4be9eec8623da999f9baaaae06df3b82ccbbf3;hb=0a2ae703192c;hp=a5991ce653dc44f2c26e668caaae464caa43cc7c;hpb=7b06615392290705bd662c2ad71d397b519346e1;p=dpdk.git diff --git a/drivers/net/mlx4/mlx4.c b/drivers/net/mlx4/mlx4.c index a5991ce653..ae4be9eec8 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 2012-2017 6WIND S.A. + * Copyright 2012-2017 Mellanox. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions @@ -54,27 +54,13 @@ #include #include #include -#include #include #include #include -/* Verbs header. */ -/* ISO C doesn't support unnamed structs/unions, disabling -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-pedantic" -#endif -#include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-pedantic" -#endif - -/* DPDK headers don't like -pedantic. */ -#ifdef PEDANTIC -#pragma GCC diagnostic ignored "-pedantic" -#endif #include #include +#include #include #include #include @@ -87,30 +73,16 @@ #include #include #include -#ifdef PEDANTIC -#pragma GCC diagnostic error "-pedantic" -#endif +#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) @@ -138,165 +110,35 @@ typedef union { (((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. */ -}; +/* 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]; -/* 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. */ +struct mlx4_conf { + uint8_t active_ports; }; -/* 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. */ - uint32_t mb_len; /* Length of a mp-issued mbuf. */ - struct mlx4_rxq_stats stats; /* RX queue counters. */ - unsigned int socket; /* CPU socket ID for allocations. */ - struct ibv_exp_res_domain *rd; /* Resource Domain. */ +/* Available parameters list. */ +const char *pmd_mlx4_init_params[] = { + MLX4_PMD_PORT_KVARG, + NULL, }; -/* TX element. */ -struct txq_elt { - struct rte_mbuf *buf; -}; +static int +mlx4_rx_intr_enable(struct rte_eth_dev *dev, uint16_t idx); -/* 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. */ - 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. */ -}; +static int +mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx); -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. */ -}; +static int +priv_rx_intr_vec_enable(struct priv *priv); -/* 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 void +priv_rx_intr_vec_disable(struct priv *priv); /** * Check if running as a secondary process. @@ -337,8 +179,7 @@ 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); } @@ -349,8 +190,7 @@ 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); } @@ -659,7 +499,15 @@ priv_get_mtu(struct priv *priv, uint16_t *mtu) static int priv_set_mtu(struct priv *priv, uint16_t mtu) { - return priv_set_sysfs_ulong(priv, "mtu", mtu); + uint16_t new_mtu; + + if (priv_set_sysfs_ulong(priv, "mtu", mtu) || + priv_get_mtu(priv, &new_mtu)) + return -1; + if (new_mtu == mtu) + return 0; + errno = EINVAL; + return -1; } /** @@ -683,7 +531,7 @@ priv_set_flags(struct priv *priv, unsigned int keep, unsigned int flags) if (priv_get_sysfs_ulong(priv, "flags", &tmp) == -1) return -1; tmp &= keep; - tmp |= flags; + tmp |= (flags & (~keep)); return priv_set_sysfs_ulong(priv, "flags", tmp); } @@ -698,7 +546,7 @@ txq_cleanup(struct txq *txq); static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, const struct rte_eth_rxconf *conf, + unsigned int socket, int inactive, const struct rte_eth_rxconf *conf, struct rte_mempool *mp); static void @@ -734,12 +582,15 @@ dev_configure(struct rte_eth_dev *dev) } if (rxqs_n == priv->rxqs_n) return 0; - if ((rxqs_n & (rxqs_n - 1)) != 0) { - ERROR("%p: invalid number of RX queues (%u)," - " must be a power of 2", - (void *)dev, rxqs_n); - return EINVAL; + 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. */ @@ -775,7 +626,7 @@ dev_configure(struct rte_eth_dev *dev) priv->rss = 1; tmp = priv->rxqs_n; priv->rxqs_n = rxqs_n; - ret = rxq_setup(dev, &priv->rxq_parent, 0, 0, NULL, NULL); + ret = rxq_setup(dev, &priv->rxq_parent, 0, 0, 0, NULL, NULL); if (!ret) return 0; /* Failure, rollback. */ @@ -995,7 +846,7 @@ txq_alloc_elts(struct txq *txq, unsigned int elts_n) } mr_linear = ibv_reg_mr(txq->priv->pd, elts_linear, sizeof(*elts_linear), - (IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_REMOTE_WRITE)); + IBV_ACCESS_LOCAL_WRITE); if (mr_linear == NULL) { ERROR("%p: unable to configure MR, ibv_reg_mr() failed", (void *)txq); @@ -1044,14 +895,20 @@ error: static void txq_free_elts(struct txq *txq) { - unsigned int i; unsigned int elts_n = txq->elts_n; + 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; + txq->elts_head = 0; + txq->elts_tail = 0; + txq->elts_comp = 0; + txq->elts_comp_cd = 0; + txq->elts_comp_cd_init = 0; txq->elts = NULL; txq->elts_linear = NULL; txq->mr_linear = NULL; @@ -1061,12 +918,17 @@ txq_free_elts(struct txq *txq) rte_free(elts_linear); if (elts == NULL) return; - for (i = 0; (i != elemof(*elts)); ++i) { - struct txq_elt *elt = &(*elts)[i]; + while (elts_tail != elts_head) { + struct txq_elt *elt = &(*elts)[elts_tail]; - if (elt->buf == NULL) - continue; + assert(elt->buf != NULL); rte_pktmbuf_free(elt->buf); +#ifndef NDEBUG + /* Poisoning. */ + memset(elt, 0x77, sizeof(*elt)); +#endif + if (++elts_tail == elts_n) + elts_tail = 0; } rte_free(elts); } @@ -1184,9 +1046,72 @@ txq_complete(struct txq *txq) return 0; } +struct mlx4_check_mempool_data { + int ret; + char *start; + char *end; +}; + +/* Called by mlx4_check_mempool() when iterating the memory chunks. */ +static void mlx4_check_mempool_cb(struct rte_mempool *mp, + void *opaque, struct rte_mempool_memhdr *memhdr, + unsigned mem_idx) +{ + struct mlx4_check_mempool_data *data = opaque; + + (void)mp; + (void)mem_idx; + + /* It already failed, skip the next chunks. */ + if (data->ret != 0) + return; + /* It is the first chunk. */ + if (data->start == NULL && data->end == NULL) { + data->start = memhdr->addr; + data->end = data->start + memhdr->len; + return; + } + if (data->end == memhdr->addr) { + data->end += memhdr->len; + return; + } + if (data->start == (char *)memhdr->addr + memhdr->len) { + data->start -= memhdr->len; + return; + } + /* Error, mempool is not virtually contigous. */ + data->ret = -1; +} + +/** + * Check if a mempool can be used: it must be virtually contiguous. + * + * @param[in] mp + * Pointer to memory pool. + * @param[out] start + * Pointer to the start address of the mempool virtual memory area + * @param[out] end + * Pointer to the end address of the mempool virtual memory area + * + * @return + * 0 on success (mempool is virtually contiguous), -1 on error. + */ +static int mlx4_check_mempool(struct rte_mempool *mp, uintptr_t *start, + uintptr_t *end) +{ + struct mlx4_check_mempool_data data; + + memset(&data, 0, sizeof(data)); + rte_mempool_mem_iter(mp, mlx4_check_mempool_cb, &data); + *start = (uintptr_t)data.start; + *end = (uintptr_t)data.end; + + return data.ret; +} + /* For best performance, this function should not be inlined. */ -static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, const struct rte_mempool *) - __attribute__((noinline)); +static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, struct rte_mempool *) + __rte_noinline; /** * Register mempool as a memory region. @@ -1200,15 +1125,21 @@ static struct ibv_mr *mlx4_mp2mr(struct ibv_pd *, const struct rte_mempool *) * Memory region pointer, NULL in case of error. */ static struct ibv_mr * -mlx4_mp2mr(struct ibv_pd *pd, const struct rte_mempool *mp) +mlx4_mp2mr(struct ibv_pd *pd, struct rte_mempool *mp) { const struct rte_memseg *ms = rte_eal_get_physmem_layout(); - uintptr_t start = mp->elt_va_start; - uintptr_t end = mp->elt_va_end; + uintptr_t start; + uintptr_t end; unsigned int i; + if (mlx4_check_mempool(mp, &start, &end) != 0) { + ERROR("mempool %p: not virtually contiguous", + (void *)mp); + return NULL; + } + DEBUG("mempool %p area start=%p end=%p size=%zu", - (const void *)mp, (void *)start, (void *)end, + (void *)mp, (void *)start, (void *)end, (size_t)(end - start)); /* Round start and end to page boundary if found in memory segments. */ for (i = 0; (i < RTE_MAX_MEMSEG) && (ms[i].addr != NULL); ++i) { @@ -1222,12 +1153,12 @@ mlx4_mp2mr(struct ibv_pd *pd, const struct rte_mempool *mp) end = RTE_ALIGN_CEIL(end, align); } DEBUG("mempool %p using start=%p end=%p size=%zu for MR", - (const void *)mp, (void *)start, (void *)end, + (void *)mp, (void *)start, (void *)end, (size_t)(end - start)); return ibv_reg_mr(pd, (void *)start, end - start, - IBV_ACCESS_LOCAL_WRITE | IBV_ACCESS_REMOTE_WRITE); + IBV_ACCESS_LOCAL_WRITE); } /** @@ -1262,7 +1193,7 @@ txq_mb2mp(struct rte_mbuf *buf) * mr->lkey on success, (uint32_t)-1 on failure. */ static uint32_t -txq_mp2mr(struct txq *txq, const struct rte_mempool *mp) +txq_mp2mr(struct txq *txq, struct rte_mempool *mp) { unsigned int i; struct ibv_mr *mr; @@ -1280,7 +1211,7 @@ txq_mp2mr(struct txq *txq, const struct rte_mempool *mp) } /* Add a new entry, register MR first. */ DEBUG("%p: discovered new memory pool \"%s\" (%p)", - (void *)txq, mp->name, (const void *)mp); + (void *)txq, mp->name, (void *)mp); mr = mlx4_mp2mr(txq->priv->pd, mp); if (unlikely(mr == NULL)) { DEBUG("%p: unable to configure MR, ibv_reg_mr() failed.", @@ -1301,12 +1232,11 @@ txq_mp2mr(struct txq *txq, const struct rte_mempool *mp) txq->mp2mr[i].mr = mr; txq->mp2mr[i].lkey = mr->lkey; DEBUG("%p: new MR lkey for MP \"%s\" (%p): 0x%08" PRIu32, - (void *)txq, mp->name, (const void *)mp, txq->mp2mr[i].lkey); + (void *)txq, mp->name, (void *)mp, txq->mp2mr[i].lkey); return txq->mp2mr[i].lkey; } struct txq_mp2mr_mbuf_check_data { - const struct rte_mempool *mp; int ret; }; @@ -1314,34 +1244,26 @@ struct txq_mp2mr_mbuf_check_data { * Callback function for rte_mempool_obj_iter() to check whether a given * mempool object looks like a mbuf. * - * @param[in, out] arg - * Context data (struct txq_mp2mr_mbuf_check_data). Contains mempool pointer - * and return value. - * @param[in] start - * Object start address. - * @param[in] end - * Object end address. + * @param[in] mp + * The mempool pointer + * @param[in] arg + * Context data (struct txq_mp2mr_mbuf_check_data). Contains the + * return value. + * @param[in] obj + * Object address. * @param index - * Unused. - * - * @return - * Nonzero value when object is not a mbuf. + * Object index, unused. */ static void -txq_mp2mr_mbuf_check(void *arg, void *start, void *end, - uint32_t index __rte_unused) +txq_mp2mr_mbuf_check(struct rte_mempool *mp, void *arg, void *obj, + uint32_t index __rte_unused) { struct txq_mp2mr_mbuf_check_data *data = arg; - struct rte_mbuf *buf = - (void *)((uintptr_t)start + data->mp->header_size); + struct rte_mbuf *buf = obj; - (void)index; /* Check whether mbuf structure fits element size and whether mempool * pointer is valid. */ - if (((uintptr_t)end >= (uintptr_t)(buf + 1)) && - (buf->pool == data->mp)) - data->ret = 0; - else + if (sizeof(*buf) > mp->elt_size || buf->pool != mp) data->ret = -1; } @@ -1355,28 +1277,16 @@ txq_mp2mr_mbuf_check(void *arg, void *start, void *end, * Pointer to TX queue structure. */ static void -txq_mp2mr_iter(const struct rte_mempool *mp, void *arg) +txq_mp2mr_iter(struct rte_mempool *mp, void *arg) { struct txq *txq = arg; struct txq_mp2mr_mbuf_check_data data = { - .mp = mp, - .ret = -1, + .ret = 0, }; - /* Discard empty mempools. */ - if (mp->size == 0) - return; /* Register mempool only if the first element looks like a mbuf. */ - rte_mempool_obj_iter((void *)mp->elt_va_start, - 1, - mp->header_size + mp->elt_size + mp->trailer_size, - 1, - mp->elt_pa, - mp->pg_num, - mp->pg_shift, - txq_mp2mr_mbuf_check, - &data); - if (data.ret) + if (rte_mempool_obj_iter(mp, txq_mp2mr_mbuf_check, &data) == 0 || + data.ret == -1) return; txq_mp2mr(txq, mp); } @@ -1585,6 +1495,10 @@ mlx4_tx_burst(void *dpdk_txq, struct rte_mbuf **pkts, uint16_t pkts_n) if (likely(elt->buf != NULL)) { struct rte_mbuf *tmp = elt->buf; +#ifndef NDEBUG + /* Poisoning. */ + memset(elt, 0x66, sizeof(*elt)); +#endif /* Faster than rte_pktmbuf_free(). */ do { struct rte_mbuf *next = NEXT(tmp); @@ -2454,6 +2368,7 @@ 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 @@ -2854,6 +2769,8 @@ rxq_cleanup(struct rxq *rxq) } 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, @@ -2889,19 +2806,25 @@ rxq_cq_to_pkt_type(uint32_t flags) if (flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) pkt_type = TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV4_PACKET, RTE_PTYPE_L3_IPV4) | + IBV_EXP_CQ_RX_OUTER_IPV4_PACKET, + RTE_PTYPE_L3_IPV4_EXT_UNKNOWN) | TRANSPOSE(flags, - IBV_EXP_CQ_RX_OUTER_IPV6_PACKET, RTE_PTYPE_L3_IPV6) | + IBV_EXP_CQ_RX_OUTER_IPV6_PACKET, + RTE_PTYPE_L3_IPV6_EXT_UNKNOWN) | TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, RTE_PTYPE_INNER_L3_IPV4) | + IBV_EXP_CQ_RX_IPV4_PACKET, + RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN) | TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, RTE_PTYPE_INNER_L3_IPV6); + IBV_EXP_CQ_RX_IPV6_PACKET, + RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN); else pkt_type = TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV4_PACKET, RTE_PTYPE_L3_IPV4) | + IBV_EXP_CQ_RX_IPV4_PACKET, + RTE_PTYPE_L3_IPV4_EXT_UNKNOWN) | TRANSPOSE(flags, - IBV_EXP_CQ_RX_IPV6_PACKET, RTE_PTYPE_L3_IPV6); + IBV_EXP_CQ_RX_IPV6_PACKET, + RTE_PTYPE_L3_IPV6_EXT_UNKNOWN); return pkt_type; } @@ -2923,25 +2846,20 @@ rxq_cq_to_ol_flags(const struct rxq *rxq, uint32_t flags) if (rxq->csum) ol_flags |= - TRANSPOSE(~flags, + TRANSPOSE(flags, IBV_EXP_CQ_RX_IP_CSUM_OK, - PKT_RX_IP_CKSUM_BAD) | - TRANSPOSE(~flags, + PKT_RX_IP_CKSUM_GOOD) | + TRANSPOSE(flags, IBV_EXP_CQ_RX_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_BAD); - /* - * PKT_RX_IP_CKSUM_BAD and PKT_RX_L4_CKSUM_BAD are used in place - * of PKT_RX_EIP_CKSUM_BAD because the latter is not functional - * (its value is 0). - */ + PKT_RX_L4_CKSUM_GOOD); if ((flags & IBV_EXP_CQ_RX_TUNNEL_PACKET) && (rxq->csum_l2tun)) ol_flags |= - TRANSPOSE(~flags, + TRANSPOSE(flags, IBV_EXP_CQ_RX_OUTER_IP_CSUM_OK, - PKT_RX_IP_CKSUM_BAD) | - TRANSPOSE(~flags, + PKT_RX_IP_CKSUM_GOOD) | + TRANSPOSE(flags, IBV_EXP_CQ_RX_OUTER_TCP_UDP_CSUM_OK, - PKT_RX_L4_CKSUM_BAD); + PKT_RX_L4_CKSUM_GOOD); return ol_flags; } @@ -3057,7 +2975,7 @@ mlx4_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) * cacheline while allocating rep. */ rte_prefetch0(seg); - rep = __rte_mbuf_raw_alloc(rxq->mp); + rep = rte_mbuf_raw_alloc(rxq->mp); if (unlikely(rep == NULL)) { /* * Unable to allocate a replacement mbuf, @@ -3086,7 +3004,6 @@ mlx4_rx_burst_sp(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) rep->ol_flags = -1; #endif assert(rep->buf_len == seg->buf_len); - assert(rep->buf_len == rxq->mb_len); /* Reconfigure sge to use rep instead of seg. */ assert(sge->lkey == rxq->mr->lkey); sge->addr = ((uintptr_t)rep->buf_addr + seg_headroom); @@ -3217,8 +3134,8 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) * Fetch initial bytes of packet descriptor into a * cacheline while allocating rep. */ - rte_prefetch0(seg); - rte_prefetch0(&seg->cacheline1); + rte_mbuf_prefetch_part1(seg); + rte_mbuf_prefetch_part2(seg); ret = rxq->if_cq->poll_length_flags(rxq->cq, NULL, NULL, &flags); if (unlikely(ret < 0)) { @@ -3256,7 +3173,7 @@ mlx4_rx_burst(void *dpdk_rxq, struct rte_mbuf **pkts, uint16_t pkts_n) if (ret == 0) break; len = ret; - rep = __rte_mbuf_raw_alloc(rxq->mp); + rep = rte_mbuf_raw_alloc(rxq->mp); if (unlikely(rep == NULL)) { /* * Unable to allocate a replacement mbuf, @@ -3268,6 +3185,8 @@ 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; } @@ -3468,7 +3387,8 @@ rxq_setup_qp_rss(struct priv *priv, struct ibv_cq *cq, uint16_t desc, 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 = priv->rxqs_n; + 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; @@ -3506,6 +3426,7 @@ rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq) 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); @@ -3514,6 +3435,7 @@ rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq) (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)); @@ -3528,9 +3450,10 @@ rxq_rehash(struct rte_eth_dev *dev, struct rxq *rxq) rxq->csum_l2tun = tmpl.csum_l2tun; } /* Enable scattered packets support for this queue if necessary. */ - if ((dev->data->dev_conf.rxmode.jumbo_frame) && + assert(mb_len >= RTE_PKTMBUF_HEADROOM); + if (dev->data->dev_conf.rxmode.enable_scatter && (dev->data->dev_conf.rxmode.max_rx_pkt_len > - (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) { + (mb_len - RTE_PKTMBUF_HEADROOM))) { tmpl.sp = 1; desc_n /= MLX4_PMD_SGE_WR_N; } else @@ -3691,6 +3614,9 @@ 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 @@ -3701,7 +3627,7 @@ skip_rtr: */ static int rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, - unsigned int socket, const struct rte_eth_rxconf *conf, + unsigned int socket, int inactive, const struct rte_eth_rxconf *conf, struct rte_mempool *mp) { struct priv *priv = dev->data->dev_private; @@ -3718,7 +3644,7 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, } attr; enum ibv_exp_query_intf_status status; struct ibv_recv_wr *bad_wr; - struct rte_mbuf *buf; + unsigned int mb_len; int ret = 0; int parent = (rxq == &priv->rxq_parent); @@ -3734,33 +3660,32 @@ rxq_setup(struct rte_eth_dev *dev, struct rxq *rxq, uint16_t desc, 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); return EINVAL; } - /* Get mbuf length. */ - buf = rte_pktmbuf_alloc(mp); - if (buf == NULL) { - ERROR("%p: unable to allocate mbuf", (void *)dev); - return ENOMEM; - } - tmpl.mb_len = buf->buf_len; - assert((rte_pktmbuf_headroom(buf) + - rte_pktmbuf_tailroom(buf)) == tmpl.mb_len); - assert(rte_pktmbuf_headroom(buf) == RTE_PKTMBUF_HEADROOM); - rte_pktmbuf_free(buf); /* 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. */ - if ((dev->data->dev_conf.rxmode.jumbo_frame) && - (dev->data->dev_conf.rxmode.max_rx_pkt_len > - (tmpl.mb_len - RTE_PKTMBUF_HEADROOM))) { + assert(mb_len >= RTE_PKTMBUF_HEADROOM); + if (dev->data->dev_conf.rxmode.max_rx_pkt_len <= + (mb_len - RTE_PKTMBUF_HEADROOM)) { + tmpl.sp = 0; + } else if (dev->data->dev_conf.rxmode.enable_scatter) { tmpl.sp = 1; desc /= MLX4_PMD_SGE_WR_N; + } 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); @@ -3786,11 +3711,22 @@ 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", @@ -3802,7 +3738,7 @@ skip_mr: DEBUG("priv->device_attr.max_sge is %d", priv->device_attr.max_sge); #ifdef RSS_SUPPORT - if (priv->rss) + if (priv->rss && !inactive) tmpl.qp = rxq_setup_qp_rss(priv, tmpl.cq, desc, parent, tmpl.rd); else @@ -3938,6 +3874,7 @@ 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()) @@ -3969,7 +3906,9 @@ mlx4_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t desc, return -ENOMEM; } } - ret = rxq_setup(dev, rxq, desc, socket, conf, mp); + if (idx >= rte_align32pow2(priv->rxqs_n + 1) >> 1) + inactive = 1; + ret = rxq_setup(dev, rxq, desc, socket, inactive, conf, mp); if (ret) rte_free(rxq); else { @@ -4019,9 +3958,15 @@ 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. * @@ -4040,6 +3985,7 @@ mlx4_dev_start(struct rte_eth_dev *dev) unsigned int i = 0; unsigned int r; struct rxq *rxq; + int ret; if (mlx4_is_secondary()) return -E_RTE_SECONDARY; @@ -4059,8 +4005,6 @@ mlx4_dev_start(struct rte_eth_dev *dev) } /* Iterate only once when RSS is enabled. */ do { - int ret; - /* Ignore nonexistent RX queues. */ if (rxq == NULL) continue; @@ -4073,22 +4017,47 @@ mlx4_dev_start(struct rte_eth_dev *dev) 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; + goto err; } while ((--r) && ((rxq = (*priv->rxqs)[++i]), i)); - priv_dev_interrupt_handler_install(priv, dev); + 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; + } priv_unlock(priv); return 0; +err: + /* 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; } /** @@ -4123,6 +4092,7 @@ mlx4_dev_stop(struct rte_eth_dev *dev) rxq = (*priv->rxqs)[0]; r = priv->rxqs_n; } + mlx4_priv_flow_stop(priv); /* Iterate only once when RSS is enabled. */ do { /* Ignore nonexistent RX queues. */ @@ -4185,9 +4155,16 @@ 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. * @@ -4250,11 +4227,97 @@ mlx4_dev_close(struct rte_eth_dev *dev) 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)); } +/** + * Change the link state (UP / DOWN). + * + * @param priv + * Pointer to Ethernet device private data. + * @param up + * Nonzero for link up, otherwise link down. + * + * @return + * 0 on success, errno value on failure. + */ +static int +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; + } else { + err = priv_set_flags(priv, ~IFF_UP, ~IFF_UP); + if (err) + return err; + dev->rx_pkt_burst = removed_rx_burst; + dev->tx_pkt_burst = removed_tx_burst; + } + return 0; +} + +/** + * DPDK callback to bring the link DOWN. + * + * @param dev + * Pointer to Ethernet device structure. + * + * @return + * 0 on success, errno value on failure. + */ +static int +mlx4_set_link_down(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + int err; + + priv_lock(priv); + err = priv_set_link(priv, 0); + priv_unlock(priv); + return err; +} + +/** + * DPDK callback to bring the link UP. + * + * @param dev + * Pointer to Ethernet device structure. + * + * @return + * 0 on success, errno value on failure. + */ +static int +mlx4_set_link_up(struct rte_eth_dev *dev) +{ + struct priv *priv = dev->data->dev_private; + int err; + + priv_lock(priv); + err = priv_set_link(priv, 1); + priv_unlock(priv); + return err; +} /** * DPDK callback to get information about the device. * @@ -4270,6 +4333,8 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) unsigned int max; char ifname[IF_NAMESIZE]; + info->pci_dev = RTE_ETH_DEV_TO_PCI(dev); + if (priv == NULL) return; priv_lock(priv); @@ -4303,6 +4368,12 @@ mlx4_dev_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *info) 0); if (priv_get_ifname(priv, &ifname) == 0) info->if_index = if_nametoindex(ifname); + info->speed_capa = + ETH_LINK_SPEED_1G | + ETH_LINK_SPEED_10G | + ETH_LINK_SPEED_20G | + ETH_LINK_SPEED_40G | + ETH_LINK_SPEED_56G; priv_unlock(priv); } @@ -4465,26 +4536,30 @@ end: * @param vmdq * VMDq pool index to associate address with (ignored). */ -static void +static int 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; + int re; if (mlx4_is_secondary()) - return; + return -ENOTSUP; (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)) + if (index >= (elemof(priv->mac) - 1)) { + re = EINVAL; goto end; - priv_mac_addr_add(priv, index, - (const uint8_t (*)[ETHER_ADDR_LEN]) - mac_addr->addr_bytes); + } + re = priv_mac_addr_add(priv, index, + (const uint8_t (*)[ETHER_ADDR_LEN]) + mac_addr->addr_bytes); end: priv_unlock(priv); + return -re; } /** @@ -4664,7 +4739,7 @@ end: } /** - * DPDK callback to retrieve physical link information (unlocked version). + * DPDK callback to retrieve physical link information. * * @param dev * Pointer to Ethernet device structure. @@ -4672,9 +4747,9 @@ 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 = mlx4_get_priv(dev); struct ethtool_cmd edata = { .cmd = ETHTOOL_GSET }; @@ -4682,6 +4757,8 @@ 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; @@ -4692,7 +4769,7 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) memset(&dev_link, 0, sizeof(dev_link)); dev_link.link_status = ((ifr.ifr_flags & IFF_UP) && (ifr.ifr_flags & IFF_RUNNING)); - ifr.ifr_data = &edata; + ifr.ifr_data = (void *)&edata; if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { WARN("ioctl(SIOCETHTOOL, ETHTOOL_GSET) failed: %s", strerror(errno)); @@ -4705,6 +4782,8 @@ mlx4_link_update_unlocked(struct rte_eth_dev *dev, int wait_to_complete) dev_link.link_speed = link_speed; dev_link.link_duplex = ((edata.duplex == DUPLEX_HALF) ? ETH_LINK_HALF_DUPLEX : ETH_LINK_FULL_DUPLEX); + dev_link.link_autoneg = !(dev->data->dev_conf.link_speeds & + ETH_LINK_SPEED_FIXED); if (memcmp(&dev_link, &dev->data->dev_link, sizeof(dev_link))) { /* Link status changed. */ dev->data->dev_link = dev_link; @@ -4714,27 +4793,9 @@ 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; -} +mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, + struct rte_pci_addr *pci_addr); /** * DPDK callback to change the MTU. @@ -4785,17 +4846,15 @@ mlx4_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu) for (i = 0; (i != priv->rxqs_n); ++i) { struct rxq *rxq = (*priv->rxqs)[i]; unsigned int max_frame_len; - int sp; if (rxq == NULL) continue; - /* Calculate new maximum frame length according to MTU and - * toggle scattered support (sp) if necessary. */ + /* Calculate new maximum frame length according to MTU. */ max_frame_len = (priv->mtu + ETHER_HDR_LEN + (ETHER_MAX_VLAN_FRAME_LEN - ETHER_MAX_LEN)); - sp = (max_frame_len > (rxq->mb_len - RTE_PKTMBUF_HEADROOM)); /* Provide new values to rxq_setup(). */ - dev->data->dev_conf.rxmode.jumbo_frame = sp; + dev->data->dev_conf.rxmode.jumbo_frame = + (max_frame_len > ETHER_MAX_LEN); dev->data->dev_conf.rxmode.max_rx_pkt_len = max_frame_len; ret = rxq_rehash(dev, rxq); if (ret) { @@ -4849,7 +4908,7 @@ mlx4_dev_get_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) if (mlx4_is_secondary()) return -E_RTE_SECONDARY; - ifr.ifr_data = ðpause; + ifr.ifr_data = (void *)ðpause; priv_lock(priv); if (priv_ifreq(priv, SIOCETHTOOL, &ifr)) { ret = errno; @@ -4899,7 +4958,7 @@ mlx4_dev_set_flow_ctrl(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf) if (mlx4_is_secondary()) return -E_RTE_SECONDARY; - ifr.ifr_data = ðpause; + ifr.ifr_data = (void *)ðpause; ethpause.autoneg = fc_conf->autoneg; if (((fc_conf->mode & RTE_FC_FULL) == RTE_FC_FULL) || (fc_conf->mode & RTE_FC_RX_PAUSE)) @@ -5047,16 +5106,67 @@ mlx4_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on) return -ret; } -static const struct eth_dev_ops mlx4_dev_ops = { - .dev_configure = mlx4_dev_configure, - .dev_start = mlx4_dev_start, - .dev_stop = mlx4_dev_stop, - .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, +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, +}; + +/** + * Manage filter operations. + * + * @param dev + * Pointer to Ethernet device structure. + * @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_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; + case RTE_ETH_FILTER_FDIR: + DEBUG("%p: filter type FDIR is not supported by this PMD", + (void *)dev); + break; + default: + ERROR("%p: filter type (%d) not supported", + (void *)dev, filter_type); + break; + } + return -ret; +} + +static const struct eth_dev_ops mlx4_dev_ops = { + .dev_configure = mlx4_dev_configure, + .dev_start = mlx4_dev_start, + .dev_stop = mlx4_dev_stop, + .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, @@ -5079,6 +5189,9 @@ static const struct eth_dev_ops mlx4_dev_ops = { .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, }; /** @@ -5118,7 +5231,7 @@ mlx4_ibv_device_to_pci_addr(const struct ibv_device *device, /* Extract information. */ if (sscanf(line, "PCI_SLOT_NAME=" - "%" SCNx16 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", + "%" SCNx32 ":%" SCNx8 ":%" SCNx8 ".%" SCNx8 "\n", &pci_addr->domain, &pci_addr->bus, &pci_addr->devid, @@ -5209,35 +5322,44 @@ mlx4_getenv_int(const char *name) static void mlx4_dev_link_status_handler(void *); static void -mlx4_dev_interrupt_handler(struct rte_intr_handle *, void *); +mlx4_dev_interrupt_handler(void *); /** - * Link status handler. + * 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 - * Nonzero if the callback process can be called immediately. + * Number of events */ static int -priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev) +priv_dev_status_handler(struct priv *priv, struct rte_eth_dev *dev, + uint32_t *events) { struct ibv_async_event event; int port_change = 0; 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) + if ((event.event_type == IBV_EVENT_PORT_ACTIVE || + event.event_type == IBV_EVENT_PORT_ERR) && + (priv->intr_conf.lsc == 1)) { port_change = 1; - else + 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); @@ -5247,7 +5369,7 @@ priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev) struct rte_eth_link *link = &dev->data->dev_link; priv->pending_alarm = 0; - mlx4_link_update_unlocked(dev, 0); + mlx4_link_update(dev, 0); if (((link->link_speed == 0) && link->link_status) || ((link->link_speed != 0) && !link->link_status)) { /* Inconsistent status, check again later. */ @@ -5255,8 +5377,9 @@ priv_dev_link_status_handler(struct priv *priv, struct rte_eth_dev *dev) rte_eal_alarm_set(MLX4_ALARM_TIMEOUT_US, mlx4_dev_link_status_handler, dev); - } else - ret = 1; + } else { + *events |= (1 << RTE_ETH_EVENT_INTR_LSC); + } } return ret; } @@ -5272,14 +5395,16 @@ 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); - ret = priv_dev_link_status_handler(priv, dev); + ret = priv_dev_status_handler(priv, dev, &events); priv_unlock(priv); - if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); + if (ret > 0 && events & (1 << RTE_ETH_EVENT_INTR_LSC)) + _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL, + NULL); } /** @@ -5291,18 +5416,32 @@ mlx4_dev_link_status_handler(void *arg) * Callback argument. */ static void -mlx4_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) +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; - (void)intr_handle; priv_lock(priv); - ret = priv_dev_link_status_handler(priv, dev); + ret = priv_dev_status_handler(priv, dev, &ev); priv_unlock(priv); - if (ret) - _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC); + 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")); + } } /** @@ -5312,20 +5451,30 @@ mlx4_dev_interrupt_handler(struct rte_intr_handle *intr_handle, void *cb_arg) * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev structure. + * @return + * 0 on success, negative errno value on failure. */ -static void +static int priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) { - if (!dev->data->dev_conf.intr_conf.lsc) - return; - rte_intr_callback_unregister(&priv->intr_handle, - mlx4_dev_interrupt_handler, - dev); - if (priv->pending_alarm) - rte_eal_alarm_cancel(mlx4_dev_link_status_handler, dev); - priv->pending_alarm = 0; + 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 = 0; + priv->intr_handle.type = RTE_INTR_HANDLE_UNKNOWN; + return ret; } /** @@ -5335,30 +5484,377 @@ priv_dev_interrupt_handler_uninstall(struct priv *priv, struct rte_eth_dev *dev) * Pointer to private structure. * @param dev * Pointer to the rte_eth_dev 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 +priv_dev_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) { - int rc, flags; + int flags; + int rc; - if (!dev->data->dev_conf.intr_conf.lsc) - return; + /* 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; - rte_intr_callback_register(&priv->intr_handle, - mlx4_dev_interrupt_handler, - dev); + 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 +priv_dev_removal_interrupt_handler_uninstall(struct priv *priv, + struct rte_eth_dev *dev) +{ + if (dev->data->dev_conf.intr_conf.rmv) { + priv->intr_conf.rmv = 0; + return priv_dev_interrupt_handler_uninstall(priv, dev); + } + 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 +priv_dev_link_interrupt_handler_uninstall(struct priv *priv, + struct rte_eth_dev *dev) +{ + int ret = 0; + + 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; +} + +/** + * Install link 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 +priv_dev_link_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) +{ + int ret; + + 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 0; +} + +/** + * Install removal 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 +priv_dev_removal_interrupt_handler_install(struct priv *priv, + struct rte_eth_dev *dev) +{ + int ret; + + 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; +} + +/** + * Allocate queue vector and fill epoll fd list for Rx interrupts. + * + * @param priv + * Pointer to private structure. + * + * @return + * 0 on success, negative on failure. + */ +static int +priv_rx_intr_vec_enable(struct priv *priv) +{ + 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 (!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; + } + 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++; + } + if (!count) + priv_rx_intr_vec_disable(priv); + else + intr_handle->nb_efd = count; + return 0; +} + +/** + * Clean up Rx interrupts handler. + * + * @param priv + * Pointer to private structure. + */ +static void +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 priv *priv = dev->data->dev_private; + struct rxq *rxq = (*priv->rxqs)[idx]; + int ret; + + if (!rxq || !rxq->channel) + ret = EINVAL; + else + ret = ibv_req_notify_cq(rxq->cq, 0); + if (ret) + WARN("unable to arm interrupt on rx queue %d", idx); + return -ret; +} + +/** + * DPDK callback for Rx queue interrupt disable. + * + * @param dev + * Pointer to Ethernet device structure. + * @param idx + * Rx queue index. + * + * @return + * 0 on success, negative on failure. + */ +static int +mlx4_rx_intr_disable(struct rte_eth_dev *dev, uint16_t idx) +{ + struct priv *priv = dev->data->dev_private; + struct rxq *rxq = (*priv->rxqs)[idx]; + struct ibv_cq *ev_cq; + void *ev_ctx; + int ret; + + 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) + WARN("unable to disable interrupt on rx queue %d", + idx); + else + ibv_ack_cq_events(rxq->cq, 1); + return -ret; +} + +/** + * Verify and store value for device argument. + * + * @param[in] key + * Key argument to verify. + * @param[in] val + * Value associated with key. + * @param out + * User data. + * + * @return + * 0 on success, negative errno value on failure. + */ +static int +mlx4_arg_parse(const char *key, const char *val, void *out) +{ + struct mlx4_conf *conf = out; + 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) { + if (tmp >= MLX4_PMD_MAX_PHYS_PORTS) { + ERROR("invalid port index %lu (max: %u)", + tmp, MLX4_PMD_MAX_PHYS_PORTS - 1); + return -EINVAL; + } + conf->active_ports |= 1 << tmp; + } else { + WARN("%s: unknown parameter", key); + return -EINVAL; + } + return 0; +} + +/** + * Parse device parameters. + * + * @param devargs + * Device arguments structure. + * + * @return + * 0 on success, negative errno value on failure. + */ +static int +mlx4_args(struct rte_devargs *devargs, struct mlx4_conf *conf) +{ + struct rte_kvargs *kvlist; + unsigned int arg_count; + int ret = 0; + int i; + + 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, + 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. @@ -5375,19 +5871,22 @@ static struct eth_driver mlx4_driver; * 0 on success, negative errno value on failure. */ static int -mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) +mlx4_pci_probe(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) { struct ibv_device **list; struct ibv_device *ibv_dev; int err = 0; struct ibv_context *attr_ctx = NULL; struct ibv_device_attr device_attr; + struct mlx4_conf conf = { + .active_ports = 0, + }; unsigned int vf; int idx; int i; (void)pci_drv; - assert(pci_drv == &mlx4_driver.pci_drv); + assert(pci_drv == &mlx4_driver); /* Get mlx4_dev[] index. */ idx = mlx4_dev_idx(&pci_dev->addr); if (idx == -1) { @@ -5401,10 +5900,8 @@ mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) 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); @@ -5436,11 +5933,11 @@ mlx4_pci_devinit(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; @@ -5452,6 +5949,15 @@ mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) goto error; INFO("%u port(s) detected", device_attr.phys_port_cnt); + if (mlx4_args(pci_dev->device.devargs, &conf)) { + ERROR("failed to process device arguments"); + goto error; + } + /* Use all ports when none are defined */ + if (conf.active_ports == 0) { + for (i = 0; i < MLX4_PMD_MAX_PHYS_PORTS; i++) + conf.active_ports |= 1 << i; + } 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); @@ -5465,6 +5971,9 @@ mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) #endif /* HAVE_EXP_QUERY_DEVICE */ struct ether_addr mac; + /* If port is not active, skip. */ + if (!(conf.active_ports & (1 << i))) + continue; #ifdef HAVE_EXP_QUERY_DEVICE exp_device_attr.comp_mask = IBV_EXP_DEVICE_ATTR_EXP_CAP_FLAGS; #ifdef RSS_SUPPORT @@ -5634,7 +6143,7 @@ mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) snprintf(name, sizeof(name), "%s port %u", ibv_get_device_name(ibv_dev), port); - eth_dev = rte_eth_dev_allocate(name, RTE_ETH_DEV_PCI); + eth_dev = rte_eth_dev_allocate(name); } if (eth_dev == NULL) { ERROR("can not allocate rte ethdev"); @@ -5670,23 +6179,32 @@ mlx4_pci_devinit(struct rte_pci_driver *pci_drv, struct rte_pci_device *pci_dev) 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->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); /* 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: @@ -5724,47 +6242,39 @@ error: static const struct rte_pci_id mlx4_pci_id_map[] = { { - .vendor_id = PCI_VENDOR_ID_MELLANOX, - .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3, - .subsystem_vendor_id = PCI_ANY_ID, - .subsystem_device_id = PCI_ANY_ID + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX3) }, { - .vendor_id = PCI_VENDOR_ID_MELLANOX, - .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3PRO, - .subsystem_vendor_id = PCI_ANY_ID, - .subsystem_device_id = PCI_ANY_ID + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX3PRO) }, { - .vendor_id = PCI_VENDOR_ID_MELLANOX, - .device_id = PCI_DEVICE_ID_MELLANOX_CONNECTX3VF, - .subsystem_vendor_id = PCI_ANY_ID, - .subsystem_device_id = PCI_ANY_ID + RTE_PCI_DEVICE(PCI_VENDOR_ID_MELLANOX, + PCI_DEVICE_ID_MELLANOX_CONNECTX3VF) }, { .vendor_id = 0 } }; -static struct eth_driver mlx4_driver = { - .pci_drv = { - .name = MLX4_DRIVER_NAME, - .id_table = mlx4_pci_id_map, - .devinit = mlx4_pci_devinit, - .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, }; /** * Driver initialization routine. */ -static int -rte_mlx4_pmd_init(const char *name, const char *args) +RTE_INIT(rte_mlx4_pmd_init); +static void +rte_mlx4_pmd_init(void) { - (void)name; - (void)args; - RTE_BUILD_BUG_ON(sizeof(wr_id_t) != sizeof(uint64_t)); /* * RDMAV_HUGEPAGES_SAFE tells ibv_fork_init() we intend to use @@ -5774,14 +6284,10 @@ rte_mlx4_pmd_init(const char *name, const char *args) */ setenv("RDMAV_HUGEPAGES_SAFE", "1", 1); ibv_fork_init(); - rte_eal_pci_register(&mlx4_driver.pci_drv); - return 0; + rte_pci_register(&mlx4_driver); } -static struct rte_driver rte_mlx4_driver = { - .type = PMD_PDEV, - .name = MLX4_DRIVER_NAME, - .init = rte_mlx4_pmd_init, -}; - -PMD_REGISTER_DRIVER(rte_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");