diff --git a/MAINTAINERS b/MAINTAINERS index 79152ae0ff..fd2f34d903 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -434,6 +434,7 @@ Intel avf M: Jingjing Wu M: Wenzhuo Lu F: drivers/net/avf/ +F: doc/guides/nics/features/avf*.ini Mellanox mlx4 M: Adrien Mazarguil diff --git a/config/common_base b/config/common_base index cff7eba8b6..f3007a1117 100644 --- a/config/common_base +++ b/config/common_base @@ -221,6 +221,10 @@ CONFIG_RTE_LIBRTE_FM10K_INC_VECTOR=y # Compile burst-oriented AVF PMD driver # CONFIG_RTE_LIBRTE_AVF_PMD=y +CONFIG_RTE_LIBRTE_AVF_DEBUG_TX=n +CONFIG_RTE_LIBRTE_AVF_DEBUG_TX_FREE=n +CONFIG_RTE_LIBRTE_AVF_DEBUG_RX=n +CONFIG_RTE_LIBRTE_AVF_16BYTE_RX_DESC=n # # Compile burst-oriented Mellanox ConnectX-3 (MLX4) PMD diff --git a/doc/guides/nics/features/avf.ini b/doc/guides/nics/features/avf.ini new file mode 100644 index 0000000000..8a294e9dae --- /dev/null +++ b/doc/guides/nics/features/avf.ini @@ -0,0 +1,22 @@ +; +; Supported features of the 'avf' network poll mode driver. +; +; Refer to default.ini for the full list of available PMD features. +; +[Features] +Queue start/stop = Y +Jumbo frame = Y +Scattered Rx = Y +TSO = Y +RSS hash = Y +CRC offload = Y +VLAN offload = Y +L3 checksum offload = Y +L4 checksum offload = Y +Packet type parsing = Y +Multiprocess aware = Y +BSD nic_uio = Y +Linux UIO = Y +Linux VFIO = Y +x86-32 = Y +x86-64 = Y diff --git a/drivers/net/avf/Makefile b/drivers/net/avf/Makefile index e172bf5cee..8d54fc9e2d 100644 --- a/drivers/net/avf/Makefile +++ b/drivers/net/avf/Makefile @@ -13,6 +13,9 @@ LDLIBS += -lrte_eal -lrte_mbuf -lrte_mempool -lrte_ring LDLIBS += -lrte_ethdev -lrte_net -lrte_kvargs -lrte_hash LDLIBS += -lrte_bus_pci +# used to dump HW descriptor for debugging +# CFLAGS += -DDEBUG_DUMP_DESC + EXPORT_MAP := rte_pmd_avf_version.map LIBABIVER := 1 diff --git a/drivers/net/avf/avf_ethdev.c b/drivers/net/avf/avf_ethdev.c index e0ee05a6c7..4597de17a9 100644 --- a/drivers/net/avf/avf_ethdev.c +++ b/drivers/net/avf/avf_ethdev.c @@ -39,6 +39,7 @@ static void avf_dev_stop(struct rte_eth_dev *dev); static void avf_dev_close(struct rte_eth_dev *dev); static void avf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info); +static const uint32_t *avf_dev_supported_ptypes_get(struct rte_eth_dev *dev); int avf_logtype_init; int avf_logtype_driver; @@ -53,6 +54,7 @@ static const struct eth_dev_ops avf_eth_dev_ops = { .dev_stop = avf_dev_stop, .dev_close = avf_dev_close, .dev_infos_get = avf_dev_info_get, + .dev_supported_ptypes_get = avf_dev_supported_ptypes_get, .rx_queue_start = avf_dev_rx_queue_start, .rx_queue_stop = avf_dev_rx_queue_stop, .tx_queue_start = avf_dev_tx_queue_start, @@ -204,9 +206,12 @@ avf_init_queues(struct rte_eth_dev *dev) if (ret != AVF_SUCCESS) break; } - /* TODO: set rx/tx function to vector/scatter/single-segment + /* set rx/tx function to vector/scatter/single-segment * according to parameters */ + avf_set_rx_function(dev); + avf_set_tx_function(dev); + return ret; } @@ -407,6 +412,23 @@ avf_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) }; } +static const uint32_t * +avf_dev_supported_ptypes_get(struct rte_eth_dev *dev) +{ + static const uint32_t ptypes[] = { + RTE_PTYPE_L2_ETHER, + RTE_PTYPE_L3_IPV4_EXT_UNKNOWN, + RTE_PTYPE_L4_FRAG, + RTE_PTYPE_L4_ICMP, + RTE_PTYPE_L4_NONFRAG, + RTE_PTYPE_L4_SCTP, + RTE_PTYPE_L4_TCP, + RTE_PTYPE_L4_UDP, + RTE_PTYPE_UNKNOWN + }; + return ptypes; +} + static int avf_check_vf_reset_done(struct avf_hw *hw) { @@ -556,7 +578,19 @@ avf_dev_init(struct rte_eth_dev *eth_dev) /* assign ops func pointer */ eth_dev->dev_ops = &avf_eth_dev_ops; + eth_dev->rx_pkt_burst = &avf_recv_pkts; + eth_dev->tx_pkt_burst = &avf_xmit_pkts; + eth_dev->tx_pkt_prepare = &avf_prep_pkts; + /* For secondary processes, we don't initialise any further as primary + * has already done this work. Only check if we need a different RX + * and TX function. + */ + if (rte_eal_process_type() != RTE_PROC_PRIMARY) { + avf_set_rx_function(eth_dev); + avf_set_tx_function(eth_dev); + return 0; + } rte_eth_copy_pci_info(eth_dev, pci_dev); hw->vendor_id = pci_dev->id.vendor_id; diff --git a/drivers/net/avf/avf_log.h b/drivers/net/avf/avf_log.h index e3f106b312..8d574d3f31 100644 --- a/drivers/net/avf/avf_log.h +++ b/drivers/net/avf/avf_log.h @@ -20,4 +20,25 @@ extern int avf_logtype_driver; PMD_DRV_LOG_RAW(level, fmt "\n", ## args) #define PMD_DRV_FUNC_TRACE() PMD_DRV_LOG(DEBUG, " >>") +#ifdef RTE_LIBRTE_AVF_DEBUG_RX +#define PMD_RX_LOG(level, fmt, args...) \ + RTE_LOG_DP(level, PMD, "%s(): " fmt "\n", __func__, ## args) +#else +#define PMD_RX_LOG(level, fmt, args...) do { } while (0) +#endif + +#ifdef RTE_LIBRTE_AVF_DEBUG_TX +#define PMD_TX_LOG(level, fmt, args...) \ + RTE_LOG_DP(level, PMD, "%s(): " fmt "\n", __func__, ## args) +#else +#define PMD_TX_LOG(level, fmt, args...) do { } while (0) +#endif + +#ifdef RTE_LIBRTE_AVF_DEBUG_TX_FREE +#define PMD_TX_FREE_LOG(level, fmt, args...) \ + RTE_LOG_DP(level, PMD, "%s(): " fmt "\n", __func__, ## args) +#else +#define PMD_TX_FREE_LOG(level, fmt, args...) do { } while (0) +#endif + #endif /* _AVF_LOG_H_ */ diff --git a/drivers/net/avf/avf_rxtx.c b/drivers/net/avf/avf_rxtx.c index 2d4fb4cd13..baccec461b 100644 --- a/drivers/net/avf/avf_rxtx.c +++ b/drivers/net/avf/avf_rxtx.c @@ -34,17 +34,11 @@ static inline int check_rx_thresh(uint16_t nb_desc, uint16_t thresh) { /* The following constraints must be satisfied: - * thresh >= AVF_RX_MAX_BURST * thresh < rxq->nb_rx_desc - * (rxq->nb_rx_desc % thresh) == 0 */ - if (thresh < AVF_RX_MAX_BURST || - thresh >= nb_desc || - (nb_desc % thresh != 0)) { - PMD_INIT_LOG(ERR, "rx_free_thresh (%u) must be less than %u, " - "greater than or equal to %u, " - "and a divisor of %u", - thresh, nb_desc, AVF_RX_MAX_BURST, nb_desc); + if (thresh >= nb_desc) { + PMD_INIT_LOG(ERR, "rx_free_thresh (%u) must be less than %u", + thresh, nb_desc); return -EINVAL; } return 0; @@ -614,3 +608,780 @@ avf_stop_queues(struct rte_eth_dev *dev) dev->data->rx_queue_state[i] = RTE_ETH_QUEUE_STATE_STOPPED; } } + +static inline void +avf_rxd_to_vlan_tci(struct rte_mbuf *mb, volatile union avf_rx_desc *rxdp) +{ + if (rte_le_to_cpu_64(rxdp->wb.qword1.status_error_len) & + (1 << AVF_RX_DESC_STATUS_L2TAG1P_SHIFT)) { + mb->ol_flags |= PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED; + mb->vlan_tci = + rte_le_to_cpu_16(rxdp->wb.qword0.lo_dword.l2tag1); + } else { + mb->vlan_tci = 0; + } +} + +/* Translate the rx descriptor status and error fields to pkt flags */ +static inline uint64_t +avf_rxd_to_pkt_flags(uint64_t qword) +{ + uint64_t flags; + uint64_t error_bits = (qword >> AVF_RXD_QW1_ERROR_SHIFT); + +#define AVF_RX_ERR_BITS 0x3f + + /* Check if RSS_HASH */ + flags = (((qword >> AVF_RX_DESC_STATUS_FLTSTAT_SHIFT) & + AVF_RX_DESC_FLTSTAT_RSS_HASH) == + AVF_RX_DESC_FLTSTAT_RSS_HASH) ? PKT_RX_RSS_HASH : 0; + + if (likely((error_bits & AVF_RX_ERR_BITS) == 0)) { + flags |= (PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD); + return flags; + } + + if (unlikely(error_bits & (1 << AVF_RX_DESC_ERROR_IPE_SHIFT))) + flags |= PKT_RX_IP_CKSUM_BAD; + else + flags |= PKT_RX_IP_CKSUM_GOOD; + + if (unlikely(error_bits & (1 << AVF_RX_DESC_ERROR_L4E_SHIFT))) + flags |= PKT_RX_L4_CKSUM_BAD; + else + flags |= PKT_RX_L4_CKSUM_GOOD; + + /* TODO: Oversize error bit is not processed here */ + + return flags; +} + +/* implement recv_pkts */ +uint16_t +avf_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts) +{ + volatile union avf_rx_desc *rx_ring; + volatile union avf_rx_desc *rxdp; + struct avf_rx_queue *rxq; + union avf_rx_desc rxd; + struct rte_mbuf *rxe; + struct rte_eth_dev *dev; + struct rte_mbuf *rxm; + struct rte_mbuf *nmb; + uint16_t nb_rx; + uint32_t rx_status; + uint64_t qword1; + uint16_t rx_packet_len; + uint16_t rx_id, nb_hold; + uint64_t dma_addr; + uint64_t pkt_flags; + static const uint32_t ptype_tbl[UINT8_MAX + 1] __rte_cache_aligned = { + /* [0] reserved */ + [1] = RTE_PTYPE_L2_ETHER, + /* [2] - [21] reserved */ + [22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_FRAG, + [23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_NONFRAG, + [24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_UDP, + /* [25] reserved */ + [26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_TCP, + [27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_SCTP, + [28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_ICMP, + /* All others reserved */ + }; + + nb_rx = 0; + nb_hold = 0; + rxq = rx_queue; + rx_id = rxq->rx_tail; + rx_ring = rxq->rx_ring; + + while (nb_rx < nb_pkts) { + rxdp = &rx_ring[rx_id]; + qword1 = rte_le_to_cpu_64(rxdp->wb.qword1.status_error_len); + rx_status = (qword1 & AVF_RXD_QW1_STATUS_MASK) >> + AVF_RXD_QW1_STATUS_SHIFT; + + /* Check the DD bit first */ + if (!(rx_status & (1 << AVF_RX_DESC_STATUS_DD_SHIFT))) + break; + AVF_DUMP_RX_DESC(rxq, rxdp, rx_id); + + nmb = rte_mbuf_raw_alloc(rxq->mp); + if (unlikely(!nmb)) { + dev = &rte_eth_devices[rxq->port_id]; + dev->data->rx_mbuf_alloc_failed++; + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed port_id=%u " + "queue_id=%u", rxq->port_id, rxq->queue_id); + break; + } + + rxd = *rxdp; + nb_hold++; + rxe = rxq->sw_ring[rx_id]; + rx_id++; + if (unlikely(rx_id == rxq->nb_rx_desc)) + rx_id = 0; + + /* Prefetch next mbuf */ + rte_prefetch0(rxq->sw_ring[rx_id]); + + /* When next RX descriptor is on a cache line boundary, + * prefetch the next 4 RX descriptors and next 8 pointers + * to mbufs. + */ + if ((rx_id & 0x3) == 0) { + rte_prefetch0(&rx_ring[rx_id]); + rte_prefetch0(rxq->sw_ring[rx_id]); + } + rxm = rxe; + rxe = nmb; + dma_addr = + rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb)); + rxdp->read.hdr_addr = 0; + rxdp->read.pkt_addr = dma_addr; + + rx_packet_len = ((qword1 & AVF_RXD_QW1_LENGTH_PBUF_MASK) >> + AVF_RXD_QW1_LENGTH_PBUF_SHIFT) - rxq->crc_len; + + rxm->data_off = RTE_PKTMBUF_HEADROOM; + rte_prefetch0(RTE_PTR_ADD(rxm->buf_addr, RTE_PKTMBUF_HEADROOM)); + rxm->nb_segs = 1; + rxm->next = NULL; + rxm->pkt_len = rx_packet_len; + rxm->data_len = rx_packet_len; + rxm->port = rxq->port_id; + rxm->ol_flags = 0; + avf_rxd_to_vlan_tci(rxm, &rxd); + pkt_flags = avf_rxd_to_pkt_flags(qword1); + rxm->packet_type = + ptype_tbl[(uint8_t)((qword1 & + AVF_RXD_QW1_PTYPE_MASK) >> AVF_RXD_QW1_PTYPE_SHIFT)]; + + if (pkt_flags & PKT_RX_RSS_HASH) + rxm->hash.rss = + rte_le_to_cpu_32(rxd.wb.qword0.hi_dword.rss); + + rxm->ol_flags |= pkt_flags; + + rx_pkts[nb_rx++] = rxm; + } + rxq->rx_tail = rx_id; + + /* If the number of free RX descriptors is greater than the RX free + * threshold of the queue, advance the receive tail register of queue. + * Update that register with the value of the last processed RX + * descriptor minus 1. + */ + nb_hold = (uint16_t)(nb_hold + rxq->nb_rx_hold); + if (nb_hold > rxq->rx_free_thresh) { + PMD_RX_LOG(DEBUG, "port_id=%u queue_id=%u rx_tail=%u " + "nb_hold=%u nb_rx=%u", + rxq->port_id, rxq->queue_id, + rx_id, nb_hold, nb_rx); + rx_id = (uint16_t)((rx_id == 0) ? + (rxq->nb_rx_desc - 1) : (rx_id - 1)); + AVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id); + nb_hold = 0; + } + rxq->nb_rx_hold = nb_hold; + + return nb_rx; +} + +/* implement recv_scattered_pkts */ +uint16_t +avf_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, + uint16_t nb_pkts) +{ + struct avf_rx_queue *rxq = rx_queue; + union avf_rx_desc rxd; + struct rte_mbuf *rxe; + struct rte_mbuf *first_seg = rxq->pkt_first_seg; + struct rte_mbuf *last_seg = rxq->pkt_last_seg; + struct rte_mbuf *nmb, *rxm; + uint16_t rx_id = rxq->rx_tail; + uint16_t nb_rx = 0, nb_hold = 0, rx_packet_len; + struct rte_eth_dev *dev; + uint32_t rx_status; + uint64_t qword1; + uint64_t dma_addr; + uint64_t pkt_flags; + + volatile union avf_rx_desc *rx_ring = rxq->rx_ring; + volatile union avf_rx_desc *rxdp; + static const uint32_t ptype_tbl[UINT8_MAX + 1] __rte_cache_aligned = { + /* [0] reserved */ + [1] = RTE_PTYPE_L2_ETHER, + /* [2] - [21] reserved */ + [22] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_FRAG, + [23] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_NONFRAG, + [24] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_UDP, + /* [25] reserved */ + [26] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_TCP, + [27] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_SCTP, + [28] = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4_EXT_UNKNOWN | + RTE_PTYPE_L4_ICMP, + /* All others reserved */ + }; + + while (nb_rx < nb_pkts) { + rxdp = &rx_ring[rx_id]; + qword1 = rte_le_to_cpu_64(rxdp->wb.qword1.status_error_len); + rx_status = (qword1 & AVF_RXD_QW1_STATUS_MASK) >> + AVF_RXD_QW1_STATUS_SHIFT; + + /* Check the DD bit */ + if (!(rx_status & (1 << AVF_RX_DESC_STATUS_DD_SHIFT))) + break; + AVF_DUMP_RX_DESC(rxq, rxdp, rx_id); + + nmb = rte_mbuf_raw_alloc(rxq->mp); + if (unlikely(!nmb)) { + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed port_id=%u " + "queue_id=%u", rxq->port_id, rxq->queue_id); + dev = &rte_eth_devices[rxq->port_id]; + dev->data->rx_mbuf_alloc_failed++; + break; + } + + rxd = *rxdp; + nb_hold++; + rxe = rxq->sw_ring[rx_id]; + rx_id++; + if (rx_id == rxq->nb_rx_desc) + rx_id = 0; + + /* Prefetch next mbuf */ + rte_prefetch0(rxq->sw_ring[rx_id]); + + /* When next RX descriptor is on a cache line boundary, + * prefetch the next 4 RX descriptors and next 8 pointers + * to mbufs. + */ + if ((rx_id & 0x3) == 0) { + rte_prefetch0(&rx_ring[rx_id]); + rte_prefetch0(rxq->sw_ring[rx_id]); + } + + rxm = rxe; + rxe = nmb; + dma_addr = + rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb)); + + /* Set data buffer address and data length of the mbuf */ + rxdp->read.hdr_addr = 0; + rxdp->read.pkt_addr = dma_addr; + rx_packet_len = (qword1 & AVF_RXD_QW1_LENGTH_PBUF_MASK) >> + AVF_RXD_QW1_LENGTH_PBUF_SHIFT; + rxm->data_len = rx_packet_len; + rxm->data_off = RTE_PKTMBUF_HEADROOM; + + /* If this is the first buffer of the received packet, set the + * pointer to the first mbuf of the packet and initialize its + * context. Otherwise, update the total length and the number + * of segments of the current scattered packet, and update the + * pointer to the last mbuf of the current packet. + */ + if (!first_seg) { + first_seg = rxm; + first_seg->nb_segs = 1; + first_seg->pkt_len = rx_packet_len; + } else { + first_seg->pkt_len = + (uint16_t)(first_seg->pkt_len + + rx_packet_len); + first_seg->nb_segs++; + last_seg->next = rxm; + } + + /* If this is not the last buffer of the received packet, + * update the pointer to the last mbuf of the current scattered + * packet and continue to parse the RX ring. + */ + if (!(rx_status & (1 << AVF_RX_DESC_STATUS_EOF_SHIFT))) { + last_seg = rxm; + continue; + } + + /* This is the last buffer of the received packet. If the CRC + * is not stripped by the hardware: + * - Subtract the CRC length from the total packet length. + * - If the last buffer only contains the whole CRC or a part + * of it, free the mbuf associated to the last buffer. If part + * of the CRC is also contained in the previous mbuf, subtract + * the length of that CRC part from the data length of the + * previous mbuf. + */ + rxm->next = NULL; + if (unlikely(rxq->crc_len > 0)) { + first_seg->pkt_len -= ETHER_CRC_LEN; + if (rx_packet_len <= ETHER_CRC_LEN) { + rte_pktmbuf_free_seg(rxm); + first_seg->nb_segs--; + last_seg->data_len = + (uint16_t)(last_seg->data_len - + (ETHER_CRC_LEN - rx_packet_len)); + last_seg->next = NULL; + } else + rxm->data_len = (uint16_t)(rx_packet_len - + ETHER_CRC_LEN); + } + + first_seg->port = rxq->port_id; + first_seg->ol_flags = 0; + avf_rxd_to_vlan_tci(first_seg, &rxd); + pkt_flags = avf_rxd_to_pkt_flags(qword1); + first_seg->packet_type = + ptype_tbl[(uint8_t)((qword1 & + AVF_RXD_QW1_PTYPE_MASK) >> AVF_RXD_QW1_PTYPE_SHIFT)]; + + if (pkt_flags & PKT_RX_RSS_HASH) + first_seg->hash.rss = + rte_le_to_cpu_32(rxd.wb.qword0.hi_dword.rss); + + first_seg->ol_flags |= pkt_flags; + + /* Prefetch data of first segment, if configured to do so. */ + rte_prefetch0(RTE_PTR_ADD(first_seg->buf_addr, + first_seg->data_off)); + rx_pkts[nb_rx++] = first_seg; + first_seg = NULL; + } + + /* Record index of the next RX descriptor to probe. */ + rxq->rx_tail = rx_id; + rxq->pkt_first_seg = first_seg; + rxq->pkt_last_seg = last_seg; + + /* If the number of free RX descriptors is greater than the RX free + * threshold of the queue, advance the Receive Descriptor Tail (RDT) + * register. Update the RDT with the value of the last processed RX + * descriptor minus 1, to guarantee that the RDT register is never + * equal to the RDH register, which creates a "full" ring situtation + * from the hardware point of view. + */ + nb_hold = (uint16_t)(nb_hold + rxq->nb_rx_hold); + if (nb_hold > rxq->rx_free_thresh) { + PMD_RX_LOG(DEBUG, "port_id=%u queue_id=%u rx_tail=%u " + "nb_hold=%u nb_rx=%u", + rxq->port_id, rxq->queue_id, + rx_id, nb_hold, nb_rx); + rx_id = (uint16_t)(rx_id == 0 ? + (rxq->nb_rx_desc - 1) : (rx_id - 1)); + AVF_PCI_REG_WRITE(rxq->qrx_tail, rx_id); + nb_hold = 0; + } + rxq->nb_rx_hold = nb_hold; + + return nb_rx; +} + +static inline int +avf_xmit_cleanup(struct avf_tx_queue *txq) +{ + struct avf_tx_entry *sw_ring = txq->sw_ring; + uint16_t last_desc_cleaned = txq->last_desc_cleaned; + uint16_t nb_tx_desc = txq->nb_tx_desc; + uint16_t desc_to_clean_to; + uint16_t nb_tx_to_clean; + + volatile struct avf_tx_desc *txd = txq->tx_ring; + + desc_to_clean_to = (uint16_t)(last_desc_cleaned + txq->rs_thresh); + if (desc_to_clean_to >= nb_tx_desc) + desc_to_clean_to = (uint16_t)(desc_to_clean_to - nb_tx_desc); + + desc_to_clean_to = sw_ring[desc_to_clean_to].last_id; + if ((txd[desc_to_clean_to].cmd_type_offset_bsz & + rte_cpu_to_le_64(AVF_TXD_QW1_DTYPE_MASK)) != + rte_cpu_to_le_64(AVF_TX_DESC_DTYPE_DESC_DONE)) { + PMD_TX_FREE_LOG(DEBUG, "TX descriptor %4u is not done " + "(port=%d queue=%d)", desc_to_clean_to, + txq->port_id, txq->queue_id); + return -1; + } + + if (last_desc_cleaned > desc_to_clean_to) + nb_tx_to_clean = (uint16_t)((nb_tx_desc - last_desc_cleaned) + + desc_to_clean_to); + else + nb_tx_to_clean = (uint16_t)(desc_to_clean_to - + last_desc_cleaned); + + txd[desc_to_clean_to].cmd_type_offset_bsz = 0; + + txq->last_desc_cleaned = desc_to_clean_to; + txq->nb_free = (uint16_t)(txq->nb_free + nb_tx_to_clean); + + return 0; +} + +/* Check if the context descriptor is needed for TX offloading */ +static inline uint16_t +avf_calc_context_desc(uint64_t flags) +{ + static uint64_t mask = PKT_TX_TCP_SEG; + + return (flags & mask) ? 1 : 0; +} + +static inline void +avf_txd_enable_checksum(uint64_t ol_flags, + uint32_t *td_cmd, + uint32_t *td_offset, + union avf_tx_offload tx_offload) +{ + /* Set MACLEN */ + *td_offset |= (tx_offload.l2_len >> 1) << + AVF_TX_DESC_LENGTH_MACLEN_SHIFT; + + /* Enable L3 checksum offloads */ + if (ol_flags & PKT_TX_IP_CKSUM) { + *td_cmd |= AVF_TX_DESC_CMD_IIPT_IPV4_CSUM; + *td_offset |= (tx_offload.l3_len >> 2) << + AVF_TX_DESC_LENGTH_IPLEN_SHIFT; + } else if (ol_flags & PKT_TX_IPV4) { + *td_cmd |= AVF_TX_DESC_CMD_IIPT_IPV4; + *td_offset |= (tx_offload.l3_len >> 2) << + AVF_TX_DESC_LENGTH_IPLEN_SHIFT; + } else if (ol_flags & PKT_TX_IPV6) { + *td_cmd |= AVF_TX_DESC_CMD_IIPT_IPV6; + *td_offset |= (tx_offload.l3_len >> 2) << + AVF_TX_DESC_LENGTH_IPLEN_SHIFT; + } + + if (ol_flags & PKT_TX_TCP_SEG) { + *td_cmd |= AVF_TX_DESC_CMD_L4T_EOFT_TCP; + *td_offset |= (tx_offload.l4_len >> 2) << + AVF_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; + return; + } + + /* Enable L4 checksum offloads */ + switch (ol_flags & PKT_TX_L4_MASK) { + case PKT_TX_TCP_CKSUM: + *td_cmd |= AVF_TX_DESC_CMD_L4T_EOFT_TCP; + *td_offset |= (sizeof(struct tcp_hdr) >> 2) << + AVF_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; + break; + case PKT_TX_SCTP_CKSUM: + *td_cmd |= AVF_TX_DESC_CMD_L4T_EOFT_SCTP; + *td_offset |= (sizeof(struct sctp_hdr) >> 2) << + AVF_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; + break; + case PKT_TX_UDP_CKSUM: + *td_cmd |= AVF_TX_DESC_CMD_L4T_EOFT_UDP; + *td_offset |= (sizeof(struct udp_hdr) >> 2) << + AVF_TX_DESC_LENGTH_L4_FC_LEN_SHIFT; + break; + default: + break; + } +} + +/* set TSO context descriptor + * support IP -> L4 and IP -> IP -> L4 + */ +static inline uint64_t +avf_set_tso_ctx(struct rte_mbuf *mbuf, union avf_tx_offload tx_offload) +{ + uint64_t ctx_desc = 0; + uint32_t cd_cmd, hdr_len, cd_tso_len; + + if (!tx_offload.l4_len) { + PMD_TX_LOG(DEBUG, "L4 length set to 0"); + return ctx_desc; + } + + /* in case of non tunneling packet, the outer_l2_len and + * outer_l3_len must be 0. + */ + hdr_len = tx_offload.l2_len + + tx_offload.l3_len + + tx_offload.l4_len; + + cd_cmd = AVF_TX_CTX_DESC_TSO; + cd_tso_len = mbuf->pkt_len - hdr_len; + ctx_desc |= ((uint64_t)cd_cmd << AVF_TXD_CTX_QW1_CMD_SHIFT) | + ((uint64_t)cd_tso_len << AVF_TXD_CTX_QW1_TSO_LEN_SHIFT) | + ((uint64_t)mbuf->tso_segsz << AVF_TXD_CTX_QW1_MSS_SHIFT); + + return ctx_desc; +} + +/* Construct the tx flags */ +static inline uint64_t +avf_build_ctob(uint32_t td_cmd, uint32_t td_offset, unsigned int size, + uint32_t td_tag) +{ + return rte_cpu_to_le_64(AVF_TX_DESC_DTYPE_DATA | + ((uint64_t)td_cmd << AVF_TXD_QW1_CMD_SHIFT) | + ((uint64_t)td_offset << + AVF_TXD_QW1_OFFSET_SHIFT) | + ((uint64_t)size << + AVF_TXD_QW1_TX_BUF_SZ_SHIFT) | + ((uint64_t)td_tag << + AVF_TXD_QW1_L2TAG1_SHIFT)); +} + +/* TX function */ +uint16_t +avf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts) +{ + volatile struct avf_tx_desc *txd; + volatile struct avf_tx_desc *txr; + struct avf_tx_queue *txq; + struct avf_tx_entry *sw_ring; + struct avf_tx_entry *txe, *txn; + struct rte_mbuf *tx_pkt; + struct rte_mbuf *m_seg; + uint16_t tx_id; + uint16_t nb_tx; + uint32_t td_cmd; + uint32_t td_offset; + uint32_t td_tag; + uint64_t ol_flags; + uint16_t nb_used; + uint16_t nb_ctx; + uint16_t tx_last; + uint16_t slen; + uint64_t buf_dma_addr; + union avf_tx_offload tx_offload = {0}; + + txq = tx_queue; + sw_ring = txq->sw_ring; + txr = txq->tx_ring; + tx_id = txq->tx_tail; + txe = &sw_ring[tx_id]; + + /* Check if the descriptor ring needs to be cleaned. */ + if (txq->nb_free < txq->free_thresh) + avf_xmit_cleanup(txq); + + for (nb_tx = 0; nb_tx < nb_pkts; nb_tx++) { + td_cmd = 0; + td_tag = 0; + td_offset = 0; + + tx_pkt = *tx_pkts++; + RTE_MBUF_PREFETCH_TO_FREE(txe->mbuf); + + ol_flags = tx_pkt->ol_flags; + tx_offload.l2_len = tx_pkt->l2_len; + tx_offload.l3_len = tx_pkt->l3_len; + tx_offload.l4_len = tx_pkt->l4_len; + tx_offload.tso_segsz = tx_pkt->tso_segsz; + + /* Calculate the number of context descriptors needed. */ + nb_ctx = avf_calc_context_desc(ol_flags); + + /* The number of descriptors that must be allocated for + * a packet equals to the number of the segments of that + * packet plus 1 context descriptor if needed. + */ + nb_used = (uint16_t)(tx_pkt->nb_segs + nb_ctx); + tx_last = (uint16_t)(tx_id + nb_used - 1); + + /* Circular ring */ + if (tx_last >= txq->nb_tx_desc) + tx_last = (uint16_t)(tx_last - txq->nb_tx_desc); + + PMD_TX_LOG(DEBUG, "port_id=%u queue_id=%u" + " tx_first=%u tx_last=%u", + txq->port_id, txq->queue_id, tx_id, tx_last); + + if (nb_used > txq->nb_free) { + if (avf_xmit_cleanup(txq)) { + if (nb_tx == 0) + return 0; + goto end_of_tx; + } + if (unlikely(nb_used > txq->rs_thresh)) { + while (nb_used > txq->nb_free) { + if (avf_xmit_cleanup(txq)) { + if (nb_tx == 0) + return 0; + goto end_of_tx; + } + } + } + } + + /* Descriptor based VLAN insertion */ + if (ol_flags & PKT_TX_VLAN_PKT) { + td_cmd |= AVF_TX_DESC_CMD_IL2TAG1; + td_tag = tx_pkt->vlan_tci; + } + + /* According to datasheet, the bit2 is reserved and must be + * set to 1. + */ + td_cmd |= 0x04; + + /* Enable checksum offloading */ + if (ol_flags & AVF_TX_CKSUM_OFFLOAD_MASK) + avf_txd_enable_checksum(ol_flags, &td_cmd, + &td_offset, tx_offload); + + if (nb_ctx) { + /* Setup TX context descriptor if required */ + volatile struct avf_tx_context_desc *ctx_txd = + (volatile struct avf_tx_context_desc *) + &txr[tx_id]; + uint16_t cd_l2tag2 = 0; + uint64_t cd_type_cmd_tso_mss = + AVF_TX_DESC_DTYPE_CONTEXT; + + txn = &sw_ring[txe->next_id]; + RTE_MBUF_PREFETCH_TO_FREE(txn->mbuf); + if (txe->mbuf) { + rte_pktmbuf_free_seg(txe->mbuf); + txe->mbuf = NULL; + } + + /* TSO enabled */ + if (ol_flags & PKT_TX_TCP_SEG) + cd_type_cmd_tso_mss |= + avf_set_tso_ctx(tx_pkt, tx_offload); + + AVF_DUMP_TX_DESC(txq, ctx_txd, tx_id); + txe->last_id = tx_last; + tx_id = txe->next_id; + txe = txn; + } + + m_seg = tx_pkt; + do { + txd = &txr[tx_id]; + txn = &sw_ring[txe->next_id]; + + if (txe->mbuf) + rte_pktmbuf_free_seg(txe->mbuf); + txe->mbuf = m_seg; + + /* Setup TX Descriptor */ + slen = m_seg->data_len; + buf_dma_addr = rte_mbuf_data_iova(m_seg); + txd->buffer_addr = rte_cpu_to_le_64(buf_dma_addr); + txd->cmd_type_offset_bsz = avf_build_ctob(td_cmd, + td_offset, + slen, + td_tag); + + AVF_DUMP_TX_DESC(txq, txd, tx_id); + txe->last_id = tx_last; + tx_id = txe->next_id; + txe = txn; + m_seg = m_seg->next; + } while (m_seg); + + /* The last packet data descriptor needs End Of Packet (EOP) */ + td_cmd |= AVF_TX_DESC_CMD_EOP; + txq->nb_used = (uint16_t)(txq->nb_used + nb_used); + txq->nb_free = (uint16_t)(txq->nb_free - nb_used); + + if (txq->nb_used >= txq->rs_thresh) { + PMD_TX_LOG(DEBUG, "Setting RS bit on TXD id=" + "%4u (port=%d queue=%d)", + tx_last, txq->port_id, txq->queue_id); + + td_cmd |= AVF_TX_DESC_CMD_RS; + + /* Update txq RS bit counters */ + txq->nb_used = 0; + } + + txd->cmd_type_offset_bsz |= + rte_cpu_to_le_64(((uint64_t)td_cmd) << + AVF_TXD_QW1_CMD_SHIFT); + AVF_DUMP_TX_DESC(txq, txd, tx_id); + } + +end_of_tx: + rte_wmb(); + + PMD_TX_LOG(DEBUG, "port_id=%u queue_id=%u tx_tail=%u nb_tx=%u", + txq->port_id, txq->queue_id, tx_id, nb_tx); + + AVF_PCI_REG_WRITE_RELAXED(txq->qtx_tail, tx_id); + txq->tx_tail = tx_id; + + return nb_tx; +} + +/* TX prep functions */ +uint16_t +avf_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts, + uint16_t nb_pkts) +{ + int i, ret; + uint64_t ol_flags; + struct rte_mbuf *m; + + for (i = 0; i < nb_pkts; i++) { + m = tx_pkts[i]; + ol_flags = m->ol_flags; + + /* Check condition for nb_segs > AVF_TX_MAX_MTU_SEG. */ + if (!(ol_flags & PKT_TX_TCP_SEG)) { + if (m->nb_segs > AVF_TX_MAX_MTU_SEG) { + rte_errno = -EINVAL; + return i; + } + } else if ((m->tso_segsz < AVF_MIN_TSO_MSS) || + (m->tso_segsz > AVF_MAX_TSO_MSS)) { + /* MSS outside the range are considered malicious */ + rte_errno = -EINVAL; + return i; + } + + if (ol_flags & AVF_TX_OFFLOAD_NOTSUP_MASK) { + rte_errno = -ENOTSUP; + return i; + } + +#ifdef RTE_LIBRTE_ETHDEV_DEBUG + ret = rte_validate_tx_offload(m); + if (ret != 0) { + rte_errno = ret; + return i; + } +#endif + ret = rte_net_intel_cksum_prepare(m); + if (ret != 0) { + rte_errno = ret; + return i; + } + } + + return i; +} + +/* choose rx function*/ +void +avf_set_rx_function(struct rte_eth_dev *dev) +{ + if (dev->data->scattered_rx) + dev->rx_pkt_burst = avf_recv_scattered_pkts; + else + dev->rx_pkt_burst = avf_recv_pkts; +} + +/* choose tx function*/ +void +avf_set_tx_function(struct rte_eth_dev *dev) +{ + dev->tx_pkt_burst = avf_xmit_pkts; + dev->tx_pkt_prepare = avf_prep_pkts; +} diff --git a/drivers/net/avf/avf_rxtx.h b/drivers/net/avf/avf_rxtx.h index e227cd1df2..cad240d9b5 100644 --- a/drivers/net/avf/avf_rxtx.h +++ b/drivers/net/avf/avf_rxtx.h @@ -19,6 +19,25 @@ #define DEFAULT_TX_RS_THRESH 32 #define DEFAULT_TX_FREE_THRESH 32 +#define AVF_MIN_TSO_MSS 256 +#define AVF_MAX_TSO_MSS 9668 +#define AVF_TSO_MAX_SEG UINT8_MAX +#define AVF_TX_MAX_MTU_SEG 8 + +#define AVF_TX_CKSUM_OFFLOAD_MASK ( \ + PKT_TX_IP_CKSUM | \ + PKT_TX_L4_MASK | \ + PKT_TX_TCP_SEG) + +#define AVF_TX_OFFLOAD_MASK ( \ + PKT_TX_VLAN_PKT | \ + PKT_TX_IP_CKSUM | \ + PKT_TX_L4_MASK | \ + PKT_TX_TCP_SEG) + +#define AVF_TX_OFFLOAD_NOTSUP_MASK \ + (PKT_TX_OFFLOAD_MASK ^ AVF_TX_OFFLOAD_MASK) + /* HW desc structure, both 16-byte and 32-byte types are supported */ #ifdef RTE_LIBRTE_AVF_16BYTE_RX_DESC #define avf_rx_desc avf_16byte_rx_desc @@ -85,6 +104,18 @@ struct avf_tx_queue { bool tx_deferred_start; /* don't start this queue in dev start */ }; +/* Offload features */ +union avf_tx_offload { + uint64_t data; + struct { + uint64_t l2_len:7; /* L2 (MAC) Header Length. */ + uint64_t l3_len:9; /* L3 (IP) Header Length. */ + uint64_t l4_len:8; /* L4 Header Length. */ + uint64_t tso_segsz:16; /* TCP TSO segment size */ + /* uint64_t unused : 24; */ + }; +}; + int avf_dev_rx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx, uint16_t nb_desc, @@ -105,6 +136,17 @@ int avf_dev_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id); int avf_dev_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id); void avf_dev_tx_queue_release(void *txq); void avf_stop_queues(struct rte_eth_dev *dev); +uint16_t avf_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, + uint16_t nb_pkts); +uint16_t avf_recv_scattered_pkts(void *rx_queue, + struct rte_mbuf **rx_pkts, + uint16_t nb_pkts); +uint16_t avf_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, + uint16_t nb_pkts); +uint16_t avf_prep_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, + uint16_t nb_pkts); +void avf_set_rx_function(struct rte_eth_dev *dev); +void avf_set_tx_function(struct rte_eth_dev *dev); static inline void avf_dump_rx_descriptor(struct avf_rx_queue *rxq, @@ -157,4 +199,15 @@ void avf_dump_tx_descriptor(const struct avf_tx_queue *txq, txq->queue_id, name, tx_id, tx_desc->buffer_addr, tx_desc->cmd_type_offset_bsz); } + +#ifdef DEBUG_DUMP_DESC +#define AVF_DUMP_RX_DESC(rxq, desc, rx_id) \ + avf_dump_rx_descriptor(rxq, desc, rx_id) +#define AVF_DUMP_TX_DESC(txq, desc, tx_id) \ + avf_dump_tx_descriptor(txq, desc, tx_id) +#else +#define AVF_DUMP_RX_DESC(rxq, desc, rx_id) do { } while (0) +#define AVF_DUMP_TX_DESC(txq, desc, tx_id) do { } while (0) +#endif + #endif /* _AVF_RXTX_H_ */