From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 8B0A4A0C48; Mon, 14 Jun 2021 21:17:03 +0200 (CEST) Received: from [217.70.189.124] (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 081BF4067E; Mon, 14 Jun 2021 21:17:03 +0200 (CEST) Received: from shelob.oktetlabs.ru (shelob.oktetlabs.ru [91.220.146.113]) by mails.dpdk.org (Postfix) with ESMTP id DE3CD4067A for ; Mon, 14 Jun 2021 21:17:01 +0200 (CEST) Received: from [192.168.1.71] (unknown [188.170.85.171]) (using TLSv1.3 with cipher TLS_AES_128_GCM_SHA256 (128/128 bits) key-exchange X25519 server-signature RSA-PSS (2048 bits) server-digest SHA256) (No client certificate requested) by shelob.oktetlabs.ru (Postfix) with ESMTPSA id 465447F514; Mon, 14 Jun 2021 22:17:01 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 shelob.oktetlabs.ru 465447F514 DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=oktetlabs.ru; s=default; t=1623698221; bh=dWStVP9lRW809M8XzDackz0xS5t+ZLA1jhK/wu7nWXQ=; h=Subject:To:References:From:Date:In-Reply-To; b=B8k87yMMU5Q/8WKuspeXmMCG+3kWBBgzyc01GtIUL2P3omheflbSKVM5xGIZu/mi2 Upm7W3PaQS1DEhnBg9R+tjrNu4AVl9XbCP2MoyI0EMTgJVrFqxw1+gekPsH3DKVNdy gLM9Vul9sGIf1JLPePClA7w14ClUFNSwUDuSehic= To: Jiawen Wu , dev@dpdk.org References: <20210602094108.1575640-1-jiawenwu@trustnetic.com> <20210602094108.1575640-21-jiawenwu@trustnetic.com> From: Andrew Rybchenko Message-ID: Date: Mon, 14 Jun 2021 22:17:00 +0300 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:78.0) Gecko/20100101 Thunderbird/78.10.0 MIME-Version: 1.0 In-Reply-To: <20210602094108.1575640-21-jiawenwu@trustnetic.com> Content-Type: text/plain; charset=utf-8; format=flowed Content-Language: en-US Content-Transfer-Encoding: 7bit Subject: Re: [dpdk-dev] [PATCH v5 20/24] net/ngbe: support bulk and scatter Rx X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" On 6/2/21 12:41 PM, Jiawen Wu wrote: > Add bulk allocation receive function, and support scattered Rx rely on > Rx offload. The patch should advertise Rx scatter offload, not earlier. Corresponding bits of the patch should be here as well. > Signed-off-by: Jiawen Wu > --- > doc/guides/nics/ngbe.rst | 1 + > drivers/net/ngbe/ngbe_ethdev.c | 15 +- > drivers/net/ngbe/ngbe_ethdev.h | 8 + > drivers/net/ngbe/ngbe_rxtx.c | 583 +++++++++++++++++++++++++++++++++ > drivers/net/ngbe/ngbe_rxtx.h | 2 + > 5 files changed, 607 insertions(+), 2 deletions(-) > > diff --git a/doc/guides/nics/ngbe.rst b/doc/guides/nics/ngbe.rst > index 04fa3e90a8..e999e0b580 100644 > --- a/doc/guides/nics/ngbe.rst > +++ b/doc/guides/nics/ngbe.rst > @@ -14,6 +14,7 @@ Features > - Checksum offload > - Jumbo frames > - Link state information > +- Scattered and gather for RX > > Prerequisites > ------------- > diff --git a/drivers/net/ngbe/ngbe_ethdev.c b/drivers/net/ngbe/ngbe_ethdev.c > index 4dab920caa..260bca0e4f 100644 > --- a/drivers/net/ngbe/ngbe_ethdev.c > +++ b/drivers/net/ngbe/ngbe_ethdev.c > @@ -112,8 +112,16 @@ eth_ngbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused) > eth_dev->rx_pkt_burst = &ngbe_recv_pkts; > eth_dev->tx_pkt_burst = &ngbe_xmit_pkts_simple; > > - if (rte_eal_process_type() != RTE_PROC_PRIMARY) > + /* > + * For secondary processes, we don't initialise any further as primary > + * has already done this work. Only check we don't need a different > + * RX and TX function. RX -> Rx, TX -> Tx > + */ > + if (rte_eal_process_type() != RTE_PROC_PRIMARY) { > + ngbe_set_rx_function(eth_dev); > + > return 0; > + } > > rte_eth_copy_pci_info(eth_dev, pci_dev); > > @@ -359,7 +367,10 @@ ngbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info) > const uint32_t * > ngbe_dev_supported_ptypes_get(struct rte_eth_dev *dev) > { > - if (dev->rx_pkt_burst == ngbe_recv_pkts) > + if (dev->rx_pkt_burst == ngbe_recv_pkts || > + dev->rx_pkt_burst == ngbe_recv_pkts_sc_single_alloc || > + dev->rx_pkt_burst == ngbe_recv_pkts_sc_bulk_alloc || > + dev->rx_pkt_burst == ngbe_recv_pkts_bulk_alloc) I don't understand why 3 flavors of the Rx are added in the single patch. It looks separate features with separate conditions to use. > return ngbe_get_supported_ptypes(); > > return NULL; > diff --git a/drivers/net/ngbe/ngbe_ethdev.h b/drivers/net/ngbe/ngbe_ethdev.h > index c0f8483eca..1e21db5e25 100644 > --- a/drivers/net/ngbe/ngbe_ethdev.h > +++ b/drivers/net/ngbe/ngbe_ethdev.h > @@ -78,6 +78,14 @@ void ngbe_dev_tx_init(struct rte_eth_dev *dev); > uint16_t ngbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, > uint16_t nb_pkts); > > +uint16_t ngbe_recv_pkts_bulk_alloc(void *rx_queue, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts); > + > +uint16_t ngbe_recv_pkts_sc_single_alloc(void *rx_queue, > + struct rte_mbuf **rx_pkts, uint16_t nb_pkts); > +uint16_t ngbe_recv_pkts_sc_bulk_alloc(void *rx_queue, > + struct rte_mbuf **rx_pkts, uint16_t nb_pkts); > + > uint16_t ngbe_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts, > uint16_t nb_pkts); > > diff --git a/drivers/net/ngbe/ngbe_rxtx.c b/drivers/net/ngbe/ngbe_rxtx.c > index 9462da5b7a..f633718237 100644 > --- a/drivers/net/ngbe/ngbe_rxtx.c > +++ b/drivers/net/ngbe/ngbe_rxtx.c > @@ -321,6 +321,257 @@ rx_desc_error_to_pkt_flags(uint32_t rx_status) > return pkt_flags; > } > > +/* > + * LOOK_AHEAD defines how many desc statuses to check beyond the > + * current descriptor. > + * It must be a pound define for optimal performance. > + * Do not change the value of LOOK_AHEAD, as the ngbe_rx_scan_hw_ring > + * function only works with LOOK_AHEAD=8. > + */ > +#define LOOK_AHEAD 8 > +#if (LOOK_AHEAD != 8) > +#error "PMD NGBE: LOOK_AHEAD must be 8\n" > +#endif > +static inline int > +ngbe_rx_scan_hw_ring(struct ngbe_rx_queue *rxq) > +{ > + volatile struct ngbe_rx_desc *rxdp; > + struct ngbe_rx_entry *rxep; > + struct rte_mbuf *mb; > + uint16_t pkt_len; > + uint64_t pkt_flags; > + int nb_dd; > + uint32_t s[LOOK_AHEAD]; > + uint32_t pkt_info[LOOK_AHEAD]; > + int i, j, nb_rx = 0; > + uint32_t status; > + > + /* get references to current descriptor and S/W ring entry */ > + rxdp = &rxq->rx_ring[rxq->rx_tail]; > + rxep = &rxq->sw_ring[rxq->rx_tail]; > + > + status = rxdp->qw1.lo.status; > + /* check to make sure there is at least 1 packet to receive */ > + if (!(status & rte_cpu_to_le_32(NGBE_RXD_STAT_DD))) > + return 0; > + > + /* > + * Scan LOOK_AHEAD descriptors at a time to determine which descriptors > + * reference packets that are ready to be received. > + */ > + for (i = 0; i < RTE_PMD_NGBE_RX_MAX_BURST; > + i += LOOK_AHEAD, rxdp += LOOK_AHEAD, rxep += LOOK_AHEAD) { > + /* Read desc statuses backwards to avoid race condition */ > + for (j = 0; j < LOOK_AHEAD; j++) > + s[j] = rte_le_to_cpu_32(rxdp[j].qw1.lo.status); > + > + rte_atomic_thread_fence(__ATOMIC_ACQUIRE); > + > + /* Compute how many status bits were set */ > + for (nb_dd = 0; nb_dd < LOOK_AHEAD && > + (s[nb_dd] & NGBE_RXD_STAT_DD); nb_dd++) > + ; > + > + for (j = 0; j < nb_dd; j++) > + pkt_info[j] = rte_le_to_cpu_32(rxdp[j].qw0.dw0); > + > + nb_rx += nb_dd; > + > + /* Translate descriptor info to mbuf format */ > + for (j = 0; j < nb_dd; ++j) { > + mb = rxep[j].mbuf; > + pkt_len = rte_le_to_cpu_16(rxdp[j].qw1.hi.len) - > + rxq->crc_len; > + mb->data_len = pkt_len; > + mb->pkt_len = pkt_len; > + mb->vlan_tci = rte_le_to_cpu_16(rxdp[j].qw1.hi.tag); > + > + /* convert descriptor fields to rte mbuf flags */ > + pkt_flags = rx_desc_status_to_pkt_flags(s[j], > + rxq->vlan_flags); > + pkt_flags |= rx_desc_error_to_pkt_flags(s[j]); > + pkt_flags |= > + ngbe_rxd_pkt_info_to_pkt_flags(pkt_info[j]); > + mb->ol_flags = pkt_flags; > + mb->packet_type = > + ngbe_rxd_pkt_info_to_pkt_type(pkt_info[j], > + rxq->pkt_type_mask); > + > + if (likely(pkt_flags & PKT_RX_RSS_HASH)) > + mb->hash.rss = > + rte_le_to_cpu_32(rxdp[j].qw0.dw1); > + } > + > + /* Move mbuf pointers from the S/W ring to the stage */ > + for (j = 0; j < LOOK_AHEAD; ++j) > + rxq->rx_stage[i + j] = rxep[j].mbuf; > + > + /* stop if all requested packets could not be received */ > + if (nb_dd != LOOK_AHEAD) > + break; > + } > + > + /* clear software ring entries so we can cleanup correctly */ > + for (i = 0; i < nb_rx; ++i) > + rxq->sw_ring[rxq->rx_tail + i].mbuf = NULL; > + > + return nb_rx; > +} > + > +static inline int > +ngbe_rx_alloc_bufs(struct ngbe_rx_queue *rxq, bool reset_mbuf) > +{ > + volatile struct ngbe_rx_desc *rxdp; > + struct ngbe_rx_entry *rxep; > + struct rte_mbuf *mb; > + uint16_t alloc_idx; > + __le64 dma_addr; > + int diag, i; > + > + /* allocate buffers in bulk directly into the S/W ring */ > + alloc_idx = rxq->rx_free_trigger - (rxq->rx_free_thresh - 1); > + rxep = &rxq->sw_ring[alloc_idx]; > + diag = rte_mempool_get_bulk(rxq->mb_pool, (void *)rxep, > + rxq->rx_free_thresh); > + if (unlikely(diag != 0)) > + return -ENOMEM; > + > + rxdp = &rxq->rx_ring[alloc_idx]; > + for (i = 0; i < rxq->rx_free_thresh; ++i) { > + /* populate the static rte mbuf fields */ > + mb = rxep[i].mbuf; > + if (reset_mbuf) > + mb->port = rxq->port_id; > + > + rte_mbuf_refcnt_set(mb, 1); > + mb->data_off = RTE_PKTMBUF_HEADROOM; > + > + /* populate the descriptors */ > + dma_addr = rte_cpu_to_le_64(rte_mbuf_data_iova_default(mb)); > + NGBE_RXD_HDRADDR(&rxdp[i], 0); > + NGBE_RXD_PKTADDR(&rxdp[i], dma_addr); > + } > + > + /* update state of internal queue structure */ > + rxq->rx_free_trigger = rxq->rx_free_trigger + rxq->rx_free_thresh; > + if (rxq->rx_free_trigger >= rxq->nb_rx_desc) > + rxq->rx_free_trigger = rxq->rx_free_thresh - 1; > + > + /* no errors */ > + return 0; > +} > + > +static inline uint16_t > +ngbe_rx_fill_from_stage(struct ngbe_rx_queue *rxq, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts) > +{ > + struct rte_mbuf **stage = &rxq->rx_stage[rxq->rx_next_avail]; > + int i; > + > + /* how many packets are ready to return? */ > + nb_pkts = (uint16_t)RTE_MIN(nb_pkts, rxq->rx_nb_avail); > + > + /* copy mbuf pointers to the application's packet list */ > + for (i = 0; i < nb_pkts; ++i) > + rx_pkts[i] = stage[i]; > + > + /* update internal queue state */ > + rxq->rx_nb_avail = (uint16_t)(rxq->rx_nb_avail - nb_pkts); > + rxq->rx_next_avail = (uint16_t)(rxq->rx_next_avail + nb_pkts); > + > + return nb_pkts; > +} > + > +static inline uint16_t > +ngbe_rx_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts) > +{ > + struct ngbe_rx_queue *rxq = (struct ngbe_rx_queue *)rx_queue; > + struct rte_eth_dev *dev = &rte_eth_devices[rxq->port_id]; > + uint16_t nb_rx = 0; > + > + /* Any previously recv'd pkts will be returned from the Rx stage */ > + if (rxq->rx_nb_avail) > + return ngbe_rx_fill_from_stage(rxq, rx_pkts, nb_pkts); > + > + /* Scan the H/W ring for packets to receive */ > + nb_rx = (uint16_t)ngbe_rx_scan_hw_ring(rxq); > + > + /* update internal queue state */ > + rxq->rx_next_avail = 0; > + rxq->rx_nb_avail = nb_rx; > + rxq->rx_tail = (uint16_t)(rxq->rx_tail + nb_rx); > + > + /* if required, allocate new buffers to replenish descriptors */ > + if (rxq->rx_tail > rxq->rx_free_trigger) { > + uint16_t cur_free_trigger = rxq->rx_free_trigger; > + > + if (ngbe_rx_alloc_bufs(rxq, true) != 0) { > + int i, j; > + > + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed port_id=%u " > + "queue_id=%u", (uint16_t)rxq->port_id, > + (uint16_t)rxq->queue_id); > + > + dev->data->rx_mbuf_alloc_failed += > + rxq->rx_free_thresh; > + > + /* > + * Need to rewind any previous receives if we cannot > + * allocate new buffers to replenish the old ones. > + */ > + rxq->rx_nb_avail = 0; > + rxq->rx_tail = (uint16_t)(rxq->rx_tail - nb_rx); > + for (i = 0, j = rxq->rx_tail; i < nb_rx; ++i, ++j) > + rxq->sw_ring[j].mbuf = rxq->rx_stage[i]; > + > + return 0; > + } > + > + /* update tail pointer */ > + rte_wmb(); > + ngbe_set32_relaxed(rxq->rdt_reg_addr, cur_free_trigger); > + } > + > + if (rxq->rx_tail >= rxq->nb_rx_desc) > + rxq->rx_tail = 0; > + > + /* received any packets this loop? */ > + if (rxq->rx_nb_avail) > + return ngbe_rx_fill_from_stage(rxq, rx_pkts, nb_pkts); > + > + return 0; > +} > + > +/* split requests into chunks of size RTE_PMD_NGBE_RX_MAX_BURST */ > +uint16_t > +ngbe_recv_pkts_bulk_alloc(void *rx_queue, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts) > +{ > + uint16_t nb_rx; > + > + if (unlikely(nb_pkts == 0)) > + return 0; > + > + if (likely(nb_pkts <= RTE_PMD_NGBE_RX_MAX_BURST)) > + return ngbe_rx_recv_pkts(rx_queue, rx_pkts, nb_pkts); > + > + /* request is relatively large, chunk it up */ > + nb_rx = 0; > + while (nb_pkts) { > + uint16_t ret, n; > + > + n = (uint16_t)RTE_MIN(nb_pkts, RTE_PMD_NGBE_RX_MAX_BURST); > + ret = ngbe_rx_recv_pkts(rx_queue, &rx_pkts[nb_rx], n); > + nb_rx = (uint16_t)(nb_rx + ret); > + nb_pkts = (uint16_t)(nb_pkts - ret); > + if (ret < n) > + break; > + } > + > + return nb_rx; > +} > + > uint16_t > ngbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, > uint16_t nb_pkts) > @@ -501,6 +752,288 @@ ngbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, > return nb_rx; > } > > +/** > + * ngbe_fill_cluster_head_buf - fill the first mbuf of the returned packet > + * > + * Fill the following info in the HEAD buffer of the Rx cluster: > + * - RX port identifier > + * - hardware offload data, if any: > + * - RSS flag & hash > + * - IP checksum flag > + * - VLAN TCI, if any > + * - error flags > + * @head HEAD of the packet cluster > + * @desc HW descriptor to get data from > + * @rxq Pointer to the Rx queue > + */ > +static inline void > +ngbe_fill_cluster_head_buf(struct rte_mbuf *head, struct ngbe_rx_desc *desc, > + struct ngbe_rx_queue *rxq, uint32_t staterr) > +{ > + uint32_t pkt_info; > + uint64_t pkt_flags; > + > + head->port = rxq->port_id; > + > + /* The vlan_tci field is only valid when PKT_RX_VLAN is > + * set in the pkt_flags field. > + */ > + head->vlan_tci = rte_le_to_cpu_16(desc->qw1.hi.tag); > + pkt_info = rte_le_to_cpu_32(desc->qw0.dw0); > + pkt_flags = rx_desc_status_to_pkt_flags(staterr, rxq->vlan_flags); > + pkt_flags |= rx_desc_error_to_pkt_flags(staterr); > + pkt_flags |= ngbe_rxd_pkt_info_to_pkt_flags(pkt_info); > + head->ol_flags = pkt_flags; > + head->packet_type = ngbe_rxd_pkt_info_to_pkt_type(pkt_info, > + rxq->pkt_type_mask); > + > + if (likely(pkt_flags & PKT_RX_RSS_HASH)) > + head->hash.rss = rte_le_to_cpu_32(desc->qw0.dw1); > +} > + > +/** > + * ngbe_recv_pkts_sc - receive handler for scatter case. > + * > + * @rx_queue Rx queue handle > + * @rx_pkts table of received packets > + * @nb_pkts size of rx_pkts table > + * @bulk_alloc if TRUE bulk allocation is used for a HW ring refilling > + * > + * Returns the number of received packets/clusters (according to the "bulk > + * receive" interface). > + */ > +static inline uint16_t > +ngbe_recv_pkts_sc(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts, > + bool bulk_alloc) > +{ > + struct ngbe_rx_queue *rxq = rx_queue; > + struct rte_eth_dev *dev = &rte_eth_devices[rxq->port_id]; > + volatile struct ngbe_rx_desc *rx_ring = rxq->rx_ring; > + struct ngbe_rx_entry *sw_ring = rxq->sw_ring; > + struct ngbe_scattered_rx_entry *sw_sc_ring = rxq->sw_sc_ring; > + uint16_t rx_id = rxq->rx_tail; > + uint16_t nb_rx = 0; > + uint16_t nb_hold = rxq->nb_rx_hold; > + uint16_t prev_id = rxq->rx_tail; > + > + while (nb_rx < nb_pkts) { > + bool eop; > + struct ngbe_rx_entry *rxe; > + struct ngbe_scattered_rx_entry *sc_entry; > + struct ngbe_scattered_rx_entry *next_sc_entry = NULL; > + struct ngbe_rx_entry *next_rxe = NULL; > + struct rte_mbuf *first_seg; > + struct rte_mbuf *rxm; > + struct rte_mbuf *nmb = NULL; > + struct ngbe_rx_desc rxd; > + uint16_t data_len; > + uint16_t next_id; > + volatile struct ngbe_rx_desc *rxdp; > + uint32_t staterr; > + > +next_desc: > + rxdp = &rx_ring[rx_id]; > + staterr = rte_le_to_cpu_32(rxdp->qw1.lo.status); > + > + if (!(staterr & NGBE_RXD_STAT_DD)) > + break; > + > + rxd = *rxdp; > + > + PMD_RX_LOG(DEBUG, "port_id=%u queue_id=%u rx_id=%u " > + "staterr=0x%x data_len=%u", > + rxq->port_id, rxq->queue_id, rx_id, staterr, > + rte_le_to_cpu_16(rxd.qw1.hi.len)); > + > + if (!bulk_alloc) { > + nmb = rte_mbuf_raw_alloc(rxq->mb_pool); > + if (nmb == NULL) { > + PMD_RX_LOG(DEBUG, "RX mbuf alloc failed " > + "port_id=%u queue_id=%u", > + rxq->port_id, rxq->queue_id); > + > + dev->data->rx_mbuf_alloc_failed++; > + break; > + } > + } else if (nb_hold > rxq->rx_free_thresh) { > + uint16_t next_rdt = rxq->rx_free_trigger; > + > + if (!ngbe_rx_alloc_bufs(rxq, false)) { > + rte_wmb(); > + ngbe_set32_relaxed(rxq->rdt_reg_addr, > + next_rdt); > + nb_hold -= rxq->rx_free_thresh; > + } else { > + PMD_RX_LOG(DEBUG, "RX bulk alloc failed " > + "port_id=%u queue_id=%u", > + rxq->port_id, rxq->queue_id); > + > + dev->data->rx_mbuf_alloc_failed++; > + break; > + } > + } > + > + nb_hold++; > + rxe = &sw_ring[rx_id]; > + eop = staterr & NGBE_RXD_STAT_EOP; > + > + next_id = rx_id + 1; > + if (next_id == rxq->nb_rx_desc) > + next_id = 0; > + > + /* Prefetch next mbuf while processing current one. */ > + rte_ngbe_prefetch(sw_ring[next_id].mbuf); > + > + /* > + * When next RX descriptor is on a cache-line boundary, > + * prefetch the next 4 RX descriptors and the next 4 pointers > + * to mbufs. > + */ > + if ((next_id & 0x3) == 0) { > + rte_ngbe_prefetch(&rx_ring[next_id]); > + rte_ngbe_prefetch(&sw_ring[next_id]); > + } > + > + rxm = rxe->mbuf; > + > + if (!bulk_alloc) { > + __le64 dma = > + rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb)); > + /* > + * Update RX descriptor with the physical address of the > + * new data buffer of the new allocated mbuf. > + */ > + rxe->mbuf = nmb; > + > + rxm->data_off = RTE_PKTMBUF_HEADROOM; > + NGBE_RXD_HDRADDR(rxdp, 0); > + NGBE_RXD_PKTADDR(rxdp, dma); > + } else { > + rxe->mbuf = NULL; > + } > + > + /* > + * Set data length & data buffer address of mbuf. > + */ > + data_len = rte_le_to_cpu_16(rxd.qw1.hi.len); > + rxm->data_len = data_len; > + > + if (!eop) { > + uint16_t nextp_id; > + > + nextp_id = next_id; > + next_sc_entry = &sw_sc_ring[nextp_id]; > + next_rxe = &sw_ring[nextp_id]; > + rte_ngbe_prefetch(next_rxe); > + } > + > + sc_entry = &sw_sc_ring[rx_id]; > + first_seg = sc_entry->fbuf; > + sc_entry->fbuf = NULL; > + > + /* > + * 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 == NULL) { > + first_seg = rxm; > + first_seg->pkt_len = data_len; > + first_seg->nb_segs = 1; > + } else { > + first_seg->pkt_len += data_len; > + first_seg->nb_segs++; > + } > + > + prev_id = rx_id; > + rx_id = next_id; > + > + /* > + * If this is not the last buffer of the received packet, update > + * the pointer to the first mbuf at the NEXTP entry in the > + * sw_sc_ring and continue to parse the RX ring. RX -> Rx > + */ > + if (!eop && next_rxe) { > + rxm->next = next_rxe->mbuf; > + next_sc_entry->fbuf = first_seg; > + goto next_desc; > + } > + > + /* Initialize the first mbuf of the returned packet */ > + ngbe_fill_cluster_head_buf(first_seg, &rxd, rxq, staterr); > + > + /* Deal with the case, when HW CRC srip is disabled. */ > + first_seg->pkt_len -= rxq->crc_len; > + if (unlikely(rxm->data_len <= rxq->crc_len)) { > + struct rte_mbuf *lp; > + > + for (lp = first_seg; lp->next != rxm; lp = lp->next) > + ; > + > + first_seg->nb_segs--; > + lp->data_len -= rxq->crc_len - rxm->data_len; > + lp->next = NULL; > + rte_pktmbuf_free_seg(rxm); > + } else { > + rxm->data_len -= rxq->crc_len; > + } > + > + /* Prefetch data of first segment, if configured to do so. */ > + rte_packet_prefetch((char *)first_seg->buf_addr + > + first_seg->data_off); > + > + /* > + * Store the mbuf address into the next entry of the array > + * of returned packets. > + */ > + rx_pkts[nb_rx++] = first_seg; > + } > + > + /* > + * Record index of the next RX descriptor to probe. > + */ > + rxq->rx_tail = rx_id; > + > + /* > + * If the number of free RX descriptors is greater than the RX free RX -> Rx twice > + * 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 situation from the > + * hardware point of view... > + */ > + if (!bulk_alloc && 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); > + > + rte_wmb(); > + ngbe_set32_relaxed(rxq->rdt_reg_addr, prev_id); > + nb_hold = 0; > + } > + > + rxq->nb_rx_hold = nb_hold; > + return nb_rx; > +} > + > +uint16_t > +ngbe_recv_pkts_sc_single_alloc(void *rx_queue, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts) > +{ > + return ngbe_recv_pkts_sc(rx_queue, rx_pkts, nb_pkts, false); > +} > + > +uint16_t > +ngbe_recv_pkts_sc_bulk_alloc(void *rx_queue, struct rte_mbuf **rx_pkts, > + uint16_t nb_pkts) > +{ > + return ngbe_recv_pkts_sc(rx_queue, rx_pkts, nb_pkts, true); > +} > + > /********************************************************************* > * > * Queue management functions > @@ -1064,6 +1597,54 @@ ngbe_dev_rx_queue_setup(struct rte_eth_dev *dev, > return 0; > } > > +void __rte_cold > +ngbe_set_rx_function(struct rte_eth_dev *dev) > +{ > + struct ngbe_adapter *adapter = NGBE_DEV_ADAPTER(dev); > + > + if (dev->data->scattered_rx) { > + /* > + * Set the scattered callback: there are bulk and > + * single allocation versions. > + */ > + if (adapter->rx_bulk_alloc_allowed) { > + PMD_INIT_LOG(DEBUG, "Using a Scattered with bulk " > + "allocation callback (port=%d).", > + dev->data->port_id); > + dev->rx_pkt_burst = ngbe_recv_pkts_sc_bulk_alloc; > + } else { > + PMD_INIT_LOG(DEBUG, "Using Regular (non-vector, " > + "single allocation) " > + "Scattered Rx callback " > + "(port=%d).", > + dev->data->port_id); > + > + dev->rx_pkt_burst = ngbe_recv_pkts_sc_single_alloc; > + } > + /* > + * Below we set "simple" callbacks according to port/queues parameters. > + * If parameters allow we are going to choose between the following > + * callbacks: > + * - Bulk Allocation > + * - Single buffer allocation (the simplest one) > + */ > + } else if (adapter->rx_bulk_alloc_allowed) { > + PMD_INIT_LOG(DEBUG, "Rx Burst Bulk Alloc Preconditions are " > + "satisfied. Rx Burst Bulk Alloc function " > + "will be used on port=%d.", > + dev->data->port_id); > + > + dev->rx_pkt_burst = ngbe_recv_pkts_bulk_alloc; > + } else { > + PMD_INIT_LOG(DEBUG, "Rx Burst Bulk Alloc Preconditions are not " > + "satisfied, or Scattered Rx is requested " > + "(port=%d).", > + dev->data->port_id); > + > + dev->rx_pkt_burst = ngbe_recv_pkts; > + } > +} > + > /* > * Initializes Receive Unit. > */ > @@ -1211,6 +1792,8 @@ ngbe_dev_rx_init(struct rte_eth_dev *dev) > wr32(hw, NGBE_SECRXCTL, rdrxctl); > } > > + ngbe_set_rx_function(dev); > + > return 0; > } > > diff --git a/drivers/net/ngbe/ngbe_rxtx.h b/drivers/net/ngbe/ngbe_rxtx.h > index d6b9127cb4..4b8596b24a 100644 > --- a/drivers/net/ngbe/ngbe_rxtx.h > +++ b/drivers/net/ngbe/ngbe_rxtx.h > @@ -298,6 +298,8 @@ struct ngbe_txq_ops { > void (*reset)(struct ngbe_tx_queue *txq); > }; > > +void ngbe_set_rx_function(struct rte_eth_dev *dev); > + > uint64_t ngbe_get_tx_port_offloads(struct rte_eth_dev *dev); > uint64_t ngbe_get_rx_queue_offloads(struct rte_eth_dev *dev); > uint64_t ngbe_get_rx_port_offloads(struct rte_eth_dev *dev); >