DPDK patches and discussions
 help / color / mirror / Atom feed
From: Ziyang Xuan <xuanziyang2@huawei.com>
To: <dev@dpdk.org>
Cc: <ferruh.yigit@intel.com>, <cloud.wangxiaoyun@huawei.com>,
	<zhouguoyang@huawei.com>, <shahar.belkar@huawei.com>,
	<luoxianjun@huawei.com>, Ziyang Xuan <xuanziyang2@huawei.com>
Subject: [dpdk-dev] [PATCH v2 09/11] net/hinic: add RX module
Date: Wed, 29 May 2019 11:50:50 +0800	[thread overview]
Message-ID: <81ad45e892c7937cc5751686e30601cd7e90598c.1559100649.git.xuanziyang2@huawei.com> (raw)
In-Reply-To: <cover.1559100649.git.xuanziyang2@huawei.com>

Add code for RX module.

Signed-off-by: Ziyang Xuan <xuanziyang2@huawei.com>
---
 drivers/net/hinic/Makefile       |   2 +-
 drivers/net/hinic/hinic_pmd_rx.c | 919 +++++++++++++++++++++++++++++++
 2 files changed, 920 insertions(+), 1 deletion(-)
 create mode 100644 drivers/net/hinic/hinic_pmd_rx.c

diff --git a/drivers/net/hinic/Makefile b/drivers/net/hinic/Makefile
index 6e19acf08..dd40216ec 100644
--- a/drivers/net/hinic/Makefile
+++ b/drivers/net/hinic/Makefile
@@ -77,7 +77,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_wq.c
 
 SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_dpdev.c
 #SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_ethdev.c
-#SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
+SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_rx.c
 #SRCS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += hinic_pmd_tx.c
 # this lib depends upon:
 DEPDIRS-$(CONFIG_RTE_LIBRTE_HINIC_PMD) += lib/librte_eal lib/librte_ether
diff --git a/drivers/net/hinic/hinic_pmd_rx.c b/drivers/net/hinic/hinic_pmd_rx.c
new file mode 100644
index 000000000..27e723764
--- /dev/null
+++ b/drivers/net/hinic/hinic_pmd_rx.c
@@ -0,0 +1,919 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Huawei Technologies Co., Ltd
+ */
+
+#include <rte_ether.h>
+#include <rte_ethdev.h>
+#include <rte_mbuf.h>
+#ifdef __ARM64_NEON__
+#include <arm_neon.h>
+#endif
+
+#include "hinic_pmd_ethdev.h"
+#include "hinic_pmd_rx.h"
+
+#ifdef HINIC_XSTAT_RXBUF_INFO
+static void hinic_rxq_buffer_done_count(struct hinic_rxq *rxq)
+{
+	u16 sw_ci, avail_pkts = 0, hit_done = 0, cqe_hole = 0;
+	u32 status;
+	volatile struct hinic_rq_cqe *rx_cqe;
+
+	for (sw_ci = 0; sw_ci < rxq->q_depth; sw_ci++) {
+		rx_cqe = &rxq->rx_cqe[sw_ci];
+
+		/* test current ci is done */
+		status = rx_cqe->status;
+		if (!HINIC_GET_RX_DONE_BE(status)) {
+			if (hit_done) {
+				cqe_hole++;
+				hit_done = 0;
+			}
+			continue;
+		}
+
+		avail_pkts++;
+		hit_done = 1;
+	}
+
+	rxq->rxq_stats.rx_avail = avail_pkts;
+	rxq->rxq_stats.rx_hole = cqe_hole;
+}
+#endif
+
+void hinic_rxq_get_stats(struct hinic_rxq *rxq, struct hinic_rxq_stats *stats)
+{
+	if (!rxq || !stats)
+		return;
+
+#ifdef HINIC_XSTAT_RXBUF_INFO
+	rxq->rxq_stats.rx_mbuf = (rxq->q_depth)
+				- HINIC_GET_RQ_FREE_WQEBBS(rxq);
+
+	hinic_rxq_buffer_done_count(rxq);
+#endif
+#ifdef HINIC_XSTAT_MBUF_USE
+	rxq->rxq_stats.left_mbuf = rxq->rxq_stats.alloc_mbuf
+				- rxq->rxq_stats.free_mbuf;
+#endif
+	memcpy(stats, &rxq->rxq_stats, sizeof(rxq->rxq_stats));
+}
+
+void hinic_rxq_stats_reset(struct hinic_rxq *rxq)
+{
+	struct hinic_rxq_stats *rxq_stats;
+
+	if (rxq == NULL)
+		return;
+
+	rxq_stats = &rxq->rxq_stats;
+	memset(rxq_stats, 0, sizeof(*rxq_stats));
+}
+
+/* mbuf alloc and free */
+static inline struct rte_mbuf *hinic_rte_rxmbuf_alloc(struct rte_mempool *mp)
+{
+	struct rte_mbuf *m;
+
+	m  = rte_mbuf_raw_alloc(mp);
+	return m;
+}
+
+static int hinic_rx_alloc_cqe(struct hinic_rxq *rxq)
+{
+	size_t cqe_mem_size;
+
+	/* allocate continuous cqe memory for saving number of memory zone */
+	cqe_mem_size = sizeof(struct hinic_rq_cqe) * rxq->q_depth;
+	rxq->cqe_start_vaddr = dma_zalloc_coherent(rxq->nic_dev, cqe_mem_size,
+						   &rxq->cqe_start_paddr,
+						   GFP_KERNEL);
+	if (!rxq->cqe_start_vaddr) {
+		PMD_DRV_LOG(ERR, "Allocate cqe dma memory failed");
+		return -ENOMEM;
+	}
+
+	rxq->rx_cqe = (struct hinic_rq_cqe *)rxq->cqe_start_vaddr;
+
+	return HINIC_OK;
+}
+
+static void hinic_rx_free_cqe(struct hinic_rxq *rxq)
+{
+	size_t cqe_mem_size;
+
+	cqe_mem_size = sizeof(struct hinic_rq_cqe) * rxq->q_depth;
+	dma_free_coherent(rxq->nic_dev, cqe_mem_size,
+			  rxq->cqe_start_vaddr,
+			  rxq->cqe_start_paddr);
+	rxq->cqe_start_vaddr = NULL;
+}
+
+static int hinic_rx_fill_wqe(struct hinic_rxq *rxq)
+{
+	hinic_nic_dev *nic_dev = rxq->nic_dev;
+	struct hinic_rq_wqe *rq_wqe;
+	dma_addr_t buf_dma_addr, cqe_dma_addr;
+	u16 pi = 0;
+	int rq_wqe_len;
+	int i;
+
+	buf_dma_addr = 0;
+	cqe_dma_addr = rxq->cqe_start_paddr;
+	for (i = 0; i < rxq->q_depth; i++) {
+		rq_wqe = (struct hinic_rq_wqe *)
+			hinic_get_rq_wqe(nic_dev->hwdev, rxq->q_id, &pi);
+		if (!rq_wqe) {
+			PMD_DRV_LOG(ERR, "Get rq wqe failed");
+			break;
+		}
+
+		hinic_prepare_rq_wqe(rq_wqe, pi, buf_dma_addr, cqe_dma_addr);
+		cqe_dma_addr +=  sizeof(struct hinic_rq_cqe);
+
+		rq_wqe_len = sizeof(struct hinic_rq_wqe);
+		hinic_cpu_to_be32(rq_wqe, rq_wqe_len);
+	}
+
+	hinic_return_rq_wqe(nic_dev->hwdev, rxq->q_id, i);
+
+	return i;
+}
+
+/* alloc cqe and prepare rqe */
+int hinic_setup_rx_resources(struct hinic_rxq *rxq)
+{
+	u64 rx_info_sz;
+	int err, pkts;
+
+	rx_info_sz = rxq->q_depth * sizeof(*rxq->rx_info);
+	rxq->rx_info = kzalloc_aligned(rx_info_sz, GFP_KERNEL);
+	if (!rxq->rx_info)
+		return -ENOMEM;
+
+	err = hinic_rx_alloc_cqe(rxq);
+	if (err) {
+		PMD_DRV_LOG(ERR, "Allocate rx cqe failed");
+		goto rx_cqe_err;
+	}
+
+	pkts = hinic_rx_fill_wqe(rxq);
+	if (pkts != rxq->q_depth) {
+		PMD_DRV_LOG(ERR, "Fill rx wqe failed");
+		err = -ENOMEM;
+		goto rx_fill_err;
+	}
+
+	return 0;
+
+rx_fill_err:
+	hinic_rx_free_cqe(rxq);
+
+rx_cqe_err:
+	kfree(rxq->rx_info);
+	rxq->rx_info = NULL;
+
+	return err;
+}
+
+void hinic_free_rx_resources(struct hinic_rxq *rxq)
+{
+	if (rxq->rx_info == NULL)
+		return;
+
+	hinic_rx_free_cqe(rxq);
+	kfree(rxq->rx_info);
+	rxq->rx_info = NULL;
+}
+
+void hinic_free_all_rx_resources(struct rte_eth_dev *eth_dev)
+{
+	u16 q_id;
+	hinic_nic_dev *nic_dev =
+		(hinic_nic_dev *)HINIC_ETH_DEV_TO_PRIVATE_NIC_DEV(eth_dev);
+
+	for (q_id = 0; q_id < nic_dev->num_rq; q_id++) {
+		eth_dev->data->rx_queues[q_id] = NULL;
+
+		if (nic_dev->rxqs[q_id] == NULL)
+			continue;
+
+		hinic_free_all_rx_skbs(nic_dev->rxqs[q_id]);
+		hinic_free_rx_resources(nic_dev->rxqs[q_id]);
+		kfree(nic_dev->rxqs[q_id]);
+		nic_dev->rxqs[q_id] = NULL;
+	}
+}
+
+void hinic_free_all_rx_mbuf(struct rte_eth_dev *eth_dev)
+{
+	u16 q_id;
+	hinic_nic_dev *nic_dev = HINIC_ETH_DEV_TO_PRIVATE_NIC_DEV(eth_dev);
+
+	for (q_id = 0; q_id < nic_dev->num_rq; q_id++)
+		hinic_free_all_rx_skbs(nic_dev->rxqs[q_id]);
+}
+
+static void hinic_recv_jumbo_pkt(struct hinic_rxq *rxq,
+				 struct rte_mbuf *head_skb,
+				 u32 remain_pkt_len)
+{
+	hinic_nic_dev *nic_dev = rxq->nic_dev;
+	struct rte_mbuf *cur_mbuf, *rxm = NULL;
+	struct hinic_rx_info *rx_info;
+	u16 sw_ci, rx_buf_len = rxq->buf_len;
+	u32 pkt_len;
+
+	while (remain_pkt_len > 0) {
+		sw_ci = hinic_get_rq_local_ci(nic_dev->hwdev, rxq->q_id);
+		rx_info = &rxq->rx_info[sw_ci];
+
+		hinic_update_rq_local_ci(nic_dev->hwdev, rxq->q_id, 1);
+
+		pkt_len = remain_pkt_len > rx_buf_len ?
+			rx_buf_len : remain_pkt_len;
+		remain_pkt_len -= pkt_len;
+
+		cur_mbuf = rx_info->mbuf;
+		cur_mbuf->data_len = (u16)pkt_len;
+		cur_mbuf->next = NULL;
+
+		head_skb->pkt_len += cur_mbuf->data_len;
+		head_skb->nb_segs++;
+#ifdef HINIC_XSTAT_MBUF_USE
+		rxq->rxq_stats.free_mbuf++;
+#endif
+
+		if (!rxm)
+			head_skb->next = cur_mbuf;
+		else
+			rxm->next = cur_mbuf;
+
+		rxm = cur_mbuf;
+	}
+}
+
+static void hinic_rss_deinit(hinic_nic_dev *nic_dev)
+{
+	u8 prio_tc[HINIC_DCB_UP_MAX] = {0};
+	(void)hinic_rss_cfg(nic_dev->hwdev, 0,
+			    nic_dev->rss_tmpl_idx, 0, prio_tc);
+}
+
+static int hinic_rss_key_init(hinic_nic_dev *nic_dev,
+			      struct rte_eth_rss_conf *rss_conf)
+{
+	u8 default_rss_key[HINIC_RSS_KEY_SIZE] = {
+			 0x6d, 0x5a, 0x56, 0xda, 0x25, 0x5b, 0x0e, 0xc2,
+			 0x41, 0x67, 0x25, 0x3d, 0x43, 0xa3, 0x8f, 0xb0,
+			 0xd0, 0xca, 0x2b, 0xcb, 0xae, 0x7b, 0x30, 0xb4,
+			 0x77, 0xcb, 0x2d, 0xa3, 0x80, 0x30, 0xf2, 0x0c,
+			 0x6a, 0x42, 0xb7, 0x3b, 0xbe, 0xac, 0x01, 0xfa};
+	u8 hashkey[HINIC_RSS_KEY_SIZE] = {0};
+	u8 tmpl_idx = nic_dev->rss_tmpl_idx;
+
+	if (rss_conf->rss_key == NULL)
+		memcpy(hashkey, default_rss_key, HINIC_RSS_KEY_SIZE);
+	else
+		memcpy(hashkey, rss_conf->rss_key, rss_conf->rss_key_len);
+
+	return hinic_rss_set_template_tbl(nic_dev->hwdev, tmpl_idx, hashkey);
+}
+
+static void hinic_fill_rss_type(struct nic_rss_type *rss_type,
+				struct rte_eth_rss_conf *rss_conf)
+{
+	u64 rss_hf = rss_conf->rss_hf;
+
+	rss_type->ipv4 = (rss_hf & (ETH_RSS_IPV4 | ETH_RSS_FRAG_IPV4)) ? 1 : 0;
+	rss_type->tcp_ipv4 = (rss_hf & ETH_RSS_NONFRAG_IPV4_TCP) ? 1 : 0;
+	rss_type->ipv6 = (rss_hf & (ETH_RSS_IPV6 | ETH_RSS_FRAG_IPV6)) ? 1 : 0;
+	rss_type->ipv6_ext = (rss_hf & ETH_RSS_IPV6_EX) ? 1 : 0;
+	rss_type->tcp_ipv6 = (rss_hf & ETH_RSS_NONFRAG_IPV6_TCP) ? 1 : 0;
+	rss_type->tcp_ipv6_ext = (rss_hf & ETH_RSS_IPV6_TCP_EX) ? 1 : 0;
+	rss_type->udp_ipv4 = (rss_hf & ETH_RSS_NONFRAG_IPV4_UDP) ? 1 : 0;
+	rss_type->udp_ipv6 = (rss_hf & ETH_RSS_NONFRAG_IPV6_UDP) ? 1 : 0;
+}
+
+static void hinic_fillout_indir_tbl(hinic_nic_dev *nic_dev, u32 *indir)
+{
+	u8 rss_queue_count = nic_dev->num_rss;
+	int i = 0, j;
+
+	if (rss_queue_count == 0) {
+		/* delete q_id from indir tbl */
+		for (i = 0; i < HINIC_RSS_INDIR_SIZE; i++)
+			indir[i] = 0xFF;	/* Invalid value in indir tbl */
+	} else {
+		while (i < HINIC_RSS_INDIR_SIZE)
+			for (j = 0; (j < rss_queue_count) &&
+			     (i < HINIC_RSS_INDIR_SIZE); j++)
+				indir[i++] = nic_dev->rx_queue_list[j];
+	}
+}
+
+static int hinic_rss_init(hinic_nic_dev *nic_dev,
+			  __attribute__((unused)) u8 *rq2iq_map,
+			  struct rte_eth_rss_conf *rss_conf)
+{
+	u32 indir_tbl[HINIC_RSS_INDIR_SIZE] = {0};
+	struct nic_rss_type rss_type = {0};
+	u8 prio_tc[HINIC_DCB_UP_MAX] = {0};
+	u8 tmpl_idx = 0xFF, num_tc = 0;
+	int err;
+
+	tmpl_idx = nic_dev->rss_tmpl_idx;
+
+	err = hinic_rss_key_init(nic_dev, rss_conf);
+	if (err)
+		return err;
+
+	if (!nic_dev->rss_indir_flag) {
+		hinic_fillout_indir_tbl(nic_dev, indir_tbl);
+		err = hinic_rss_set_indir_tbl(nic_dev->hwdev, tmpl_idx,
+					      indir_tbl);
+		if (err)
+			return err;
+	}
+
+	hinic_fill_rss_type(&rss_type, rss_conf);
+	err = hinic_set_rss_type(nic_dev->hwdev, tmpl_idx, rss_type);
+	if (err)
+		return err;
+
+	err = hinic_rss_set_hash_engine(nic_dev->hwdev, tmpl_idx,
+					HINIC_RSS_HASH_ENGINE_TYPE_TOEP);
+	if (err)
+		return err;
+
+	return hinic_rss_cfg(nic_dev->hwdev, 1, tmpl_idx, num_tc, prio_tc);
+}
+
+static void hinic_add_rq_to_rx_queue_list(hinic_nic_dev *nic_dev, u16 queue_id)
+{
+	u8 rss_queue_count = nic_dev->num_rss;
+
+	RTE_ASSERT(rss_queue_count <= (RTE_DIM(nic_dev->rx_queue_list) - 1));
+
+	nic_dev->rx_queue_list[rss_queue_count] = queue_id;
+	nic_dev->num_rss++;
+}
+
+/**
+ * hinic_setup_num_qps - determine num_qps from rss_tmpl_id
+ * @nic_dev: pointer to the private ethernet device
+ * Return: 0 on Success, error code otherwise.
+ **/
+static int hinic_setup_num_qps(hinic_nic_dev *nic_dev)
+{
+	int err, i;
+
+	if (!(nic_dev->flags & ETH_MQ_RX_RSS_FLAG)) {
+		nic_dev->flags &= ~ETH_MQ_RX_RSS_FLAG;
+		nic_dev->num_rss = 0;
+		if (nic_dev->num_rq > 1) {
+			/* get rss template id */
+			err = hinic_rss_template_alloc(nic_dev->hwdev,
+						       &nic_dev->rss_tmpl_idx);
+			if (err) {
+				PMD_DRV_LOG(WARNING, "Alloc rss template failed");
+				return err;
+			}
+			nic_dev->flags |= ETH_MQ_RX_RSS_FLAG;
+			for (i = 0; i < nic_dev->num_rq; i++)
+				hinic_add_rq_to_rx_queue_list(nic_dev, i);
+		}
+	}
+
+	return 0;
+}
+
+static void hinic_destroy_num_qps(hinic_nic_dev *nic_dev)
+{
+	if (nic_dev->flags & ETH_MQ_RX_RSS_FLAG) {
+		if (hinic_rss_template_free(nic_dev->hwdev,
+					    nic_dev->rss_tmpl_idx))
+			PMD_DRV_LOG(WARNING, "Free rss template failed");
+
+		nic_dev->flags &= ~ETH_MQ_RX_RSS_FLAG;
+	}
+}
+
+static int hinic_config_mq_rx_rss(hinic_nic_dev *nic_dev, bool on)
+{
+	int ret = 0;
+
+	if (on) {
+		ret = hinic_setup_num_qps(nic_dev);
+		if (ret)
+			PMD_DRV_LOG(ERR, "Setup num_qps failed");
+	} else {
+		hinic_destroy_num_qps(nic_dev);
+	}
+
+	return ret;
+}
+
+int hinic_config_mq_mode(struct rte_eth_dev *dev, bool on)
+{
+	hinic_nic_dev *nic_dev = HINIC_ETH_DEV_TO_PRIVATE_NIC_DEV(dev);
+	struct rte_eth_conf *dev_conf = &dev->data->dev_conf;
+	int ret = 0;
+
+	switch (dev_conf->rxmode.mq_mode) {
+	case ETH_MQ_RX_RSS:
+		ret = hinic_config_mq_rx_rss(nic_dev, on);
+		break;
+	default:
+		break;
+	}
+
+	return ret;
+}
+
+int hinic_rx_configure(struct rte_eth_dev *dev)
+{
+	hinic_nic_dev *nic_dev = HINIC_ETH_DEV_TO_PRIVATE_NIC_DEV(dev);
+	struct rte_eth_rss_conf rss_conf =
+		dev->data->dev_conf.rx_adv_conf.rss_conf;
+	u32 csum_en = 0;
+	int err;
+
+	if (nic_dev->flags & ETH_MQ_RX_RSS_FLAG) {
+		if (rss_conf.rss_hf == 0) {
+			rss_conf.rss_hf = HINIC_RSS_OFFLOAD_ALL;
+		} else if ((rss_conf.rss_hf & HINIC_RSS_OFFLOAD_ALL) == 0) {
+			PMD_DRV_LOG(ERR, "Do not support rss offload all");
+			goto rss_config_err;
+		}
+
+		err = hinic_rss_init(nic_dev, NULL, &rss_conf);
+		if (err) {
+			PMD_DRV_LOG(ERR, "Init rss failed");
+			goto rss_config_err;
+		}
+	}
+
+	/* Enable both L3/L4 rx checksum offload */
+	if (dev->data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_CHECKSUM)
+		csum_en = HINIC_RX_CSUM_OFFLOAD_EN;
+
+	err = hinic_set_rx_csum_offload(nic_dev->hwdev, csum_en);
+	if (err)
+		goto rx_csum_ofl_err;
+
+	return 0;
+
+rx_csum_ofl_err:
+rss_config_err:
+	hinic_destroy_num_qps(nic_dev);
+
+	return HINIC_ERROR;
+}
+
+void hinic_rx_remove_configure(struct rte_eth_dev *dev)
+{
+	hinic_nic_dev *nic_dev = HINIC_ETH_DEV_TO_PRIVATE_NIC_DEV(dev);
+
+	if (nic_dev->flags & ETH_MQ_RX_RSS_FLAG) {
+		hinic_rss_deinit(nic_dev);
+		hinic_destroy_num_qps(nic_dev);
+	}
+}
+
+void hinic_free_all_rx_skbs(struct hinic_rxq *rxq)
+{
+	hinic_nic_dev *nic_dev = (hinic_nic_dev *)rxq->nic_dev;
+	struct hinic_rx_info *rx_info;
+	int free_wqebbs =
+		hinic_get_rq_free_wqebbs(nic_dev->hwdev, rxq->q_id) + 1;
+	volatile struct hinic_rq_cqe *rx_cqe;
+	u16 ci;
+
+	while (free_wqebbs++ < rxq->q_depth) {
+		ci = hinic_get_rq_local_ci(nic_dev->hwdev, rxq->q_id);
+
+		rx_cqe = &rxq->rx_cqe[ci];
+
+		/* clear done bit */
+		rx_cqe->status = 0;
+
+		rx_info = &rxq->rx_info[ci];
+#ifdef HINIC_XSTAT_MBUF_USE
+		hinic_rx_free_mbuf(rxq, rx_info->mbuf);
+#else
+		hinic_rx_free_mbuf(rx_info->mbuf);
+#endif
+		rx_info->mbuf = NULL;
+
+		hinic_update_rq_local_ci(nic_dev->hwdev, rxq->q_id, 1);
+	}
+}
+
+static inline struct hinic_rq_wqe *hinic_get_rearm_rq_wqe(struct hinic_rxq *rxq,
+							  u16 *prod_idx)
+{
+	u32 cur_pi;
+	struct hinic_wq *wq = rxq->wq;
+
+	/* record current pi */
+	cur_pi = MASKED_WQE_IDX(wq, wq->prod_idx);
+
+	/* update next pi and delta */
+	wq->prod_idx += 1;
+	wq->delta -= 1;
+
+	/* return current pi */
+	*prod_idx = cur_pi;
+	return (struct hinic_rq_wqe *)WQ_WQE_ADDR(wq, cur_pi);
+}
+
+/* performance: byteorder swap m128i */
+static inline void hinic_rq_cqe_be_to_cpu32(void *dst_le32,
+					    volatile void *src_be32)
+{
+#ifndef __ARM64_NEON__
+	volatile __m128i *wqe_be = (volatile __m128i *)src_be32;
+	__m128i *wqe_le = (__m128i *)dst_le32;
+	__m128i shuf_mask =  _mm_set_epi8(12, 13, 14, 15, 8, 9, 10,
+					  11, 4, 5, 6, 7, 0, 1, 2, 3);
+
+	/* l2nic just use first 128 bits */
+	wqe_le[0] = _mm_shuffle_epi8(wqe_be[0], shuf_mask);
+#else
+	volatile uint8x16_t *wqe_be = (volatile uint8x16_t *)src_be32;
+	uint8x16_t *wqe_le = (uint8x16_t *)dst_le32;
+	const uint8x16_t shuf_mask = {3, 2, 1, 0, 7, 6, 5, 4, 11, 10,
+					9, 8, 15, 14, 13, 12};
+
+	/* l2nic just use first 128 bits */
+	wqe_le[0] = vqtbl1q_u8(wqe_be[0], shuf_mask);
+#endif
+}
+
+static inline uint64_t hinic_rx_rss_hash(uint32_t offload_type,
+					 uint32_t cqe_hass_val,
+					 uint32_t *rss_hash)
+{
+	uint32_t rss_type;
+
+	rss_type = HINIC_GET_RSS_TYPES(offload_type);
+	if (likely(rss_type != 0)) {
+		*rss_hash = cqe_hass_val;
+		return PKT_RX_RSS_HASH;
+	}
+
+	return 0;
+}
+
+static inline uint64_t hinic_rx_csum(uint32_t status, struct hinic_rxq *rxq)
+{
+	uint32_t checksum_err;
+	uint64_t flags;
+
+	/* most case checksum is ok */
+	checksum_err = HINIC_GET_RX_CSUM_ERR(status);
+	if (likely(checksum_err == 0))
+		return (PKT_RX_IP_CKSUM_GOOD | PKT_RX_L4_CKSUM_GOOD);
+
+	/* If BYPASS bit set, all other status indications should be ignored */
+	if (unlikely(HINIC_CSUM_ERR_BYPASSED(checksum_err)))
+		return PKT_RX_IP_CKSUM_UNKNOWN;
+
+	flags = 0;
+
+	/* IP checksum error */
+	if (HINIC_CSUM_ERR_IP(checksum_err))
+		flags |= PKT_RX_IP_CKSUM_BAD;
+	else
+		flags |= PKT_RX_IP_CKSUM_GOOD;
+
+	/* L4 checksum error */
+	if (HINIC_CSUM_ERR_L4(checksum_err))
+		flags |= PKT_RX_L4_CKSUM_BAD;
+	else
+		flags |= PKT_RX_L4_CKSUM_GOOD;
+
+	if (unlikely(HINIC_CSUM_ERR_OTHER(checksum_err)))
+		flags = PKT_RX_L4_CKSUM_NONE;
+
+	rxq->rxq_stats.errors++;
+
+	return flags;
+}
+
+static inline uint64_t hinic_rx_vlan(uint32_t offload_type, uint32_t vlan_len,
+				     uint16_t *vlan_tci)
+{
+	uint16_t vlan_tag;
+
+	vlan_tag = HINIC_GET_RX_VLAN_TAG(vlan_len);
+	if (!HINIC_GET_RX_VLAN_OFFLOAD_EN(offload_type) || 0 == vlan_tag) {
+		*vlan_tci = 0;
+		return 0;
+	}
+
+	*vlan_tci = vlan_tag;
+
+	return PKT_RX_VLAN | PKT_RX_VLAN_STRIPPED;
+}
+
+static inline uint64_t hinic_rx_pkt_type(uint32_t offload_type)
+{
+	uint32_t pkt_type, pkt_idx;
+	static const uint32_t pkt_type_table[RQ_CQE_PKT_TYPES_L2_MASK + 1]
+	__rte_cache_aligned = {
+		[3] =  RTE_PTYPE_L3_IPV4,
+		[4] =  RTE_PTYPE_L3_IPV4_EXT,
+		[5] =  RTE_PTYPE_L4_FRAG,
+		[7] =  RTE_PTYPE_L3_IPV6,
+		[9] =  RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_SCTP,
+		[10] = RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_UDP,
+		[11] = RTE_PTYPE_TUNNEL_VXLAN,
+		[13] = RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP,
+		[14] = RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP,
+		[15] = RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP,
+		[16] = RTE_PTYPE_TUNNEL_NVGRE,
+		[65] = RTE_PTYPE_L4_ICMP,
+		[66] = RTE_PTYPE_L4_ICMP,
+		[76] = RTE_PTYPE_L2_ETHER_LLDP,
+		[81] = RTE_PTYPE_L2_ETHER_ARP,
+		/* All others reserved */
+	};
+	pkt_idx = HINIC_GET_PKT_TYPES(offload_type);
+
+	/* Unknown type */
+	if (unlikely(pkt_idx == 0))
+		return RTE_PTYPE_UNKNOWN;
+
+	/* if hardware report index not correct set l2 ether as default */
+	pkt_type = RTE_PTYPE_L2_ETHER;
+	pkt_type |= pkt_type_table[HINIC_PKT_TYPES_L2(pkt_idx)];
+
+	return pkt_type;
+}
+
+static inline u32 hinic_rx_alloc_mbuf_bulk(struct hinic_rxq *rxq,
+					   struct rte_mbuf **mbufs,
+					   u32 exp_mbuf_cnt)
+{
+	int rc;
+	u32 avail_cnt;
+
+	rc = rte_pktmbuf_alloc_bulk(rxq->mb_pool, mbufs, exp_mbuf_cnt);
+	if (likely(rc == HINIC_OK)) {
+		avail_cnt = exp_mbuf_cnt;
+	} else {
+		avail_cnt = 0;
+		rxq->rxq_stats.rx_nombuf += exp_mbuf_cnt;
+	}
+#ifdef HINIC_XSTAT_MBUF_USE
+	rxq->rxq_stats.alloc_mbuf += avail_cnt;
+#endif
+	return avail_cnt;
+}
+
+#ifdef HINIC_XSTAT_MBUF_USE
+void hinic_rx_free_mbuf(struct hinic_rxq *rxq, struct rte_mbuf *m)
+{
+	rte_pktmbuf_free(m);
+	rxq->rxq_stats.free_mbuf++;
+}
+#else
+void hinic_rx_free_mbuf(struct rte_mbuf *m)
+{
+	rte_pktmbuf_free(m);
+}
+#endif
+
+static struct rte_mbuf *hinic_rx_alloc_mbuf(struct hinic_rxq *rxq,
+					dma_addr_t *dma_addr)
+{
+	struct rte_mbuf *mbuf;
+
+	mbuf = hinic_rte_rxmbuf_alloc(rxq->mb_pool);
+	if (unlikely(!mbuf))
+		return NULL;
+
+	*dma_addr = rte_mbuf_data_iova_default(mbuf);
+
+#ifdef HINIC_XSTAT_MBUF_USE
+	rxq->rxq_stats.alloc_mbuf++;
+#endif
+
+	return mbuf;
+}
+
+static inline void hinic_rearm_rxq_mbuf(struct hinic_rxq *rxq)
+{
+	u16 pi;
+	u32 i, free_wqebbs, rearm_wqebbs, exp_wqebbs;
+	dma_addr_t dma_addr;
+	struct hinic_rq_wqe *rq_wqe;
+	struct rte_mbuf **rearm_mbufs;
+
+	/* check free wqebb fo rearm */
+	free_wqebbs = HINIC_GET_RQ_FREE_WQEBBS(rxq);
+	if (unlikely(free_wqebbs < rxq->rx_free_thresh))
+		return;
+
+	/* get rearm mbuf array */
+	pi = HINIC_GET_RQ_LOCAL_PI(rxq);
+	rearm_mbufs = (struct rte_mbuf **)(&rxq->rx_info[pi]);
+
+	/* check rxq free wqebbs turn around */
+	if (unlikely(pi > rxq->rxinfo_align_end))
+		exp_wqebbs = rxq->q_depth - pi;
+	else
+		exp_wqebbs = rxq->rx_free_thresh;
+
+	/* alloc mbuf in bulk */
+	rearm_wqebbs = hinic_rx_alloc_mbuf_bulk(rxq, rearm_mbufs, exp_wqebbs);
+	if (unlikely(rearm_wqebbs == 0))
+		return;
+
+	/* rearm rx mbuf */
+	rq_wqe = (struct hinic_rq_wqe *)WQ_WQE_ADDR(rxq->wq, (u32)pi);
+	for (i = 0; i < rearm_wqebbs; i++) {
+		dma_addr = rte_mbuf_data_iova_default(rearm_mbufs[i]);
+		rq_wqe->buf_desc.addr_high =
+					cpu_to_be32(upper_32_bits(dma_addr));
+		rq_wqe->buf_desc.addr_low =
+					cpu_to_be32(lower_32_bits(dma_addr));
+		rq_wqe++;
+	}
+	rxq->wq->prod_idx += rearm_wqebbs;
+	rxq->wq->delta -= rearm_wqebbs;
+
+	/* update rq hw_pi */
+	rte_wmb();
+	HINIC_UPDATE_RQ_HW_PI(rxq, pi + rearm_wqebbs);
+}
+
+void hinic_rx_alloc_pkts(struct hinic_rxq *rxq)
+{
+	hinic_nic_dev *nic_dev = rxq->nic_dev;
+	struct hinic_rq_wqe *rq_wqe;
+	struct hinic_rx_info *rx_info;
+	struct rte_mbuf *mb;
+	dma_addr_t dma_addr;
+	u16 pi = 0;
+	int i, free_wqebbs;
+
+	free_wqebbs = HINIC_GET_RQ_FREE_WQEBBS(rxq);
+	for (i = 0; i < free_wqebbs; i++) {
+		mb = hinic_rx_alloc_mbuf(rxq, &dma_addr);
+		if (unlikely(!mb)) {
+			rxq->rxq_stats.rx_nombuf++;
+			break;
+		}
+
+		rq_wqe = (struct hinic_rq_wqe *)
+			hinic_get_rq_wqe(nic_dev->hwdev, rxq->q_id, &pi);
+		if (unlikely(!rq_wqe)) {
+#ifdef HINIC_XSTAT_MBUF_USE
+			hinic_rx_free_mbuf(rxq, mb);
+#else
+			hinic_rx_free_mbuf(mb);
+#endif
+			break;
+		}
+
+		/* fill buffer address only */
+		rq_wqe->buf_desc.addr_high =
+				cpu_to_be32(upper_32_bits(dma_addr));
+		rq_wqe->buf_desc.addr_low =
+				cpu_to_be32(lower_32_bits(dma_addr));
+
+		rx_info = &rxq->rx_info[pi];
+		rx_info->mbuf = mb;
+	}
+
+	if (likely(i > 0)) {
+		rte_wmb();
+		HINIC_UPDATE_RQ_HW_PI(rxq, pi + 1);
+	}
+}
+
+u16 hinic_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, u16 nb_pkts)
+{
+	struct rte_mbuf *rxm;
+	struct hinic_rxq *rxq = (struct hinic_rxq *)rx_queue;
+	struct hinic_rx_info *rx_info;
+	volatile struct hinic_rq_cqe *rx_cqe;
+	u16 rx_buf_len, pkts = 0;
+	u16 sw_ci, ci_mask, wqebb_cnt = 0;
+	u32 pkt_len, status, vlan_len;
+	u64 rx_bytes = 0;
+#ifdef HINIC_XSTAT_PROF_RX
+	uint64_t t1 = rte_get_tsc_cycles();
+	uint64_t t2;
+#endif
+	struct hinic_rq_cqe cqe;
+	u32 offload_type, rss_hash;
+
+	rx_buf_len = rxq->buf_len;
+
+	/* 1. get polling start ci */
+	ci_mask = HINIC_GET_RQ_WQE_MASK(rxq);
+	sw_ci = HINIC_GET_RQ_LOCAL_CI(rxq);
+
+	while (pkts < nb_pkts) {
+		 /* 2. current ci is done */
+		rx_cqe = &rxq->rx_cqe[sw_ci];
+		status = rx_cqe->status;
+		if (!HINIC_GET_RX_DONE_BE(status))
+			break;
+
+		/* read other cqe member after status */
+		rte_rmb();
+
+		/* convert cqe and get packet length */
+		hinic_rq_cqe_be_to_cpu32(&cqe, (volatile void *)rx_cqe);
+		vlan_len = cqe.vlan_len;
+
+		rx_info = &rxq->rx_info[sw_ci];
+		rxm = rx_info->mbuf;
+
+		/* 3. next ci point and prefetch */
+		sw_ci++;
+		sw_ci &= ci_mask;
+
+		/* prefetch next mbuf first 64B */
+		rte_prefetch0(rxq->rx_info[sw_ci].mbuf);
+
+		/* 4. jumbo frame process */
+		pkt_len = HINIC_GET_RX_PKT_LEN(vlan_len);
+		if (likely(pkt_len <= rx_buf_len)) {
+			rxm->data_len = pkt_len;
+			rxm->pkt_len = pkt_len;
+			wqebb_cnt++;
+		} else {
+			rxm->data_len = rx_buf_len;
+			rxm->pkt_len = rx_buf_len;
+
+			/* if jumbo use multi-wqebb update ci,
+			 * recv_jumbo_pkt will also update ci
+			 */
+			HINIC_UPDATE_RQ_LOCAL_CI(rxq, wqebb_cnt + 1);
+			wqebb_cnt = 0;
+			hinic_recv_jumbo_pkt(rxq, rxm, pkt_len - rx_buf_len);
+			sw_ci = HINIC_GET_RQ_LOCAL_CI(rxq);
+		}
+
+		/* 5. vlan/checksum/rss/pkt_type/gro offload */
+		rxm->data_off = RTE_PKTMBUF_HEADROOM;
+		rxm->port = rxq->port_id;
+		offload_type = cqe.offload_type;
+
+		/* vlan offload */
+		rxm->ol_flags |= hinic_rx_vlan(offload_type, vlan_len,
+					       &rxm->vlan_tci);
+
+		/* checksum offload */
+		rxm->ol_flags |= hinic_rx_csum(cqe.status, rxq);
+
+		/* rss hash offload */
+		rss_hash = cqe.rss_hash;
+		rxm->ol_flags |= hinic_rx_rss_hash(offload_type, rss_hash,
+						   &rxm->hash.rss);
+
+		/* packet type parser offload */
+		rxm->packet_type = hinic_rx_pkt_type(offload_type);
+
+		/* 6. clear done bit */
+		rx_cqe->status = 0;
+
+		rx_bytes += pkt_len;
+		rx_pkts[pkts++] = rxm;
+	}
+
+	if (pkts) {
+		/* 7. update ci */
+		HINIC_UPDATE_RQ_LOCAL_CI(rxq, wqebb_cnt);
+
+		/* do packet stats */
+		rxq->rxq_stats.packets += pkts;
+		rxq->rxq_stats.bytes += rx_bytes;
+#ifdef HINIC_XSTAT_MBUF_USE
+		rxq->rxq_stats.free_mbuf += pkts;
+#endif
+	}
+
+#ifdef HINIC_XSTAT_RXBUF_INFO
+	rxq->rxq_stats.burst_pkts = pkts;
+#endif
+
+	/* 8. rearm mbuf to rxq */
+	hinic_rearm_rxq_mbuf(rxq);
+
+#ifdef HINIC_XSTAT_PROF_RX
+	/* do profiling stats */
+	t2 = rte_get_tsc_cycles();
+	rxq->rxq_stats.app_tsc = t1 - rxq->prof_rx_end_tsc;
+	rxq->prof_rx_end_tsc = t2;
+	rxq->rxq_stats.pmd_tsc = t2 - t1;
+#endif
+
+	return pkts;
+}
-- 
2.18.0


  parent reply	other threads:[~2019-05-29  3:39 UTC|newest]

Thread overview: 13+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-05-29  3:45 [dpdk-dev] [PATCH v2 00/11] A new net PMD - hinic Ziyang Xuan
2019-05-29  3:46 ` [dpdk-dev] [PATCH v2 01/11] net/hinic/base: add registers for Huawei Hi1822 NIC Ziyang Xuan
2019-05-29  3:48 ` [dpdk-dev] [PATCH v2 02/11] net/hinic/base: add command channels code Ziyang Xuan
2019-05-29  3:49 ` [dpdk-dev] [PATCH v2 03/11] net/hinic/base: add mgmt module interactive code Ziyang Xuan
2019-05-29  3:49 ` [dpdk-dev] [PATCH v2 04/11] net/hinic/base: add code about hardware operation Ziyang Xuan
2019-05-29  3:49 ` [dpdk-dev] [PATCH v2 05/11] net/hinic/base: add eqs and context code Ziyang Xuan
2019-05-29  3:50 ` [dpdk-dev] [PATCH v2 06/11] net/hinic/base: add code for nic business Ziyang Xuan
2019-05-29  3:50 ` [dpdk-dev] [PATCH v2 07/11] net/hinic/base: add various headers Ziyang Xuan
2019-05-29  3:50 ` [dpdk-dev] [PATCH v2 08/11] net/hinic: add hinic PMD build and doc files Ziyang Xuan
2019-05-29  4:59   ` Stephen Hemminger
2019-05-29  3:50 ` Ziyang Xuan [this message]
2019-05-29  3:51 ` [dpdk-dev] [PATCH v2 10/11] net/hinic: add TX module Ziyang Xuan
2019-05-29  3:51 ` [dpdk-dev] [PATCH v2 11/11] net/hinic: add support for basic device operations Ziyang Xuan

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=81ad45e892c7937cc5751686e30601cd7e90598c.1559100649.git.xuanziyang2@huawei.com \
    --to=xuanziyang2@huawei.com \
    --cc=cloud.wangxiaoyun@huawei.com \
    --cc=dev@dpdk.org \
    --cc=ferruh.yigit@intel.com \
    --cc=luoxianjun@huawei.com \
    --cc=shahar.belkar@huawei.com \
    --cc=zhouguoyang@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).