From: Jiawen Wu <jiawenwu@trustnetic.com>
To: dev@dpdk.org
Cc: Jiawen Wu <jiawenwu@trustnetic.com>
Subject: [dpdk-dev] [PATCH v7 17/19] net/ngbe: add simple Rx flow
Date: Tue, 6 Jul 2021 17:55:43 +0800 [thread overview]
Message-ID: <20210706095545.10776-18-jiawenwu@trustnetic.com> (raw)
In-Reply-To: <20210706095545.10776-1-jiawenwu@trustnetic.com>
Initialize device with the simplest receive function.
Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
---
drivers/net/ngbe/ngbe_ethdev.c | 1 +
drivers/net/ngbe/ngbe_ethdev.h | 3 +
drivers/net/ngbe/ngbe_rxtx.c | 169 +++++++++++++++++++++++++++++++++
drivers/net/ngbe/ngbe_rxtx.h | 73 ++++++++++++++
4 files changed, 246 insertions(+)
diff --git a/drivers/net/ngbe/ngbe_ethdev.c b/drivers/net/ngbe/ngbe_ethdev.c
index 2a40dbd184..e5e2f44454 100644
--- a/drivers/net/ngbe/ngbe_ethdev.c
+++ b/drivers/net/ngbe/ngbe_ethdev.c
@@ -137,6 +137,7 @@ eth_ngbe_dev_init(struct rte_eth_dev *eth_dev, void *init_params __rte_unused)
PMD_INIT_FUNC_TRACE();
eth_dev->dev_ops = &ngbe_eth_dev_ops;
+ eth_dev->rx_pkt_burst = &ngbe_recv_pkts;
if (rte_eal_process_type() != RTE_PROC_PRIMARY)
return 0;
diff --git a/drivers/net/ngbe/ngbe_ethdev.h b/drivers/net/ngbe/ngbe_ethdev.h
index 1d71d28023..ff366d6a5e 100644
--- a/drivers/net/ngbe/ngbe_ethdev.h
+++ b/drivers/net/ngbe/ngbe_ethdev.h
@@ -99,6 +99,9 @@ int ngbe_dev_tx_queue_start(struct rte_eth_dev *dev, uint16_t tx_queue_id);
int ngbe_dev_tx_queue_stop(struct rte_eth_dev *dev, uint16_t tx_queue_id);
+uint16_t ngbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts);
+
void ngbe_set_ivar_map(struct ngbe_hw *hw, int8_t direction,
uint8_t queue, uint8_t msix_vector);
diff --git a/drivers/net/ngbe/ngbe_rxtx.c b/drivers/net/ngbe/ngbe_rxtx.c
index 2622f562c4..e72c74869a 100644
--- a/drivers/net/ngbe/ngbe_rxtx.c
+++ b/drivers/net/ngbe/ngbe_rxtx.c
@@ -15,6 +15,175 @@
#include "ngbe_ethdev.h"
#include "ngbe_rxtx.h"
+/*
+ * Prefetch a cache line into all cache levels.
+ */
+#define rte_ngbe_prefetch(p) rte_prefetch0(p)
+
+/*********************************************************************
+ *
+ * Rx functions
+ *
+ **********************************************************************/
+uint16_t
+ngbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+ uint16_t nb_pkts)
+{
+ struct ngbe_rx_queue *rxq;
+ volatile struct ngbe_rx_desc *rx_ring;
+ volatile struct ngbe_rx_desc *rxdp;
+ struct ngbe_rx_entry *sw_ring;
+ struct ngbe_rx_entry *rxe;
+ struct rte_mbuf *rxm;
+ struct rte_mbuf *nmb;
+ struct ngbe_rx_desc rxd;
+ uint64_t dma_addr;
+ uint32_t staterr;
+ uint16_t pkt_len;
+ uint16_t rx_id;
+ uint16_t nb_rx;
+ uint16_t nb_hold;
+
+ nb_rx = 0;
+ nb_hold = 0;
+ rxq = rx_queue;
+ rx_id = rxq->rx_tail;
+ rx_ring = rxq->rx_ring;
+ sw_ring = rxq->sw_ring;
+ struct rte_eth_dev *dev = &rte_eth_devices[rxq->port_id];
+ while (nb_rx < nb_pkts) {
+ /*
+ * The order of operations here is important as the DD status
+ * bit must not be read after any other descriptor fields.
+ * rx_ring and rxdp are pointing to volatile data so the order
+ * of accesses cannot be reordered by the compiler. If they were
+ * not volatile, they could be reordered which could lead to
+ * using invalid descriptor fields when read from rxd.
+ */
+ rxdp = &rx_ring[rx_id];
+ staterr = rxdp->qw1.lo.status;
+ if (!(staterr & rte_cpu_to_le_32(NGBE_RXD_STAT_DD)))
+ break;
+ rxd = *rxdp;
+
+ /*
+ * End of packet.
+ *
+ * If the NGBE_RXD_STAT_EOP flag is not set, the Rx packet
+ * is likely to be invalid and to be dropped by the various
+ * validation checks performed by the network stack.
+ *
+ * Allocate a new mbuf to replenish the RX ring descriptor.
+ * If the allocation fails:
+ * - arrange for that Rx descriptor to be the first one
+ * being parsed the next time the receive function is
+ * invoked [on the same queue].
+ *
+ * - Stop parsing the Rx ring and return immediately.
+ *
+ * This policy do not drop the packet received in the Rx
+ * descriptor for which the allocation of a new mbuf failed.
+ * Thus, it allows that packet to be later retrieved if
+ * mbuf have been freed in the mean time.
+ * As a side effect, holding Rx descriptors instead of
+ * systematically giving them back to the NIC may lead to
+ * Rx ring exhaustion situations.
+ * However, the NIC can gracefully prevent such situations
+ * to happen by sending specific "back-pressure" flow control
+ * frames to its peer(s).
+ */
+ PMD_RX_LOG(DEBUG,
+ "port_id=%u queue_id=%u rx_id=%u ext_err_stat=0x%08x pkt_len=%u",
+ (uint16_t)rxq->port_id, (uint16_t)rxq->queue_id,
+ (uint16_t)rx_id, (uint32_t)staterr,
+ (uint16_t)rte_le_to_cpu_16(rxd.qw1.hi.len));
+
+ 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",
+ (uint16_t)rxq->port_id,
+ (uint16_t)rxq->queue_id);
+ dev->data->rx_mbuf_alloc_failed++;
+ break;
+ }
+
+ nb_hold++;
+ rxe = &sw_ring[rx_id];
+ rx_id++;
+ if (rx_id == rxq->nb_rx_desc)
+ rx_id = 0;
+
+ /* Prefetch next mbuf while processing current one. */
+ rte_ngbe_prefetch(sw_ring[rx_id].mbuf);
+
+ /*
+ * When next Rx descriptor is on a cache-line boundary,
+ * prefetch the next 4 Rx descriptors and the next 8 pointers
+ * to mbufs.
+ */
+ if ((rx_id & 0x3) == 0) {
+ rte_ngbe_prefetch(&rx_ring[rx_id]);
+ rte_ngbe_prefetch(&sw_ring[rx_id]);
+ }
+
+ rxm = rxe->mbuf;
+ rxe->mbuf = nmb;
+ dma_addr = rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb));
+ NGBE_RXD_HDRADDR(rxdp, 0);
+ NGBE_RXD_PKTADDR(rxdp, dma_addr);
+
+ /*
+ * Initialize the returned mbuf.
+ * setup generic mbuf fields:
+ * - number of segments,
+ * - next segment,
+ * - packet length,
+ * - Rx port identifier.
+ */
+ pkt_len = (uint16_t)(rte_le_to_cpu_16(rxd.qw1.hi.len));
+ rxm->data_off = RTE_PKTMBUF_HEADROOM;
+ rte_packet_prefetch((char *)rxm->buf_addr + rxm->data_off);
+ rxm->nb_segs = 1;
+ rxm->next = NULL;
+ rxm->pkt_len = pkt_len;
+ rxm->data_len = pkt_len;
+ rxm->port = rxq->port_id;
+
+ /*
+ * Store the mbuf address into the next entry of the array
+ * of returned packets.
+ */
+ 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 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...
+ */
+ 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",
+ (uint16_t)rxq->port_id, (uint16_t)rxq->queue_id,
+ (uint16_t)rx_id, (uint16_t)nb_hold,
+ (uint16_t)nb_rx);
+ rx_id = (uint16_t)((rx_id == 0) ?
+ (rxq->nb_rx_desc - 1) : (rx_id - 1));
+ ngbe_set32(rxq->rdt_reg_addr, rx_id);
+ nb_hold = 0;
+ }
+ rxq->nb_rx_hold = nb_hold;
+ return nb_rx;
+}
+
+
/*********************************************************************
*
* Queue management functions
diff --git a/drivers/net/ngbe/ngbe_rxtx.h b/drivers/net/ngbe/ngbe_rxtx.h
index 668588fbf8..edfba8a672 100644
--- a/drivers/net/ngbe/ngbe_rxtx.h
+++ b/drivers/net/ngbe/ngbe_rxtx.h
@@ -51,6 +51,77 @@ struct ngbe_rx_desc {
#define NGBE_RXD_HDRADDR(rxd, v) \
(((volatile __le64 *)(rxd))[1] = cpu_to_le64(v))
+/* @ngbe_rx_desc.dw0 */
+#define NGBE_RXD_RSSTYPE(dw) RS(dw, 0, 0xF)
+#define NGBE_RSSTYPE_NONE 0
+#define NGBE_RSSTYPE_IPV4TCP 1
+#define NGBE_RSSTYPE_IPV4 2
+#define NGBE_RSSTYPE_IPV6TCP 3
+#define NGBE_RSSTYPE_IPV4SCTP 4
+#define NGBE_RSSTYPE_IPV6 5
+#define NGBE_RSSTYPE_IPV6SCTP 6
+#define NGBE_RSSTYPE_IPV4UDP 7
+#define NGBE_RSSTYPE_IPV6UDP 8
+#define NGBE_RSSTYPE_FDIR 15
+#define NGBE_RXD_SECTYPE(dw) RS(dw, 4, 0x3)
+#define NGBE_RXD_SECTYPE_NONE LS(0, 4, 0x3)
+#define NGBE_RXD_SECTYPE_IPSECESP LS(2, 4, 0x3)
+#define NGBE_RXD_SECTYPE_IPSECAH LS(3, 4, 0x3)
+#define NGBE_RXD_TPIDSEL(dw) RS(dw, 6, 0x7)
+#define NGBE_RXD_PTID(dw) RS(dw, 9, 0xFF)
+#define NGBE_RXD_RSCCNT(dw) RS(dw, 17, 0xF)
+#define NGBE_RXD_HDRLEN(dw) RS(dw, 21, 0x3FF)
+#define NGBE_RXD_SPH MS(31, 0x1)
+
+/* @ngbe_rx_desc.dw1 */
+/** bit 0-31, as rss hash when **/
+#define NGBE_RXD_RSSHASH(rxd) ((rxd)->qw0.dw1)
+
+/** bit 0-31, as ip csum when **/
+#define NGBE_RXD_IPID(rxd) ((rxd)->qw0.hi.ipid)
+#define NGBE_RXD_CSUM(rxd) ((rxd)->qw0.hi.csum)
+
+/* @ngbe_rx_desc.dw2 */
+#define NGBE_RXD_STATUS(rxd) ((rxd)->qw1.lo.status)
+/** bit 0-1 **/
+#define NGBE_RXD_STAT_DD MS(0, 0x1) /* Descriptor Done */
+#define NGBE_RXD_STAT_EOP MS(1, 0x1) /* End of Packet */
+/** bit 2-31, when EOP=0 **/
+#define NGBE_RXD_NEXTP_RESV(v) LS(v, 2, 0x3)
+#define NGBE_RXD_NEXTP(dw) RS(dw, 4, 0xFFFF) /* Next Descriptor */
+/** bit 2-31, when EOP=1 **/
+#define NGBE_RXD_PKT_CLS_MASK MS(2, 0x7) /* Packet Class */
+#define NGBE_RXD_PKT_CLS_TC_RSS LS(0, 2, 0x7) /* RSS Hash */
+#define NGBE_RXD_PKT_CLS_FLM LS(1, 2, 0x7) /* FDir Match */
+#define NGBE_RXD_PKT_CLS_SYN LS(2, 2, 0x7) /* TCP Sync */
+#define NGBE_RXD_PKT_CLS_5TUPLE LS(3, 2, 0x7) /* 5 Tuple */
+#define NGBE_RXD_PKT_CLS_ETF LS(4, 2, 0x7) /* Ethertype Filter */
+#define NGBE_RXD_STAT_VLAN MS(5, 0x1) /* IEEE VLAN Packet */
+#define NGBE_RXD_STAT_UDPCS MS(6, 0x1) /* UDP xsum calculated */
+#define NGBE_RXD_STAT_L4CS MS(7, 0x1) /* L4 xsum calculated */
+#define NGBE_RXD_STAT_IPCS MS(8, 0x1) /* IP xsum calculated */
+#define NGBE_RXD_STAT_PIF MS(9, 0x1) /* Non-unicast address */
+#define NGBE_RXD_STAT_EIPCS MS(10, 0x1) /* Encap IP xsum calculated */
+#define NGBE_RXD_STAT_VEXT MS(11, 0x1) /* Multi-VLAN */
+#define NGBE_RXD_STAT_IPV6EX MS(12, 0x1) /* IPv6 with option header */
+#define NGBE_RXD_STAT_LLINT MS(13, 0x1) /* Pkt caused LLI */
+#define NGBE_RXD_STAT_1588 MS(14, 0x1) /* IEEE1588 Time Stamp */
+#define NGBE_RXD_STAT_SECP MS(15, 0x1) /* Security Processing */
+#define NGBE_RXD_STAT_LB MS(16, 0x1) /* Loopback Status */
+/*** bit 17-30, when PTYPE=IP ***/
+#define NGBE_RXD_STAT_BMC MS(17, 0x1) /* PTYPE=IP, BMC status */
+#define NGBE_RXD_ERR_HBO MS(23, 0x1) /* Header Buffer Overflow */
+#define NGBE_RXD_ERR_EIPCS MS(26, 0x1) /* Encap IP header error */
+#define NGBE_RXD_ERR_SECERR MS(27, 0x1) /* macsec or ipsec error */
+#define NGBE_RXD_ERR_RXE MS(29, 0x1) /* Any MAC Error */
+#define NGBE_RXD_ERR_L4CS MS(30, 0x1) /* TCP/UDP xsum error */
+#define NGBE_RXD_ERR_IPCS MS(31, 0x1) /* IP xsum error */
+#define NGBE_RXD_ERR_CSUM(dw) RS(dw, 30, 0x3)
+
+/* @ngbe_rx_desc.dw3 */
+#define NGBE_RXD_LENGTH(rxd) ((rxd)->qw1.hi.len)
+#define NGBE_RXD_VLAN(rxd) ((rxd)->qw1.hi.tag)
+
/*****************************************************************************
* Transmit Descriptor
*****************************************************************************/
@@ -81,6 +152,8 @@ struct ngbe_tx_desc {
#define RX_RING_SZ ((NGBE_RING_DESC_MAX + RTE_PMD_NGBE_RX_MAX_BURST) * \
sizeof(struct ngbe_rx_desc))
+#define rte_packet_prefetch(p) rte_prefetch1(p)
+
#define RTE_NGBE_REGISTER_POLL_WAIT_10_MS 10
#define RTE_NGBE_WAIT_100_US 100
--
2.21.0.windows.1
next prev parent reply other threads:[~2021-07-06 9:58 UTC|newest]
Thread overview: 32+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-06 9:55 [dpdk-dev] [PATCH v7 00/19] net: ngbe PMD Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 01/19] net/ngbe: add build and doc infrastructure Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 02/19] net/ngbe: support probe and remove Jiawen Wu
2021-07-07 13:48 ` Andrew Rybchenko
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 03/19] net/ngbe: add log type and error type Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 04/19] net/ngbe: define registers Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 05/19] net/ngbe: set MAC type and LAN ID with device initialization Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 06/19] net/ngbe: init and validate EEPROM Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 07/19] net/ngbe: add HW initialization Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 08/19] net/ngbe: identify PHY and reset PHY Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 09/19] net/ngbe: store MAC address Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 10/19] net/ngbe: support link update Jiawen Wu
2021-07-07 13:50 ` Andrew Rybchenko
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 11/19] net/ngbe: setup the check PHY link Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 12/19] net/ngbe: add Rx queue setup and release Jiawen Wu
2021-07-07 13:56 ` Andrew Rybchenko
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 13/19] net/ngbe: add Tx " Jiawen Wu
2021-07-07 14:00 ` Andrew Rybchenko
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 14/19] net/ngbe: add device start and stop operations Jiawen Wu
2021-07-07 14:04 ` Andrew Rybchenko
2021-07-08 6:09 ` Jiawen Wu
2021-07-08 7:36 ` Andrew Rybchenko
2021-07-08 7:50 ` Jiawen Wu
2021-07-08 8:07 ` Andrew Rybchenko
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 15/19] net/ngbe: add Tx queue start and stop Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 16/19] net/ngbe: add Rx " Jiawen Wu
2021-07-06 9:55 ` Jiawen Wu [this message]
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 18/19] net/ngbe: add simple Tx flow Jiawen Wu
2021-07-06 9:55 ` [dpdk-dev] [PATCH v7 19/19] net/ngbe: support to close and reset device Jiawen Wu
2021-07-07 14:12 ` Andrew Rybchenko
2021-07-08 8:56 ` Jiawen Wu
2021-07-07 14:13 ` [dpdk-dev] [PATCH v7 00/19] net: ngbe PMD Andrew Rybchenko
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=20210706095545.10776-18-jiawenwu@trustnetic.com \
--to=jiawenwu@trustnetic.com \
--cc=dev@dpdk.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.