DPDK patches and discussions
 help / color / mirror / Atom feed
From: Jie Hai <haijie1@huawei.com>
To: <dev@dpdk.org>
Cc: <lihuisong@huawei.com>, <fengchengwen@huawei.com>,
	<liuyonglong@huawei.com>, <huangdengdui@huawei.com>,
	<ferruh.yigit@amd.com>, <haijie1@huawei.com>
Subject: [PATCH] net/hns3: fix Rx packet truncation when KEEP CRC enabled
Date: Tue, 6 Feb 2024 09:10:30 +0800	[thread overview]
Message-ID: <20240206011030.2007689-1-haijie1@huawei.com> (raw)

From: Dengdui Huang <huangdengdui@huawei.com>

When KEEP_CRC offload is enabled, some packets will be truncated and
the CRC is still be stripped in following cases:
1. For HIP08 hardware, the packet type is TCP and the length
   is less than or equal to 60B.
2. For other hardwares, the packet type is IP and the length
   is less than or equal to 60B.

So driver has to recaculate positively the CRC of these packets
on above cases.

In addition, to avoid impacting performance, KEEP_CRC is not
supported when NEON or SVE algorithm is used.

Fixes: 8973d7c4ca12 ("net/hns3: support keeping CRC")
Cc: stable@dpdk.org

Signed-off-by: Dengdui Huang <huangdengdui@huawei.com>
Signed-off-by: Jie Hai <haijie1@huawei.com>
---
 drivers/net/hns3/hns3_rxtx.c          | 116 ++++++++++++++++++++------
 drivers/net/hns3/hns3_rxtx.h          |   9 ++
 drivers/net/hns3/hns3_rxtx_vec.c      |   3 +-
 drivers/net/hns3/hns3_rxtx_vec_neon.h |  19 -----
 drivers/net/hns3/hns3_rxtx_vec_sve.c  |   3 +-
 5 files changed, 103 insertions(+), 47 deletions(-)

diff --git a/drivers/net/hns3/hns3_rxtx.c b/drivers/net/hns3/hns3_rxtx.c
index 7e636a0a2e99..6e9b6db843f9 100644
--- a/drivers/net/hns3/hns3_rxtx.c
+++ b/drivers/net/hns3/hns3_rxtx.c
@@ -11,6 +11,7 @@
 #include <rte_io.h>
 #include <rte_net.h>
 #include <rte_malloc.h>
+#include <rte_net_crc.h>
 #if defined(RTE_ARCH_ARM64)
 #include <rte_cpuflags.h>
 #include <rte_vect.h>
@@ -1732,8 +1733,9 @@ hns3_rx_buf_len_calc(struct rte_mempool *mp, uint16_t *rx_buf_len)
 }
 
 static int
-hns3_rxq_conf_runtime_check(struct hns3_hw *hw, uint16_t buf_size,
-				uint16_t nb_desc)
+hns3_rxq_conf_runtime_check(struct hns3_hw *hw,
+			    const struct rte_eth_rxconf *conf,
+			    uint16_t buf_size, uint16_t nb_desc)
 {
 	struct rte_eth_dev *dev = &rte_eth_devices[hw->data->port_id];
 	eth_rx_burst_t pkt_burst = dev->rx_pkt_burst;
@@ -1766,6 +1768,14 @@ hns3_rxq_conf_runtime_check(struct hns3_hw *hw, uint16_t buf_size,
 			return -EINVAL;
 		}
 	}
+
+	if ((conf->offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) &&
+	    pkt_burst != hns3_recv_pkts_simple &&
+	    pkt_burst != hns3_recv_scattered_pkts) {
+		hns3_err(hw, "KEEP_CRC offload is not supported in the current rx function.");
+		return -EINVAL;
+	}
+
 	return 0;
 }
 
@@ -1802,7 +1812,7 @@ hns3_rx_queue_conf_check(struct hns3_hw *hw, const struct rte_eth_rxconf *conf,
 	}
 
 	if (hw->data->dev_started) {
-		ret = hns3_rxq_conf_runtime_check(hw, *buf_size, nb_desc);
+		ret = hns3_rxq_conf_runtime_check(hw, conf, *buf_size, nb_desc);
 		if (ret) {
 			hns3_err(hw, "Rx queue runtime setup fail.");
 			return ret;
@@ -1917,6 +1927,8 @@ hns3_rx_queue_setup(struct rte_eth_dev *dev, uint16_t idx, uint16_t nb_desc,
 	memset(&rxq->err_stats, 0, sizeof(struct hns3_rx_bd_errors_stats));
 	memset(&rxq->dfx_stats, 0, sizeof(struct hns3_rx_dfx_stats));
 
+	if (hw->revision < PCI_REVISION_ID_HIP09_A)
+		rxq->keep_crc_fail_only_tcp = true;
 	/* CRC len set here is used for amending packet length */
 	if (dev->data->dev_conf.rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)
 		rxq->crc_len = RTE_ETHER_CRC_LEN;
@@ -2400,6 +2412,47 @@ hns3_rx_ptp_timestamp_handle(struct hns3_rx_queue *rxq, struct rte_mbuf *mbuf,
 	pf->rx_timestamp = timestamp;
 }
 
+static inline bool
+hns3_need_recalculate_crc(struct hns3_rx_queue *rxq, struct rte_mbuf *m)
+{
+	uint32_t ptype = m->packet_type;
+
+	if (m->pkt_len > HNS3_KEEP_CRC_OK_MIN_PKT_LEN)
+		return false;
+
+	if (!(RTE_ETH_IS_IPV4_HDR(ptype) || RTE_ETH_IS_IPV6_HDR(ptype)))
+		return false;
+
+	if (rxq->keep_crc_fail_only_tcp)
+		return (ptype & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_TCP;
+
+	return true;
+}
+
+static inline void
+hns3_recalculate_crc(struct hns3_rx_queue *rxq, struct rte_mbuf *m)
+{
+	char *append_data;
+	uint32_t crc;
+
+	crc = rte_net_crc_calc(rte_pktmbuf_mtod(m, void *),
+			       m->data_len, RTE_NET_CRC32_ETH);
+
+	/*
+	 * The hns3 driver requires that mbuf size must be at least 512B.
+	 * When CRC is stripped by hardware, the pkt_len must be less than
+	 * or equal to 60B. Therefore, the space of the mbuf is enough
+	 * to insert the CRC.
+	 *
+	 * In addition, after CRC is stripped by hardware, pkt_len and data_len
+	 * do not contain the CRC length. Therefore, after CRC data is appended
+	 * by PMD again, both pkt_len and data_len add the CRC length.
+	 */
+	append_data = rte_pktmbuf_append(m, rxq->crc_len);
+	/* The CRC data is binary data and does not care about the byte order. */
+	rte_memcpy(append_data, (void *)&crc, rxq->crc_len);
+}
+
 uint16_t
 hns3_recv_pkts_simple(void *rx_queue,
 		      struct rte_mbuf **rx_pkts,
@@ -2470,8 +2523,7 @@ hns3_recv_pkts_simple(void *rx_queue,
 		rxdp->rx.bd_base_info = 0;
 
 		rxm->data_off = RTE_PKTMBUF_HEADROOM;
-		rxm->pkt_len = (uint16_t)(rte_le_to_cpu_16(rxd.rx.pkt_len)) -
-				rxq->crc_len;
+		rxm->pkt_len = rte_le_to_cpu_16(rxd.rx.pkt_len);
 		rxm->data_len = rxm->pkt_len;
 		rxm->port = rxq->port_id;
 		rxm->hash.rss = rte_le_to_cpu_32(rxd.rx.rss_hash);
@@ -2492,10 +2544,16 @@ hns3_recv_pkts_simple(void *rx_queue,
 			goto pkt_err;
 
 		rxm->packet_type = hns3_rx_calc_ptype(rxq, l234_info, ol_info);
-
 		if (rxm->packet_type == RTE_PTYPE_L2_ETHER_TIMESYNC)
 			rxm->ol_flags |= RTE_MBUF_F_RX_IEEE1588_PTP;
 
+		if (unlikely(rxq->crc_len > 0)) {
+			if (hns3_need_recalculate_crc(rxq, rxm))
+				hns3_recalculate_crc(rxq, rxm);
+			rxm->pkt_len -= rxq->crc_len;
+			rxm->data_len -= rxq->crc_len;
+		}
+
 		hns3_rxd_to_vlan_tci(rxq, rxm, l234_info, &rxd);
 
 		/* Increment bytes counter  */
@@ -2662,10 +2720,10 @@ hns3_recv_scattered_pkts(void *rx_queue,
 
 		rxm->data_off = RTE_PKTMBUF_HEADROOM;
 		rxm->data_len = rte_le_to_cpu_16(rxd.rx.size);
+		rxm->next = NULL;
 
 		if (!(bd_base_info & BIT(HNS3_RXD_FE_B))) {
 			last_seg = rxm;
-			rxm->next = NULL;
 			continue;
 		}
 
@@ -2679,23 +2737,6 @@ hns3_recv_scattered_pkts(void *rx_queue,
 		 */
 		first_seg->pkt_len = rte_le_to_cpu_16(rxd.rx.pkt_len);
 
-		/*
-		 * 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 -= rxq->crc_len;
-			recalculate_data_len(first_seg, last_seg, rxm, rxq,
-				rxm->data_len);
-		}
-
 		first_seg->port = rxq->port_id;
 		first_seg->hash.rss = rte_le_to_cpu_32(rxd.rx.rss_hash);
 		first_seg->ol_flags = RTE_MBUF_F_RX_RSS_HASH;
@@ -2721,10 +2762,35 @@ hns3_recv_scattered_pkts(void *rx_queue,
 
 		first_seg->packet_type = hns3_rx_calc_ptype(rxq,
 						l234_info, ol_info);
-
 		if (first_seg->packet_type == RTE_PTYPE_L2_ETHER_TIMESYNC)
 			rxm->ol_flags |= RTE_MBUF_F_RX_IEEE1588_PTP;
 
+		/*
+		 * 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.
+		 *
+		 * In addition, the CRC is still stripped for a kind of packets:
+		 * 1. All IP-TCP packet whose the length is less than and equal
+		 *    to 60 Byte (no CRC) on HIP08 hardware.
+		 * 2. All IP packet whose the length is less than and equal to
+		 *    60 Byte (no CRC) on other hardware.
+		 * In this case, the PMD calculates the CRC and appends it to
+		 * mbuf.
+		 */
+		if (unlikely(rxq->crc_len > 0)) {
+			if (hns3_need_recalculate_crc(rxq, first_seg))
+				hns3_recalculate_crc(rxq, first_seg);
+			first_seg->pkt_len -= rxq->crc_len;
+			recalculate_data_len(first_seg, last_seg, rxm, rxq,
+				rxm->data_len);
+		}
+
 		hns3_rxd_to_vlan_tci(rxq, first_seg, l234_info, &rxd);
 
 		/* Increment bytes counter */
diff --git a/drivers/net/hns3/hns3_rxtx.h b/drivers/net/hns3/hns3_rxtx.h
index e2ad42bb8e30..8b70a8238ae0 100644
--- a/drivers/net/hns3/hns3_rxtx.h
+++ b/drivers/net/hns3/hns3_rxtx.h
@@ -178,6 +178,8 @@
 		(HNS3_TXD_VLD_CMD | HNS3_TXD_FE_CMD | HNS3_TXD_DEFAULT_BDTYPE)
 #define HNS3_TXD_SEND_SIZE_SHIFT	16
 
+#define HNS3_KEEP_CRC_OK_MIN_PKT_LEN	60
+
 enum hns3_pkt_l2t_type {
 	HNS3_L2_TYPE_UNICAST,
 	HNS3_L2_TYPE_MULTICAST,
@@ -328,6 +330,13 @@ struct hns3_rx_queue {
 	/* 4 if RTE_ETH_RX_OFFLOAD_KEEP_CRC offload set, 0 otherwise */
 	uint8_t crc_len;
 
+	/*
+	 * Indicate the range of packet type that fail to keep CRC.
+	 * For HIP08 hardware, only the IP-TCP packet type is included.
+	 * For other hardwares, all IP packet type is included.
+	 */
+	uint8_t keep_crc_fail_only_tcp:1;
+
 	/*
 	 * Indicate whether ignore the outer VLAN field in the Rx BD reported
 	 * by the Hardware. Because the outer VLAN is the PVID if the PVID is
diff --git a/drivers/net/hns3/hns3_rxtx_vec.c b/drivers/net/hns3/hns3_rxtx_vec.c
index 9708ec614e02..bf37ce51b1ad 100644
--- a/drivers/net/hns3/hns3_rxtx_vec.c
+++ b/drivers/net/hns3/hns3_rxtx_vec.c
@@ -185,7 +185,8 @@ hns3_rx_check_vec_support(struct rte_eth_dev *dev)
 	struct rte_eth_rxmode *rxmode = &dev->data->dev_conf.rxmode;
 	uint64_t offloads_mask = RTE_ETH_RX_OFFLOAD_TCP_LRO |
 				 RTE_ETH_RX_OFFLOAD_VLAN |
-				 RTE_ETH_RX_OFFLOAD_TIMESTAMP;
+				 RTE_ETH_RX_OFFLOAD_TIMESTAMP |
+				 RTE_ETH_RX_OFFLOAD_KEEP_CRC;
 
 	if (dev->data->scattered_rx)
 		return -ENOTSUP;
diff --git a/drivers/net/hns3/hns3_rxtx_vec_neon.h b/drivers/net/hns3/hns3_rxtx_vec_neon.h
index 0dc6b9f0a22d..60ec501a2ab6 100644
--- a/drivers/net/hns3/hns3_rxtx_vec_neon.h
+++ b/drivers/net/hns3/hns3_rxtx_vec_neon.h
@@ -148,14 +148,6 @@ hns3_recv_burst_vec(struct hns3_rx_queue *__restrict rxq,
 		8, 9, 10, 11,	         /* rx.rss_hash to rte_mbuf.hash.rss */
 	};
 
-	uint16x8_t crc_adjust = {
-		0, 0,         /* ignore pkt_type field */
-		rxq->crc_len, /* sub crc on pkt_len */
-		0,            /* ignore high-16bits of pkt_len */
-		rxq->crc_len, /* sub crc on data_len */
-		0, 0, 0,      /* ignore non-length fields */
-	};
-
 	/* compile-time verifies the shuffle mask */
 	RTE_BUILD_BUG_ON(offsetof(struct rte_mbuf, pkt_len) !=
 			 offsetof(struct rte_mbuf, rx_descriptor_fields1) + 4);
@@ -171,7 +163,6 @@ hns3_recv_burst_vec(struct hns3_rx_queue *__restrict rxq,
 		uint8x16_t pkt_mb1, pkt_mb2, pkt_mb3, pkt_mb4;
 		uint64x2_t mbp1, mbp2;
 		uint16x4_t bd_vld = {0};
-		uint16x8_t tmp;
 		uint64_t stat;
 
 		/* calc how many bd valid */
@@ -225,16 +216,6 @@ hns3_recv_burst_vec(struct hns3_rx_queue *__restrict rxq,
 		pkt_mb3 = vqtbl2q_u8(pkt_mbuf3, shuf_desc_fields_msk);
 		pkt_mb4 = vqtbl2q_u8(pkt_mbuf4, shuf_desc_fields_msk);
 
-		/* 4 packets remove crc */
-		tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb1), crc_adjust);
-		pkt_mb1 = vreinterpretq_u8_u16(tmp);
-		tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb2), crc_adjust);
-		pkt_mb2 = vreinterpretq_u8_u16(tmp);
-		tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb3), crc_adjust);
-		pkt_mb3 = vreinterpretq_u8_u16(tmp);
-		tmp = vsubq_u16(vreinterpretq_u16_u8(pkt_mb4), crc_adjust);
-		pkt_mb4 = vreinterpretq_u8_u16(tmp);
-
 		/* save packet info to rx_pkts mbuf */
 		vst1q_u8((void *)&sw_ring[pos + 0].mbuf->rx_descriptor_fields1,
 			 pkt_mb1);
diff --git a/drivers/net/hns3/hns3_rxtx_vec_sve.c b/drivers/net/hns3/hns3_rxtx_vec_sve.c
index 8aa4448558cf..67c87f570e8a 100644
--- a/drivers/net/hns3/hns3_rxtx_vec_sve.c
+++ b/drivers/net/hns3/hns3_rxtx_vec_sve.c
@@ -36,8 +36,7 @@ hns3_desc_parse_field_sve(struct hns3_rx_queue *rxq,
 		/* init rte_mbuf.rearm_data last 64-bit */
 		rx_pkts[i]->ol_flags = RTE_MBUF_F_RX_RSS_HASH;
 		rx_pkts[i]->hash.rss = rxdp[i].rx.rss_hash;
-		rx_pkts[i]->pkt_len = rte_le_to_cpu_16(rxdp[i].rx.pkt_len) -
-					rxq->crc_len;
+		rx_pkts[i]->pkt_len = rte_le_to_cpu_16(rxdp[i].rx.pkt_len);
 		rx_pkts[i]->data_len = rx_pkts[i]->pkt_len;
 
 		l234_info = rxdp[i].rx.l234_info;
-- 
2.30.0


             reply	other threads:[~2024-02-06  1:15 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-02-06  1:10 Jie Hai [this message]
2024-02-07 14:15 ` Ferruh Yigit
2024-02-20  3:58   ` Jie Hai
2024-02-23 13:53     ` Ferruh Yigit
2024-02-26  3:16       ` Jie Hai
2024-02-26 16:43         ` Ferruh Yigit
2024-02-28  2:27           ` huangdengdui
2024-02-28 13:07             ` Ferruh Yigit
2024-02-29  3:58               ` huangdengdui
2024-02-29  9:25                 ` Ferruh Yigit
2024-03-01  6:55                   ` huangdengdui
2024-03-01 11:10                     ` Ferruh Yigit
2024-03-08 11:36                       ` Jie Hai
2024-03-22  6:28                         ` Jie Hai

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=20240206011030.2007689-1-haijie1@huawei.com \
    --to=haijie1@huawei.com \
    --cc=dev@dpdk.org \
    --cc=fengchengwen@huawei.com \
    --cc=ferruh.yigit@amd.com \
    --cc=huangdengdui@huawei.com \
    --cc=lihuisong@huawei.com \
    --cc=liuyonglong@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).