From: Chengwen Feng <fengcheng...@huawei.com>

The vector path processes multiple descriptors at a time and modifies
the corresponding mbufs accordingly. When the mbufs cannot be applied
for, the previous patch [1] only clears the first descriptor's valid
bit, which is not enough. In this patch, the mbufs points to the
fake_mbuf to prevent overwrite.

[1] 'commit 01843ab2f2fc ("net/hns3: fix crash for NEON and SVE")'

Fixes: a3d4f4d291d7 ("net/hns3: support NEON Rx")
Fixes: 01843ab2f2fc ("net/hns3: fix crash for NEON and SVE")
Cc: sta...@dpdk.org

Signed-off-by: Chengwen Feng <fengcheng...@huawei.com>
Signed-off-by: Dengdui Huang <huangdeng...@huawei.com>
---
 drivers/net/hns3/hns3_rxtx_vec.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/hns3/hns3_rxtx_vec.h b/drivers/net/hns3/hns3_rxtx_vec.h
index 9018e79c2f..6afa2fb814 100644
--- a/drivers/net/hns3/hns3_rxtx_vec.h
+++ b/drivers/net/hns3/hns3_rxtx_vec.h
@@ -109,8 +109,12 @@ hns3_rxq_rearm_mbuf(struct hns3_rx_queue *rxq)
                /*
                 * Clear VLD bit for the first descriptor rearmed in case
                 * of going to receive packets later.
+                * And also point mbufs to fake_mbuf to prevent modification
+                * of the mbuf field during vector packet receiving.
                 */
                rxdp[0].rx.bd_base_info = 0;
+               for (i = 0; i < HNS3_VECTOR_RX_OFFSET_TABLE_LEN; i++)
+                       rxep[i].mbuf = &rxq->fake_mbuf;
                rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
                return;
        }
-- 
2.33.0

Reply via email to