Bug 1238 - net/mlx5: RxQ failure with vectorized MPRQ
Summary: net/mlx5: RxQ failure with vectorized MPRQ
Status: UNCONFIRMED
Alias: None
Product: DPDK
Classification: Unclassified
Component: ethdev (show other bugs)
Version: 23.03
Hardware: x86 Linux
: Normal normal
Target Milestone: ---
Assignee: dev
URL:
Depends on:
Blocks:
 
Reported: 2023-05-26 02:31 CEST by Dmitry Kozlyuk
Modified: 2023-05-26 02:31 CEST (History)
0 users


Attachments
Add an attachment (proposed patch, testcase, etc.)

Description Dmitry Kozlyuk 2023-05-26 02:31:02 CEST 
CPU: Intel(R) Xeon(R) Gold 6238R CPU @ 2.20GHz
NIC: Mellanox Technologies MT28800 Family [ConnectX-5 Ex]
OFED: MLNX_OFED_LINUX-23.04-0.5.3.3
FW: 16.35.2000
Kernel: 5.10.0-21-amd64

Commit 633684e0 "net/mlx5: fix error CQE dumping for vectorized Rx" presumably introduced a regression: Rx queues eventually stop with "CQ error on CQN 0x57d, syndrome 0x1" in dmesg. There is unfortunately no repro with testpmd yet. The error happens under certain CPU load from an app and with a different delay for each RxQ, which suggests that this may be a concurrency issue.
Note You need to log in before you can comment on or make changes to this bug.