Skip to contentSkip to conversation

Many PMD's provide receive timestamp but are missing the read_clock ethdev operation

DPDK/ethdev Bug 1960 [email protected] 1 participant 1 message Jun 30

lore
Share & tools
lore
#0 lore raw

http://bugs.dpdk.org/show_bug.cgi?id=1960

Bug ID: 1960
Summary: Many PMD's provide receive timestamp but are missing
the read_clock ethdev operation
Product: DPDK
Version: 22.03
Hardware: All
OS: All
Status: UNCONFIRMED
Severity: normal
Priority: Normal
Component: ethdev
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---

The implementation of receive timestamp is inconsistent in DPDK. If a driver advertises RX timestamp offload it must also have the facility to tell application what units that timestamp is in. Some searching showed.

Summary

You're absolutely right - many DPDK drivers that advertise
RTE_ETH_RX_OFFLOAD_TIMESTAMP capability don't implement the
read_clock ethdev operation properly. Here's the breakdown:

Drivers with PROPER implementation (have both timestamp support and
read_clock):

- cnxk - Marvell OCTEON
- mlx5 - Mellanox/NVIDIA
- pcap - Software pcap driver
- intel/e1000 (igb, igc) - Intel 1G NICs
- intel/ice - Intel E810 NICs

Drivers MISSING read_clock implementation despite supporting timestamps:

- af_packet - Linux AF_PACKET
- ark - Atomic Rules Arkville
- bnxt - Broadcom NetXtreme
- dpaa2 - NXP DPAA2
- ena - Amazon ENA
- failsafe - Fail-safe PMD
- hns3 - Huawei HiSilicon
- nfb - Cesnet NFB
- sxe2 - Wangxun SXE2000
- xsc - Yunsilicon XSC
- intel/cpfl - Intel Infrastructure Processing Unit
- intel/iavf - Intel Adaptive Virtual Function
- intel/idpf - Intel Infrastructure Data Path Function

Signature · 2 lines
You are receiving this mail because:
You are the assignee for the bug.