Callback or hook into completion queues?

Callback or hook into completion queues?

[fwefew 4t4tg]

[-- Attachment #1: Type: text/plain, Size: 774 bytes --]

Once I call rte_eth_tx_burst() (Mellanox Connect5-LX) is there a way to
inspect or get a callback when transmitted packets go into the NIC's
completion queue?

This is related to the earlier question on timestamping with rdtsc().
Ideally I'd take the timestamp as soon (close) to the time the packet is on
the wire.

I've looked at tx callbacks however this is invoked as the packet is about
to go into the "hardware queue for transmission" meaning there's lot's of
work + serialization of packet's data to electrical signals at NIC
bandwidth to come before the packet is on the wire.

The ideal time to get run rdtsc() is when the NIC delivers a completion
event to a CQ for packets sent.

Presumably there's something in the mlx5 driver or perhaps DMA library to
do this?

[-- Attachment #2: Type: text/html, Size: 942 bytes --]

Re: Callback or hook into completion queues?

[Stephen Hemminger]

On Tue, 7 Mar 2023 17:07:41 -0500
fwefew 4t4tg <7532yahoo@gmail.com> wrote:

> Once I call rte_eth_tx_burst() (Mellanox Connect5-LX) is there a way to
> inspect or get a callback when transmitted packets go into the NIC's
> completion queue?

Not in DPDK.

> 
> This is related to the earlier question on timestamping with rdtsc().
> Ideally I'd take the timestamp as soon (close) to the time the packet is on
> the wire.


> 
> I've looked at tx callbacks however this is invoked as the packet is about
> to go into the "hardware queue for transmission" meaning there's lot's of
> work + serialization of packet's data to electrical signals at NIC
> bandwidth to come before the packet is on the wire.
> 
> The ideal time to get run rdtsc() is when the NIC delivers a completion
> event to a CQ for packets sent.
> 
> Presumably there's something in the mlx5 driver or perhaps DMA library to
> do this?

Not that I have ever seen.