On 2026-02-05 08:54:34 [+0100], Kurt Kanzenbach wrote:
> Retrieve Tx timestamp directly from interrupt handler for i210.
> 
> The current implementation uses schedule_work() which is executed by the
> system work queue to retrieve Tx timestamps. This increases latency and can
> lead to timeouts in case of heavy system load. i210 is often used in
> industrial systems, where timestamp timeouts can be fatal.
> 
> Therefore, fetch the timestamp directly from the interrupt handler.
> 
> The work queue code stays for all other NICs supported by igb.
> 
> Tested on Intel i210 and i350.
> 
> Signed-off-by: Kurt Kanzenbach <[email protected]>

Reviewed-by: Sebastian Andrzej Siewior <[email protected]>

IMHO this is a compromise with Miroslav where he observed less PTP
timestamps on the i350. While testing I did not get near Miroslav's
difference but there was a small change. I don't understand *why*
because the current workqueue usage reads the timestamp on the same CPU
on which the interrupt occurred. Doing it directly just avoids the
context switch. This feels beneficial.

Sebastian

Reply via email to