On Tue, Mar 24, 2020 at 05:34:34PM +0200, Grygorii Strashko wrote:
I tested both ways and kept this version as i'v not seen any degradation,
but, of course, i'll redo the test (or may be you can advise what test to run).
Measure the time delay from when the frame arrives in the stack until
that frame+RxTimestamp arrives in the application. I expect the round
about way via kthread takes longer.
My thoughts were - network stack might not immediately deliver packet to the application
The network stack always delivers the packet, but you artificially
delay that delivery by calling netif_receive_skb() later on from
cpts_match_rx_ts().
and PTP worker can be tuned (pri and smp_affinity),
That won't avoid the net softirq.
resulted code will be more structured,
I am afraid people will copy this pattern in new drivers. It really
does not make much sense.