Re: [PATCH v2 net-next 1/6] net: sched: propagate "skip_sw" flag to offload for flower and matchall

From: Vladimir Oltean
Date: Tue Oct 22 2024 - 13:36:22 EST


On Sun, Oct 20, 2024 at 06:59:12PM +0300, Ido Schimmel wrote:
> Possibly a stupid question given I don't remember all the details of the
> TC offload, but is there a reason not to put the 'skip_sw' indication in
> 'struct flow_cls_common_offload' and initialize the new field as part of
> tc_cls_common_offload_init()?
>
> Seems like it won't require patching every classifier and will also work
> for the re-offload case (e.g., fl_reoffload())?

I think you forgot more about tc than I ever knew. The answer isn't
more complicated than "I didn't think about it". I've tested this
simpler proposal and will send v3 using it. Thanks.