Re: [Intel-wired-lan] [next-queue v2 2/4] i40e: Record number TXes cleaned during NAPI

[Jesse Brandeburg <jesse.brandeburg@intel.com>]

On 10/6/2022 10:32 AM, Joe Damato wrote:
> Sorry, but I don't see the value in the second param. NAPI decides what to
> do based on nb_pkts. That's the only parameter that matters for the purpose
> of NAPI going into poll mode or not, right?
> 
> If so: I don't see any reason why a second parameter is necessary.

Sridhar and I talked about this offline. We agree now that you can just 
proceed with the single parameter.

> 
> As I mentioned earlier: if it's just that the name of the parameter isn't
> right (e.g., you want it to be 'tx_processed' instead of 'tx_cleaned') then
> that's an easy fix; I'll just change the name.

I think the name change isn't necessary, since we're not going to extend 
this patch with full XDP events printed (see below)

> 
> It doesn't seem helpful to have xsk_frames as an out parameter for
> i40e_napi_poll tracepoint; that value is not used to determine anything
> about i40e's NAPI.
> 
>> I am not completely clear on the reasoning behind setting clean_complete
>> based on number of packets transmitted in case of XDP.
>>>
>>>> That might reduce the complexity a bit, and will probably still be pretty
>>>> useful for people tuning their non-XDP workloads.
>>
>> This option is fine too.
> 
> I'll give Jesse a chance to weigh in before I proceed with spinning a v3.

I'm ok with the patch you have now, that shows nb_pkts because it's the 
input to the polling decision. We can add the detail about XDP transmits 
cleaned in a later series or patch that is by someone who wants the XDP 
details in the napi poll context.

_______________________________________________
Intel-wired-lan mailing list
Intel-wired-lan@osuosl.org
https://lists.osuosl.org/mailman/listinfo/intel-wired-lan