Re: [PATCH bpf-next v2 01/10] net: introduce preferred busy-polling

["Björn Töpel" <bjorn.topel@intel.com>]

On 2020-11-16 17:19, Björn Töpel wrote:
> On 2020-11-16 17:04, Jakub Kicinski wrote:
>> On Mon, 16 Nov 2020 12:04:07 +0100 Björn Töpel wrote:
>>> @@ -6771,6 +6806,19 @@ static int napi_poll(struct napi_struct *n, 
>>> struct list_head *repoll)
>>>       if (likely(work < weight))
>>>           goto out_unlock;
>>> +    /* The NAPI context has more processing work, but busy-polling
>>> +     * is preferred. Exit early.
>>> +     */
>>> +    if (napi_prefer_busy_poll(n)) {
>>> +        if (napi_complete_done(n, work)) {
>>> +            /* If timeout is not set, we need to make sure
>>> +             * that the NAPI is re-scheduled.
>>> +             */
>>> +            napi_schedule(n);
>>> +        }
>>> +        goto out_unlock;
>>> +    }
>>
>> Why is this before the disabled check?
>>
> 
> This path is when the budget was exceeded (more work to be done). If the
> prefer flag is set, the napi loop is exited prematurely. We check the
> return value for napi_complete_done, to make sure that there is actually
> a TO value set.
> 
> Uhm, maybe I not following what you mean by the "disabled check".
>

Ok, too little coffee.

Yeah, maybe it would make sense to have the disabled check *before* the
"prefer"-checking.


Björn