Re: [RFC v2] net: sched: implement TCQ_F_CAN_BYPASS for lockless qdisc

[Yunsheng Lin]

On 2021/3/17 6:48, Cong Wang wrote:
> On Mon, Mar 15, 2021 at 2:07 PM Jakub Kicinski <kuba@xxxxxxxxxx> wrote:
>>
>> I thought pfifo was supposed to be "lockless" and this change
>> re-introduces a lock between producer and consumer, no?
> 
> It has never been truly lockless, it uses two spinlocks in the ring buffer
> implementation, and it introduced a q->seqlock recently, with this patch
> now we have priv->lock, 4 locks in total. So our "lockless" qdisc ends
> up having more locks than others. ;) I don't think we are going to a
> right direction...

Yes, we have 4 locks in total, but lockless qdisc only use two locks
in this patch, which are priv->lock and q->seqlock.

The qdisc at least uses two locks, which is qdisc_lock(q) and q->busylock,
which seems to have bigger contention when concurrent accessing to the
same qdisc.

If we want to reduce the total number of lock, we can use qdisc_lock(q)
for lockless qdisc and remove q->seqlock:)

> 
> Thanks.
> 
> .
>