On 2021/4/13 11:26, Hillf Danton wrote: > On Tue, 13 Apr 2021 10:56:42 Yunsheng Lin wrote: >> On 2021/4/13 10:21, Hillf Danton wrote: >>> On Mon, 12 Apr 2021 20:00:43 Yunsheng Lin wrote: >>>> >>>> Yes, the below patch seems to fix the data race described in >>>> the commit log. >>>> Then what is the difference between my patch and your patch below:) >>> >>> Hehe, this is one of the tough questions over a bounch of weeks. >>> >>> If a seqcount can detect the race between skb enqueue and dequeue then we >>> cant see any excuse for not rolling back to the point without NOLOCK. >> >> I am not sure I understood what you meant above. >> >> As my understanding, the below patch is essentially the same as >> your previous patch, the only difference I see is it uses qdisc->pad >> instead of __QDISC_STATE_NEED_RESCHEDULE. >> >> So instead of proposing another patch, it would be better if you >> comment on my patch, and make improvement upon that. >> > Happy to do that after you show how it helps revert NOLOCK.
Actually I am not going to revert NOLOCK, but add optimization to it if the patch fixes the packet stuck problem. Is there any reason why you want to revert it? > > . >
