Re: [Patch net] net_sched: fix a missing rcu barrier in mini_qdisc_pair_swap()

Thu, Dec 21, 2017 at 09:54:06PM CET, xiyou.wangcong@gmail.com wrote:

On Thu, Dec 21, 2017 at 11:01 AM, Cong Wang [off-list ref] wrote:

quoted

On Thu, Dec 21, 2017 at 1:03 AM, Jiri Pirko [off-list ref] wrote:

quoted

But again, we don't we just free qdisc in call_rcu and avoid the
barrier?

Non-sense again. Why qdisc code should be adjusted for your
miniq code? It is your own responsibility to take care of this shit.
Don't spread it out of minq.

Also, in case you believe call_rcu to free qdisc is queued after
the call_rcu in miniq, you are wrong again:

I certainly don't :) Not sure why you think I do.

https://www.kernel.org/pub/linux/kernel/people/paulmck/Answers/RCU/RCUCBordering.html

The rcu callbacks don't guarantee FIFO ordering.