Re: This breaks netem use cases

[Cong Wang <xiyou.wangcong@gmail.com>]

On Thu, Jul 10, 2025 at 10:26:46AM +0200, Paolo Abeni wrote:
> On 7/8/25 9:42 PM, Cong Wang wrote:
> > (Cc LKML for more audience, since this clearly breaks potentially useful
> > use cases)
> > 
> > On Tue, Jul 08, 2025 at 04:43:26PM +0000, William Liu wrote:
> >> netem_enqueue's duplication prevention logic breaks when a netem
> >> resides in a qdisc tree with other netems - this can lead to a
> >> soft lockup and OOM loop in netem_dequeue, as seen in [1].
> >> Ensure that a duplicating netem cannot exist in a tree with other
> >> netems.
> > 
> > As I already warned in your previous patchset, this breaks the following
> > potentially useful use case:
> > 
> > sudo tc qdisc add dev eth0 root handle 1: mq
> > sudo tc qdisc add dev eth0 parent 1:1 handle 10: netem duplicate 100%
> > sudo tc qdisc add dev eth0 parent 1:2 handle 20: netem duplicate 100%
> > 
> > I don't see any logical problem of such use case, therefore we should
> > consider it as valid, we can't break it.
> 
> My understanding is that even the solution you proposed breaks a
> currently accepted configuration:
> 
> https://lore.kernel.org/netdev/CAM0EoMmBdZBzfUAms5-0hH5qF5ODvxWfgqrbHaGT6p3-uOD6vg@mail.gmail.com/

Maybe it is not obvious, my patch does not reject users' setup. It
probably has bugs, I am more than just happy to address any bugs in the
next iteration (like for any patch), in fact it is my obligation.

My appologize if I misled any of you to believe my patch is bug-free or
perfect, it is never the case.

For Jamal's patch, it is his intention to break users' setup, and this
won't change during any iteration.

They are significantly different.

> 
> I call them (both the linked one and the inline one) 'configurations'
> instead of 'use-cases' because I don't see how any of them could have
> real users, other than: https://xkcd.com/1172/.

Please let me know if you have any other way to use netem duplication on
a multiqueue NIC _directly_ without worrying about the global spinlock.

I bet you have none. Either you need to use it indirectly (attaching to
a non-root) or you have to face the global spinlock (aka without using
mq).

I am open to your education. :)

Thanks a lot!