Re: [PATCH RFC net 1/1] net/sched: Fix mirred to self recursion

From: Davide Caratti <dcaratti@redhat.com>
To: Eric Dumazet <edumazet@google.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>,
	davem@davemloft.net, kuba@kernel.org,  pabeni@redhat.com,
	jiri@resnulli.us, xiyou.wangcong@gmail.com,
	 netdev@vger.kernel.org, renmingshuai@huawei.com,
	 Victor Nogueira <victor@mojatatu.com>
Subject: Re: [PATCH RFC net 1/1] net/sched: Fix mirred to self recursion
Date: Tue, 16 Apr 2024 11:28:36 +0200	[thread overview]
Message-ID: <CAKa-r6tLA82W9kMJABKG1fKfBiWjsD3nkaSwwzhnRD0O8QfWwA@mail.gmail.com> (raw)
In-Reply-To: <CANn89iJQZ5R=Cct494W0DbNXR3pxOj54zDY7bgtFFCiiC1abDg@mail.gmail.com>

hi Eric, thanks for looking at this!

On Tue, Apr 16, 2024 at 11:14 AM Eric Dumazet <edumazet@google.com> wrote:
>
> On Tue, Apr 16, 2024 at 10:05 AM Davide Caratti <dcaratti@redhat.com> wrote:

[...]

(FTR, the discussion started off-list :) more context at
https://github.com/multipath-tcp/mptcp_net-next/issues/451 )

> > I tried a similar approach some months ago [1],  but  _ like Eric
> > noticed  _ it might slowdown qdisc_destroy() too much because of the
> > call to synchronize_rcu(). Maybe the key unregistering can be done
> > later (e.g. when the qdisc is freed) ?
> >
> > [1] https://lore.kernel.org/netdev/73065927a49619fcd60e5b765c929f899a66cd1a.1701853200.git.dcaratti@redhat.com/
> >
>
> Hmmm, I think I missed that lockdep_unregister_key() was a NOP unless
> CONFIG_LOCKDEP=y

yes, the slowdown is there only on debug kernels.

> Sorry for this, can you repost your patch ?

Sure, but please leave me some time to understand what to do with HTB
(you were right at [1] and I didn't notice
htb_set_lockdep_class_child()  [2] at that time).

thanks!

[1] https://lore.kernel.org/netdev/CANn89iLXjstLx-=hFR0Rhav462+9pH3JTyE45t+nyiszKKCPTQ@mail.gmail.com/
[2] https://elixir.bootlin.com/linux/latest/source/net/sched/sch_htb.c#L1042
-- 
davide