[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <7ce1a0dba3cc100e6f73a7499b407176a99c0aa9.camel@redhat.com>
Date: Tue, 23 Apr 2024 11:21:23 +0200
From: Paolo Abeni <pabeni@...hat.com>
To: Davide Caratti <dcaratti@...hat.com>, Eric Dumazet
<edumazet@...gle.com>, Saeed Mahameed <saeedm@...dia.com>, Tariq Toukan
<tariqt@...dia.com>
Cc: netdev@...r.kernel.org, renmingshuai@...wei.com, jiri@...nulli.us,
xiyou.wangcong@...il.com, xmu@...hat.com, Christoph Paasch
<cpaasch@...le.com>, Jamal Hadi Salim <jhs@...atatu.com>, Maxim
Mikityanskiy <maxim@...valent.com>, Victor Nogueira <victor@...atatu.com>
Subject: Re: [PATCH net-next v2] net/sched: fix false lockdep warning on
qdisc root lock
On Thu, 2024-04-18 at 16:01 +0200, Davide Caratti wrote:
> hello,
>
> On Thu, Apr 18, 2024 at 3:50 PM Davide Caratti <dcaratti@...hat.com> wrote:
> >
>
> [...]
>
> > This happens when TC does a mirred egress redirect from the root qdisc of
> > device A to the root qdisc of device B. As long as these two locks aren't
> > protecting the same qdisc, they can be acquired in chain: add a per-qdisc
> > lockdep key to silence false warnings.
> > This dynamic key should safely replace the static key we have in sch_htb:
> > it was added to allow enqueueing to the device "direct qdisc" while still
> > holding the qdisc root lock.
> >
> > v2: don't use static keys anymore in HTB direct qdiscs (thanks Eric Dumazet)
>
> I didn't have the correct setup to test HTB offload, so any feedback
> for the HTB part is appreciated. On a debug kernel the extra time
> taken to register / de-register dynamic lockdep keys is very evident
> (more when qdisc are created: the time needed for "tc qdisc add ..."
> becomes an order of magnitude bigger, while the time for "tc qdisc del
> ..." doubles).
@Eric: why do you think the lockdep slowdown would be critical? We
don't expect to see lockdep in production, right?
Enabling lockdep will defeat most/all cacheline optimization moving
around all fields after a lock, performances should be significantly
impacted anyway.
WDYT?
The HTB bits looks safe to me, but it would be great if someone @nvidia
could actually test it (AFAICS mlx5 is the only user of such
annotation).
Thanks!
Paolo
Powered by blists - more mailing lists