[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <aLgNvJ3vBtgD9Mq0@gondor.apana.org.au>
Date: Wed, 3 Sep 2025 17:43:24 +0800
From: Herbert Xu <herbert@...dor.apana.org.au>
To: "Paul E. McKenney" <paulmck@...nel.org>
Cc: Masami Hiramatsu <mhiramat@...nel.org>,
Steven Rostedt <rostedt@...dmis.org>,
Menglong Dong <dongml2@...natelecom.cn>,
mathieu.desnoyers@...icios.com, linux-kernel@...r.kernel.org,
linux-trace-kernel@...r.kernel.org,
kernel test robot <oliver.sang@...el.com>, tgraf@...g.ch,
linux-crypto@...r.kernel.org
Subject: Re: [PATCH] tracing: fprobe: fix suspicious rcu usage in fprobe_entry
On Mon, Sep 01, 2025 at 08:00:15AM -0700, Paul E. McKenney wrote:
>
> If this is happening often enough, it would be easy for me to create an
> rcu_dereference_all_check() that allows all forms of vanilla RCU readers
> (but not, for example, SRCU readers), but with only two use cases,
> it is not clear to me that this is an overall win.
Hi Paul:
Please create such a helper. Because the alternative is for me
to do something like this in rhashtable:
#define rht_dereference_rcu(p, ht) \
rcu_dereference_check(p, lockdep_rht_mutex_is_held(ht) || \
rcu_read_lock_any_held())
This really makes no sense because rcu_read_lock_any_held is an
internal RCU implementation detail and has nothing to do with
rhashtable.
rhashtable is just a middle-man like RCU. The actual context
(be it vanilla, bh or sched RCU) used is entirely up to the user.
Actually what puzzles me is why can't we just get rid of the
bh and sched variants of rcu_dereference? After all, there is
only one synchronize_rcu/call_rcu and it supports all three
variants.
Cheers,
--
Email: Herbert Xu <herbert@...dor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Powered by blists - more mailing lists