Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 28 Sep 2022 18:46:42 +0900
From:   Sergey Senozhatsky <senozhatsky@chromium.org>
To:     Petr Mladek <pmladek@suse.com>
Cc:     John Ogness <john.ogness@linutronix.de>,
        Sergey Senozhatsky <senozhatsky@chromium.org>,
        Steven Rostedt <rostedt@goodmis.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Subject: Re: [PATCH printk 06/18] printk: Protect [un]register_console() with
 a mutex
Message-ID: <YzQYArVKyyxxidxn@google.com>
References: <20220924000454.3319186-1-john.ogness@linutronix.de>
 <20220924000454.3319186-7-john.ogness@linutronix.de>
 <YzMT27FVllY3u05k@alley>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YzMT27FVllY3u05k@alley>
Precedence: bulk

On (22/09/27 17:16), Petr Mladek wrote:
[..]
> > +static int console_unregister_locked(struct console *console);
> > +
> >  /*
> >   * The console driver calls this routine during kernel initialization
> >   * to register the console printing procedure with printk() and to
> > @@ -3107,13 +3148,14 @@ void register_console(struct console *newcon)
> >  	bool realcon_enabled = false;
> >  	int err;
> >  
> > -	for_each_console(con) {
> > +	console_list_lock();
> 
> Hmm, the new mutex is really nasty. It has very strange semantic.
> It makes the locking even more complicated.

[..]

I fully agree with everything you said. This lock nesting made me
scratch my head wondering was it previous CPU hotplug code that had
multiple nested locks or was it something else?

> Anyway, I would like to avoid adding console_mutex. From my POV,
> it is a hack that complicates the code. Taking console_lock()
> should be enough. Using rcu walk would be good enough.
> 
> Do I miss something, please?
> 
> Or is this part of some strategy to remove console_sem later, please?

So I can only explain what potential I saw in list lock: the idea
that third party that iterates over consoles lists does not stop
entire console output machinery, and, moreover, that third party
does not flush pending messages once it's done with the business
it had to do under console_sem. E.g. it can be a systemd or any
other user-space process doing something with /dev/tty, which can
suddenly stop all consoles output (console_lock()) and then also
has to flush pending kernel messages (console_unlock()). Was this
goal, however, fully achieved - no, a third party that wants to
->flags &= ~CON_ENABLED a particular console still stops the entire
console output (and flushes pending messages, unless handover-ed).

I like what you suggested with srcu.