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;
From:   John Ogness <john.ogness@linutronix.de>
To:     Daniel Thompson <daniel.thompson@linaro.org>
Cc:     Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <senozhatsky@chromium.org>,
        Steven Rostedt <rostedt@goodmis.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        linux-kernel@vger.kernel.org,
        Jason Wessel <jason.wessel@windriver.com>,
        Douglas Anderson <dianders@chromium.org>,
        Aaron Tomlin <atomlin@redhat.com>,
        Luis Chamberlain <mcgrof@kernel.org>,
        kgdb-bugreport@lists.sourceforge.net
Subject: Re: [PATCH printk v3 15/40] kdb: use srcu console list iterator
In-Reply-To: <20221109085325.wiub564iqnewvczb@ash.lan>
References: <20221107141638.3790965-1-john.ogness@linutronix.de>
 <20221107141638.3790965-16-john.ogness@linutronix.de>
 <20221109085325.wiub564iqnewvczb@ash.lan>
Date:   Wed, 09 Nov 2022 10:33:55 +0106
Message-ID: <87wn848okk.fsf@jogness.linutronix.de>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Hi Daniel,

On 2022-11-09, Daniel Thompson <daniel.thompson@linaro.org> wrote:
>> +	/*
>> +	 * The console_srcu_read_lock() only provides safe console list
>> +	 * traversal. The use of the ->write() callback relies on all other
>> +	 * CPUs being stopped at the moment and console drivers being able to
>> +	 * handle reentrance when @oops_in_progress is set. (Note that there
>> +	 * is no guarantee for either criteria.)
>> +	 */
>
> The debugger entry protocol does ensure that other CPUs are either
> stopped or unresponsive. In the case where the other CPU is unresponsive
> (e.g. timed out after being asked to stop) then there is a "real" printk()
> issued prior to any of the above interference with the console system to
> the developer driving the debugger gets as much clue as we can offer them
> about what is going on (typically this is emitted from regular interrupt
> context).
>
> Given this comment is part of the debugger code then for the
> oops_in_progress hack it might be more helpful to describe what
> the developer in front the debugger needs to do to have the most
> reliable debug session possible.
>
>   There is no guarantee that every console drivers can handle reentrance
>   in this way; the developer deploying the debugger is responsible for
>   ensuring that the console drivers they have selected handle reentrance
>   appropriately.

Thanks for the explanation. I will change the comment to:

	/*
	 * The console_srcu_read_lock() only provides safe console list
	 * traversal. The use of the ->write() callback relies on all other
	 * CPUs being stopped at the moment and console drivers being able to
	 * handle reentrance when @oops_in_progress is set.
	 *
	 * There is no guarantee that every console driver can handle
	 * reentrance in this way; the developer deploying the debugger
	 * is responsible for ensuring that the console drivers they
	 * have selected handle reentrance appropriately.
	 */

John