Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Date:   Fri, 1 May 2020 10:44:54 +0900
To:     Alper Nebi Yasak <alpernebiyasak@gmail.com>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Jiri Slaby <jslaby@suse.com>, Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
        linux-serial@vger.kernel.org, Steven Rostedt <rostedt@goodmis.org>,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>,
        Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        Arvind Sankar <nivedita@alum.mit.edu>,
        Benjamin Herrenschmidt <benh@kernel.crashing.org>,
        "David S. Miller" <davem@davemloft.net>,
        Feng Tang <feng.tang@intel.com>
Subject: Re: [RFC PATCH v2 1/3] printk: Add function to set console to
 preferred console's driver
Message-ID: <20200501014454.GB288759@jagdpanzerIV.localdomain>
References: <20200430161438.17640-1-alpernebiyasak@gmail.com>
 <20200430161438.17640-2-alpernebiyasak@gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20200430161438.17640-2-alpernebiyasak@gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On (20/04/30 19:14), Alper Nebi Yasak wrote:
> Currently, add_preferred_console sets a preferred console, but doesn't
> actually change /dev/console to match it. That part is handled within
> register_device, where a newly registered console driver will be set as
> /dev/console if it matches the preferred console.
> 
> However, if the relevant driver is already registered, the only way to
> set it as /dev/console is by un-registering and re-registering it.

Hmm. Preferred console selection is very fragile, there are too many
setups and workarounds that even minor tweaks introduce regressions
oftentimes.

We have, by the way, a pending patchset which changes the same
are - preferred console selection.

git://git.kernel.org/pub/scm/linux/kernel/git/pmladek/printk.git for-5.7-preferred-console

[..]
> An example is the xenfb_make_preferred_console() function:
> 
> 	console_lock();
> 	for_each_console(c) {
> 		if (!strcmp(c->name, "tty") && c->index == 0)
> 			break;
> 	}
> 	console_unlock();
> 	if (c) {
> 		unregister_console(c);
> 		c->flags |= CON_CONSDEV;
> 		c->flags &= ~CON_PRINTBUFFER; /* don't print again */
> 		register_console(c);
> 	}

I didn't know about this code.

> The code above was introduced in commit 9e124fe16ff2 ("xen: Enable
> console tty by default in domU if it's not a dummy"). In short, it's aim
> is to set VT as the preferred console only after a working framebuffer
> is registered and thus VT is not the dummy device.
> 
> This patch introduces an update_console_to_preferred function that
> handles the necessary /dev/console change. With this change, the example
> above can be replaced with:
> 
> 	console_lock();
> 	add_preferred_console("tty", 0, NULL);
> 	update_console_to_preferred();
> 	console_unlock();
> 
> More importantly, these two calls can be moved to vt.c in order to bump
> its priority when a non-dummy backend for it is introduced, solving that
> problem in general.

Let me take a look over the weekend.

	-ss