Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Wed, 11 Dec 2019 11:01:49 +0900
From:   Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
To:     Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc:     Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>,
        linux-kernel@vger.kernel.org, Petr Mladek <pmladek@suse.com>,
        Sergey Senozhatsky <sergey.senozhatsky@gmail.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [RFC/PATCH] printk: Fix preferred console selection with
 multiple matches
Message-ID: <20191211020149.GN88619@google.com>
References: <b8131bf32a5572352561ec7f2457eb61cc811390.camel@kernel.crashing.org>
 <20191210080154.GJ88619@google.com>
 <98df321d16adb67c5579ac4b67d845fc0c2c97df.camel@kernel.crashing.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <98df321d16adb67c5579ac4b67d845fc0c2c97df.camel@kernel.crashing.org>
User-Agent: Mutt/1.10.1 (2018-07-13)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On (19/12/11 09:26), Benjamin Herrenschmidt wrote:
[..]
> No not exactly. Architectures/platforms use add_preferred_console()
> (such as arm64 with ACPI but powerpc at least does it too) based on
> various factors to select a reasonable "default" for that specific
> platform. Without that the kernel will basically default to the first
> one to register which may not be what you want.
> 
> The command line ones however want to override the defaults (provided
> they exist, ie, it's possible that whever is specified on the command
> line doesn't actually exist, and thus shall be ignored. That typically
> happens when there is either no match or ->setup fails).
> 
> > Hmm.
> > 
> > The patch may affect setups where alias matching is expected to
> > happen. E.g.:
> > 
> > 	console=uartFOO,BAR
> > 
> > Is 8250 the only console that does alias matching?
> 
> Why would the patch affect this negatively ? Today we stop on the first
> match, mark the driver enabled, and make it preferred if the match
> index matches preferred_console.

As far as I know, ->match() does not only match but also does ->setup().
If we have two console list entries that match (one via aliasing and one
via exact match) then the console driver is setup twice. Do all console
drivers handle it? [double setup]

If we could perform simple alias matching, without ->setup() call, and
exact matching (strcmp()), and then, if newcon would match two entries,
we would pick up the last matching entry and configure newcon only once.

This changes the order, tho.

[..]
>  - Another match that is marked preferred_console, in which case in
> addition to being enabled, the newly registered console will also be
> made the default console (ie, first in the list with CONSDEV set). This
> is actually what we want ! IE. The console matches the last specified
> one on the command line.

Well, it still looks to me that what you want is to "ignore alias
match and prefer exact match".

	-ss