Re: [PATCH v3 10/11] printk: Try enable preferred consoles only when there are any

[Petr Mladek]

On Tue 2026-06-02 10:53:11, Petr Mladek wrote:
> try_enable_preferred_console() used to be always called because it
> had several hidden effects, namely:
> 
> - enabled Braille consoles which were ignored by "preferred_dev_console"
>   because they were not associated with /dev/console.
> 
> - returned success when a console was pre-enabled using CON_ENABLED
>   flag.
> 
> - returned success when a console was enabled by default because
>   try_enable_default_console() did not return success.
> 
> The first two hidden effects were removed in previous patches. The 2nd fix
> actually helps even the 3rd case. try_enable_default_console() sets
> CON_ENABLED flag on success and is later handled as pre-enabled.
> 
> Prevent any future hidden effects and call try_enable_preferred_console()
> only when some console is preferred.
> 
> No behavior change.
> 
> Signed-off-by: Petr Mladek <pmladek@xxxxxxxx>
> ---
>  kernel/printk/printk.c | 46 +++++++++++++++++++++---------------------
>  1 file changed, 23 insertions(+), 23 deletions(-)
> 
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index a4eaf157cece..d62901a13bf7 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -4099,35 +4099,35 @@ static int try_enable_console(struct console *newcon)
>  			return err;
>  	}
>  
> -	/*
> -	 * See if we want to enable this console driver by default.
> -	 *
> -	 * Nope when a console is preferred by the command line, device
> -	 * tree, or SPCR.
> -	 *
> -	 * The first real console with tty binding (driver) wins. More
> -	 * consoles might get enabled before the right one is found.
> -	 *
> -	 * Note that a console with tty binding will have CON_CONSDEV
> -	 * flag set and will be first in the list.
> -	 */
> -	if (preferred_dev_console < 0) {
> +	if (preferred_dev_console >= 0) {
> +		/* See if this console matches one we selected on the command line */
> +		err = try_enable_preferred_console(newcon, true);
> +		if (err != -ENOENT)
> +			return err;
> +
> +		/* If not, try to match against the platform default(s) */
> +		err = try_enable_preferred_console(newcon, false);
> +		if (err != -ENOENT)
> +			return err;
> +	} else {
> +		/*
> +		 * See if we want to enable this console driver by default.
> +		 *
> +		 * Nope when a console is preferred by the command line, device
> +		 * tree, or SPCR.
> +		 *
> +		 * The first real console with tty binding (driver) wins. More
> +		 * consoles might get enabled before the right one is found.
> +		 *
> +		 * Note that a console with tty binding will have CON_CONSDEV
> +		 * flag set and will be first in the list.
> +		 */
>  		if (hlist_empty(&console_list) || !console_first()->device ||
>  		    console_first()->flags & CON_BOOT) {
>  			try_enable_default_console(newcon);
>  		}

This will require initializing err variable at the beginnit to prevent
returning a random value. I am going to fix it in v4.

Note that I am working on v4 because we want to fix handling pre-enabled
consoles in a first patch. It will prevent an out-of-bound access,
see
https://lore.kernel.org/all/7sq4tr2nmlz32tvkf6vpsghv6exvqfghsrlvywjcqihzsqqbf7@bspclmti5xg4/

>  	}
>  
> -	/* See if this console matches one we selected on the command line */
> -	err = try_enable_preferred_console(newcon, true);
> -	if (err != -ENOENT)
> -		return err;
> -
> -	/* If not, try to match against the platform default(s) */
> -	err = try_enable_preferred_console(newcon, false);
> -	if (err != -ENOENT)
> -		return err;
> -
>  	/*
>  	 * Some consoles, such as pstore and netconsole, can be enabled even
>  	 * without matching. Accept them at this stage when they had a chance

Best Regards,
Petr

PS: You might want to ignore v3. I am going to send v4 tomorrow after
    I double check the reshufling of the patches.