Re: [RESEND][PATCH 1/2] console: add new CONSOLE_ACTIVATE_ALL_FALLBACK option

[Ahmad Fatoum <a.fatoum@pengutronix.de>]

On 7/30/24 10:09, Sascha Hauer wrote:
> On Fri, Jul 26, 2024 at 02:26:22PM +0200, Ahmad Fatoum wrote:
>> We already have three CONSOLE_ACTIVATE options and every one of them has
>> drawbacks:
>>
>>   - ACTIVATE_ALL: May write barebox log to external devices like MCUs
>>     that don't expect it
>>
>>   - ACTIVATE_FIRST: Not applicable for most systems that probe from
>>     device tree, where the order of probe is not necessarily fixed,
>>     so what console is first may change over updates
>>
>>   - ACTIVATE_NONE: has a misleading name and may leave the user without
>>     any consoles at all if nothing else activates a console
>>
>> Let's add a new option and make it the default, which avoids all these
>> issues: Like ACTIVATE_NONE, it expects board code, DT or environment to
>> enable a console and if none of them do it falls back to activating all
>> consoles, so the user isn't kept in the dark with an error instructing
>> the user to fix this.
>>
>> Signed-off-by: Ahmad Fatoum <a.fatoum@pengutronix.de>
>> ---
>> Sent out second patch only by mistake instead of both...
>> ---
>>  common/Kconfig   | 19 ++++++++++++++++++-
>>  common/console.c | 27 +++++++++++++++++++++++++++
>>  2 files changed, 45 insertions(+), 1 deletion(-)
>>
>> diff --git a/common/Kconfig b/common/Kconfig
>> index 31360892aeef..2dda5ce5743a 100644
>> --- a/common/Kconfig
>> +++ b/common/Kconfig
>> @@ -812,7 +812,7 @@ endchoice
>>  choice
>>  	prompt "Console activation strategy"
>>  	depends on CONSOLE_FULL
>> -	default CONSOLE_ACTIVATE_FIRST
>> +	default CONSOLE_ACTIVATE_ALL_FALLBACK
>>  
>>  config CONSOLE_ACTIVATE_FIRST
>>  	bool
>> @@ -831,6 +831,23 @@ config CONSOLE_ACTIVATE_ALL
>>  	  Only the first registered console will have the full startup
>>  	  log though.
>>  
>> +config CONSOLE_ACTIVATE_ALL_FALLBACK
>> +	bool
>> +	prompt "activate all consoles as fallback"
>> +	help
>> +	  This option is similar to CONFIG_CONSOLE_ACTIVATE_NONE in that it
>> +	  leaves consoles disabled on startup. If by the end of barebox
>> +	  startup, no consoles have been activated via board code, device
>> +	  tree or environment, barebox will enable all registered consoles
>> +	  as fallback, so the user has a chance to see output.
>> +
>> +	  This will be indicated by a fat error, so the user knows that
>> +	  the configuration needs to be fixed. If you don't see any
>> +	  output at all, consider trying again after enabling
>> +	  CONFIG_CONSOLE_ACTIVATE_ALL, so consoles are activated immediately
>> +	  at registration time and/or with CONFIG_DEBUG_LL, so barebox output
>> +	  is written even before console drivers were registered.
>> +
>>  config CONSOLE_ACTIVATE_NONE
>>  	prompt "leave all consoles disabled"
>>  	bool
>> diff --git a/common/console.c b/common/console.c
>> index 73b4c4d4db01..e83a3e1e2d7f 100644
>> --- a/common/console.c
>> +++ b/common/console.c
>> @@ -450,6 +450,33 @@ int console_unregister(struct console_device *cdev)
>>  }
>>  EXPORT_SYMBOL(console_unregister);
>>  
>> +static int console_activate_all_fallback(void)
>> +{
>> +	int activate = CONSOLE_STDIOE;
>> +	struct console_device *cdev;
>> +
>> +	for_each_console(cdev) {
>> +		if (cdev->f_active & (CONSOLE_STDOUT | CONSOLE_STDERR))
>> +			return 0;
>> +	}
>> +
>> +	if (IS_ENABLED(CONFIG_CONSOLE_DISABLE_INPUT))
>> +		activate &= ~CONSOLE_STDIN;
>> +
>> +	for_each_console(cdev)
>> +		console_set_active(cdev, activate);
>> +
>> +	/*
>> +	 * This is last resort, so the user is not kept in the dark.
>> +	 * Writing to all consoles is a bad idea as the devices at the
>> +	 * other side might get confused by it, thus the error log level.
>> +	 */
>> +	pr_err("No consoles were activated. Activating all consoles as fallback!\n");
>> +
>> +	return 0;
>> +}
>> +postenvironment_initcall(console_activate_all_fallback);
> 
> Shouldn't this code only run when CONSOLE_ACTIVATE_ALL_FALLBACK is
> enabled?

Of course, sorry about that. Just sent out v3.

> 
> Sascha
>