Re: [PATCH] Docs: printk: add 'console=null|""' to admin/kernel-parameters

[Petr Mladek]

On Mon 2022-02-14 16:56:15, Randy Dunlap wrote:
> Tell about 'console=null|""' and how to use it.
> 
> Signed-off-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> Cc: Petr Mladek <pmladek@xxxxxxxx>
> Cc: Sergey Senozhatsky <senozhatsky@xxxxxxxxxxxx>
> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
> Cc: John Ogness <john.ogness@xxxxxxxxxxxxx>
> Cc: Jonathan Corbet <corbet@xxxxxxx>
> Cc: linux-doc@xxxxxxxxxxxxxxx
> ---
>  Documentation/admin-guide/kernel-parameters.txt |    6 ++++++
>  1 file changed, 6 insertions(+)
> 
> --- linux-next-20220214.orig/Documentation/admin-guide/kernel-parameters.txt
> +++ linux-next-20220214/Documentation/admin-guide/kernel-parameters.txt
> @@ -724,6 +724,12 @@
>  		hvc<n>	Use the hypervisor console device <n>. This is for
>  			both Xen and PowerPC hypervisors.
>  
> +		{ null | "" }
> +			Use to disable console output, i.e., to have kernel
> +			console messages discarded.
> +			This must be the first (or only) console= string
> +			used on the kernel command line.

It must be the only console= parameter on the command line. Otherwise,
the other consoles get enabled as well.

It might make sense to detect this situation and print a warning or
so. Nobody has sent a patch for this so far.

But there is even bigger problem. The default console is also used
as stdin/stdout/stderr for the init process. It might fail when there
is no console driver associated with it.

The problem with stdin/stdout/stderr does not happen when
CONFIG_NULL_TTY is enabled and ttynull driver is available.
Unfortunately, it can't be enabled by default because it can
be used by mistake, see the commit a91bd6223ecd46addc71e
("Revert "init/console: Use ttynull as a fallback when there
is no console").

And there is still a mystery that has not been explained yet,
see https://lore.kernel.org/r/a46e9a26-5b9f-f14c-26be-0b4d41fa7429@xxxxxxxxxxxx

On the positive note. console=null mostly works. All the problems are
hard to reproduce.


Now, what to do with this patch. I would suggest two changes:

1. Replace "must be the first (or only)" with "must be the only"

2. Mention that it is suggested to enable CONFIG_NULL_TTY that
   will avoid problems with stdin/stdout/stderr of the init process.
   But it might cause the ttynull might be used even when a real
   console is configured. And that more details can be found
   in the commit a91bd6223ecd46addc71e ("Revert "init/console:
   Use ttynull as a fallback when there is no console").

   It might be enough to mention this in the commit message.

Best Regards,
Petr