On Thu, Nov 14, 2019 at 8:33 PM Sergey Senozhatsky
<sergey.senozhatsky.work@xxxxxxxxx> wrote:
>
> Gosh, that part of printk is really complex.
>
> On (19/11/14 10:57), Petr Mladek wrote:
> > For a proper solution we would need to match boot and real
> > consoles that write messages into the physical device.
> > But I am afraid that there is no support for this.
>
> Wouldn't those have same tty driver?
>
> ---
>
> kernel/printk/printk.c | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index f1b08015d3fa..a84cb20acf42 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -2690,6 +2690,19 @@ static int __init keep_bootcon_setup(char *str)
>
> early_param("keep_bootcon", keep_bootcon_setup);
>
> +static bool known_console_driver(struct console *newcon)
> +{
> + struct console *con;
> +
> + for_each_console(con) {
> + if (!(con->flags & CON_ENABLED))
> + continue;
> + if (con->device && con->device == newcon->device)
> + return true;
> + }
> + return false;
> +}
> +
> /*
> * The console driver calls this routine during kernel initialization
> * to register the console printing procedure with printk() and to
> @@ -2828,6 +2841,9 @@ void register_console(struct console *newcon)
> if (newcon->flags & CON_EXTENDED)
> nr_ext_console_drivers++;
>
> + if (known_console_driver(newcon))
> + newcon->flags &= ~CON_PRINTBUFFER;
> +
> if (newcon->flags & CON_PRINTBUFFER) {
> /*
> * console_unlock(); will print out the buffered messages
Thanks. It also needs to be cleared when the second console driver is
registered (of the same type, boot or normal), not just when a normal
con replaces a bootconsole. A simple way of avoiding the problem I'm
seeing is to not even set the CON_PRINTBUFFER flag on my consoles. It
skips the replay and the output on all consoles looks fine. The flag
is only used by register_console(), although I don't think that is the
intended usage? There are no console drivers that do this.
