Re: [RFC][PATCH v9 1/2] printk: Make printk() completely async

[Petr Mladek <pmladek@suse.com>]

On Fri 2016-04-01 23:10:04, Sergey Senozhatsky wrote:
> This patch makes printk() completely asynchronous (similar to what
> 
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index bfbf284..2e50c48 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -2722,6 +2761,47 @@ static int __init printk_late_init(void)
>  late_initcall(printk_late_init);
>  
>  #if defined CONFIG_PRINTK
> +static int printk_kthread_func(void *data)
> +{
> +	while (1) {
> +		set_current_state(TASK_INTERRUPTIBLE);
> +		if (!printk_kthread_need_flush_console)
> +			schedule();
> +
> +		__set_current_state(TASK_RUNNING);
> +
> +		console_lock();
> +		console_unlock();
> +		/*
> +		 * Avoid an infinite loop when console_unlock() cannot
> +		 * access consoles, e.g. because console_suspended is
> +		 * true. schedule(), someone else will print the messages
> +		 * from resume_console().
> +		 */
> +		printk_kthread_need_flush_console = false;

You need to move this assigment right above the console_lock()/console_unlock()
calls. Otherwise, there is a race:

CPU0:				CPU1

printk_kthread_func()

  console_unlock()

				printk()

				printk_kthread_need_flush_console = true;
				wake_up_process(printk_kthread);

  printk_kthread_need_flush_console = false;

  set_current_state(TASK_INTERRUPTIBLE);
  if (!printk_kthread_need_flush_console)
    schedule();

=> sleeping without processing the last message.

You could safely set it false before calling
console_lock()/console_unlock() because you are calling the commands
that are requested by the flag.

With the above described change:

Reviewed-by: Petr Mladek <pmladek@suse.com>

Best Regards,
Petr