Re: [PATCH] panic.c: export panic_on_oops

[Ingo Molnar <mingo@elte.hu>]

* Simon Kagstrom <simon.kagstrom@netinsight.net> wrote:

> With my patch it instead works like this:
> 
> - mtdoops_console_write continuously writes messages to the buffer, but
>   never calls mtdoops_console_sync() itself.
> 
> - mtdoops_console_sync (i.e., the ->unblank() callback) will schedule
>   work if oops_in_progress is set.
> 
> - if we have a panic, it will call mtdoops_write directly (if
>   mtd->panic_write is set, otherwise we are out of luck). This is also
>   the code path on oopses in interrupt context.
> 
> So the workqueue only gets used on unblank() from oopses. I think the 
> second implementation is simpler, but it also changes the behavior of 
> mtdoops a bit to include messages before the oops/panic as well.

The main printk principle is simplicity:

- The simpler the printk codepath, the higher the chances that we still
  are within the window of opportunity to get anything out to the user.

- Furthermore, the simpler the printk codepath, the larger the window of
  opportunity is to begin with: we rely on less external state, so we
  have a smaller surface of interaction that might break printk.

In that sense, i think your modified workqueue use is less wrong than 
what is in current mainline, but i'm afraid it's still wrong.

Why use a workqueue on unblank()? Why use a workqueue _at all_? (If we 
piggyback to any kernel thread we could as well piggyback to syslog 
itself - and we all know how often syslog fails at capturing oopses.)

The mtdoops console driver only seems to act if there's an emergency - 
and in emergencies we really _never_ want to do complex things like 
using a workqueue thread.

( Sidenote: i proffer that we dont want to use a workqueue in the 
  'regular' printk case either - but that seems to be irrelevant here as 
  mtdoops does not seem to save / care about regular non-emergency 
  printks. )

	Ingo