Re: [PATCH 05/37] kdb: core for kgdb back end

[Peter Zijlstra <peterz@infradead.org>]

On Mon, 2009-12-28 at 10:33 -0600, Jason Wessel wrote:
> Peter Zijlstra wrote:
> > On Wed, 2009-12-23 at 15:19 -0600, Jason Wessel wrote:
> >> +kdb_prio(char *name, struct rt_prio_array *array, kdb_printf_t xxx_printf,
> [..]
> >> +kdb_runqueue(unsigned long cpu, kdb_printf_t xxx_printf)
> [..]
> > 
> > So what is this code supposed to do? It looks like it simply cannot be
> > correct.
> > 
> 
> Beyond the fact that the code may not be right, it is a small part of
> a kdb module which is not included, and these functions can
> immediately disappear.  For reference, the original kdb module was
> called "kdbm_sched" and I believe it was intended as an aid to help
> debug the scheduler and process priorities in earlier versions of the
> kernel.
> 
> I have removed all this code and reduced the patch to kernel/sched.c
> to the following. 

Awesome. Please let me know if in the future you find the need for
similar helpers, I'm sure we can figure out something that'll work as
intended.

>  The reason we need the attached patch is to be able to call
> curr_task() for kdb to perform some basic per cpu task analysis while
> the system is stopped.

Right, so this set_curr_task() thing is way scary, maybe we can simplify
things to only expose curr_task()?

> --- a/kernel/sched.c
> +++ b/kernel/sched.c
> @@ -9784,9 +9784,9 @@ void normalize_rt_tasks(void)
>  
>  #endif /* CONFIG_MAGIC_SYSRQ */
>  
> -#ifdef CONFIG_IA64
> +#if defined(CONFIG_IA64) || defined(CONFIG_KGDB_KDB)
>  /*
> - * These functions are only useful for the IA64 MCA handling.
> + * These functions are only useful for the IA64 MCA handling, or kdb.
>   *
>   * They can only be called when the whole system has been
>   * stopped - every CPU needs to be quiescent, and no scheduling