Re: [PATCH v3 3/4] printk: nbcon: Allow KDB to acquire the NBCON context

[Marcos Paulo de Souza <mpdesouza@suse.com>]

On Fri, 2025-09-05 at 16:58 +0206, John Ogness wrote:
> On 2025-09-02, Marcos Paulo de Souza <mpdesouza@suse.com> wrote:
> > diff --git a/kernel/printk/nbcon.c b/kernel/printk/nbcon.c
> > index
> > ff218e95a505fd10521c2c4dfb00ad5ec5773953..352235a0eb4a484caccf86d3a
> > 57d1a149218ecec 100644
> > --- a/kernel/printk/nbcon.c
> > +++ b/kernel/printk/nbcon.c
> > @@ -255,6 +258,7 @@ static int
> > nbcon_context_try_acquire_direct(struct nbcon_context *ctxt,
> >  		 * interrupted by the panic CPU while printing.
> >  		 */
> >  		if (other_cpu_in_panic() &&
> > +		    READ_ONCE(kdb_printf_cpu) !=
> > raw_smp_processor_id() &&
> 
> This needs some sort of ifdef CONFIG_KGDB_KDB wrapped around it.
> Maybe
> it would be cleaner to have some macro.

Ouch.. that's true!

> 
> #ifdef CONFIG_KGDB_KDB
> #define KGDB_IS_ACTIVE() (READ_ONCE(kdb_printf_cpu) ==
> raw_smp_processor_id())
> #else
> #define KGDB_IS_ACTIVE() false
> #endif
> 
> and then something like:
> 
> 	if (other_cpu_in_panic() &&
> 	    !KGDB_IS_ACTIVE() &&
> 	    (!is_reacquire || cur->unsafe_takeover)) {
> 		return -EPERM;
> 	}
> 
> Or however you prefer. We need to compile for !CONFIG_KGDB_KDB.

I like the idea! Let me prepare it for v4.

> 
> John