Hi Petr,

Currently in linux-next we have commit b955f0eead54 ("printk: nbcon:
Fix nbcon_cpu_emergency_flush() when preemptible"). But as you
pointed out [0], there is no need to disable preemption at all if
we only need to know if the current CPU is in emergency state.

Here is an alternative patch that correctly addresses the issue by
avoiding the migration checks.

Since the previous patch is not yet mainline, perhaps you want to
use this one instead in order to avoid a revert. Up to you.

John Ogness

[0] https://lore.kernel.org/lkml/Zn7b8g1HtuTIAwyi@pathway.suse.cz

John Ogness (1):
  printk: nbcon: do not require migration disabled for
    nbcon_get_cpu_emergency_nesting()

 kernel/printk/nbcon.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)


base-commit: c51cf7315db1392af58b909d1004133f6dbf0bf6
-- 
2.39.2

On Fri 2024-07-19 15:54:17, John Ogness wrote:
> Hi Petr,
> 
> Currently in linux-next we have commit b955f0eead54 ("printk: nbcon:
> Fix nbcon_cpu_emergency_flush() when preemptible"). But as you
> pointed out [0], there is no need to disable preemption at all if
> we only need to know if the current CPU is in emergency state.
> 
> Here is an alternative patch that correctly addresses the issue by
> avoiding the migration checks.
> 
> Since the previous patch is not yet mainline, perhaps you want to
> use this one instead in order to avoid a revert. Up to you.

Makes sense. I have replaced the commit b955f0eead54 ("printk: nbcon:
Fix nbcon_cpu_emergency_flush() when preemptible") with this new patch
in printk/linux.git, branch rework/write-atomic.

> [0] https://lore.kernel.org/lkml/Zn7b8g1HtuTIAwyi@pathway.suse.cz
> 
> John Ogness (1):
>   printk: nbcon: do not require migration disabled for
>     nbcon_get_cpu_emergency_nesting()
> 
>  kernel/printk/nbcon.c | 9 ++++++++-
>  1 file changed, 8 insertions(+), 1 deletion(-)

Best Regards,
Petr