I was puzzled why there is a preempt_disable(). Turns out it is due to
the per-CPU usage of the previous irq_work, as introduced in commit
74876a98a87a1 ("printk: Wake up klogd using irq_work"), where the or and
enqueue must happen on the same CPU.

Since we don't have this anymore, it can go.

Remove the preempt_disable from __wake_up_klogd().

Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---

I'm sorry, I noticed this after starring at it once I sent the patch. I
could resend and merge it into the previous if it is preferred.

 kernel/printk/printk.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 7c5a220e84323..0fd6c37e2b333 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -4623,7 +4623,6 @@ static void __wake_up_klogd(int val)
 	if (WARN_ON_ONCE(console_irqwork_blocked))
 		return;
 
-	preempt_disable();
 	/*
 	 * Guarantee any new records can be seen by tasks preparing to wait
 	 * before this context checks if the wait queue is empty.
@@ -4640,8 +4639,6 @@ static void __wake_up_klogd(int val)
 
 	if (val & PRINTK_PENDING_OUTPUT)
 		irq_work_queue(&pending_output_work);
-
-	preempt_enable();
 }
 
 /**
-- 
2.55.0

On 2026-09-11 12:51:28 [+0200], To linux-kernel@vger.kernel.org wrote:
> I'm sorry, I noticed this after starring at it once I sent the patch. I
> could resend and merge it into the previous if it is preferred.

Now that I forgot to close the editor and did stare again, didn't we
want to rip out console_irqwork_blocked()?

Sebastian

Hi Sebastian,

I am responding for completeness...

On 2026-09-11, Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote:
> didn't we want to rip out console_irqwork_blocked()?

As I mention in a previous email [0], it provides other significant
benefits. So we do not want to rip it out, but at some point we should
rename it for its "true" purpose.

John

[0] https://lore.kernel.org/lkml/87wltbo97m.fsf@jogness.linutronix.de

On 2026-09-21 18:53:36 [+0206], John Ogness wrote:
> Hi Sebastian,
Hi John,

> I am responding for completeness...
> 
> On 2026-09-11, Sebastian Andrzej Siewior <bigeasy@linutronix.de> wrote:
> > didn't we want to rip out console_irqwork_blocked()?
> 
> As I mention in a previous email [0], it provides other significant
> benefits. So we do not want to rip it out, but at some point we should
> rename it for its "true" purpose.

Then please do so. Right now it looks like a workaround for tegra20
which has been replaced with the lazy-irq-work workaround.

> John

Sebastian