Re: [PATCH tty v11 1/2] serial: 8250: Switch to nbcon console, take 2

[Sebastian Andrzej Siewior]

On 2026-08-26 09:35:49 [+0200], Petr Mladek wrote:
> > As for this 8250 "switch to nbcon" series, I am uncertain how to
> > proceed. Is it really printk's job (and, by extension, the console
> > driver's job) to decipher when it is allowed to queue irq_work?
> 
> I wonder how other architectures hanle this situation. I wonder
> if they just ignore the pending IRQs.

The interrupts should be disabled and the enqueued irq_work should be
delayed until after the interrupts are enabled again while entering
idle. An interrupt is probably what pulls the CPU out of idle.
Is it possible to get some kind of crash dump or anything else to figure
out the state of CPU while it hangs?
If that irq_work is the problem that making nbcon_irq_work LAZY should
delay it, masking the problem.

Either way, I don't see why raising irq_work should be a problem. That
interrupt has to trigger once interrupts are enabled again. However, if
this is missed then further irq_work enqueues won't trigger an
interrupt which might be a problem if something relies on it.

So the question what is doing while it makes no progress and does not
fully boot.

> Best Regards,
> Petr

Sebastian