Re: [PATCH] hung_task: configurable hung-task stacktrace loglevel

[John Ogness <john.ogness@linutronix.de>]

On 2025-04-25, Petr Mladek <pmladek@suse.com> wrote:
> I am afraid that manipulating log levels is a lost fight because
> different people might have different opinion about how various
> messages are important.

Wasn't that the whole point of Sergey's patch? To make it configurable?

I must admit that I am not happy with the patch. Mostly because it is
too specific. And I am not sure if we really want to try to make it all
dynamic with a report API either. At least we need to think about it
more carefully.

One thing that crossed my mind was that we have enter/exit markers for
emergency mode, which should be used whenever something "bad" happens. I
am wondering if a fixed loglevel could be configured for all messages
stored by a CPU in emergency mode. This might also encourage developers
to track down and mark more emergency sections. For the nbcon series, I
really only picked a few obvious ones, but I am sure there are more.

In other words, I would prefer to recycle the emergenceny enter/exit
markers rather than introduce new ones. (Unless we are also talking
about reports that are totally normal and acceptable during runtime.)

John