Re: [PATCH 2/2] sched: mark PRINTK_DEFERRED_CONTEXT_MASK in __schedule()

[Peter Zijlstra]

On Mon, Sep 28, 2020 at 10:00:22AM +0200, Peter Zijlstra wrote:
> On Mon, Sep 28, 2020 at 12:52:44AM -0700, Joe Perches wrote:
> > On Mon, 2020-09-28 at 09:32 +0200, Peter Zijlstra wrote:
> > > On Mon, Sep 28, 2020 at 12:11:30AM +0800, Chengming Zhou wrote:
> > > > The WARN_ON/WARN_ON_ONCE with rq lock held in __schedule() should be
> > > > deferred by marking the PRINTK_DEFERRED_CONTEXT_MASK, or will cause
> > > > deadlock on rq lock in the printk path.
> > []
> > > NAK printk_deferred is an abomination, kill that.
> > 
> > Didn't you introduce it?
> > Should you be complaining to yourself?
> 
> Yeah. I should've hacked around it then I suppose. Still, no reason to
> proliferate that crap. Afaik the new printk should be able to deal with
> this at some point.
> 
> The thing we're not going to do it add a cache-miss to schedule just
> because.

Also, using deferred for WARN/BUG is fundamentally wrong, you want that
splat _NOW_, there might not be a later.