Re: [PATCH v2 3/3] printk/nmi: Prevent deadlock when accessing the main log buffer in NMI
From: Petr Mladek
Date: Fri Jun 29 2018 - 03:25:26 EST
On Fri 2018-06-29 10:47:03, Sergey Senozhatsky wrote:
> On (06/28/18 11:41), Petr Mladek wrote:
> > >
> > > A side note: This nesting also handles recursive printk-s for us.
> > >
> > > NMI:
> > > printk_nmi_enter
> > > ftrace_dump
> > > printk_nmi_direct_enter
> > > vprintk_func
> > > spin_lock(logbuf_lock)
> > > vprintk_store
> > > vsprintf
> > > WARN_ON
> > > vprintk_func
> > > vprintk_nmi
> >
> > Uff, it seems that the current design is "good" at lest from some
> > points of view.
>
> yep yep
>
> > > > + len = vprintk_store(0, LOGLEVEL_DEFAULT, NULL, 0, fmt, args);
> > > > + raw_spin_unlock(&logbuf_lock);
> > > > + defer_console();
> > > > + return len;
> > > > + }
> > >
> > > So, maybe, something a bit better than defer_console().
> >
> > I am not super happy with the name either. But wakeup_console(),
> > schedule_console(), or queue_console() looked confusing.
>
> Hmm. defer_console() makes me think that we are dealing with that
> fbcon=nodefer and deferred console takeover thing here.
>
>
> So I summon Mr. Rostedt!
>
> Does schedule_console_output() look bad?
> What about defer_console_output()?
I am fine with both. I slightly prefer defer_console_output()
because I have "schedule" associated with deadlocks in this
code path (context) ;-)
Best Regards,
Petr