Re: [PATCH v4] panic: add an option to replay all the printk message in buffer

[Sergey Senozhatsky]

On (04/29/19 13:44), Petr Mladek wrote:
> On Sat 2019-04-27 02:16:40, Sergey Senozhatsky wrote:
> > On (04/27/19 01:43), Sergey Senozhatsky wrote:
> > [..]
> > > > The console waiter logic is effective but it does not always
> > > > work. The current console owner must be calling the console
> > > > drivers.
> > > >
> > > > >   Hmm, we might have a bit of a problem here, maybe.
> > > >
> > > > Hmm, the printk() might wait forever when NMI stopped
> > > > the current console owner in the console driver code
> > > > or with the logbuf_lock taken.
> > > 
> > > I guess this is why we re-init logbuf lock from panic,
> > > however, we don't do anything with the console_owner.
> 
> > > > The console waiter logic might get solved by clearing
> > > > the console_owner in console_flush_on_panic(). It can't
> > > > be much worse, we already ignore console_lock() there, ...
> > 
> > Hmm, or maybe we are fine... console_waiter logic should work
> > before we send out stop IPI/NMI from panic CPU. When we call
> > flush_on_panic() console_unlock() clears console_owner, so
> > panic_print_sys_info() should not deadlock on console_owner.
> 
> Good point!
> 
> > It's probably only problematic if we kill a console_owner
> > CPU and then try to printk() (from smp_send_stop()) before
> > we do flush_on_panic()->console_unlock().
> 
> Yup. There are called several functions between smp_send_stop()
> and console_flush_on_panic().
> 
> The question is if it is worth a code complication. We could
> never 100% guarantee that printk() would work in panic().
> I more and more understand what Peter Zijlstra means
> by the duct taping.

Agreed.

	-ss