Re: [PATCH v5] panic: Avoid the extra noise dmesg

From: Petr Mladek
Date: Fri Mar 01 2019 - 03:54:06 EST


On Fri 2019-03-01 12:11:31, Feng Tang wrote:
> Hi Andrew,
>
> Thanks for the review!
>
> On Thu, Feb 28, 2019 at 12:00:14PM -0800, Andrew Morton wrote:
> > On Fri, 22 Feb 2019 14:09:59 +0800 Feng Tang <feng.tang@xxxxxxxxx> wrote:
> >
> > > When kernel panic happens, it will first print the panic call stack,
> > > then the ending msg like:
> > >
> > > [ 35.743249] ---[ end Kernel panic - not syncing: Fatal exception
> > > [ 35.749975] ------------[ cut here ]------------
> > >
> > > The above message are very useful for debugging.
> > >
> > > But if system is configured to not reboot on panic, say the "panic_timeout"
> > > parameter equals 0, it will likely print out many noisy message like
> > > WARN() call stack for each and every CPU except the panic one, messages
> > > like below:
> > >
> > > WARNING: CPU: 1 PID: 280 at kernel/sched/core.c:1198 set_task_cpu+0x183/0x190
> > > Call Trace:
> > > <IRQ>
> > > try_to_wake_up
> > > default_wake_function
> > > autoremove_wake_function
> > > __wake_up_common
> > > __wake_up_common_lock
> > > __wake_up
> > > wake_up_klogd_work_func
> > > irq_work_run_list
> > > irq_work_tick
> > > update_process_times
> > > tick_sched_timer
> > > __hrtimer_run_queues
> > > hrtimer_interrupt
> > > smp_apic_timer_interrupt
> > > apic_timer_interrupt
> >
> > It's a fairly ugly-looking patch but I am inclined to agree.
>
> Yes, it's ugly :) we've changed 3 methods to tackle this.
>
> > The panicing CPU is spinning and blinking a LED and all CPUs have
> > interrupts enabled and the system is known to be in a messed up state.
> > All sorts of kernel code could emit all sorts of output in such
> > circumstances. So a global printk-killing knob seems appropriate.
> >
> > Thoughts:
> >
> > - why do the suppression in vprintk_emit()? Doing it right at entry
> > to printk() seems cleaner, more explicit?
>
> Yes, I put it in printk() in one earlier post, and Petr suggested to
> put it into vprintk_emit so that it works for all printk() interfaces,
> like the devkmsg_write -> printk_emit -> vprintk_emit path.

Yes, there are more printk interfaces. The check in vprintk_emit()
allows to calm down also prink_deferred() and dev_printk().

> > - Other code sites may wish to suppress all printks. Perhaps
> > `panic_suppress_printk' should just be called `suppress_printk'?
> Ok, then I'll move the definition from panic.c to printk code.

This change looks fine to me.

Best Regards,
Petr

PS: I am sorry for the late review. I have spent many days with
reviewing a proposal of rewrite of printk() internals.