Re: [PATCH 1/3] watchdog/softlockup: Preserve original timestamp when touching watchdog externally

[Petr Mladek]

On Mon 2019-10-21 14:42:14, Peter Zijlstra wrote:
> On Mon, Aug 19, 2019 at 12:47:30PM +0200, Petr Mladek wrote:
> > Some bug report included the same softlockups in flush_tlb_kernel_range()
> > in regular intervals. Unfortunately was not clear if there was a progress
> > or not.
> > 
> > The situation can be simulated with a simply busy loop:
> > 
> > 	while (true)
> > 	      cpu_relax();
> > 
> > The softlockup detector produces:
> > 
> > [  168.277520] watchdog: BUG: soft lockup - CPU#1 stuck for 22s! [cat:4865]
> > [  196.277604] watchdog: BUG: soft lockup - CPU#1 stuck for 22s! [cat:4865]
> > [  236.277522] watchdog: BUG: soft lockup - CPU#1 stuck for 23s! [cat:4865]
> > 
> > One would expect only one softlockup report or several reports with
> > an increased duration.
> 
> Let's just say our expectations differ.
> 
> > The result is that each softlockup is reported only once unless
> > another process get scheduled:
> > 
> > [  320.248948] watchdog: BUG: soft lockup - CPU#2 stuck for 26s! [cat:4916]
> 
> Which would greatly confuse me; as the above would have me think the
> situation got resolved (no more lockups reported) even though it is
> still very much stuck there.
> 
> IOW, I don't see how this makes anything better. You're removing
> information.

The 2nd patch brings back the regular report but with correctly
counted time (stuck for XXs).

I split it into two patches because I was not sure what would be
preferred behavior. I prefer the regular reports as well.

Best Regards,
Petr