Re: [PATCH RFC v6 00/21] DEPT(Dependency Tracker)

[Hyeonggon Yoo <42.hyeyoo@gmail.com>]

On Fri, May 06, 2022 at 09:11:35AM +0900, Byungchul Park wrote:
> Linus wrote:
> >
> > On Wed, May 4, 2022 at 1:19 AM Byungchul Park <byungchul.park@lge.com> wrote:
> > >
> > > Hi Linus and folks,
> > >
> > > I've been developing a tool for detecting deadlock possibilities by
> > > tracking wait/event rather than lock(?) acquisition order to try to
> > > cover all synchonization machanisms.
> > 
> > So what is the actual status of reports these days?
> > 
> > Last time I looked at some reports, it gave a lot of false positives
> > due to mis-understanding prepare_to_sleep().
> 
> Yes, it was. I handled the case in the following way:
> 
> 1. Stage the wait at prepare_to_sleep(), which might be used at commit.
>    Which has yet to be an actual wait that Dept considers.
> 2. If the condition for sleep is true, the wait will be committed at
>    __schedule(). The wait becomes an actual one that Dept considers.
> 3. If the condition is false and the task gets back to TASK_RUNNING,
>    clean(=reset) the staged wait.
> 
> That way, Dept only works with what actually hits to __schedule() for
> the waits through sleep.
> 
> > For this all to make sense, it would need to not have false positives
> > (or at least a very small number of them together with a way to sanely
> 
> Yes. I agree with you. I got rid of them that way I described above.
>

IMHO DEPT should not report what lockdep allows (Not talking about
wait events). I mean lockdep allows some kind of nested locks but
DEPT reports them.

When I was collecting reports from DEPT on varous configurations,
Most of them was report of down_write_nested(), which is allowed in
lockdep.

DEPT should not report at least what we know it's not a real deadlock.
Otherwise there will be reports that is never fixed, which is quite
unpleasant and reporters cannot examine all of them if it's real deadlock
or not.

> > get rid of them), and also have a track record of finding things that
> > lockdep doesn't.
> 
> I have some reports that wait_for_completion or waitqueue is involved.
> It's worth noting those are not tracked by Lockdep. I'm checking if
> those are true positive or not. I will share those reports once I get
> more convinced for that.
> 
> > Maybe such reports have been sent out with the current situation, and
> > I haven't seen them.
> 
> Dept reports usually have been sent to me privately, not in LKML. As I
> told you, I'm planning to share them.
> 
> 	Byungchul
> 
> > 
> >                  Linus
> > 

-- 
Thanks,
Hyeonggon