Re: [PATCH] driver core: add wait event for deferred probe

[Grant Likely <grant.likely@secretlab.ca>]

On Fri, 15 Feb 2013 00:58:23 +0800, Haojian Zhuang <haojian.zhuang@linaro.org> wrote:
> On 15 February 2013 00:50, Arnd Bergmann <arnd@arndb.de> wrote:
> > On Thursday 14 February 2013, Haojian Zhuang wrote:
> >> On 14 February 2013 23:57, Arnd Bergmann <arnd@arndb.de> wrote:
> >> > On Thursday 14 February 2013, Haojian Zhuang wrote:
> >> >> If you can change it into code in below, it could work. Otherwise, it
> >> >> always fails.
> >> >>         driver_deferred_probe_enable = true;
> >> >>         driver_deferred_probe_trigger();
> >> >> +      deferred_probe_work_func(NULL);
> >> >>         return 0;
> >> >>
> >> >> Because deferred_probe_work_func() depends on that deferred_probe is added
> >> >> into deferred_probe_active_list. If driver_deferred_probe_trigger() isn't called
> >> >> first, the deferred uart probe can't be added into active list. So even you call
> >> >> work_func at here, it doesn't help.
> >> >>
> >> >
> >> > Would that not cause two instances of the work function to run at the same time?
> >> > That sounds like a source for a lot of problems.
> >> >
> >> >         Arnd
> >>
> >> Two instances of the work function? I'm sorry that I don't
> >> understanding your meaning.
> >> Could you help explain your question?
> >
> > I mean you end up calling the work function directly, while it gets run as part
> > of the work queue on a different CPU at the same time. I just noticed that
> > there is actually locking in place in deferred_probe_work_func that prevents
> > any actual bugs, but you are still adding extra overhead here.
> >
> > Maybe just add
> >
> >         flush_workqueue(deferred_wq);
> >
> > here?
> >
> >         Arnd
> 
> It's fine to me. Since both of them are flushing workqueue.
> 
> Tested-by: <haojian.zhuang@linaro.org>

Hahaha. I just came to the same conclusion. I'll craft a proper patch
and send it off.

g.