Re: [PATCHv3 3/4] drivers/base: clean up the usage of devices_kset_move_last()

[Pingfan Liu]

On Wed, Jul 4, 2018 at 6:18 PM Rafael J. Wysocki <rjw@xxxxxxxxxxxxx> wrote:
>
> On Wednesday, July 4, 2018 6:40:09 AM CEST Pingfan Liu wrote:
> > On Tue, Jul 3, 2018 at 10:28 PM Rafael J. Wysocki <rjw@xxxxxxxxxxxxx> wrote:
> > >
> > > On Tuesday, July 3, 2018 8:50:41 AM CEST Pingfan Liu wrote:
> > > > Clean up the referring to the code in commit 52cdbdd49853 ("driver core:
> > > > correct device's shutdown order"). So later we can revert it safely.
> > > >
> > > > Cc: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> > > > Cc: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> > > > Cc: Grygorii Strashko <grygorii.strashko@xxxxxx>
> > > > Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx>
> > > > Cc: Bjorn Helgaas <helgaas@xxxxxxxxxx>
> > > > Cc: Dave Young <dyoung@xxxxxxxxxx>
> > > > Cc: linux-pci@xxxxxxxxxxxxxxx
> > > > Cc: linuxppc-dev@xxxxxxxxxxxxxxxx
> > > > Signed-off-by: Pingfan Liu <kernelfans@xxxxxxxxx>
> > > > ---
> > > >  drivers/base/core.c | 7 -------
> > > >  1 file changed, 7 deletions(-)
> > > >
> > > > diff --git a/drivers/base/core.c b/drivers/base/core.c
> > > > index 684b994..db3deb8 100644
> > > > --- a/drivers/base/core.c
> > > > +++ b/drivers/base/core.c
> > > > @@ -127,13 +127,6 @@ static int device_reorder_to_tail(struct device *dev, void *not_used)
> > > >  {
> > > >       struct device_link *link;
> > > >
> > > > -     /*
> > > > -      * Devices that have not been registered yet will be put to the ends
> > > > -      * of the lists during the registration, so skip them here.
> > > > -      */
> > > > -     if (device_is_registered(dev))
> > > > -             devices_kset_move_last(dev);
> > > > -
> > > >       if (device_pm_initialized(dev))
> > > >               device_pm_move_last(dev);
> > >
> > > You can't do this.
> > >
> > > If you do it, that will break power management in some situations.
> > >
> > Could you shed light on it? I had a quick browsing of pm code, but it
> > is a big function, and I got lost in it.
> > If the above code causes failure, then does it imply that the seq in
> > devices_kset should be the same as dpm_list?
>
> Generally, yes it should.
>
> > But in device_shutdown(), it only intersect with pm by
> > pm_runtime_get_noresume(dev) and pm_runtime_barrier(dev). How do these
> > function affect the seq in dpm_list?
>
> They are not related to dpm_list directly.
>
> However, if you shut down a supplier device before its consumer and that
> involves power management, then the consumer shutdown may fail and lock up
> the system
>
Ah, get your point. The patch in this series "[PATCHv3 2/4]
drivers/base: utilize device tree info to shutdown devices" still obey
the shutdown order "parent<-child" and "supplier<-consumer". It just
utilizes device-tree info to achieve this, since it turns out not easy
to maintain such order in devices_kset. As I described in the commit
log of [2/4], it needs two nested recursion, and should consider the
breakage of devices_kset's spinlock.

> I asked you elsewhere to clearly describe the problem you are trying to
> address.  Please do that in the first place.
>
OK, I will reply your question in [0/4]

Thanks,
Pingfan