Re: [PATCH v5 1/2] remoteproc: qcom: Add per subsystem SSR notification

[Bjorn Andersson]

On Tue 23 Jun 18:41 PDT 2020, rishabhb@xxxxxxxxxxxxxx wrote:

> On 2020-06-23 14:45, Alex Elder wrote:
> > On 6/22/20 8:04 PM, Rishabh Bhatnagar wrote:
> > > Currently there is a single notification chain which is called
> > > whenever any
> > > remoteproc shuts down. This leads to all the listeners being
> > > notified, and
> > > is not an optimal design as kernel drivers might only be interested in
> > > listening to notifications from a particular remoteproc. Create a
> > > global
> > > list of remoteproc notification info data structures. This will hold
> > > the
> > > name and notifier_list information for a particular remoteproc. The
> > > API
> > > to register for notifications will use name argument to retrieve the
> > > notification info data structure and the notifier block will be
> > > added to
> > > that data structure's notification chain. Also move from blocking
> > > notifier
> > > to srcu notifer based implementation to support dynamic notifier head
> > > creation.
> > > 
> > > Signed-off-by: Siddharth Gupta <sidgup@xxxxxxxxxxxxxx>
> > > Signed-off-by: Rishabh Bhatnagar <rishabhb@xxxxxxxxxxxxxx>
> > 
> > Sorry, a few more comments, but I think your next one will
> > likely be fine.
> > 
> > General:
> > - SSR subsystems are added but never removed.  Note that
> >   "qcom_common.o" can be built as a module, and if that
> >   module were ever removed, memory allocated for these
> >   subsystems would be leaked.
> Hi Alex,
> Thank you for reviewing this patchset quickly. This point was
> brought up by Bjorn and it was decided that I will push another patch on
> top in which I'll do the cleanup during module exit.
> > - Will a remoteproc subdev (and in particular, an SSR subdev)
> >   ever be removed?  What happens to entities that have
> >   registered for SSR notifications in that case?
> In practice it should never be removed. If it is clients will
> never get notification about subsystem shutdown/powerup.

Given that clients make direct function calls into qcom_common.ko,
qcom_common.ko would not be possible to rmmod until all clients has been
rmmod'ed. As such there shouldn't be any remaining listeners, or
subdevices, when this happens.

Regards,
Bjorn