Hi James,
On Wed, Apr 22, 2020 at 06:32:09PM -0700, James Smart wrote:
> On 4/15/2020 11:19 AM, Daniel Wagner wrote:
> > > +void
> > > +efc_node_transition(struct efc_node *node,
> > > + void *(*state)(struct efc_sm_ctx *,
> > > + enum efc_sm_event, void *), void *data)
> > > +{
> > > + struct efc_sm_ctx *ctx = &node->sm;
> > > +
> > > + if (ctx->current_state == state) {
> > > + efc_node_post_event(node, EFC_EVT_REENTER, data);
> > > + } else {
> > > + efc_node_post_event(node, EFC_EVT_EXIT, data);
> > > + ctx->current_state = state;
> > > + efc_node_post_event(node, EFC_EVT_ENTER, data);
> > > + }
> >
> > Why does efc_node_transition not need to take the efc->lock as in
> > efc_remote_node_cb? How are the state machine state transitions
> > serialized?
>
> efc_remote_node_cb is a callback called from outside the statemachine, so it
> needs to take the lock. efc_node_transition is called from within the
> statemachine, after the lock is taken. In general the lock is taken upon
> entering the statemachine and released before exiting. There isn't granular
> locking within the statemachine.
Thanks for explaining. I find such short explanation extremly helpful
when looking at code. It might be obvious but not for everyone :)
Could you add this as comment somehwere?
Thanks,
Daniel
