[PATCH 05/11] nvmet-fcloop: track tport with ref counting

[Daniel Wagner]

On Fri, Feb 28, 2025 at 09:30:42AM +0100, Daniel Wagner wrote:
> On Fri, Feb 28, 2025 at 08:27:40AM +0100, Hannes Reinecke wrote:
> > >   	if (!tport->remoteport) {
> > >   		tls_req->status = -ECONNREFUSED;
> > >   		spin_lock(&tport->lock);
> > >   		list_add_tail(&tls_req->ls_list, &tport->ls_list);
> > >   		spin_unlock(&tport->lock);
> > > -		queue_work(nvmet_wq, &tport->ls_work);
> > > +		if (queue_work(nvmet_wq, &tport->ls_work))
> > > +			fcloop_tport_get(tport);
> > 
> > Don't you need to remove the 'tls_req' from the list, too, seeing that
> > it'll never be transferred?
> 
> Good point. I'll add the error handling.

In fact, I think a WARN_ONCE is the better choice here, as the element
should not be on the list (an error in the code) and queue_work will only
return false if the work item is already scheduled.