Re: [PATCH 5/5] cgroup: separate out cgroup_attach_proc error handling code

[Mandeep Singh Baines <msb-F7+t8E8rja9g9hUCZPvPmw@public.gmane.org>]

Tejun Heo (tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org) wrote:
> Hello,
> 
> On Tue, Dec 20, 2011 at 03:14:33PM -0800, Mandeep Singh Baines wrote:
> > @@ -2067,9 +2067,10 @@ int cgroup_attach_proc(struct cgroup *cgrp, struct task_struct *leader)
> >  	read_unlock(&tasklist_lock);
> >  
> >  	/* methods shouldn't be called if no task is actually migrating */
> > -	retval = 0;
> > -	if (!group_size)
> > +	if (!group_size) {
> > +		retval = 0;
> >  		goto out_free_group_list;
> > +	}
> 
> Eh... I don't think this is an improvement.  It's just different.
> 

The main benefit is that the comment is directly above the code its
describing but I can drop this part of the change.

> > @@ -2126,20 +2127,20 @@ int cgroup_attach_proc(struct cgroup *cgrp, struct task_struct *leader)
> >  	 */
> >  	synchronize_rcu();
> >  	cgroup_wakeup_rmdir_waiter(cgrp);
> > -	retval = 0;
> > +	flex_array_free(group);
> > +	return 0;
> 
> Hmm... maybe goto out_free_group_list?  Duplicating cleanup on success
> and failure paths can lead future updaters forget one of them.  The
> exit path in this function isn't pretty but I don't think the proposed
> patch improves it either.
> 

Should I drop the patch or add the goto? Its 5/5 so easy enough to drop
since nothing else depends on it.

> Thanks.
> 
> -- 
> tejun