Re: [PATCH 02/11] Task watchers: Register process events task watcher

[Matt Helsley <matthltc@us.ibm.com>]

On Tue, 2006-06-13 at 19:39 -0500, Chase Venters wrote:
> On Tuesday 13 June 2006 18:54, Matt Helsley wrote:
> 
> > +static int cn_proc_watch_task(struct notifier_block *nb, unsigned long
> > val, +			      void *t)
> > +{
> > +	struct task_struct *task = t;
> 
> Why the copy?

	It shouldn't result in a copy. Since t is a void* I don't think any
additional instructions or stack space are required. This should have
zero runtime cost while improving clarity of the code. It's also needed
in the next patch.

> > +	int rc = NOTIFY_OK;
> > +
> > +	switch (get_watch_event(val)) {
> > +	case WATCH_TASK_CLONE:
> > +		proc_fork_connector(task);
> > +		break;
> > +	case WATCH_TASK_EXEC:
> > +		proc_exec_connector(task);
> > +		break;
> > +	case WATCH_TASK_UID:
> > +		proc_id_connector(task, PROC_EVENT_UID);
> > +		break;
> > +	case WATCH_TASK_GID:
> > +		proc_id_connector(task, PROC_EVENT_GID);
> > +		break;
> > +	case WATCH_TASK_EXIT:
> > +		proc_exit_connector(task);
> > +		break;
> > +	default: /* we don't care about WATCH_TASK_INIT|FREE because we
> > +		    don't keep per-task info */
> > +		rc = NOTIFY_DONE; /* ignore all other notifications */
> > +		break;
> > +	}
> > +	return rc;
> > +}
> > +
> 
> >  /*
> >   * cn_proc_init - initialization entry point
> >   *
> >   * Adds the connector callback to the connector driver.
> >   */
> > @@ -219,11 +259,16 @@ static int __init cn_proc_init(void)
> >  	int err;
> >
> >  	if ((err = cn_add_callback(&cn_proc_event_id, "cn_proc",
> >  	 			   &cn_proc_mcast_ctl))) {
> >  		printk(KERN_WARNING "cn_proc failed to register\n");
> > -		return err;
> > +		goto out;
> >  	}
> > -	return 0;
> > +
> > +	err = register_task_watcher(&cn_proc_nb);
> > +	if (err != 0)
> 
> if (err)

I don't see any benefit to changing this. Care to elaborate on why this
is important?

> Thanks,
> Chase