Re: [PATCH 1/2] async: Handle kthread_run() return codes.

[Cornelia Huck <cornelia.huck@de.ibm.com>]

On Tue, 13 Jan 2009 20:34:34 +0000,
Arjan van de Ven <arjan@infradead.org> wrote:

> >  		ec = atomic_read(&entry_count);
> >  
> >  		while (tc < ec && tc < MAX_THREADS) {
> > -			kthread_run(async_thread, NULL, "async/%i",
> > tc);
> > +			if (IS_ERR(kthread_run(async_thread, NULL,
> > "async/%i",
> > +					       tc)))
> > +				/* Try again later. */
> > +				goto schedule;
> 
> I do not like this recovery to be honest. an msleep() followed by a
> "continue" is probably much better.

Will change.

> 
> 
> > @@ -330,7 +333,9 @@ static int async_manager_thread(void *un
> >  static int __init async_init(void)
> >  {
> >  	if (async_enabled)
> > -		kthread_run(async_manager_thread, NULL, "async/mgr");
> > +		if (IS_ERR(kthread_run(async_manager_thread, NULL,
> > +				       "async/mgr")))
> > +			async_enabled = 0;
> 
> hmm maybe; it might make more sense to set it to 0 here, and have the
> thread itself set the variable to 1..... that way we KNOW the thread is
> running for sure..

That would look a bit strange to me.

setup_async sets async_enabled -> async_init unsets it again ->
async_manager_thread sets it again

If anything, we could use two variables (use_async and async_enabled),
but that smells of overengeneering to me...