Re: [PATCH v4 1/3] dax: Add an enum for specifying dax wakup mode

[Matthew Wilcox <willy@infradead.org>]

On Mon, Apr 26, 2021 at 01:52:17PM -0400, Vivek Goyal wrote:
> On Mon, Apr 26, 2021 at 02:46:32PM +0100, Matthew Wilcox wrote:
> > On Fri, Apr 23, 2021 at 09:07:21AM -0400, Vivek Goyal wrote:
> > > +enum dax_wake_mode {
> > > +	WAKE_NEXT,
> > > +	WAKE_ALL,
> > > +};
> > 
> > Why define them in this order when ...
> > 
> > > @@ -196,7 +207,7 @@ static void dax_wake_entry(struct xa_state *xas, void *entry, bool wake_all)
> > >  	 * must be in the waitqueue and the following check will see them.
> > >  	 */
> > >  	if (waitqueue_active(wq))
> > > -		__wake_up(wq, TASK_NORMAL, wake_all ? 0 : 1, &key);
> > > +		__wake_up(wq, TASK_NORMAL, mode == WAKE_ALL ? 0 : 1, &key);
> > 
> > ... they're used like this?  This is almost as bad as
> > 
> > enum bool {
> > 	true,
> > 	false,
> > };
> 
> Hi Matthew,
> 
> So you prefer that I should switch order of WAKE_NEXT and WAKE_ALL? 
> 
> enum dax_wake_mode {
> 	WAKE_ALL,
> 	WAKE_NEXT,
> };

That, yes.

> And then do following to wake task.
> 
> if (waitqueue_active(wq))
> 	__wake_up(wq, TASK_NORMAL, mode, &key);

No, the third argument to __wake_up() is a count, not an enum.  It just so
happens that '0' means 'all' and we only ever wake up 1 and not, say, 5.
So the logical way to define the enum is ALL, NEXT which _just happens
to match_ the usage of __wake_up().