Re: [PATCH 1/2] Add round_jiffies_up and related routines

[Jens Axboe <jens.axboe@oracle.com>]

On Wed, Nov 05 2008, Alan Stern wrote:
> On Wed, 5 Nov 2008, Tejun Heo wrote:
> 
> > Alan Stern wrote:
> > > This patch (as1158) adds round_jiffies_up() and friends.  These
> > > routines work like the analogous round_jiffies() functions, except
> > > that they will never round down.
> > > 
> > > The new routines will be useful for timeouts where we don't care
> > > exactly when the timer expires, provided it doesn't expire too soon.
> > > 
> > > Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
> > 
> > Heh... I have exactly the same patches but mines were named
> > round_up_jiffies().
> 
> To an American, "round_up_jiffies" sounds like something a cowboy might 
> do.  :-)
> 
> I haven't bothered to look throughout the kernel to see where 
> round_jiffies_up() could be used.  Have you done this?

Heh, I do agree :-)

> > > +	unsigned long j0 = jiffies;
> > > +
> > > +	barrier();	/* Prevent the compiler from aliasing j0 and jiffies */
> > > +	return round_jiffies_common(j + j0, cpu, false) - j0;
> > 
> > jiffies is volatile.  No need for explicit barrier,
> 
> I didn't realize that.  Good, it makes things easier.
> 
> > but this part is
> > necessary for correct operation as if jiffies go up by two the
> > calculation will wrap and the returned value will be very large.  I
> > think this fix deserves a separate patch and proper explanation.
> 
> How about if I remove the barrier() call?  Should this new code still
> go in a separate patch?

I think it's fine as-is without the barrier. Can you resend it as such,
makes it easier to merge up (plus, it does need a new signed-off-by).

-- 
Jens Axboe