Re: [PATCH v2] rcu: Don't return a value from rcu_assign_pointer()

From: Paul E. McKenney
Date: Mon May 27 2019 - 15:27:18 EST


On Mon, May 27, 2019 at 10:57:52AM -0700, Joe Perches wrote:
> On Mon, 2019-05-27 at 10:49 -0700, Paul E. McKenney wrote:
> > On Mon, May 27, 2019 at 10:21:22AM -0700, Joe Perches wrote:
> > > On Mon, 2019-05-27 at 09:10 -0700, Paul E. McKenney wrote:
> > > > On Mon, May 27, 2019 at 10:49:57AM +0200, Andrea Parri wrote:
> > > > > Quoting Paul [1]:
> > > > >
> > > > > "Given that a quick (and perhaps error-prone) search of the uses
> > > > > of rcu_assign_pointer() in v5.1 didn't find a single use of the
> > > > > return value, let's please instead change the documentation and
> > > > > implementation to eliminate the return value."
> > > > >
> > > > > [1] https://lkml.kernel.org/r/20190523135013.GL28207@xxxxxxxxxxxxx
> > > >
> > > > Queued, thank you!
> > > >
> > > > Adding the checkpatch maintainers on CC as well. The "do { } while
> > > > (0)" prevents the return value from being used, by design. Given the
> > > > checkpatch complaint, is there some better way to achieve this?
> > >
> > > Not sure what the checkpatch complaint is here.
> >
> > Checkpatch seems to want at least two statements in each
> > "do { } while (0)" macro definition:
> >
> > WARNING: Single statement macros should not use a do {} while (0) loop
> >
> > > Reading the link above, there seems to be a compiler warning.
> >
> > The compiler warning is a theoretical issue that is being fixed by this
> > patch, and the patch is giving the checkpatch warning.
> >
> > > Perhaps a statement expression macro with no return value?
> > >
> > > #define rcu_assign_pointer(p, v) ({ (p) = (v); ; })
> >
> > This is at best an acquired taste for me...
>
> Another ugly possibility could be:
>
> #define rcu_assign_pointer(p, v) do {if (1) (p) = (v); } while (0)

And, not to be left out, another ugly possibility might be:

#define rcu_assign_pointer(p, v) ((void)((p) = (v)))

> Possibly the best option would be to ignore checkpatch here
> and just add a comment above the use.

Works for me!

Thanx, Paul