Re: linux-next: build warning after merge of the final tree (Linus' tree related - vai vfs tree)

[Jan Kara <jack@suse.cz>]

On Mon 09-09-13 09:45:52, Olof Johansson wrote:
> On Mon, Sep 9, 2013 at 7:39 AM, Jan Kara <jack@suse.cz> wrote:
> > On Sun 08-09-13 20:21:54, Olof Johansson wrote:
> >> On Fri, Sep 06, 2013 at 10:52:52AM +0200, Geert Uytterhoeven wrote:
> >> > On Fri, Sep 6, 2013 at 9:19 AM, Stephen Rothwell <sfr@canb.auug.org.au> wrote:
> >> > > After merging the final tree, today's linux-next build (arm defconfig)
> >> > > produced this warning:
> >> > >
> >> > > fs/direct-io.c: In function 'sb_init_dio_done_wq':
> >> > > fs/direct-io.c:557:2: warning: value computed is not used [-Wunused-value]
> >> > >
> >> > > This is:
> >> > >
> >> > >         cmpxchg(&sb->s_dio_done_wq, NULL, wq);
> >> > >
> >> > > Introduced by commit 7b7a8665edd8 ("direct-io: Implement generic deferred
> >> > > AIO completions").
> >> >
> >> > This happens for include/asm-generic/cmpxchg.h and several other
> >> > arch-specific implementations that cast the return value of cmpxchg()
> >> > like
> >> >
> >> > #define cmpxchg(ptr, o, n)   ((__typeof__(*(ptr)))__cmpxchg(....
> >> >
> >> > If the caller of cmpxchg() doesn't use the return value, we get a
> >> > compiler warning,
> >> > at least with some versions of gcc.
> >> >
> >> > Any idea how to fix this once and for good?
> >>
> >> Should it be fixed? Chances are that the caller needs to do actions
> >> depending on if the change happened, and checking the value afterwards
> >> is inherently racy.
> >>
> >> For this specific fs/direct-io.c case it seems to be safe since the
> >> workqueue is only ever set and never cleared, but it might still be a
> >> good idea to do:
> >   I'm not against this change - feel free to add my:
> > Reviewed-by: Jan Kara <jack@suse.cz>
> >
> >   to it and merge it. However I maintain there are valid usecases where you
> > do not care about the return value so warning about it doesn't seem right.
> 
> Yes, similar to (some) __must_check-annotated functions.
> 
> > OTOH thinking about it some more I agree we have other precedents where
> > sometimes-correct-often-bugs constructs are warned about and I can see how
> > people can consider cmpxchg() to be that case. But in that case we should:
> >   a) be consistent among architectures about the warning
> >   b) comment at cmpxchg definition that you are supposed to check its
> >      return value. If you really know what you are doing, you can cast the
> >      return value to (void) and comment why it's safe.
> 
> Unfortunately cmpxchg() is a #define since it needs to use __typeof__.
> Marking it __must_check isn't feasible. I'm open for better
> suggestions.
  Well, if you explicitely type return value of cmpxchg() to its type like
some architectures do, then recent gccs will complain when the return value
isn't used. But I agree this probably isn't very futureproof.

								Honza
-- 
Jan Kara <jack@suse.cz>
SUSE Labs, CR