Re: [PATCH v2 3/3] cpuset: PF_SPREAD_PAGE and PF_SPREAD_SLAB should beatomic flags

[Tejun Heo <tj-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org>]

On Tue, Sep 23, 2014 at 07:55:48PM +0900, Tetsuo Handa wrote:
> Zefan Li wrote:
> > Tetsuo reported a hard-to-reproduce kernel crash on RHEL6, which happend
> 
> s/happend/happened/
> 
> > @@ -1972,6 +1973,14 @@ static inline void memalloc_noio_restore(unsigned int flags)
> >  TASK_PFA_TEST(NO_NEW_PRIVS, no_new_privs)
> >  TASK_PFA_SET(NO_NEW_PRIVS, no_new_privs)
> >  
> > +TASK_PFA_TEST(SPREAD_PAGE, spread_page)
> > +TASK_PFA_SET(SPREAD_PAGE, spread_page)
> > +TASK_PFA_CLEAR(SPREAD_PAGE, spread_page)
> > +
> > +TASK_PFA_TEST(SPREAD_SLAB, spread_slab)
> > +TASK_PFA_SET(SPREAD_SLAB, spread_slab)
> > +TASK_PFA_CLEAR(SPREAD_SLAB, spread_slab)
> > +
> 
> I wonder how adding 3 macro lines differs from 3 inlined functions.
> Personally, from LXR (source code browser) point of view, inlined functions
> are more friendly than macros. Also, I wonder about the cost of extracting
> macros in a file which is likely included by every file but referenced
> by few files. Speak of SPREAD_PAGE and SPREAD_SLAB, they should be defined
> as inlined functions in include/linux/cpuset.h rather than as macros in
> include/linux/sched.h ?

I think sched.h is fine along w/ inlines for other flags but yeah we
might be better off just open-coding them.

Thanks.

-- 
tejun