Re: [PATCH] cpuset: Remove unused 'struct cpuset*' variable

[Michal Hocko]

On Sat 26-11-16 08:42:40, Li Zefan wrote:
> On 2016/11/25 17:46, Arnd Bergmann wrote:
> > On Friday, November 25, 2016 1:46:04 PM CET Zefan Li wrote:
> >> On 2016/11/25 12:55, Kirtika Ruchandani wrote:
> >>> 'struct cpuset* cs' that is set but not used, was introduced in commit
> >>> 1f7dd3e5a6e4 ("cgroup: fix handling of multi-destination migration from subtree_control enabling").
> >>> cpuset_cancel_attach() uses css_cs(css) instead. Compiling with W=1
> >>> gives the folllowing harmless warning, which we'd like to fix to
> >>> reduce the noise with W=1 in the kernel.
> >>>
> >>> kernel/cpuset.c: In function âcpuset_cancel_attachâ:
> >>> kernel/cpuset.c:1502:17: warning: variable âcsâ set but not used [-Wunused-but-set-variable]
> >>>   struct cpuset *cs;
> >>>                  ^
> >>>
> >>> Fixes: 1f7dd3e5a6e4 ("cgroup: fix handling of multi-destination migration from subtree_control enabling").
> >>
> >> This isn't a bug, so I don't think this tag is proper.
> > 
> > I think it's ok since the changelog makes it clear that the
> > warning is harmless. It's still useful information to know
> > what commit introduced the warning, and the warning is fixed
> > by this patch.
> > 
> 
> People like stable tree maintainers use scripts to find out bug fixes
> that needs to be backported to older kernels, and those scripts tracks
> the Fixes tag. No doubt this patch doesn't require backporting, so
> it's better avoid using this tag.

I would disagree here. Randomly picking up fixes just because they are
Fixing some commit is just too dangerous for the stable trees. Fixes tag
should tell what was the culprit of the issue fixed by the patch,
nothing more and nothing less.

-- 
Michal Hocko
SUSE Labs