Re: [PATCH] resource, kunit: add dependency on SPARSEMEM

[Dan Williams <dan.j.williams@intel.com>]

Huang, Ying wrote:
> Geert Uytterhoeven <geert@linux-m68k.org> writes:
> 
> > Hi Huang,
> >
> > On Tue, Sep 24, 2024 at 3:25 AM Huang, Ying <ying.huang@intel.com> wrote:
> >> Guenter Roeck <linux@roeck-us.net> writes:
> >> > On 9/23/24 05:58, Geert Uytterhoeven wrote:
> >> >> Hi Günter,
> >> >> On Mon, Sep 23, 2024 at 12:50 AM Guenter Roeck <linux@roeck-us.net>
> >> >> wrote:
> >> >>> Building allmodconfig images on systems with SPARSEMEM=n results in
> >> >>> the following message.
> >> >>>
> >> >>> WARNING: unmet direct dependencies detected for GET_FREE_REGION
> >> >>>    Depends on [n]: SPARSEMEM [=n]
> >> >>>    Selected by [m]:
> >> >>>    - RESOURCE_KUNIT_TEST [=m] && RUNTIME_TESTING_MENU [=y] && KUNIT [=m]
> >
> >> After Linus' fix for PHYSMEM_END, GET_FREE_REGION doesn't need to depend
> >> on SPARSEMEM anymore.  So, I think we can remove the dependency.  Can
> >> you check whether the following patch work for you on top of latest
> >> upstream kernel (with Linus' fix).
> >
> > Yes it does, thanks!
> >
> > One remaining issue is that RESOURCE_KUNIT_TEST selects GET_FREE_REGION.
> > IMHO merely enabling a test should not enable extra functionality
> > in the kernel.  Can the individual test(s) that do depend on
> > GET_FREE_REGION be protected by #ifdef CONFIG_GET_FREE_REGION instead?
> 
> After checking GET_FREE_REGION, I don't think that it's a special
> functionality.  I guess it's selectable because it depends on SPARSEMEM
> and to reduce code size.
> 
> Hi, Dan, please correct me if I'm wrong here.

Right, the only reason it is selectable is just to be mindful is
micro-bloat in kernel/resource.c for the small number of drivers that
need that call.

> So, to reduce #ifdef in .c file as much as possible and make code
> simpler, I prefer to select it for RESOURCE_KUNIT_TEST.

I agree with the result, but for a different rationale.

RESOURCE_KUNIT_TEST is simply another "driver" that needs
GET_FREE_REGION. So while I agree with the general idea that "enabling a
test should not enable extra functionality", in this case the *test* is
the extra functionality and GET_FREE_REGION comes along for the ride.