Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <99cf531fb8c7e8155edbad56817c8c0208e793c2.camel@svanheule.net>
Subject: Re: [PATCH v3 1/4] cpumask: Fix invalid uniprocessor mask assumption
From:   Sander Vanheule <sander@svanheule.net>
To:     Andy Shevchenko <andriy.shevchenko@linux.intel.com>,
        kernel test robot <lkp@intel.com>
Cc:     Peter Zijlstra <peterz@infradead.org>,
        Yury Norov <yury.norov@gmail.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Valentin Schneider <vschneid@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Marco Elver <elver@google.com>, kbuild-all@lists.01.org,
        Linux Memory Management List <linux-mm@kvack.org>,
        linux-kernel@vger.kernel.org
Date:   Wed, 08 Jun 2022 19:59:24 +0200
In-Reply-To: <Yp3ZsoPV0yx2TKo3@smile.fi.intel.com>
References: <cc1f7d3334348cccbf9fde091015a802f379a9be.1654410109.git.sander@svanheule.net>
         <202206060858.wA0FOzRy-lkp@intel.com> <Yp3ZsoPV0yx2TKo3@smile.fi.intel.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.44.2 (3.44.2-1.fc36) 
MIME-Version: 1.0
Precedence: bulk

On Mon, 2022-06-06 at 13:40 +0300, Andy Shevchenko wrote:
> On Mon, Jun 06, 2022 at 08:48:05AM +0800, kernel test robot wrote:
> > Hi Sander,
> >=20
> > I love your patch! Yet something to improve:
> >=20
> > [auto build test ERROR on akpm-mm/mm-everything]
> > [also build test ERROR on linus/master v5.18 next-20220603]
> > [If your patch is applied to the wrong git tree, kindly drop us a note.
> > And when submitting patch, we suggest to use '--base' as documented in
> > https://git-scm.com/docs/git-format-patch]
> >=20
> > url:=C2=A0=C2=A0=C2=A0
> > https://github.com/intel-lab-lkp/linux/commits/Sander-Vanheule/cpumask-=
Fix-invalid-uniprocessor-assumptions/20220606-004959
> > base:=C2=A0=C2=A0 https://git.kernel.org/pub/scm/linux/kernel/git/akpm/=
mm.git=C2=A0mm-everything
> > config: i386-randconfig-a009
> > (https://download.01.org/0day-ci/archive/20220606/202206060858.wA0FOzRy=
-lkp@intel.com/config)
> > compiler: gcc-11 (Debian 11.3.0-1) 11.3.0
> > reproduce (this is a W=3D1 build):
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 # https://github.com/intel-l=
ab-lkp/linux/commit/37b3f10c4604ea299b454f39ac5ba3cad903ae16
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 git remote add linux-review =
https://github.com/intel-lab-lkp/linux
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 git fetch --no-tags linux-re=
view Sander-Vanheule/cpumask-Fix-invalid-uniprocessor-
> > assumptions/20220606-004959
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 git checkout 37b3f10c4604ea2=
99b454f39ac5ba3cad903ae16
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 # save the config file
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 mkdir build_dir && cp config=
 build_dir/.config
> > =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 make W=3D1 O=3Dbuild_dir ARC=
H=3Di386 SHELL=3D/bin/bash
> >=20
> > If you fix the issue, kindly add following tag where applicable
> > Reported-by: kernel test robot <lkp@intel.com>
> >=20
> > All errors (new ones prefixed by >>):
> >=20
> > =C2=A0=C2=A0 ld: arch/x86/kernel/cpu/cacheinfo.o: in function `__cache_=
amd_cpumap_setup':
> > =C2=A0=C2=A0 arch/x86/kernel/cpu/cacheinfo.c:890: undefined reference t=
o `cpu_llc_shared_map'
> > > > ld: arch/x86/kernel/cpu/cacheinfo.c:895: undefined reference to `cp=
u_llc_shared_map'
>=20
> Seems like somewhere we need stubs for UP builds for those cache related =
functions.
>=20

I think I finally figured out what's going on here.

cpu_llc_shared_map is always declared with DECLARE_PER_CPU_READ_MOSTLY, but=
 defined in
arch/x86/kernel/smpboot.c which only builds on CONFIG_SMP=3Dy.

cpu_llc_shared_map is accessed in a for_each_cpu loop:
for_each_cpu(i, cpu_llc_shared_mask(cpu))

The old (wrong) UP implementation would ignore the mask, so cpu_llc_shared_=
map access was optimised
out and the linker would never see that symbol.

Adding a stub for the two inline functions in arch/x86/include/smp.h won't =
be sufficient I'm afraid.
But I think we can safely make the following assumptions (nobody complained=
 before):
 * anything using cpumask_first_zero(), cpumask_next_zero(), and for_each_c=
pu_not() was expecting an
   empty mask on UP builds,
 * anything else would have expected a filled mask on UP builds.

I'll think about how to identify all the possible cases, but may not be abl=
e to spend a lot of time
on this in the two coming weeks. Any suggestions, or alternative solutions,=
 are of course welcome.

Best,
Sander