Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::3:8 as permitted sender) client-ip=2620:137:e000::3:8;
From:   Michael Ellerman <mpe@ellerman.id.au>
To:     Linus Walleij <linus.walleij@linaro.org>,
        Nicholas Piggin <npiggin@gmail.com>,
        Christophe Leroy <christophe.leroy@csgroup.eu>
Cc:     linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org,
        kernel test robot <lkp@intel.com>,
        Linus Walleij <linus.walleij@linaro.org>
Subject: Re: [PATCH] powerpc: Fix signature of pfn_to_kaddr()
In-Reply-To: <20231106-virt-to-pfn-fix-ppc-v1-1-93197a7ccab4@linaro.org>
References: <20231106-virt-to-pfn-fix-ppc-v1-1-93197a7ccab4@linaro.org>
Date:   Tue, 07 Nov 2023 16:57:40 +1100
Message-ID: <87ttpyw1ez.fsf@mail.lhotse>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Linus Walleij <linus.walleij@linaro.org> writes:
> There is a const in the returned value from pfn_to_kaddr()
> but there are consumers that want to modify the result
> and the generic function pfn_to_virt() in <asm-generic/page.h>
> does allow this, so let's relax this requirement and do not
> make the returned value const.
>
> Reported-by: kernel test robot <lkp@intel.com>
> Closes: https://lore.kernel.org/oe-kbuild-all/202311061940.4pBrm44u-lkp@i=
ntel.com/
=20
I'm struggling to connect the removal of const with those bug reports.
It looks like all those warnings are about 0xc000000000000000 being
outside the range of unsigned long when building 32-bit.

Is it the right bug report link?

The current signature of:

  static inline const void *pfn_to_kaddr(unsigned long pfn) ...

seems OK to me.

It allows code like:

  const void *p =3D pfn_to_kaddr(pfn);
  p++;

But errors for:

  const void *p =3D pfn_to_kaddr(pfn);
  unsigned long *q =3D p;
  *q =3D 0;

  error: initialization discards =E2=80=98const=E2=80=99 qualifier from poi=
nter target type


Having said that it looks like almost every caller of pfn_to_kaddr()
casts the result to unsigned long, so possibly that would be the better
return type in terms of the actual usage. Although that would conflict
with __va() which returns void * :/

cheers