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;
MIME-Version: 1.0
References: <20230801-bitwise-v1-1-799bec468dc4@google.com>
In-Reply-To: <20230801-bitwise-v1-1-799bec468dc4@google.com>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Tue, 1 Aug 2023 18:07:08 -0700
Message-ID: <CAHk-=wgkC80Ey0Wyi3zHYexUmteeDL3hvZrp=EpMrDccRGmMwA@mail.gmail.com>
Subject: Re: [PATCH] word-at-a-time: use the same return type for has_zero
 regardless of endianness
To:     ndesaulniers@google.com
Cc:     Arnd Bergmann <arnd@arndb.de>,
        Nathan Chancellor <nathan@kernel.org>,
        Tom Rix <trix@redhat.com>, linux-arch@vger.kernel.org,
        linux-kernel@vger.kernel.org, llvm@lists.linux.dev
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, 1 Aug 2023 at 15:22, <ndesaulniers@google.com> wrote:
>
> Compiling big-endian targets with Clang produces the diagnostic:
>
> fs/namei.c:2173:13: warning: use of bitwise '|' with boolean operands
> [-Wbitwise-instead-of-logical]
> } while (!(has_zero(a, &adata, &constants) | has_zero(b, &bdata, &constants)));

Gaah.

Yes, I think that 'has_zero()' should return the 'unsigned long' bits
on big-endian too, because I do think we always want the bit ops, and
turn it into a boolean only at the very end.

> It appears that when has_zero was introduced, two definitions were
> produced with different signatures (in particular different return types).

Big-endian was kind of a later addition, and while that file is called
"generic", it's really "little-endian has an easier time of this all,
but let's do the 'generic' file for the more complicated case".

Who ends up being affected by this? Powerpc does its own
word-at-a-time thing because the big-endian case is nasty and you can
do better with special instructions that they have.

Who else is even BE any more? Some old 32-bit arm setup?

I think the patch is fine, but I guess I'd like to know that people
who are affected actually don't see any code generation changes (or
possibly see improvements from not turning it into a bool until later)

            Linus