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: <514ccb3b1a2fbd37f238f169eef0f40c89d3e3dd.camel@xry111.site>
Subject: Re: [PATCH] LoongArch: Add unaligned access support
From:   Xi Ruoyao <xry111@xry111.site>
To:     Arnd Bergmann <arnd@arndb.de>, Huacai Chen <chenhuacai@kernel.org>
Cc:     Huacai Chen <chenhuacai@loongson.cn>, loongarch@lists.linux.dev,
        Xuefeng Li <lixuefeng@loongson.cn>,
        Tiezhu Yang <yangtiezhu@loongson.cn>,
        guoren <guoren@kernel.org>, WANG Xuerui <kernel@xen0n.name>,
        Jiaxun Yang <jiaxun.yang@flygoat.com>,
        linux-kernel@vger.kernel.org
Date:   Mon, 17 Oct 2022 16:13:47 +0800
In-Reply-To: <b97afda4-143c-46d2-a6af-dd6a700472ec@app.fastmail.com>
References: <20221016133418.2122777-1-chenhuacai@loongson.cn>
         <506fe4e5-a203-48e6-84a6-f70133be15dd@app.fastmail.com>
         <CAAhV-H7UJDgtY4NfF7-5+TbNEbec7XOpvS87H=fPad4KK0KLaw@mail.gmail.com>
         <b97afda4-143c-46d2-a6af-dd6a700472ec@app.fastmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.46.0 
MIME-Version: 1.0
Precedence: bulk

On Mon, 2022-10-17 at 09:38 +0200, Arnd Bergmann wrote:

> > Some unaligned accesses are observed from the kernel network stack, it
> > seems related to whether the packet aligns to IP header or MAC header.
>=20
> This is usually a bug in the device driver. It's a fairly common bug
> since the network driver has to ensure the alignment is correct, but
> it's usually fixable, and fixing it results in better performance on
> machines that support unaligned access as well.

Or, maybe a GCC bug is causing -mstrict-align not implemented correctly.

> > And, gcc has a -mstrict-align parameter, if without this, there are
> > unaligned instructions.
>=20
> Does this default to strict or non-strict mode? Usually gcc does not
> allow to turn this off on architectures that have no hardware support
> for unaligned access.

On LoongArch the unaligned access support is optional.  An
implementation is allowed to implement it or not.  The software can
determine if it's supported by a CPUCFG instruction.

I think -march=3Dla264 will turn off strict align, but it's not added into
GCC yet.

The GCC default is -mno-strict-align.  I expressed my concern about this
decision when I reviewed the GCC port, but at last they just kept the
decision.  But the kernel already sets -mstrict-align in CFLAGS anyway.

--=20
Xi Ruoyao <xry111@xry111.site>
School of Aerospace Science and Technology, Xidian University