Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190312220752.128141-3-anup.patel@wdc.com> <mhng-3d9a3b8a-2d5d-4db8-b8b0-4ef410a9ad52@palmer-si-x1c4>
In-Reply-To: <mhng-3d9a3b8a-2d5d-4db8-b8b0-4ef410a9ad52@palmer-si-x1c4>
From:   Anup Patel <anup@brainfault.org>
Date:   Wed, 10 Apr 2019 09:40:43 +0530
Message-ID: <CAAhSdy2MQmPte7=YaVw5Ae5vWgQYcHX-zGiEBXP1a15=CkXzwg@mail.gmail.com>
Subject: Re: [PATCH 2/3] RISC-V: Make setup_vm() independent of GCC code model
To:     Palmer Dabbelt <palmer@sifive.com>
Cc:     Anup Patel <Anup.Patel@wdc.com>, Albert Ou <aou@eecs.berkeley.edu>,
        Atish Patra <Atish.Patra@wdc.com>,
        Paul Walmsley <paul.walmsley@sifive.com>,
        Christoph Hellwig <hch@infradead.org>,
        Mike Rapoport <rppt@linux.ibm.com>,
        linux-riscv@lists.infradead.org,
        "linux-kernel@vger.kernel.org List" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Tue, Apr 9, 2019 at 10:17 PM Palmer Dabbelt <palmer@sifive.com> wrote:
>
> On Tue, 12 Mar 2019 15:08:16 PDT (-0700), Anup Patel wrote:
> > The setup_vm() must access kernel symbols in a position independent way
> > because it will be called from head.S with MMU off.
> >
> > If we compile kernel with cmodel=medany then PC-relative addressing will
> > be used in setup_vm() to access kernel symbols so it works perfectly fine.
> >
> > Although, if we compile kernel with cmodel=medlow then either absolute
> > addressing or PC-relative addressing (based on whichever requires fewer
> > instructions) is used to access kernel symbols in setup_vm(). This can
> > break setup_vm() whenever any absolute addressing is used to access
> > kernel symbols.
> >
> > With the movement of setup_vm() from kernel/setup.c to mm/init.c, the
> > setup_vm() is now broken for cmodel=medlow but it works perfectly fine
> > for cmodel=medany.
> >
> > This patch fixes setup_vm() and makes it independent of GCC code model
> > by accessing kernel symbols relative to kernel load address instead of
> > assuming PC-relative addressing.
>
> I think we ended up with a cleaner solution as 387181dcdb6c ("RISC-V: Always
> compile mm/init.c with cmodel=medany and notrace"), but let me know if I missed
> something here.

Yes, please ignore this patch. I have dropped it in latest version of
this patch series.

Regards,
Anup