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: <20221124094845.1907443-1-debug@rivosinc.com> <20230720001852.GA572618@google.com>
In-Reply-To: <20230720001852.GA572618@google.com>
From:   Guo Ren <guoren@kernel.org>
Date:   Thu, 20 Jul 2023 23:06:12 +0800
Message-ID: <CAJF2gTRFvSvEvQeDugdp73o7w4ArdtQ99JScEbLkaLnFcftVcA@mail.gmail.com>
Subject: Re: [PATCH v2] riscv: VMAP_STACK overflow detection thread-safe
To:     Sami Tolvanen <samitolvanen@google.com>
Cc:     Deepak Gupta <debug@rivosinc.com>, palmer@dabbelt.com,
        linux-kernel@vger.kernel.org, linux-riscv@lists.infradead.org,
        Jisheng Zhang <jszhang@kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Jul 20, 2023 at 8:19=E2=80=AFAM Sami Tolvanen <samitolvanen@google.=
com> wrote:
>
> Hi Deepak,
>
> On Thu, Nov 24, 2022 at 01:48:45AM -0800, Deepak Gupta wrote:
> > commit 31da94c25aea ("riscv: add VMAP_STACK overflow detection") added
> > support for CONFIG_VMAP_STACK. If overflow is detected, CPU switches to
> > `shadow_stack` temporarily before switching finally to per-cpu
> > `overflow_stack`.
> >
> > If two CPUs/harts are racing and end up in over flowing kernel stack, o=
ne
> > or both will end up corrupting each other state because `shadow_stack` =
is
> > not per-cpu. This patch optimizes per-cpu overflow stack switch by
> > directly picking per-cpu `overflow_stack` and gets rid of `shadow_stack=
`.
>
> Are you planning on resending this patch? I see it didn't gain much
> traction last time, but this looks like a much cleaner solution for
> selecting the overflow stack than having a `shadow_stack` and calling
> to C to compute the per-CPU offset. The asm_per_cpu macro also would
> come in handy when implementing CONFIG_SHADOW_CALL_STACK, which we'd
> like to have on RISC-V too.
I remember we ended up with an atomic lock mechanism instead of percpu
offset, so what's the benefit of percpu style in overflow_stack path?

>
> Sami


--=20
Best Regards
 Guo Ren