Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210524172433.015b3b6b@xhacker.debian> <20210524172529.3d23c3e7@xhacker.debian>
In-Reply-To: <20210524172529.3d23c3e7@xhacker.debian>
From:   Marco Elver <elver@google.com>
Date:   Mon, 24 May 2021 12:36:34 +0200
Message-ID: <CANpmjNOVikz=u90-xQKzWGxbH_ov5R_EkuG6ZLqVAkjkgw8Z2Q@mail.gmail.com>
Subject: Re: [PATCH 1/2] kfence: allow providing __kfence_pool in arch
 specific way
To:     Jisheng Zhang <Jisheng.Zhang@synaptics.com>
Cc:     Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>,
        Alexander Potapenko <glider@google.com>,
        Dmitry Vyukov <dvyukov@google.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        LKML <linux-kernel@vger.kernel.org>,
        kasan-dev <kasan-dev@googlegroups.com>,
        Linux Memory Management List <linux-mm@kvack.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, 24 May 2021 at 11:26, Jisheng Zhang <Jisheng.Zhang@synaptics.com> wrote:
> Some architectures may want to allocate the __kfence_pool differently
> for example, allocate the __kfence_pool earlier before paging_init().
> We also delay the memset() to kfence_init_pool().
>
> Signed-off-by: Jisheng Zhang <Jisheng.Zhang@synaptics.com>
> ---
>  mm/kfence/core.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/mm/kfence/core.c b/mm/kfence/core.c
> index e18fbbd5d9b4..65f0210edb65 100644
> --- a/mm/kfence/core.c
> +++ b/mm/kfence/core.c
> @@ -430,6 +430,8 @@ static bool __init kfence_init_pool(void)
>         if (!__kfence_pool)
>                 return false;
>
> +       memset(__kfence_pool, 0, KFENCE_POOL_SIZE);
> +

Use memzero_explicit().

Also, for the arm64 case, is delaying the zeroing relevant? You still
call kfence_alloc_pool() in patch 2/2, and zeroing it on
memblock_alloc() is not wrong, correct?

Essentially if there's not going to be any benefit to us doing the
zeroing ourselves, I'd simply leave it as-is and keep using
memblock_alloc(). And if there's some odd architecture that doesn't
even want to use kfence_alloc_pool(), they could just zero the memory
themselves. But we really should use kfence_alloc_pool(), because
otherwise it'll just become unmaintainable if on changes to
kfence_alloc_pool() we have to go and find other special architectures
that don't use it and adjust them, too.

Thanks,
-- Marco

>         if (!arch_kfence_init_pool())
>                 goto err;
>
> @@ -645,10 +647,10 @@ static DECLARE_DELAYED_WORK(kfence_timer, toggle_allocation_gate);
>
>  void __init kfence_alloc_pool(void)
>  {
> -       if (!kfence_sample_interval)
> +       if (!kfence_sample_interval || __kfence_pool)
>                 return;
>
> -       __kfence_pool = memblock_alloc(KFENCE_POOL_SIZE, PAGE_SIZE);
> +       __kfence_pool = memblock_alloc_raw(KFENCE_POOL_SIZE, PAGE_SIZE);
>
>         if (!__kfence_pool)
>                 pr_err("failed to allocate pool\n");
> --
> 2.31.0
>