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;
Date:   Tue, 14 Sep 2021 17:00:56 +0100
From:   Mark Rutland <mark.rutland@arm.com>
To:     Amit Daniel Kachhap <amit.kachhap@arm.com>
Cc:     linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org,
        Vincenzo Frascino <Vincenzo.Frascino@arm.com>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Will Deacon <will@kernel.org>
Subject: Re: [PATCH] arm64/traps: Avoid unnecessary kernel/user pointer
 conversion
Message-ID: <20210914160056.GA35239@C02TD0UTHF1T.local>
References: <20210914152742.27047-1-amit.kachhap@arm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210914152742.27047-1-amit.kachhap@arm.com>
Precedence: bulk

On Tue, Sep 14, 2021 at 08:57:42PM +0530, Amit Daniel Kachhap wrote:
> Annotating a pointer from kernel to __user and then back again might
> confuse sparse. In call_undef_hook() it can be avoided by not using the
> intermediate user pointer variable.

When you say "might confuse sparse", does it complain today? If so, can
you include an example of what goes wrong?

> Note: This patch adds no functional changes to code.
> 
> Cc: Catalin Marinas <catalin.marinas@arm.com>
> Cc: Will Deacon <will@kernel.org>
> Signed-off-by: Amit Daniel Kachhap <amit.kachhap@arm.com>
> ---
>  arch/arm64/kernel/traps.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/arm64/kernel/traps.c b/arch/arm64/kernel/traps.c
> index b03e383d944a..357d10a8bbf5 100644
> --- a/arch/arm64/kernel/traps.c
> +++ b/arch/arm64/kernel/traps.c
> @@ -404,7 +404,8 @@ static int call_undef_hook(struct pt_regs *regs)
>  
>  	if (!user_mode(regs)) {
>  		__le32 instr_le;
> -		if (get_kernel_nofault(instr_le, (__force __le32 *)pc))
> +		if (get_kernel_nofault(instr_le,
> +				       (__le32 *)instruction_pointer(regs)))

Can we make `pc` an unsigned long, instead?

It'd be nice to handle all three cases consistently, even if that means
adding __force to the two user cases.

Thanks,
Mark.

>  			goto exit;
>  		instr = le32_to_cpu(instr_le);
>  	} else if (compat_thumb_mode(regs)) {
> -- 
> 2.17.1
>