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:   Wed, 10 Jun 2020 08:03:45 -0700
From:   Kees Cook <keescook@chromium.org>
To:     Xiaoming Ni <nixiaoming@huawei.com>
Cc:     ebiederm@xmission.com, ak@linux.intel.com,
        alex.huangjianhui@huawei.com, linzichang@huawei.com,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] sysctl: Delete the code of sys_sysctl
Message-ID: <202006100803.87D14D1CC7@keescook>
References: <1591683605-8585-1-git-send-email-nixiaoming@huawei.com>
 <202006090839.6EDB4BA@keescook>
 <21a9d711-af31-f7bd-1143-af2ca5b77ba9@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <21a9d711-af31-f7bd-1143-af2ca5b77ba9@huawei.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Jun 10, 2020 at 10:17:49PM +0800, Xiaoming Ni wrote:
> On 2020/6/9 23:40, Kees Cook wrote:
> > On Tue, Jun 09, 2020 at 02:20:05PM +0800, Xiaoming Ni wrote:
> > > Since the commit 61a47c1ad3a4dc ("sysctl: Remove the sysctl system call"),
> > > sys_sysctl has lost its actual role: any input can only return an error.
> > > 
> > > Delete the code and return -ENOSYS directly at the function entry
> > > 
> > > Signed-off-by: Xiaoming Ni <nixiaoming@huawei.com>
> > 
> > Looks right to me.
> > 
> > Reviewed-by: Kees Cook <keescook@chromium.org>
> > 
> > Should this be taken a step further and just remove the syscall entirely
> > and update the per-arch tables with the ENOSYS hole?
> > 
> > -Kees
> > 
> Searching for git log, I found a commit record that deleted syscall:
> commit f5b94099739722 ("All Arch: remove linkage for sys_nfsservctl system
> call"). Could I use sys_ni_syscall to implement the hole as in the example
> here?
> E.g:
> diff --git a/arch/arm/tools/syscall.tbl b/arch/arm/tools/syscall.tbl
> index 7b3832d..f36fda6 100644
> --- a/arch/arm/tools/syscall.tbl
> +++ b/arch/arm/tools/syscall.tbl
> @@ -162,7 +162,7 @@
>  146    common  writev                  sys_writev
>  147    common  getsid                  sys_getsid
>  148    common  fdatasync               sys_fdatasync
> -149        common  _sysctl                 sys_sysctl
> +149  common  _sysctl                 sys_ni_syscall
>  150    common  mlock                   sys_mlock
>  151    common  munlock                 sys_munlock
>  152    common  mlockall                sys_mlockall
> diff --git a/arch/arm64/include/asm/unistd32.h
> b/arch/arm64/include/asm/unistd32.h
> index f8dafe9..ca41bb7 100644
> --- a/arch/arm64/include/asm/unistd32.h
> +++ b/arch/arm64/include/asm/unistd32.h
> @@ -308,8 +308,8 @@
>  __SYSCALL(__NR_getsid, sys_getsid)
>  #define __NR_fdatasync 148
>  __SYSCALL(__NR_fdatasync, sys_fdatasync)
> -#define __NR__sysctl 149
> -__SYSCALL(__NR__sysctl, compat_sys_sysctl)
> +                 /* 149 was sys_sysctl */
> +__SYSCALL(149, sys_ni_syscall)
>  #define __NR_mlock 150
>  __SYSCALL(__NR_mlock, sys_mlock)
>  #define __NR_munlock 151
> 
> 
> In this case, I need to modify a lot of code in v2.

Yeah, that looks like a good example.

> Can I add "Reviewed-by:
> Kees Cook <keescook@chromium.org>" to the v2 patch?

No, it'll be very different. I'm still a fan of the change, but send v2
and I can review that separately. Thanks!

-- 
Kees Cook