Re: [PATCH] syscall_user_dispatch: Use CONFIG_SYSCTL for sysctl guard

Joel Granados <[email protected]> Fri, 14 Aug 2026 12:55:54 +0200

Newsgroups	org.kernel.vger.linux-next,org.kernel.vger.linux-doc,org.kernel.vger.linux-kernel
Message-ID	<mnqdw5ldn7xeif6xyfkf3m2stl42np47dlhqj5tt5czur6n3mt@aufccz5rwjhx>

On Thu, Aug 13, 2026 at 08:20:39AM +0200, Karl Mehltretter wrote:
> sysctl-next removes CONFIG_PROC_SYSCTL because it mirrors CONFIG_SYSCTL.
> When combined with the syscall user dispatch sysctl in linux-next, the
> stale guard prevents registration of kernel.syscall_user_dispatch.
> 
> syscall_user_dispatch_allowed defaults to true. SUD therefore remains
> available, but administrators cannot disable new activations.
> 
> Use CONFIG_SYSCTL for the guard and documentation.
> 
> Fixes: 5b6e32ba7b59 ("syscall_user_dispatch: Add kernel.syscall_user_dispatch sysctl")
> Assisted-by: Codex:gpt-5.6-sol
> Signed-off-by: Karl Mehltretter <[email protected]>
> ---
> The conflicting sysctl-next commit is currently:
> 
>   8d75c338f0bc ("sysctl: remove CONFIG_PROC_SYSCTL, it just mirrors CONFIG_SYSCTL")
> 
> This is a cross-tree integration fix intended for the tip tree. It has
> no functional effect in tip/master, but restores the sysctl when
> combined with sysctl-next.
> 
> Verified in linux-next with an x86_64 object build using
> CONFIG_SYSCALL_USER_DISPATCH=y and CONFIG_SYSCTL=y.
> 
>  Documentation/admin-guide/sysctl/kernel.rst | 2 +-
>  kernel/entry/syscall_user_dispatch.c        | 4 ++--
>  2 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/Documentation/admin-guide/sysctl/kernel.rst b/Documentation/admin-guide/sysctl/kernel.rst
> index b6328cd0f43e9..ffea61d448ebb 100644
> --- a/Documentation/admin-guide/sysctl/kernel.rst
> +++ b/Documentation/admin-guide/sysctl/kernel.rst
> @@ -1416,7 +1416,7 @@ Controls whether userspace may arm Syscall User Dispatch via
>    ==  ===================================================================
>  
>  Only present when the kernel is built with ``CONFIG_SYSCALL_USER_DISPATCH``
> -and ``CONFIG_PROC_SYSCTL``.
> +and ``CONFIG_SYSCTL``.
>  
>  
>  sysctl_writes_strict
> diff --git a/kernel/entry/syscall_user_dispatch.c b/kernel/entry/syscall_user_dispatch.c
> index 2002c7aae4358..59c861866941d 100644
> --- a/kernel/entry/syscall_user_dispatch.c
> +++ b/kernel/entry/syscall_user_dispatch.c
> @@ -178,7 +178,7 @@ int syscall_user_dispatch_set_config(struct task_struct *task, unsigned long siz
>  					      (char __user *)(uintptr_t)cfg.selector);
>  }
>  
> -#ifdef CONFIG_PROC_SYSCTL
> +#ifdef CONFIG_SYSCTL
>  static const struct ctl_table syscall_user_dispatch_sysctls[] = {
>  	{
>  		.procname	= "syscall_user_dispatch",
> @@ -195,4 +195,4 @@ static int __init syscall_user_dispatch_sysctl_init(void)
>  	return 0;
>  }
>  late_initcall(syscall_user_dispatch_sysctl_init);
> -#endif /* CONFIG_PROC_SYSCTL */
> +#endif /* CONFIG_SYSCTL */
Reviewed-by: Joel Granados <[email protected]>

This looks good to me. It still depends on sysctl-next getting upstream.
I don't see why that would not happen, but it is worth checking after
the merge window if this change is still valid.

Best

Joel

signature.asc (application/pgp-signature, 659 B)