Re: [PATCH] vt: add mode validation in vt_setactivate

[Greg KH]

On Sun, May 03, 2026 at 03:46:15AM +0800, Wang Zihan wrote:
> The vt_setactivate() function accepts any mode value without validation,
> while VT_SETMODE correctly rejects invalid values (only VT_AUTO and
> VT_PROCESS are valid).
> 
> This allows users to set invalid mode values (e.g., 0xFF) which bypasses
> VT_PROCESS signal handling and causes undefined VT switching behavior.
> 
> Fix this by adding the same validation as VT_SETMODE.
> 
> Signed-off-by: Wang Zihan <jiyu03@xxxxxx>
> ---
>  drivers/tty/vt/vt_ioctl.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/tty/vt/vt_ioctl.c b/drivers/tty/vt/vt_ioctl.c
> index 28993a3d0acb..b8787283a0fa 100644
> --- a/drivers/tty/vt/vt_ioctl.c
> +++ b/drivers/tty/vt/vt_ioctl.c
> @@ -596,6 +596,8 @@ static int vt_setactivate(struct vt_setactivate __user *sa)
>  		return -EFAULT;
>  	if (vsa.console == 0 || vsa.console > MAX_NR_CONSOLES)
>  		return -ENXIO;
> +	if (vsa.mode.mode != VT_AUTO && vsa.mode.mode != VT_PROCESS)
> +		return -EINVAL;

Are you sure this isn't intentional?  How was this tested?  If an
"invalid" mode is set, what happens?

thanks,

greg k-h