Re: [PATCH v4] [x86] detect and report lack of NX protections

["H. Peter Anvin" <hpa@zytor.com>]

On 11/09/2009 02:10 PM, Kees Cook wrote:
> It is possible for x86_64 systems to lack the NX bit (see check_efer())
> either due to the hardware lacking support or the BIOS having turned
> off the CPU capability, so NX status should be reported.  Additionally,
> anyone booting NX-capable CPUs in 32bit mode without PAE will lack NX
> functionality, so this change provides feedback for that case as well.
> 
> v2: use "Alert:" instead of "Warning:" to avoid confusion with WARN_ON()
> v3: use "Notice:" instead of "Alert:" to avoid confusion with KERN_ALERT,
>     and switch to KERN_NOTICE, in keeping with its use for "normal but
>     significant condition" messages.
> v4: check that _NX_PAGE is non-zero to avoid setting nx_enabled accidentally.
> 
> Signed-off-by: Kees Cook <kees.cook@canonical.com>
> ---
>  arch/x86/mm/init.c     |   10 ++++++++++
>  arch/x86/mm/setup_nx.c |    3 +++
>  2 files changed, 13 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/x86/mm/init.c b/arch/x86/mm/init.c
> index 73ffd55..d98b43a 100644
> --- a/arch/x86/mm/init.c
> +++ b/arch/x86/mm/init.c
> @@ -149,6 +149,16 @@ unsigned long __init_refok init_memory_mapping(unsigned long start,
>  	set_nx();
>  	if (nx_enabled)
>  		printk(KERN_INFO "NX (Execute Disable) protection: active\n");
> +	else if (cpu_has_pae)
> +#if defined(CONFIG_X86_64) || defined(CONFIG_X86_PAE)
> +		/* PAE kernel, PAE CPU, without NX */
> +		printk(KERN_NOTICE "Notice: NX (Execute Disable) protection "
> +		       "missing in CPU or disabled in BIOS!\n");
> +#else
> +		/* 32bit non-PAE kernel, PAE CPU */
> +		printk(KERN_NOTICE "Notice: NX (Execute Disable) protection "
> +		       "cannot be enabled: non-PAE kernel!\n");
> +#endif
>  
>  	/* Enable PSE if available */
>  	if (cpu_has_pse)
> diff --git a/arch/x86/mm/setup_nx.c b/arch/x86/mm/setup_nx.c
> index 513d8ed..1b93231 100644
> --- a/arch/x86/mm/setup_nx.c
> +++ b/arch/x86/mm/setup_nx.c
> @@ -53,6 +53,9 @@ void __init set_nx(void)
>  #else
>  void set_nx(void)
>  {
> +	/* notice if _PAGE_NX exists and was removed during check_efer() */
> +	if (_PAGE_NX && ((__supported_pte_mask & _PAGE_NX) == _PAGE_NX))
> +		nx_enabled = 1;
>  }
>  #endif
>  

The second clause can only get executed if CONFIG_X86_PAE is unset,
which in turn means _PAGE_NX == 0... so that piece of code is meaningless.

It also looks to me that there is no message distinguishing the case
when nx_enabled == 1 but disable_nx == 1, and instead we say NX is
"active" when in fact it is disabled in the kernel.

	-hpa