Re: [PATCH v10 03/15] x86/alternatives: Disable LASS when patching kernel alternatives

["Edgecombe, Rick P" <rick.p.edgecombe@intel.com>]

It's not just used for alternatives anymore. bpf, kprobes, etc use it too. Maybe
drop "alternatives" from the subject?

On Mon, 2025-10-06 at 23:51 -0700, Sohil Mehta wrote:
> For patching, the kernel initializes a temporary mm area in the lower
> half of the address range. Disable LASS enforcement by toggling the
> RFLAGS.AC bit during patching to avoid triggering a #GP fault.
> 
> Introduce LASS-specific stac()/clac() helpers along with comments to
> clarify their usage versus the existing stac()/clac() helpers for SMAP.
> 
> Text poking functions used while patching kernel alternatives use the
> standard memcpy()/memset(). However, objtool complains about calling
> dynamic functions within an AC=1 region.
> 
> One workaround is to add memcpy() and memset() to the list of functions
> allowed by objtool. However, that would provide a blanket exemption for
> all usages of memcpy() and memset().
> 
> Instead, replace the standard memcpy() and memset() calls in the text
> poking functions with their unoptimized inline versions. Considering
> that patching is usually small, there is no performance impact expected.
> 
> Signed-off-by: Sohil Mehta <sohil.mehta@intel.com>
> ---
> v10:
>  - Revert to the inline functions instead of open-coding in assembly.
>  - Simplify code comments.
> ---
>  arch/x86/include/asm/smap.h   | 35 +++++++++++++++++++++++++++++++++--
>  arch/x86/kernel/alternative.c | 18 ++++++++++++++++--
>  2 files changed, 49 insertions(+), 4 deletions(-)
> 
> diff --git a/arch/x86/include/asm/smap.h b/arch/x86/include/asm/smap.h
> index 4f84d421d1cf..3ecb4b0de1f9 100644
> --- a/arch/x86/include/asm/smap.h
> +++ b/arch/x86/include/asm/smap.h
> @@ -23,18 +23,49 @@
>  
>  #else /* __ASSEMBLER__ */
>  
> +/*
> + * The CLAC/STAC instructions toggle the enforcement of X86_FEATURE_SMAP
> + * and X86_FEATURE_LASS.
> + *
> + * SMAP enforcement is based on the _PAGE_BIT_USER bit in the page
> + * tables. The kernel is not allowed to touch pages with the bit set
> + * unless the AC bit is set.
> + *
> + * Use stac()/clac() when accessing userspace (_PAGE_USER) mappings,
> + * regardless of location.
> + *
> + * Note: a barrier is implicit in alternative().
> + */
> +
>  static __always_inline void clac(void)
>  {
> -	/* Note: a barrier is implicit in alternative() */
>  	alternative("", "clac", X86_FEATURE_SMAP);
>  }
>  
>  static __always_inline void stac(void)
>  {
> -	/* Note: a barrier is implicit in alternative() */
>  	alternative("", "stac", X86_FEATURE_SMAP);
>  }
>  
> +/*
> + * LASS enforcement is based on bit 63 of the virtual address. The
> + * kernel is not allowed to touch memory in the lower half of the
> + * virtual address space unless the AC bit is set.
> + *
> + * Use lass_stac()/lass_clac() when accessing kernel mappings
> + * (!_PAGE_USER) in the lower half of the address space.
> + */
> +

The above variant has "a barrier is implicit in alternative", is it not needed
here too? Actually, not sure what that comment is trying to convey anyway.

> +static __always_inline void lass_clac(void)
> +{
> +	alternative("", "clac", X86_FEATURE_LASS);
> +}
> +
> +static __always_inline void lass_stac(void)
> +{
> +	alternative("", "stac", X86_FEATURE_LASS);
> +}
> +

Not a strong opinion, but the naming of stac()/clac() lass_stac()/lass_clac() is
a bit confusing to me. stac/clac instructions now has LASS and SMAP behavior.
Why keep the smap behavior implicit and give LASS a special variant?

The other odd aspect is that calling stac()/clac() is needed for LASS in some
places too, right? But stac()/clac() depend on X86_FEATURE_SMAP not
X86_FEATURE_LASS. A reader might wonder, why do we not need the lass variant
there too.

I'd expect in the real world LASS won't be found without SMAP. Maybe it could be
worth just improving the comment around stac()/clac() to include some nod that
it is doing LASS stuff too, or that it relies on that USER mappings are only
found in the lower half, but KERNEL mappings are not only found upper half.

>  static __always_inline unsigned long smap_save(void)
>  {
>  	unsigned long flags;
> diff --git a/arch/x86/kernel/alternative.c b/arch/x86/kernel/alternative.c
> index 79ae9cb50019..dc90b421d760 100644
> --- a/arch/x86/kernel/alternative.c
> +++ b/arch/x86/kernel/alternative.c
> @@ -2409,16 +2409,30 @@ void __init_or_module text_poke_early(void *addr, const void *opcode,
>  __ro_after_init struct mm_struct *text_poke_mm;
>  __ro_after_init unsigned long text_poke_mm_addr;
>  
> +/*
> + * Text poking creates and uses a mapping in the lower half of the
> + * address space. Relax LASS enforcement when accessing the poking
> + * address.
> + *
> + * Also, objtool enforces a strict policy of "no function calls within
> + * AC=1 regions". Adhere to the policy by using inline versions of
> + * memcpy()/memset() that will never result in a function call.

Is "Also, ..." here really a separate issue? What is the connection to
lass_stac/clac()?

> + */
> +
>  static void text_poke_memcpy(void *dst, const void *src, size_t len)
>  {
> -	memcpy(dst, src, len);
> +	lass_stac();
> +	__inline_memcpy(dst, src, len);
> +	lass_clac();
>  }
>  
>  static void text_poke_memset(void *dst, const void *src, size_t len)
>  {
>  	int c = *(const int *)src;
>  
> -	memset(dst, c, len);
> +	lass_stac();
> +	__inline_memset(dst, c, len);
> +	lass_clac();
>  }
>  
>  typedef void text_poke_f(void *dst, const void *src, size_t len);