Re: [RFC PATCH 01/13] list: introduce speculative safe list_for_each_entry()

[Greg Kroah-Hartman]

On Thu, Feb 17, 2022 at 07:48:17PM +0100, Jakob Koschel wrote:
> list_for_each_entry() selects either the correct value (pos) or a safe
> value for the additional mispredicted iteration (NULL) for the list
> iterator.
> list_for_each_entry() calls select_nospec(), which performs
> a branch-less select.
> 
> On x86, this select is performed via a cmov. Otherwise, it's performed
> via various shift/mask/etc. operations.
> 
> Kasper Acknowledgements: Jakob Koschel, Brian Johannesmeyer, Kaveh
> Razavi, Herbert Bos, Cristiano Giuffrida from the VUSec group at VU
> Amsterdam.
> 
> Co-developed-by: Brian Johannesmeyer <bjohannesmeyer@xxxxxxxxx>
> Signed-off-by: Brian Johannesmeyer <bjohannesmeyer@xxxxxxxxx>
> Signed-off-by: Jakob Koschel <jakobkoschel@xxxxxxxxx>
> ---
>  arch/x86/include/asm/barrier.h | 12 ++++++++++++
>  include/linux/list.h           |  3 ++-
>  include/linux/nospec.h         | 16 ++++++++++++++++
>  3 files changed, 30 insertions(+), 1 deletion(-)
> 
> diff --git a/arch/x86/include/asm/barrier.h b/arch/x86/include/asm/barrier.h
> index 35389b2af88e..722797ad74e2 100644
> --- a/arch/x86/include/asm/barrier.h
> +++ b/arch/x86/include/asm/barrier.h
> @@ -48,6 +48,18 @@ static inline unsigned long array_index_mask_nospec(unsigned long index,
>  /* Override the default implementation from linux/nospec.h. */
>  #define array_index_mask_nospec array_index_mask_nospec
>  
> +/* Override the default implementation from linux/nospec.h. */
> +#define select_nospec(cond, exptrue, expfalse)				\
> +({									\
> +	typeof(exptrue) _out = (exptrue);				\
> +									\
> +	asm volatile("test %1, %1\n\t"					\
> +	    "cmove %2, %0"						\
> +	    : "+r" (_out)						\
> +	    : "r" (cond), "r" (expfalse));				\
> +	_out;								\
> +})
> +
>  /* Prevent speculative execution past this barrier. */
>  #define barrier_nospec() alternative("", "lfence", X86_FEATURE_LFENCE_RDTSC)
>  
> diff --git a/include/linux/list.h b/include/linux/list.h
> index dd6c2041d09c..1a1b39fdd122 100644
> --- a/include/linux/list.h
> +++ b/include/linux/list.h
> @@ -636,7 +636,8 @@ static inline void list_splice_tail_init(struct list_head *list,
>   */
>  #define list_for_each_entry(pos, head, member)				\
>  	for (pos = list_first_entry(head, typeof(*pos), member);	\
> -	     !list_entry_is_head(pos, head, member);			\
> +	    ({ bool _cond = !list_entry_is_head(pos, head, member);	\
> +	     pos = select_nospec(_cond, pos, NULL); _cond; }); \
>  	     pos = list_next_entry(pos, member))
>  

You are not "introducing" a new macro for this, you are modifying the
existing one such that all users of it now have the select_nospec() call
in it.

Is that intentional?  This is going to hit a _lot_ of existing entries
that probably do not need it at all.

Why not just create list_for_each_entry_nospec()?

thanks,

greg k-h