[dpdk-dev v2 1/2] eal: Add rte_consttime_memsq() to prevent timing attacks memcmp.

[Stephen Hemminger]

On Fri, 26 Sep 2025 15:49:04 +0000
Kai Ji <kai.ji at intel.com> wrote:

> Bugzilla ID: 1773
> Cc: stable at dpdk.org
> 
> [0] https://bugs.dpdk.org/show_bug.cgi?id=1773
> 
> Signed-off-by: Kai Ji <kai.ji at intel.com>
> ---
>  lib/eal/include/rte_common.h | 34 ++++++++++++++++++++++++++++++++++
>  1 file changed, 34 insertions(+)
> 
> diff --git a/lib/eal/include/rte_common.h b/lib/eal/include/rte_common.h
> index 9e7d84f929..ddbba083be 100644
> --- a/lib/eal/include/rte_common.h
> +++ b/lib/eal/include/rte_common.h
> @@ -700,6 +700,40 @@ rte_is_aligned(const void * const __rte_restrict ptr, const unsigned int align)
>  	return ((uintptr_t)ptr & (align - 1)) == 0;
>  }
>  
> +/**
> + * Constant-time memory inequality comparison.
> + *
> + * This function compares two memory regions in constant time, making it
> + * resistant to timing side-channel attacks. The execution time depends only
> + * on the length parameter, not on the actual data values being compared.
> + *
> + * This is particularly important for cryptographic operations where timing
> + * differences could leak information about secret keys, passwords, or other
> + * sensitive data.
> + *
> + * @param a
> + *   Pointer to the first memory region to compare
> + * @param b
> + *   Pointer to the second memory region to compare
> + * @param n
> + *   Number of bytes to compare
> + * @return
> + *   false if the memory regions are identical, true if they differ
> + */
> +static inline bool
> +rte_consttime_memneq(const void *a, const void *b, size_t n)
> +{
> +	const volatile uint8_t *pa = (const volatile uint8_t *)a;
> +	const volatile uint8_t *pb = (const volatile uint8_t *)b;
> +	uint8_t result = 0;
> +	size_t i;
> +
> +	for (i = 0; i < n; i++)
> +		result |= pa[i] ^ pb[i];
> +
> +	return result != 0;
> +}
> +
>  /*********** Macros for compile type checks ********/
>  
>  /* Workaround for toolchain issues with missing C11 macro in FreeBSD */

Also need functional tests for any new function in the functional test suites.
Just some basic tests, suggest using random data, and lengths; validate with memcmp.
And/or see what freebsd already has.