Re: [RFC PATCH] target/arm: use x86 intrinsics to implement AES instructions

[Richard Henderson <richard.henderson@linaro.org>]

On 5/30/23 06:52, Ard Biesheuvel wrote:
> ARM intrinsics for AES deviate from the x86 ones in the way they cover
> the different stages of each round, and so mapping one to the other is
> not entirely straight-forward. However, with a bit of care, we can still
> use the x86 ones to emulate the ARM ones, which makes them constant time
> (which is an important property in crypto) and substantially more
> efficient.
> 
> Cc: Peter Maydell <peter.maydell@linaro.org>
> Cc: Alex Bennée <alex.bennee@linaro.org>
> Cc: Richard Henderson <richard.henderson@linaro.org>
> Cc: Philippe Mathieu-Daudé <f4bug@amsat.org>
> Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
> ---
> Suggestions welcome on how to make this more generic across targets and
> compilers etc.
> 
>   target/arm/tcg/crypto_helper.c | 43 ++++++++++++++++++++
>   1 file changed, 43 insertions(+)
> 
> diff --git a/target/arm/tcg/crypto_helper.c b/target/arm/tcg/crypto_helper.c
> index d28690321f..961112b6bd 100644
> --- a/target/arm/tcg/crypto_helper.c
> +++ b/target/arm/tcg/crypto_helper.c
> @@ -18,10 +18,32 @@
>   #include "crypto/sm4.h"
>   #include "vec_internal.h"
>   
> +#ifdef __x86_64
> +#pragma GCC target ("aes")

This doesn't work with clang.

What does work is __attribute__((__target__("aes"))), which requires that you pull those 
little code blocks out to separate functions to be annotated.  I believe they'll be 
inlined afterward, but not really relevant for your improvement.


> +#include <cpuid.h>
> +#include <wmmintrin.h>
> +
> +static bool have_aes(void)
> +{
> +    static int cpuid_have_aes = -1;
> +
> +    if (cpuid_have_aes == -1) {
> +        unsigned int eax, ebx, ecx, edx;
> +        int ret = __get_cpuid(0x1, &eax, &ebx, &ecx, &edx);
> +
> +        cpuid_have_aes = ret && (ecx & bit_AES);
> +    }
> +    return cpuid_have_aes > 0;
> +}
> +#endif

See "host/include/i386/host/cpuinfo.h" and "util/cpuinfo-i386.c".

I'll have a closer look at the other AES uses to see what might be sharable.


r~