Re: [PATCH v3 03/10] qemu/atomic: Add aligned_{int64,uint64}_t types

[Peter Maydell]

On Sat, 17 Jul 2021 at 20:24, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> Use it to avoid some clang-12 -Watomic-alignment errors,
> forcing some structures to be aligned and as a pointer when
> we have ensured that the address is aligned.
>
> Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  accel/tcg/atomic_template.h |  4 ++--
>  include/qemu/atomic.h       | 14 +++++++++++++-
>  include/qemu/stats64.h      |  2 +-
>  softmmu/timers-state.h      |  2 +-
>  linux-user/hppa/cpu_loop.c  |  2 +-
>  util/qsp.c                  |  4 ++--
>  6 files changed, 20 insertions(+), 8 deletions(-)

> diff --git a/linux-user/hppa/cpu_loop.c b/linux-user/hppa/cpu_loop.c
> index 3aaaf3337c..82d8183821 100644
> --- a/linux-user/hppa/cpu_loop.c
> +++ b/linux-user/hppa/cpu_loop.c
> @@ -82,7 +82,7 @@ static abi_ulong hppa_lws(CPUHPPAState *env)
>                  o64 = *(uint64_t *)g2h(cs, old);
>                  n64 = *(uint64_t *)g2h(cs, new);
>  #ifdef CONFIG_ATOMIC64
> -                r64 = qatomic_cmpxchg__nocheck((uint64_t *)g2h(cs, addr),
> +                r64 = qatomic_cmpxchg__nocheck((aligned_uint64_t *)g2h(cs, 
> addr),
>                                                 o64, n64);

This cast is OK, but it took me a while to verify that:
 * we check that 'addr' is 8-aligned further up in this function
 * we check that guest_base is at least page-aligned in
   probe_guest_base(), and there's no way to avoid that function
   getting called if you specify a guest-base value on the command line

Is it worth asserting that the value we get back from g2h() really
is 8-aligned before casting ?

Anyway,
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

thanks
-- PMM