Re: [PATCH v2 12/12] linux-user/alpha: Set r20 secondary return value

[Laurent Vivier]

Le 25/10/2019 à 13:39, Richard Henderson a écrit :
> This value is not, as far as I know, used by any linux software,
> but it is set by the kernel and is part of the ABI.
> 
> Signed-off-by: Richard Henderson <address@hidden>
> ---
> v2: Do not set the parent secondary return if SETTLS.
> ---
>  linux-user/alpha/target_cpu.h | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/linux-user/alpha/target_cpu.h b/linux-user/alpha/target_cpu.h
> index dd25e18f47..ad408ab5cc 100644
> --- a/linux-user/alpha/target_cpu.h
> +++ b/linux-user/alpha/target_cpu.h
> @@ -27,10 +27,19 @@ static inline void cpu_clone_regs_child(CPUAlphaState 
> *env, target_ulong newsp,
>      }
>      env->ir[IR_V0] = 0;
>      env->ir[IR_A3] = 0;
> +    env->ir[IR_A4] = 1;  /* OSF/1 secondary return: child */
>  }
>  
>  static inline void cpu_clone_regs_parent(CPUAlphaState *env, unsigned flags)
>  {
> +    /*
> +     * OSF/1 secondary return: parent
> +     * Note that the kernel does not do this if SETTLS, because the
> +     * settls argument register is still live after copy_thread.
> +     */
> +    if (!(flags & CLONE_SETTLS)) {
> +        env->ir[IR_A4] = 0;
> +    }
>  }
>  
>  static inline void cpu_set_tls(CPUAlphaState *env, target_ulong newtls)
> 

Reviewed-by: Laurent Vivier <address@hidden>