Re: [PATCH] target/rx: swap stack pointers on clrpsw/setpsw instruction

[Yoshinori Sato <ysato@users.sourceforge.jp>]

On Sat, 16 Apr 2022 12:20:09 +0900,
Tomoaki Kawada wrote:
> 
> The control register field PSW.U determines which stack pointer register
> (ISP or USP) is mapped as R0. In QEMU, this is implemented by having a
> value copied between ISP or USP and R0 whenever PSW.U is updated or
> access to ISP/USP is made by an mvtc/mvic instruction. However, this
> update process was incorrectly omitted in the clrpsw/setpsw (clear/set
> PSW) instructions, causing stack pointers to go out-of-sync.
> 
> This patch updates the clrpsw/setpsw translator to handle PSW.U updates
> correctly and fix this problem.
> 
> Signed-off-by: Tomoaki Kawada <i@yvt.jp>
> ---
>  target/rx/translate.c | 16 ++++++++++++++++
>  1 file changed, 16 insertions(+)
> 
> diff --git a/target/rx/translate.c b/target/rx/translate.c
> index 5db8f79a82..c282433fb7 100644
> --- a/target/rx/translate.c
> +++ b/target/rx/translate.c
> @@ -2135,6 +2135,7 @@ enum {
>  
>  static inline void clrsetpsw(DisasContext *ctx, int cb, int val)
>  {
> +    TCGv z;
>      if (cb < 8) {
>          switch (cb) {
>          case PSW_C:
> @@ -2160,7 +2161,22 @@ static inline void clrsetpsw(DisasContext *ctx, int cb, int val)
>              ctx->base.is_jmp = DISAS_UPDATE;
>              break;
>          case PSW_U:
> +            z = tcg_const_i32(0);
> +
> +            /* (PSW.U ? USP : ISP) = R0 */
> +            tcg_gen_movcond_i32(TCG_COND_NE, cpu_usp,
> +                                cpu_psw_u, z, cpu_sp, cpu_usp);
> +            tcg_gen_movcond_i32(TCG_COND_EQ, cpu_isp,
> +                                cpu_psw_u, z, cpu_sp, cpu_isp);
> +
> +            /* Set PSW.U */
>              tcg_gen_movi_i32(cpu_psw_u, val);
> +
> +            /* R0 = (PSW.U ? USP : ISP) */
> +            tcg_gen_movcond_i32(TCG_COND_NE, cpu_sp,
> +                                cpu_psw_u, z, cpu_usp, cpu_isp);
> +
> +            tcg_temp_free(z);
>              break;
>          default:
>              qemu_log_mask(LOG_GUEST_ERROR, "Invalid distination %d", cb);
> -- 
> 2.35.1
> 

Overall looks good.
I have same comment as Richard.
Can you fix it like his comment?

Thaks.

-- 
Yosinori Sato