Re: [Qemu-devel] [PATCH 2/2 for-2.6] target-sparc: fix Trap Based Address Register behavior for sparc64

[Mark Cave-Ayland]

On 14/04/16 10:29, Artyom Tarasenko wrote:

> Accoding the chapter 7.6 Trap Processing of the SPARC Architecture Manual v9,
> the Trap Based Address Register is not modified as a trap is taken.
> 
> This fix allows booting FreeBSD-10.3-RELEASE-sparc64.
> 
> Signed-off-by: Artyom Tarasenko <address@hidden>
> ---
>  target-sparc/int64_helper.c | 5 ++---
>  1 file changed, 2 insertions(+), 3 deletions(-)
> 
> diff --git a/target-sparc/int64_helper.c b/target-sparc/int64_helper.c
> index aa876cd..29360fa 100644
> --- a/target-sparc/int64_helper.c
> +++ b/target-sparc/int64_helper.c
> @@ -158,9 +158,8 @@ void sparc_cpu_do_interrupt(CPUState *cs)
>      } else if ((intno & 0x1c0) == TT_FILL) {
>          cpu_set_cwp(env, cpu_cwp_inc(env, env->cwp + 1));
>      }
> -    env->tbr &= ~0x7fffULL;
> -    env->tbr |= ((env->tl > 1) ? 1 << 14 : 0) | (intno << 5);
> -    env->pc = env->tbr;
> +    env->pc = env->tbr  & ~0x7fffULL;
> +    env->pc |= ((env->tl > 1) ? 1 << 14 : 0) | (intno << 5);
>      env->npc = env->pc + 4;
>      cs->exception_index = -1;
>  }
> 

Gosh yes, this is entirely obvious when you consider that taking a trap
shouldn't actually alter the register containing the address of the trap
table in memory :(

Again the patch applies with fuzz and checkpatch.pl complains about
line-endings but if you can rebase/resend then you can add my
Reviewed-by. Given that such a trivial fix unlocks FreeBSD then I'm okay
with this in 2.6 too.


ATB,

Mark.