Re: [PATCH v3 5/7] target/arm: Honor the HCR_EL2.TPCP bit

[Peter Maydell]

On Tue, 18 Feb 2020 at 19:10, Richard Henderson
<address@hidden> wrote:
>
> This bit traps EL1 access to cache maintenance insns that operate
> to the point of coherency or persistence.
>
> Signed-off-by: Richard Henderson <address@hidden>
> ---
>  target/arm/helper.c | 39 +++++++++++++++++++++++++++++++--------
>  1 file changed, 31 insertions(+), 8 deletions(-)
>
> diff --git a/target/arm/helper.c b/target/arm/helper.c
> index 52b6e68659..ed34d4200f 100644
> --- a/target/arm/helper.c
> +++ b/target/arm/helper.c
> @@ -4314,6 +4314,28 @@ static CPAccessResult aa64_cacheop_access(CPUARMState 
> *env,
>      return CP_ACCESS_OK;
>  }
>
> +static CPAccessResult aa64_cacheop_poc_access(CPUARMState *env,
> +                                              const ARMCPRegInfo *ri,
> +                                              bool isread)
> +{
> +    /* Cache invalidate/clean to Point of Coherency or Persistence...  */
> +    switch (arm_current_el(env)) {
> +    case 0:
> +        /* ... EL0 must UNDEF unless SCTLR_EL1.UCI is set.  */
> +        if (!(arm_sctlr(env, 0) & SCTLR_UCI)) {
> +            return CP_ACCESS_TRAP;
> +        }
> +        break;

If we're at EL0 and SCTLR.UCI is set, then we should fall
through to honour the HCR_EL2.TPCP check, not always-succeed.

> +    case 1:
> +        /* ... EL1 must trap to EL2 if HCR_EL2.TPCP is set.  */
> +        if (arm_hcr_el2_eff(env) & HCR_TPCP) {
> +            return CP_ACCESS_TRAP_EL2;
> +        }
> +        break;
> +    }
> +    return CP_ACCESS_OK;
> +}

thanks
-- PMM