Re: [PATCH 7/8] target/arm: Implement FEAT_ECV CNTPOFF_EL2 handling

[Peter Maydell]

On Fri, 1 Mar 2024 at 21:54, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 3/1/24 08:32, Peter Maydell wrote:
> > +static uint64_t gt_phys_raw_cnt_offset(CPUARMState *env)
> > +{
> > +    if ((env->cp15.scr_el3 & SCR_ECVEN) &&
> > +        FIELD_EX64(env->cp15.cnthctl_el2, CNTHCTL, ECV) &&
> > +        arm_is_el2_enabled(env) &&
> > +        (arm_hcr_el2_eff(env) & (HCR_E2H | HCR_TGE)) != (HCR_E2H | 
> > HCR_TGE)) {
>
> arm_hcr_el2_eff checks arm_is_el2_enabled and returns 0 if disabled.

Yes, and if it returns 0 then the E2H|TGE bits will not be E2H|TGE,
and so we'll incorrectly apply the CNTPOFF value. We can only elide
the arm_is_el2_enabled() test if we're checking for some HCR bit
being 1. (I also initially thought the arm_is_el2_enabled() check was
redundant and then found it was not :-))

-- PMM