On 1/31/20 5:06 AM, Peter Maydell wrote:
> On Wed, 29 Jan 2020 at 23:56, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> Reviewed-by: Alex Bennée <alex.bennee@linaro.org>
>> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>> ---
>> target/arm/cpu.h | 7 -------
>> target/arm/helper.c | 6 +++++-
>> 2 files changed, 5 insertions(+), 8 deletions(-)
>>
>> diff --git a/target/arm/cpu.h b/target/arm/cpu.h
>> index 0477cef1ea..239c9eb783 100644
>> --- a/target/arm/cpu.h
>> +++ b/target/arm/cpu.h
>> @@ -1417,13 +1417,6 @@ static inline void xpsr_write(CPUARMState *env, uint32_t val, uint32_t mask)
>> #define HCR_ATA (1ULL << 56)
>> #define HCR_DCT (1ULL << 57)
>>
>> -/*
>> - * When we actually implement ARMv8.1-VHE we should add HCR_E2H to
>> - * HCR_MASK and then clear it again if the feature bit is not set in
>> - * hcr_write().
>> - */
>> -#define HCR_MASK ((1ULL << 34) - 1)
>> -
>> #define SCR_NS (1U << 0)
>> #define SCR_IRQ (1U << 1)
>> #define SCR_FIQ (1U << 2)
>> diff --git a/target/arm/helper.c b/target/arm/helper.c
>> index 19a57a17da..f5ce05fdf3 100644
>> --- a/target/arm/helper.c
>> +++ b/target/arm/helper.c
>> @@ -4721,7 +4721,8 @@ static const ARMCPRegInfo el3_no_el2_v8_cp_reginfo[] = {
>> static void hcr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
>> {
>> ARMCPU *cpu = env_archcpu(env);
>> - uint64_t valid_mask = HCR_MASK;
>> + /* Begin with bits defined in base ARMv8.0. */
>> + uint64_t valid_mask = MAKE_64BIT_MASK(0, 34);
>>
>> if (arm_feature(env, ARM_FEATURE_EL3)) {
>> valid_mask &= ~HCR_HCD;
>> @@ -4735,6 +4736,9 @@ static void hcr_write(CPUARMState *env, const ARMCPRegInfo *ri, uint64_t value)
>> */
>> valid_mask &= ~HCR_TSC;
>> }
>> + if (cpu_isar_feature(aa64_vh, cpu)) {
>> + valid_mask |= HCR_E2H;
>> + }
>> if (cpu_isar_feature(aa64_lor, cpu)) {
>> valid_mask |= HCR_TLOR;
>> }
>
> Should HCR_E2H be in the list of bits for which we do
> a tlb_flush () in hcr_write()? (Currently we do this for
> VM, PTW and DC.) Given some of the later TLB-flushing
> changes have code that is "we only need to flush these
> TLB indexes when this register is written if E2H==1",
> it makes it easier to be sure we have the right behaviour
> if we don't need to think through scenarios of "write
> to the register, then set E2H==1"...
I don't think so, because when E2H changes, we change mmu_idx entirely.
Between E2 and E20_2 for el2, and (if TGE is also set) E10_0 and E20_0 for el0.
r~