[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH 14/16] target/arm: Implement ESB instruction
From: |
Peter Maydell |
Subject: |
Re: [PATCH 14/16] target/arm: Implement ESB instruction |
Date: |
Mon, 11 Apr 2022 17:18:39 +0100 |
On Sat, 9 Apr 2022 at 01:18, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> Check for and defer any pending virtual SError.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
> target/arm/helper.h | 1 +
> target/arm/a32.decode | 16 +++++++++-----
> target/arm/t32.decode | 18 +++++++--------
> target/arm/op_helper.c | 45 ++++++++++++++++++++++++++++++++++++++
> target/arm/translate-a64.c | 7 ++++++
> target/arm/translate.c | 10 +++++++++
> 6 files changed, 82 insertions(+), 15 deletions(-)
>
> diff --git a/target/arm/helper.h b/target/arm/helper.h
> index b463d9343b..bb7f901668 100644
> --- a/target/arm/helper.h
> +++ b/target/arm/helper.h
> @@ -54,6 +54,7 @@ DEF_HELPER_1(wfe, void, env)
> DEF_HELPER_1(yield, void, env)
> DEF_HELPER_1(pre_hvc, void, env)
> DEF_HELPER_2(pre_smc, void, env, i32)
> +DEF_HELPER_1(esb, void, env)
>
> DEF_HELPER_3(cpsr_write, void, env, i32, i32)
> DEF_HELPER_2(cpsr_write_eret, void, env, i32)
> diff --git a/target/arm/a32.decode b/target/arm/a32.decode
> index fcd8cd4f7d..f2ca480949 100644
> --- a/target/arm/a32.decode
> +++ b/target/arm/a32.decode
> @@ -187,13 +187,17 @@ SMULTT .... 0001 0110 .... 0000 .... 1110
> .... @rd0mn
>
> {
> {
> - YIELD ---- 0011 0010 0000 1111 ---- 0000 0001
> - WFE ---- 0011 0010 0000 1111 ---- 0000 0010
> - WFI ---- 0011 0010 0000 1111 ---- 0000 0011
> + [
> + YIELD ---- 0011 0010 0000 1111 ---- 0000 0001
> + WFE ---- 0011 0010 0000 1111 ---- 0000 0010
> + WFI ---- 0011 0010 0000 1111 ---- 0000 0011
>
> - # TODO: Implement SEV, SEVL; may help SMP performance.
> - # SEV ---- 0011 0010 0000 1111 ---- 0000 0100
> - # SEVL ---- 0011 0010 0000 1111 ---- 0000 0101
> + # TODO: Implement SEV, SEVL; may help SMP performance.
> + # SEV ---- 0011 0010 0000 1111 ---- 0000 0100
> + # SEVL ---- 0011 0010 0000 1111 ---- 0000 0101
> +
> + ESB ---- 0011 0010 0000 1111 ---- 0001 0000
> + ]
Why don't we decode bits [11:8] here? I see it's the same
as YIELD/WFE/WFI, but I'm not sure why we're not decoding
those bits in those insns either...
>
> # The canonical nop ends in 00000000, but the whole of the
> # rest of the space executes as nop if otherwise unsupported.
> diff --git a/target/arm/t32.decode b/target/arm/t32.decode
> index 78fadef9d6..f21ad0167a 100644
> --- a/target/arm/t32.decode
> +++ b/target/arm/t32.decode
> @@ -364,17 +364,17 @@ CLZ 1111 1010 1011 ---- 1111 .... 1000
> .... @rdm
> [
> # Hints, and CPS
> {
> - YIELD 1111 0011 1010 1111 1000 0000 0000 0001
> - WFE 1111 0011 1010 1111 1000 0000 0000 0010
> - WFI 1111 0011 1010 1111 1000 0000 0000 0011
> + [
> + YIELD 1111 0011 1010 1111 1000 0000 0000 0001
> + WFE 1111 0011 1010 1111 1000 0000 0000 0010
> + WFI 1111 0011 1010 1111 1000 0000 0000 0011
>
> - # TODO: Implement SEV, SEVL; may help SMP performance.
> - # SEV 1111 0011 1010 1111 1000 0000 0000 0100
> - # SEVL 1111 0011 1010 1111 1000 0000 0000 0101
> + # TODO: Implement SEV, SEVL; may help SMP performance.
> + # SEV 1111 0011 1010 1111 1000 0000 0000 0100
> + # SEVL 1111 0011 1010 1111 1000 0000 0000 0101
>
> - # For M-profile minimal-RAS ESB can be a NOP, which is the
> - # default behaviour since it is in the hint space.
> - # ESB 1111 0011 1010 1111 1000 0000 0001 0000
> + ESB 1111 0011 1010 1111 1000 0000 0001 0000
> + ]
>
> # The canonical nop ends in 0000 0000, but the whole rest
> # of the space is "reserved hint, behaves as nop".
> diff --git a/target/arm/op_helper.c b/target/arm/op_helper.c
> index 70b42b55fd..f50424b301 100644
> --- a/target/arm/op_helper.c
> +++ b/target/arm/op_helper.c
> @@ -972,3 +972,48 @@ void HELPER(probe_access)(CPUARMState *env, target_ulong
> ptr,
> access_type, mmu_idx, ra);
> }
> }
> +
> +void HELPER(esb)(CPUARMState *env)
> +{
> + /*
> + * QEMU does not have a source of physical SErrors, so we are
> + * only concerned with virtual SErrors.
> + *
> + * During translation, we have already checked: RAS enabled,
> + * EL2 present (enabled check done in arm_hcr_el2_eff), and
> + * PSTATE.EL in {EL0, EL1}. This function corresponds to
> + * AArch64.vESBOperation(), noting that the AArch32 version
> + * is not functionally different.
> + */
> + uint64_t hcr = arm_hcr_el2_eff(env);
> + bool enabled = !(hcr & HCR_TGE) && (hcr & HCR_AMO);
> + bool pending = enabled && (hcr & HCR_VSE);
> + bool masked = (env->daif & PSTATE_A);
> +
> + /* If VSE pending and masked, defer the exception. */
> + if (pending && masked) {
> + uint32_t syndrome;
> +
> + if (arm_el_is_aa64(env, 1)) {
> + /* Copy across IDS and ISS from VSESR. */
> + syndrome = env->cp15.vsesr_el2 & 0x1ffffff;
> + } else {
> + ARMMMUFaultInfo fi = { .type = ARMFault_AsyncExternal };
> +
> + if (extended_addresses_enabled(env)) {
> + syndrome = arm_fi_to_lfsc(&fi);
> + } else {
> + syndrome = arm_fi_to_sfsc(&fi);
> + }
> + /* Copy across AET and ExT from VSESR. */
> + syndrome |= env->cp15.vsesr_el2 & 0xd000;
> + }
> +
> + /* Set VDISR_EL2.A along with the syndrome. */
> + env->cp15.vdisr_el2 = syndrome | (1u << 31);
> +
> + /* Clear pending virtual SError */
> + env->cp15.hcr_el2 &= ~HCR_VSE;
> + cpu_reset_interrupt(env_cpu(env), CPU_INTERRUPT_VSERR);
> + }
> +}
> diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
> index 9333d7be41..cc54dff83c 100644
> --- a/target/arm/translate-a64.c
> +++ b/target/arm/translate-a64.c
> @@ -1469,6 +1469,13 @@ static void handle_hint(DisasContext *s, uint32_t insn,
> gen_helper_autib(cpu_X[17], cpu_env, cpu_X[17], cpu_X[16]);
> }
> break;
> + case 0b10000: /* ESB */
> + if (dc_isar_feature(aa64_ras, s) &&
> + arm_dc_feature(s, ARM_FEATURE_EL2) &&
> + s->current_el <= 1) {
> + gen_helper_esb(cpu_env);
> + }
> + break;
> case 0b11000: /* PACIAZ */
> if (s->pauth_active) {
> gen_helper_pacia(cpu_X[30], cpu_env, cpu_X[30],
> diff --git a/target/arm/translate.c b/target/arm/translate.c
> index bf2196b9e2..b42ca53d99 100644
> --- a/target/arm/translate.c
> +++ b/target/arm/translate.c
> @@ -6275,6 +6275,16 @@ static bool trans_WFI(DisasContext *s, arg_WFI *a)
> return true;
> }
>
> +static bool trans_ESB(DisasContext *s, arg_ESB *a)
> +{
> + if (dc_isar_feature(aa32_ras, s) &&
> + arm_dc_feature(s, ARM_FEATURE_EL2) &&
> + s->current_el <= 1) {
This is doing the right thing for M-profile but only rather
indirectly because it happens to get caught by the FEATURE_EL2
check. I think it would be safer to explicitly check for
not-M-profile (which then gives you a place to put the
"For M-profile minimal-RAS ESB can be a NOP" comment that got
removed above).
> + gen_helper_esb(cpu_env);
> + }
> + return true;
I think a comment noting that without RAS we must NOP would
be useful here.
> +}
Otherwise
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
thanks
-- PMM
- [PATCH 07/16] target/arm: Use field names for manipulating EL2 and EL3 modes, (continued)
- [PATCH 07/16] target/arm: Use field names for manipulating EL2 and EL3 modes, Richard Henderson, 2022/04/08
- [PATCH 09/16] target/arm: Enable FEAT_Debugv8p4 for -cpu max, Richard Henderson, 2022/04/08
- [PATCH 12/16] target/arm: Enable SCR and HCR bits for RAS, Richard Henderson, 2022/04/08
- [PATCH 11/16] target/arm: Add minimal RAS registers, Richard Henderson, 2022/04/08
- [PATCH 14/16] target/arm: Implement ESB instruction, Richard Henderson, 2022/04/08
- Re: [PATCH 14/16] target/arm: Implement ESB instruction,
Peter Maydell <=
[PATCH 15/16] target/arm: Enable FEAT_RAS for -cpu max, Richard Henderson, 2022/04/08
[PATCH 10/16] target/arm: Add isar_feature_{aa64,any}_ras, Richard Henderson, 2022/04/08
[PATCH 13/16] target/arm: Implement virtual SError exceptions, Richard Henderson, 2022/04/08