[RFC PATCH v5 00/22] target/arm: Implement FEAT_NMI and FEAT_GICv3_NMI

[RFC PATCH v5 00/22] target/arm: Implement FEAT_NMI and FEAT_GICv3_NMI

[Jinjie Ruan via]

This patch set implements FEAT_NMI and FEAT_GICv3_NMI for armv8. These
introduce support for a new category of interrupts in the architecture
which we can use to provide NMI like functionality.

There are two modes for using this FEAT_NMI. When PSTATE.ALLINT or
PSTATE.SP & SCTLR_ELx.SCTLR_SPINTMASK is set, any entry to ELx causes all
interrupts including those with superpriority to be masked on entry to ELn
until the mask is explicitly removed by software or hardware. PSTATE.ALLINT
can be managed by software using the new register control ALLINT.ALLINT.
Independent controls are provided for this feature at each EL, usage at EL1
should not disrupt EL2 or EL3.

I have tested it with the following linux patches which try to support
FEAT_NMI in linux kernel:

>-------https://lore.kernel.org/linux-arm-kernel/Y4sH5qX5bK9xfEBp@lpieralisi/T/#mb4ba4a2c045bf72c10c2202c1dd1b82d3240dc88

In the test, SGI, PPI and SPI interrupts can all be set to have super priority
to be converted to a hardware NMI interrupt. The SGI is tested with kernel
IPI as NMI framework, softlockup, hardlockup and kgdb test cases, and the PPI
interrupt is tested with "perf top" command with hardware NMI enabled, and
the SPI interrupt is tested with a custom test module, in which NMI interrupts
can be received and sent normally.

However, the Virtual NMI(VNMI) has not been tested.

         +-------------------------------------------------+
         |               Distributor                       |
         +-------------------------------------------------+
             SPI |  NMI                         |  NMI
                \/                            \/
            +--------+                     +-------+
            | Redist |                     | Redist|
            +--------+                     +-------+
            SGI  | NMI                     PPI | NMI
                \/                            \/
          +-------------+             +---------------+
          |CPU Interface|   ...       | CPU Interface |
          +-------------+             +---------------+
               | NMI                         | NMI
              \/                            \/
            +-----+                       +-----+
            |  PE |                       |  PE |
            +-----+                       +-----+

Changes in v5:
- Remove the comment for ALLINT in cpu.h.
- Merge allint_check() to msr_i_allint to clear the ALLINT MSR (immediate)
  implementation.
- Rename msr_i_allint() to msr_set_allint_el1() to make it clearer.
- Drop the & 1 in trans_MSR_i_ALLINT().
- Add Reviewed-by.

Changes in v4:
- Handle VNMI within the CPU and the GIC.
- Keep PSTATE.ALLINT in env->pstate but not env->allint.
- Fix the ALLINT MSR (immediate) decodetree implementation.
- Accept NMI unconditionally for arm_cpu_has_work() but add comment.
- Improve nmi mask in arm_excp_unmasked().
- Make the GICR_INMIR0 and GICD_INMIR implementation more clearer.
- Improve ICC_NMIAR1_EL1 implementation
- Extract gicv3_get_priority() to avoid priority code repetition.
- Add Reviewed-by.

Changes in v3:
- Remove the FIQ NMI.
- Adjust the patches Sequence.
- Reomve the patch "Set pstate.ALLINT in arm_cpu_reset_hold".
- Check whether support FEAT_NMI and FEAT_GICv3 for FEAT_GICv3_NMI.
- With CPU_INTERRUPT_NMI, both CPSR_I and ISR_IS must be set.
- Not include NMI logic when FEAT_NMI or SCTLR_ELx.NMI not enabled.
- Refator nmi mask in arm_excp_unmasked().
- Add VNMI definitions, add HCRX_VINMI and HCRX_VFNMI support in HCRX_EL2.
- Add Reviewed-by and Acked-by.
- Update the commit message.

Changes in v2:
- Break up the patches so that each one does only one thing.
- Remove the command line option and just implement it in "max" cpu.

Jinjie Ruan (22):
  target/arm: Handle HCR_EL2 accesses for bits introduced with FEAT_NMI
  target/arm: Add PSTATE.ALLINT
  target/arm: Add support for FEAT_NMI, Non-maskable Interrupt
  target/arm: Implement ALLINT MSR (immediate)
  target/arm: Support MSR access to ALLINT
  target/arm: Add support for Non-maskable Interrupt
  target/arm: Add support for NMI in arm_phys_excp_target_el()
  target/arm: Handle IS/FS in ISR_EL1 for NMI
  target/arm: Handle PSTATE.ALLINT on taking an exception
  hw/arm/virt: Wire NMI and VNMI irq lines from GIC to CPU
  hw/intc/arm_gicv3: Add external IRQ lines for NMI
  target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()
  hw/intc/arm_gicv3: Add irq superpriority information
  hw/intc/arm_gicv3_redist: Implement GICR_INMIR0
  hw/intc/arm_gicv3: Implement GICD_INMIR
  hw/intc: Enable FEAT_GICv3_NMI Feature
  hw/intc/arm_gicv3: Add NMI handling CPU interface registers
  hw/intc/arm_gicv3: Implement NMI interrupt prioirty
  hw/intc/arm_gicv3: Report the NMI interrupt in gicv3_cpuif_update()
  hw/intc/arm_gicv3: Report the VNMI interrupt
  target/arm: Add FEAT_NMI to max
  hw/arm/virt: Add FEAT_GICv3_NMI feature support in virt GIC

 docs/system/arm/emulation.rst      |  1 +
 hw/arm/virt.c                      | 25 ++++++++-
 hw/intc/arm_gicv3.c                | 71 +++++++++++++++++++++---
 hw/intc/arm_gicv3_common.c         |  7 +++
 hw/intc/arm_gicv3_cpuif.c          | 77 ++++++++++++++++++++++++--
 hw/intc/arm_gicv3_dist.c           | 36 ++++++++++++
 hw/intc/arm_gicv3_redist.c         | 19 +++++++
 hw/intc/gicv3_internal.h           |  6 ++
 hw/intc/trace-events               |  1 +
 include/hw/intc/arm_gic_common.h   |  2 +
 include/hw/intc/arm_gicv3_common.h |  7 +++
 target/arm/cpu-features.h          |  5 ++
 target/arm/cpu-qom.h               |  4 +-
 target/arm/cpu.c                   | 88 ++++++++++++++++++++++++++++--
 target/arm/cpu.h                   |  7 +++
 target/arm/helper.c                | 73 +++++++++++++++++++++++++
 target/arm/internals.h             | 13 +++++
 target/arm/tcg/a64.decode          |  1 +
 target/arm/tcg/cpu64.c             |  1 +
 target/arm/tcg/helper-a64.c        | 12 ++++
 target/arm/tcg/helper-a64.h        |  1 +
 target/arm/tcg/translate-a64.c     | 18 ++++++
 22 files changed, 454 insertions(+), 21 deletions(-)

-- 
2.34.1

[RFC PATCH v5 01/22] target/arm: Handle HCR_EL2 accesses for bits introduced with FEAT_NMI

[Jinjie Ruan via]

FEAT_NMI defines another three new bits in HCRX_EL2: TALLINT, HCRX_VINMI and
HCRX_VFNMI. When the feature is enabled, allow these bits to be written in
HCRX_EL2.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Update the comment for FEAT_NMI in hcrx_write().
- Update the commit message, s/thress/three/g.
v3:
- Add Reviewed-by.
- Add HCRX_VINMI and HCRX_VFNMI support in HCRX_EL2.
- Upate the commit messsage.
---
 target/arm/cpu-features.h | 5 +++++
 target/arm/helper.c       | 5 +++++
 2 files changed, 10 insertions(+)

diff --git a/target/arm/cpu-features.h b/target/arm/cpu-features.h
index 7567854db6..2ad1179be7 100644
--- a/target/arm/cpu-features.h
+++ b/target/arm/cpu-features.h
@@ -681,6 +681,11 @@ static inline bool isar_feature_aa64_sme(const ARMISARegisters *id)
     return FIELD_EX64(id->id_aa64pfr1, ID_AA64PFR1, SME) != 0;
 }
 
+static inline bool isar_feature_aa64_nmi(const ARMISARegisters *id)
+{
+    return FIELD_EX64(id->id_aa64pfr1, ID_AA64PFR1, NMI) != 0;
+}
+
 static inline bool isar_feature_aa64_tgran4_lpa2(const ARMISARegisters *id)
 {
     return FIELD_SEX64(id->id_aa64mmfr0, ID_AA64MMFR0, TGRAN4) >= 1;
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 90c4fb72ce..affa493141 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -6056,6 +6056,11 @@ static void hcrx_write(CPUARMState *env, const ARMCPRegInfo *ri,
         valid_mask |= HCRX_MSCEN | HCRX_MCE2;
     }
 
+    /* FEAT_NMI adds TALLINT, VINMI and VFNMI */
+    if (cpu_isar_feature(aa64_nmi, env_archcpu(env))) {
+        valid_mask |= HCRX_TALLINT | HCRX_VINMI | HCRX_VFNMI;
+    }
+
     /* Clear RES0 bits.  */
     env->cp15.hcrx_el2 = value & valid_mask;
 }
-- 
2.34.1

[RFC PATCH v5 02/22] target/arm: Add PSTATE.ALLINT

[Jinjie Ruan via]

When PSTATE.ALLINT is set, an IRQ or FIQ interrupt that is targeted to
ELx, with or without superpriority is masked.

As Richard suggested, place ALLINT bit in PSTATE in env->pstate.

With the change to pstate_read/write, exception entry
and return are automatically handled.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v5:
- Remove the ALLINT comment, as it is covered by "all other bits".
- Add Reviewed-by.
v4:
- Keep PSTATE.ALLINT in env->pstate but not env->allint.
- Update the commit message.
v3:
- Remove ALLINT dump in aarch64_cpu_dump_state().
- Update the commit message.
---
 target/arm/cpu.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 63f31e0d98..abb453f733 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -1543,6 +1543,7 @@ FIELD(VTCR, SL2, 33, 1)
 #define PSTATE_D (1U << 9)
 #define PSTATE_BTYPE (3U << 10)
 #define PSTATE_SSBS (1U << 12)
+#define PSTATE_ALLINT (1U << 13)
 #define PSTATE_IL (1U << 20)
 #define PSTATE_SS (1U << 21)
 #define PSTATE_PAN (1U << 22)
-- 
2.34.1

[RFC PATCH v5 03/22] target/arm: Add support for FEAT_NMI, Non-maskable Interrupt

[Jinjie Ruan via]

Add support for FEAT_NMI. NMI (FEAT_NMI) is an mandatory feature in
ARMv8.8-A and ARM v9.3-A.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v3:
- Add Reviewed-by.
- Adjust to before the MSR patches.
---
 target/arm/internals.h | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/target/arm/internals.h b/target/arm/internals.h
index 50bff44549..fee65caba5 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -1078,6 +1078,9 @@ static inline uint32_t aarch64_pstate_valid_mask(const ARMISARegisters *id)
     if (isar_feature_aa64_mte(id)) {
         valid |= PSTATE_TCO;
     }
+    if (isar_feature_aa64_nmi(id)) {
+        valid |= PSTATE_ALLINT;
+    }
 
     return valid;
 }
-- 
2.34.1

[RFC PATCH v5 04/22] target/arm: Implement ALLINT MSR (immediate)

[Jinjie Ruan via]

Add ALLINT MSR (immediate) to decodetree, in which the CRm is 0b000x. The
EL0 check is necessary to ALLINT, and the EL1 check is necessary when
imm == 1. So implement it inline for EL2/3, or EL1 with imm==0. Avoid the
unconditional write to pc and use raise_exception_ra to unwind.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v5:
- Drop the & 1 in trans_MSR_i_ALLINT().
- Simplify and merge msr_i_allint() and allint_check().
- Rename msr_i_allint() to msr_set_allint_el1().
v4:
- Fix the ALLINT MSR (immediate) decodetree implementation.
- Remove arm_is_el2_enabled() check in allint_check().
- Update env->allint to env->pstate.
- Only call allint_check() when imm == 1.
- Simplify the allint_check() to not pass "op" and extract.
- Implement it inline for EL2/3, or EL1 with imm==0.
- Pass (a->imm & 1) * PSTATE_ALLINT (i64) to simplfy the ALLINT set/clear.
v3:
- Remove EL0 check in allint_check().
- Add TALLINT check for EL1 in allint_check().
- Remove unnecessarily arm_rebuild_hflags() in msr_i_allint helper.
---
 target/arm/tcg/a64.decode      |  1 +
 target/arm/tcg/helper-a64.c    | 12 ++++++++++++
 target/arm/tcg/helper-a64.h    |  1 +
 target/arm/tcg/translate-a64.c | 18 ++++++++++++++++++
 4 files changed, 32 insertions(+)

diff --git a/target/arm/tcg/a64.decode b/target/arm/tcg/a64.decode
index 8a20dce3c8..0e7656fd15 100644
--- a/target/arm/tcg/a64.decode
+++ b/target/arm/tcg/a64.decode
@@ -207,6 +207,7 @@ MSR_i_DIT       1101 0101 0000 0 011 0100 .... 010 11111 @msr_i
 MSR_i_TCO       1101 0101 0000 0 011 0100 .... 100 11111 @msr_i
 MSR_i_DAIFSET   1101 0101 0000 0 011 0100 .... 110 11111 @msr_i
 MSR_i_DAIFCLEAR 1101 0101 0000 0 011 0100 .... 111 11111 @msr_i
+MSR_i_ALLINT    1101 0101 0000 0 001 0100 000 imm:1 000 11111
 MSR_i_SVCR      1101 0101 0000 0 011 0100 0 mask:2 imm:1 011 11111
 
 # MRS, MSR (register), SYS, SYSL. These are all essentially the
diff --git a/target/arm/tcg/helper-a64.c b/target/arm/tcg/helper-a64.c
index ebaa7f00df..7818537890 100644
--- a/target/arm/tcg/helper-a64.c
+++ b/target/arm/tcg/helper-a64.c
@@ -66,6 +66,18 @@ void HELPER(msr_i_spsel)(CPUARMState *env, uint32_t imm)
     update_spsel(env, imm);
 }
 
+void HELPER(msr_set_allint_el1)(CPUARMState *env)
+{
+    /* ALLINT update to PSTATE. */
+    if (arm_hcrx_el2_eff(env) & HCRX_TALLINT) {
+        raise_exception_ra(env, EXCP_UDEF,
+                           syn_aa64_sysregtrap(0, 1, 0, 4, 1, 0x1f, 0),
+                           exception_target_el(env), GETPC());
+    }
+
+    env->pstate |= PSTATE_ALLINT;
+}
+
 static void daif_check(CPUARMState *env, uint32_t op,
                        uint32_t imm, uintptr_t ra)
 {
diff --git a/target/arm/tcg/helper-a64.h b/target/arm/tcg/helper-a64.h
index 575a5dab7d..0518165399 100644
--- a/target/arm/tcg/helper-a64.h
+++ b/target/arm/tcg/helper-a64.h
@@ -22,6 +22,7 @@ DEF_HELPER_FLAGS_1(rbit64, TCG_CALL_NO_RWG_SE, i64, i64)
 DEF_HELPER_2(msr_i_spsel, void, env, i32)
 DEF_HELPER_2(msr_i_daifset, void, env, i32)
 DEF_HELPER_2(msr_i_daifclear, void, env, i32)
+DEF_HELPER_1(msr_set_allint_el1, void, env)
 DEF_HELPER_3(vfp_cmph_a64, i64, f16, f16, ptr)
 DEF_HELPER_3(vfp_cmpeh_a64, i64, f16, f16, ptr)
 DEF_HELPER_3(vfp_cmps_a64, i64, f32, f32, ptr)
diff --git a/target/arm/tcg/translate-a64.c b/target/arm/tcg/translate-a64.c
index 340265beb0..14e2b35b28 100644
--- a/target/arm/tcg/translate-a64.c
+++ b/target/arm/tcg/translate-a64.c
@@ -2036,6 +2036,24 @@ static bool trans_MSR_i_DAIFCLEAR(DisasContext *s, arg_i *a)
     return true;
 }
 
+static bool trans_MSR_i_ALLINT(DisasContext *s, arg_i *a)
+{
+    if (!dc_isar_feature(aa64_nmi, s) || s->current_el == 0) {
+        return false;
+    }
+
+    if (a->imm == 0) {
+        clear_pstate_bits(PSTATE_ALLINT);
+    } else if (s->current_el > 1) {
+        set_pstate_bits(PSTATE_ALLINT);
+    } else {
+        gen_helper_msr_set_allint_el1(tcg_env);
+    }
+
+    s->base.is_jmp = DISAS_TOO_MANY;
+    return true;
+}
+
 static bool trans_MSR_i_SVCR(DisasContext *s, arg_MSR_i_SVCR *a)
 {
     if (!dc_isar_feature(aa64_sme, s) || a->mask == 0) {
-- 
2.34.1

Re: [RFC PATCH v5 04/22] target/arm: Implement ALLINT MSR (immediate)

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> +static bool trans_MSR_i_ALLINT(DisasContext *s, arg_i *a)
> +{
> +    if (!dc_isar_feature(aa64_nmi, s) || s->current_el == 0) {
> +        return false;
> +    }
> +
> +    if (a->imm == 0) {
> +        clear_pstate_bits(PSTATE_ALLINT);
> +    } else if (s->current_el > 1) {
> +        set_pstate_bits(PSTATE_ALLINT);
> +    } else {
> +        gen_helper_msr_set_allint_el1(tcg_env);
> +    }
> +
> +    s->base.is_jmp = DISAS_TOO_MANY;
> +    return true;
> +}

I just noticed one final item: for imm == 0, we need

     /* Exit the cpu loop to re-evaluate pending IRQs. */
     s->base.is_jmp = DISAS_UPDATE_EXIT;

like trans_MSR_i_DAIFCLEAR.


r~

[RFC PATCH v5 05/22] target/arm: Support MSR access to ALLINT

[Jinjie Ruan via]

Support ALLINT msr access as follow:
	mrs <xt>, ALLINT	// read allint
	msr ALLINT, <xt>	// write allint with imm

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v5:
- Add Reviewed-by.
v4:
- Remove arm_is_el2_enabled() check in allint_check().
- Change to env->pstate instead of env->allint.
v3:
- Remove EL0 check in aa64_allint_access() which alreay checks in .access
  PL1_RW.
- Use arm_hcrx_el2_eff() in aa64_allint_access() instead of env->cp15.hcrx_el2.
- Make ALLINT msr access function controlled by aa64_nmi.
---
 target/arm/helper.c | 34 ++++++++++++++++++++++++++++++++++
 1 file changed, 34 insertions(+)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index affa493141..497b6e4bdf 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -4618,6 +4618,36 @@ static void aa64_daif_write(CPUARMState *env, const ARMCPRegInfo *ri,
     env->daif = value & PSTATE_DAIF;
 }
 
+static void aa64_allint_write(CPUARMState *env, const ARMCPRegInfo *ri,
+                              uint64_t value)
+{
+    env->pstate = (env->pstate & ~PSTATE_ALLINT) | (value & PSTATE_ALLINT);
+}
+
+static uint64_t aa64_allint_read(CPUARMState *env, const ARMCPRegInfo *ri)
+{
+    return env->pstate & PSTATE_ALLINT;
+}
+
+static CPAccessResult aa64_allint_access(CPUARMState *env,
+                                         const ARMCPRegInfo *ri, bool isread)
+{
+    if (arm_current_el(env) == 1 && (arm_hcrx_el2_eff(env) & HCRX_TALLINT)) {
+        return CP_ACCESS_TRAP_EL2;
+    }
+    return CP_ACCESS_OK;
+}
+
+static const ARMCPRegInfo nmi_reginfo[] = {
+    { .name = "ALLINT", .state = ARM_CP_STATE_AA64,
+      .opc0 = 3, .opc1 = 0, .opc2 = 0, .crn = 4, .crm = 3,
+      .type = ARM_CP_NO_RAW,
+      .access = PL1_RW, .accessfn = aa64_allint_access,
+      .fieldoffset = offsetof(CPUARMState, pstate),
+      .writefn = aa64_allint_write, .readfn = aa64_allint_read,
+      .resetfn = arm_cp_reset_ignore },
+};
+
 static uint64_t aa64_pan_read(CPUARMState *env, const ARMCPRegInfo *ri)
 {
     return env->pstate & PSTATE_PAN;
@@ -9724,6 +9754,10 @@ void register_cp_regs_for_features(ARMCPU *cpu)
     if (cpu_isar_feature(aa64_nv2, cpu)) {
         define_arm_cp_regs(cpu, nv2_reginfo);
     }
+
+    if (cpu_isar_feature(aa64_nmi, cpu)) {
+        define_arm_cp_regs(cpu, nmi_reginfo);
+    }
 #endif
 
     if (cpu_isar_feature(any_predinv, cpu)) {
-- 
2.34.1

[RFC PATCH v5 06/22] target/arm: Add support for Non-maskable Interrupt

[Jinjie Ruan via]

This only implements the external delivery method via the GICv3.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Accept NMI unconditionally for arm_cpu_has_work() but add comment.
- Change from & to && for EXCP_IRQ or EXCP_FIQ.
- Refator nmi mask in arm_excp_unmasked().
- Also handle VNMI in arm_cpu_exec_interrupt() and arm_cpu_set_irq().
- Rename virtual to Virtual.
v3:
- Not include CPU_INTERRUPT_NMI when FEAT_NMI not enabled
- Add ARM_CPU_VNMI.
- Refator nmi mask in arm_excp_unmasked().
- Test SCTLR_ELx.NMI for ALLINT mask for NMI.
---
 target/arm/cpu-qom.h   |  4 +-
 target/arm/cpu.c       | 88 +++++++++++++++++++++++++++++++++++++++---
 target/arm/cpu.h       |  4 ++
 target/arm/helper.c    |  2 +
 target/arm/internals.h | 10 +++++
 5 files changed, 101 insertions(+), 7 deletions(-)

diff --git a/target/arm/cpu-qom.h b/target/arm/cpu-qom.h
index 8e032691db..e0c9e18036 100644
--- a/target/arm/cpu-qom.h
+++ b/target/arm/cpu-qom.h
@@ -36,11 +36,13 @@ DECLARE_CLASS_CHECKERS(AArch64CPUClass, AARCH64_CPU,
 #define ARM_CPU_TYPE_SUFFIX "-" TYPE_ARM_CPU
 #define ARM_CPU_TYPE_NAME(name) (name ARM_CPU_TYPE_SUFFIX)
 
-/* Meanings of the ARMCPU object's four inbound GPIO lines */
+/* Meanings of the ARMCPU object's six inbound GPIO lines */
 #define ARM_CPU_IRQ 0
 #define ARM_CPU_FIQ 1
 #define ARM_CPU_VIRQ 2
 #define ARM_CPU_VFIQ 3
+#define ARM_CPU_NMI 4
+#define ARM_CPU_VNMI 5
 
 /* For M profile, some registers are banked secure vs non-secure;
  * these are represented as a 2-element array where the first element
diff --git a/target/arm/cpu.c b/target/arm/cpu.c
index 5fa86bc8d5..ad6e6200f6 100644
--- a/target/arm/cpu.c
+++ b/target/arm/cpu.c
@@ -122,6 +122,13 @@ void arm_restore_state_to_opc(CPUState *cs,
 }
 #endif /* CONFIG_TCG */
 
+/*
+ * With SCTLR_ELx.NMI == 0, IRQ with Superpriority is masked identically with
+ * IRQ without Superpriority. Moreover, if the GIC is configured so that
+ * FEAT_GICv3_NMI is only set if FEAT_NMI is set, then we won't ever see
+ * CPU_INTERRUPT_*NMI anyway. So we might as well accept NMI here
+ * unconditionally.
+ */
 static bool arm_cpu_has_work(CPUState *cs)
 {
     ARMCPU *cpu = ARM_CPU(cs);
@@ -129,6 +136,7 @@ static bool arm_cpu_has_work(CPUState *cs)
     return (cpu->power_state != PSCI_OFF)
         && cs->interrupt_request &
         (CPU_INTERRUPT_FIQ | CPU_INTERRUPT_HARD
+         | CPU_INTERRUPT_NMI | CPU_INTERRUPT_VNMI
          | CPU_INTERRUPT_VFIQ | CPU_INTERRUPT_VIRQ | CPU_INTERRUPT_VSERR
          | CPU_INTERRUPT_EXITTB);
 }
@@ -668,6 +676,7 @@ static inline bool arm_excp_unmasked(CPUState *cs, unsigned int excp_idx,
     CPUARMState *env = cpu_env(cs);
     bool pstate_unmasked;
     bool unmasked = false;
+    bool allIntMask = false;
 
     /*
      * Don't take exceptions if they target a lower EL.
@@ -678,13 +687,31 @@ static inline bool arm_excp_unmasked(CPUState *cs, unsigned int excp_idx,
         return false;
     }
 
+    if (cpu_isar_feature(aa64_nmi, env_archcpu(env)) &&
+        env->cp15.sctlr_el[target_el] & SCTLR_NMI && cur_el == target_el) {
+        allIntMask = env->pstate & PSTATE_ALLINT ||
+                     ((env->cp15.sctlr_el[target_el] & SCTLR_SPINTMASK) &&
+                      (env->pstate & PSTATE_SP));
+    }
+
     switch (excp_idx) {
+    case EXCP_NMI:
+        pstate_unmasked = !allIntMask;
+        break;
+
+    case EXCP_VNMI:
+        if ((!(hcr_el2 & HCR_IMO) && !(hcr_el2 & HCR_FMO)) ||
+             (hcr_el2 & HCR_TGE)) {
+            /* VNMIs(VIRQs or VFIQs) are only taken when hypervized.  */
+            return false;
+        }
+        return !allIntMask;
     case EXCP_FIQ:
-        pstate_unmasked = !(env->daif & PSTATE_F);
+        pstate_unmasked = (!(env->daif & PSTATE_F)) && (!allIntMask);
         break;
 
     case EXCP_IRQ:
-        pstate_unmasked = !(env->daif & PSTATE_I);
+        pstate_unmasked = (!(env->daif & PSTATE_I)) && (!allIntMask);
         break;
 
     case EXCP_VFIQ:
@@ -692,13 +719,13 @@ static inline bool arm_excp_unmasked(CPUState *cs, unsigned int excp_idx,
             /* VFIQs are only taken when hypervized.  */
             return false;
         }
-        return !(env->daif & PSTATE_F);
+        return !(env->daif & PSTATE_F) && (!allIntMask);
     case EXCP_VIRQ:
         if (!(hcr_el2 & HCR_IMO) || (hcr_el2 & HCR_TGE)) {
             /* VIRQs are only taken when hypervized.  */
             return false;
         }
-        return !(env->daif & PSTATE_I);
+        return !(env->daif & PSTATE_I) && (!allIntMask);
     case EXCP_VSERR:
         if (!(hcr_el2 & HCR_AMO) || (hcr_el2 & HCR_TGE)) {
             /* VIRQs are only taken when hypervized.  */
@@ -804,6 +831,24 @@ static bool arm_cpu_exec_interrupt(CPUState *cs, int interrupt_request)
 
     /* The prioritization of interrupts is IMPLEMENTATION DEFINED. */
 
+    if (cpu_isar_feature(aa64_nmi, env_archcpu(env))) {
+        if (interrupt_request & CPU_INTERRUPT_NMI) {
+            excp_idx = EXCP_NMI;
+            target_el = arm_phys_excp_target_el(cs, excp_idx, cur_el, secure);
+            if (arm_excp_unmasked(cs, excp_idx, target_el,
+                                  cur_el, secure, hcr_el2)) {
+                goto found;
+            }
+        }
+        if (interrupt_request & CPU_INTERRUPT_VNMI) {
+            excp_idx = EXCP_VNMI;
+            target_el = arm_phys_excp_target_el(cs, excp_idx, cur_el, secure);
+            if (arm_excp_unmasked(cs, excp_idx, target_el,
+                                  cur_el, secure, hcr_el2)) {
+                goto found;
+            }
+        }
+    }
     if (interrupt_request & CPU_INTERRUPT_FIQ) {
         excp_idx = EXCP_FIQ;
         target_el = arm_phys_excp_target_el(cs, excp_idx, cur_el, secure);
@@ -900,6 +945,31 @@ void arm_cpu_update_vfiq(ARMCPU *cpu)
     }
 }
 
+void arm_cpu_update_vnmi(ARMCPU *cpu)
+{
+    /*
+     * Update the interrupt level for VNMI, which is the logical OR of
+     * the HCRX_EL2.VINMI or HCRX_EL2.VFNMI bit and the input line level from
+     * the GIC.
+     */
+    CPUARMState *env = &cpu->env;
+    CPUState *cs = CPU(cpu);
+
+    bool new_state = ((env->cp15.hcr_el2 & HCR_VI) &&
+                      (env->cp15.hcrx_el2 & HCRX_VINMI)) ||
+                     ((env->cp15.hcr_el2 & HCR_VF) &&
+                      (env->cp15.hcrx_el2 & HCRX_VFNMI)) ||
+        (env->irq_line_state & CPU_INTERRUPT_VNMI);
+
+    if (new_state != ((cs->interrupt_request & CPU_INTERRUPT_VNMI) != 0)) {
+        if (new_state) {
+            cpu_interrupt(cs, CPU_INTERRUPT_VNMI);
+        } else {
+            cpu_reset_interrupt(cs, CPU_INTERRUPT_VNMI);
+        }
+    }
+}
+
 void arm_cpu_update_vserr(ARMCPU *cpu)
 {
     /*
@@ -929,7 +999,9 @@ static void arm_cpu_set_irq(void *opaque, int irq, int level)
         [ARM_CPU_IRQ] = CPU_INTERRUPT_HARD,
         [ARM_CPU_FIQ] = CPU_INTERRUPT_FIQ,
         [ARM_CPU_VIRQ] = CPU_INTERRUPT_VIRQ,
-        [ARM_CPU_VFIQ] = CPU_INTERRUPT_VFIQ
+        [ARM_CPU_VFIQ] = CPU_INTERRUPT_VFIQ,
+        [ARM_CPU_NMI] = CPU_INTERRUPT_NMI,
+        [ARM_CPU_VNMI] = CPU_INTERRUPT_VNMI
     };
 
     if (!arm_feature(env, ARM_FEATURE_EL2) &&
@@ -955,8 +1027,12 @@ static void arm_cpu_set_irq(void *opaque, int irq, int level)
     case ARM_CPU_VFIQ:
         arm_cpu_update_vfiq(cpu);
         break;
+    case ARM_CPU_VNMI:
+        arm_cpu_update_vnmi(cpu);
+        break;
     case ARM_CPU_IRQ:
     case ARM_CPU_FIQ:
+    case ARM_CPU_NMI:
         if (level) {
             cpu_interrupt(cs, mask[irq]);
         } else {
@@ -1355,7 +1431,7 @@ static void arm_cpu_initfn(Object *obj)
          */
         qdev_init_gpio_in(DEVICE(cpu), arm_cpu_kvm_set_irq, 4);
     } else {
-        qdev_init_gpio_in(DEVICE(cpu), arm_cpu_set_irq, 4);
+        qdev_init_gpio_in(DEVICE(cpu), arm_cpu_set_irq, 6);
     }
 
     qdev_init_gpio_out(DEVICE(cpu), cpu->gt_timer_outputs,
diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index abb453f733..6aa9f1e9ba 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -60,6 +60,8 @@
 #define EXCP_DIVBYZERO      23   /* v7M DIVBYZERO UsageFault */
 #define EXCP_VSERR          24
 #define EXCP_GPC            25   /* v9 Granule Protection Check Fault */
+#define EXCP_NMI            26
+#define EXCP_VNMI           27
 /* NB: add new EXCP_ defines to the array in arm_log_exception() too */
 
 #define ARMV7M_EXCP_RESET   1
@@ -79,6 +81,8 @@
 #define CPU_INTERRUPT_VIRQ  CPU_INTERRUPT_TGT_EXT_2
 #define CPU_INTERRUPT_VFIQ  CPU_INTERRUPT_TGT_EXT_3
 #define CPU_INTERRUPT_VSERR CPU_INTERRUPT_TGT_INT_0
+#define CPU_INTERRUPT_NMI   CPU_INTERRUPT_TGT_EXT_4
+#define CPU_INTERRUPT_VNMI  CPU_INTERRUPT_TGT_EXT_0
 
 /* The usual mapping for an AArch64 system register to its AArch32
  * counterpart is for the 32 bit world to have access to the lower
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 497b6e4bdf..4b4c8e279d 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -10628,6 +10628,8 @@ void arm_log_exception(CPUState *cs)
             [EXCP_DIVBYZERO] = "v7M DIVBYZERO UsageFault",
             [EXCP_VSERR] = "Virtual SERR",
             [EXCP_GPC] = "Granule Protection Check",
+            [EXCP_NMI] = "NMI",
+            [EXCP_VNMI] = "Virtual NMI"
         };
 
         if (idx >= 0 && idx < ARRAY_SIZE(excnames)) {
diff --git a/target/arm/internals.h b/target/arm/internals.h
index fee65caba5..4ff19b1b44 100644
--- a/target/arm/internals.h
+++ b/target/arm/internals.h
@@ -958,6 +958,16 @@ void arm_cpu_update_virq(ARMCPU *cpu);
  */
 void arm_cpu_update_vfiq(ARMCPU *cpu);
 
+/**
+ * arm_cpu_update_vnmi: Update CPU_INTERRUPT_VNMI bit in cs->interrupt_request
+ *
+ * Update the CPU_INTERRUPT_VNMI bit in cs->interrupt_request, following
+ * a change to either the input VNMI line from the GIC or the HCRX_EL2.VINMI
+ * or the HCRX_EL2.VFNMI.
+ * Must be called with the BQL held.
+ */
+void arm_cpu_update_vnmi(ARMCPU *cpu);
+
 /**
  * arm_cpu_update_vserr: Update CPU_INTERRUPT_VSERR bit
  *
-- 
2.34.1

Re: [RFC PATCH v5 06/22] target/arm: Add support for Non-maskable Interrupt

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> @@ -900,6 +945,31 @@ void arm_cpu_update_vfiq(ARMCPU *cpu)
>       }
>   }
>   
> +void arm_cpu_update_vnmi(ARMCPU *cpu)
> +{
> +    /*
> +     * Update the interrupt level for VNMI, which is the logical OR of
> +     * the HCRX_EL2.VINMI or HCRX_EL2.VFNMI bit and the input line level from
> +     * the GIC.
> +     */
> +    CPUARMState *env = &cpu->env;
> +    CPUState *cs = CPU(cpu);
> +
> +    bool new_state = ((env->cp15.hcr_el2 & HCR_VI) &&
> +                      (env->cp15.hcrx_el2 & HCRX_VINMI)) ||
> +                     ((env->cp15.hcr_el2 & HCR_VF) &&
> +                      (env->cp15.hcrx_el2 & HCRX_VFNMI)) ||

Need to use arm_hcr_el2_eff and arm_hcrx_el2_eff.  I see this is an existing error from 
the other functions too.

> +        (env->irq_line_state & CPU_INTERRUPT_VNMI);
> +
> +    if (new_state != ((cs->interrupt_request & CPU_INTERRUPT_VNMI) != 0)) {
> +        if (new_state) {
> +            cpu_interrupt(cs, CPU_INTERRUPT_VNMI);
> +        } else {
> +            cpu_reset_interrupt(cs, CPU_INTERRUPT_VNMI);
> +        }
> +    }
> +}

This is incomplete, as you need additional changes within the other two functions to not 
raise IRQ when VINMI is set, etc.


r~

Re: [RFC PATCH v5 06/22] target/arm: Add support for Non-maskable Interrupt

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> +    bool new_state = ((env->cp15.hcr_el2 & HCR_VI) &&
> +                      (env->cp15.hcrx_el2 & HCRX_VINMI)) ||
> +                     ((env->cp15.hcr_el2 & HCR_VF) &&
> +                      (env->cp15.hcrx_el2 & HCRX_VFNMI)) ||
> +        (env->irq_line_state & CPU_INTERRUPT_VNMI);

Because the GIC cannot signal an FIQ with superpriority, I think you should not include VF 
&& VFNMI in CPU_INTERRUPT_VNMI.

See comments for patch 8.


r~

Re: [RFC PATCH v5 06/22] target/arm: Add support for Non-maskable Interrupt

[Peter Maydell]

On Thu, 29 Feb 2024 at 23:02, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 2/29/24 03:10, Jinjie Ruan via wrote:
> > +    bool new_state = ((env->cp15.hcr_el2 & HCR_VI) &&
> > +                      (env->cp15.hcrx_el2 & HCRX_VINMI)) ||
> > +                     ((env->cp15.hcr_el2 & HCR_VF) &&
> > +                      (env->cp15.hcrx_el2 & HCRX_VFNMI)) ||
> > +        (env->irq_line_state & CPU_INTERRUPT_VNMI);
>
> Because the GIC cannot signal an FIQ with superpriority, I think you should not include VF
> && VFNMI in CPU_INTERRUPT_VNMI.

The GIC can't, but a hypervisor can -- it just sets the
VF and VFNMI bits if it wants one. (Architecturally, the CPU
has a FIQ-with-superpriority, it's only the GIC that doesn't.)

thanks
-- PMM

Re: [RFC PATCH v5 06/22] target/arm: Add support for Non-maskable Interrupt

[Richard Henderson]

On 3/19/24 07:03, Peter Maydell wrote:
> On Thu, 29 Feb 2024 at 23:02, Richard Henderson
> <richard.henderson@linaro.org> wrote:
>>
>> On 2/29/24 03:10, Jinjie Ruan via wrote:
>>> +    bool new_state = ((env->cp15.hcr_el2 & HCR_VI) &&
>>> +                      (env->cp15.hcrx_el2 & HCRX_VINMI)) ||
>>> +                     ((env->cp15.hcr_el2 & HCR_VF) &&
>>> +                      (env->cp15.hcrx_el2 & HCRX_VFNMI)) ||
>>> +        (env->irq_line_state & CPU_INTERRUPT_VNMI);
>>
>> Because the GIC cannot signal an FIQ with superpriority, I think you should not include VF
>> && VFNMI in CPU_INTERRUPT_VNMI.
> 
> The GIC can't, but a hypervisor can -- it just sets the
> VF and VFNMI bits if it wants one. (Architecturally, the CPU
> has a FIQ-with-superpriority, it's only the GIC that doesn't.)

Yes, I know.

The point was not to mix (irq from cpu or gic) with (fiq from cpu) so that we can 
correctly determine superpriority later.

Another way would have been to add an fiq-with-superpriority CPU_INTERRUPT bit, but since 
there's only one way to get that at present, I thought the extra bit was overkill.


r~

[RFC PATCH v5 07/22] target/arm: Add support for NMI in arm_phys_excp_target_el()

[Jinjie Ruan via]

According to Arm GIC section 4.6.3 Interrupt superpriority, the interrupt
with superpriority is always IRQ, never FIQ, so handle NMI same as IRQ in
arm_phys_excp_target_el().

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Add Reviewed-by.
v3:
- Remove nmi_is_irq flag in CPUARMState.
- Handle NMI same as IRQ in arm_phys_excp_target_el().
---
 target/arm/helper.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index 4b4c8e279d..7cdc90e9e3 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -10570,6 +10570,7 @@ uint32_t arm_phys_excp_target_el(CPUState *cs, uint32_t excp_idx,
     hcr_el2 = arm_hcr_el2_eff(env);
     switch (excp_idx) {
     case EXCP_IRQ:
+    case EXCP_NMI:
         scr = ((env->cp15.scr_el3 & SCR_IRQ) == SCR_IRQ);
         hcr = hcr_el2 & HCR_IMO;
         break;
-- 
2.34.1

[RFC PATCH v5 08/22] target/arm: Handle IS/FS in ISR_EL1 for NMI

[Jinjie Ruan via]

Add IS and FS bit in ISR_EL1 and handle the read. With CPU_INTERRUPT_NMI or
CPU_INTERRUPT_VNMI, both CPSR_I and ISR_IS must be set. With
CPU_INTERRUPT_VFIQ and HCRX_EL2.VFNMI set, both CPSR_F and ISR_FS must be set.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4；
- Also handle VNMI.
v3:
- CPU_INTERRUPT_NMI do not set FIQ, so remove it.
- With CPU_INTERRUPT_NMI, both CPSR_I and ISR_IS must be set.
---
 target/arm/cpu.h    |  2 ++
 target/arm/helper.c | 13 +++++++++++++
 2 files changed, 15 insertions(+)

diff --git a/target/arm/cpu.h b/target/arm/cpu.h
index 6aa9f1e9ba..4f9a8127f9 100644
--- a/target/arm/cpu.h
+++ b/target/arm/cpu.h
@@ -1474,6 +1474,8 @@ FIELD(CPTR_EL3, TCPAC, 31, 1)
 #define CPSR_N (1U << 31)
 #define CPSR_NZCV (CPSR_N | CPSR_Z | CPSR_C | CPSR_V)
 #define CPSR_AIF (CPSR_A | CPSR_I | CPSR_F)
+#define ISR_FS (1U << 9)
+#define ISR_IS (1U << 10)
 
 #define CPSR_IT (CPSR_IT_0_1 | CPSR_IT_2_7)
 #define CACHED_CPSR_BITS (CPSR_T | CPSR_AIF | CPSR_GE | CPSR_IT | CPSR_Q \
diff --git a/target/arm/helper.c b/target/arm/helper.c
index 7cdc90e9e3..ac44498537 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -2018,15 +2018,28 @@ static uint64_t isr_read(CPUARMState *env, const ARMCPRegInfo *ri)
         if (cs->interrupt_request & CPU_INTERRUPT_VIRQ) {
             ret |= CPSR_I;
         }
+        if (cs->interrupt_request & CPU_INTERRUPT_VNMI) {
+            ret |= ISR_IS;
+            ret |= CPSR_I;
+        }
     } else {
         if (cs->interrupt_request & CPU_INTERRUPT_HARD) {
             ret |= CPSR_I;
         }
+
+        if (cs->interrupt_request & CPU_INTERRUPT_NMI) {
+            ret |= ISR_IS;
+            ret |= CPSR_I;
+        }
     }
 
     if (hcr_el2 & HCR_FMO) {
         if (cs->interrupt_request & CPU_INTERRUPT_VFIQ) {
             ret |= CPSR_F;
+
+            if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
+                ret |= ISR_FS;
+            }
         }
     } else {
         if (cs->interrupt_request & CPU_INTERRUPT_FIQ) {
-- 
2.34.1

Re: [RFC PATCH v5 08/22] target/arm: Handle IS/FS in ISR_EL1 for NMI

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
>       if (hcr_el2 & HCR_FMO) {
>           if (cs->interrupt_request & CPU_INTERRUPT_VFIQ) {
>               ret |= CPSR_F;
> +
> +            if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
> +                ret |= ISR_FS;
> +            }

VFIO can be raised one of two ways: from the GIC and from HCR_EL2.VF.
But superpriority can only be added with HCRX_EL2.VFNMI.

You need to verify that HCR_EL2.VF is set before checking VFNMI.


r~

[RFC PATCH v5 09/22] target/arm: Handle PSTATE.ALLINT on taking an exception

[Jinjie Ruan via]

Set or clear PSTATE.ALLINT on taking an exception to ELx according to the
SCTLR_ELx.SPINTMASK bit.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v3:
- Add Reviewed-by.
---
 target/arm/helper.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index ac44498537..b796dbdf21 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -11539,6 +11539,15 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
         }
     }
 
+    if (cpu_isar_feature(aa64_nmi, cpu) &&
+        (env->cp15.sctlr_el[new_el] & SCTLR_NMI)) {
+        if (!(env->cp15.sctlr_el[new_el] & SCTLR_SPINTMASK)) {
+            new_mode |= PSTATE_ALLINT;
+        } else {
+            new_mode &= ~PSTATE_ALLINT;
+        }
+    }
+
     pstate_write(env, PSTATE_DAIF | new_mode);
     env->aarch64 = true;
     aarch64_restore_sp(env, new_el);
-- 
2.34.1

[RFC PATCH v5 10/22] hw/arm/virt: Wire NMI and VNMI irq lines from GIC to CPU

[Jinjie Ruan via]

Wire the new NMI and VNMI interrupt line from the GIC to each CPU.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Add Reviewed-by.
v3:
- Also add VNMI wire.
---
 hw/arm/virt.c | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 0af1943697..2d4a187fd5 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -804,7 +804,8 @@ static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
 
     /* Wire the outputs from each CPU's generic timer and the GICv3
      * maintenance interrupt signal to the appropriate GIC PPI inputs,
-     * and the GIC's IRQ/FIQ/VIRQ/VFIQ interrupt outputs to the CPU's inputs.
+     * and the GIC's IRQ/FIQ/VIRQ/VFIQ/NMI/VNMI interrupt outputs to the
+     * CPU's inputs.
      */
     for (i = 0; i < smp_cpus; i++) {
         DeviceState *cpudev = DEVICE(qemu_get_cpu(i));
@@ -848,6 +849,10 @@ static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
                            qdev_get_gpio_in(cpudev, ARM_CPU_VIRQ));
         sysbus_connect_irq(gicbusdev, i + 3 * smp_cpus,
                            qdev_get_gpio_in(cpudev, ARM_CPU_VFIQ));
+        sysbus_connect_irq(gicbusdev, i + 4 * smp_cpus,
+                           qdev_get_gpio_in(cpudev, ARM_CPU_NMI));
+        sysbus_connect_irq(gicbusdev, i + 5 * smp_cpus,
+                           qdev_get_gpio_in(cpudev, ARM_CPU_VNMI));
     }
 
     fdt_add_gic_node(vms);
-- 
2.34.1

[RFC PATCH v5 11/22] hw/intc/arm_gicv3: Add external IRQ lines for NMI

[Jinjie Ruan via]

Augment the GICv3's QOM device interface by adding one
new set of sysbus IRQ line, to signal NMI to each CPU.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Add Reviewed-by.
v3:
- Add support for VNMI.
---
 hw/intc/arm_gicv3_common.c         | 6 ++++++
 include/hw/intc/arm_gic_common.h   | 2 ++
 include/hw/intc/arm_gicv3_common.h | 2 ++
 3 files changed, 10 insertions(+)

diff --git a/hw/intc/arm_gicv3_common.c b/hw/intc/arm_gicv3_common.c
index cb55c72681..c52f060026 100644
--- a/hw/intc/arm_gicv3_common.c
+++ b/hw/intc/arm_gicv3_common.c
@@ -299,6 +299,12 @@ void gicv3_init_irqs_and_mmio(GICv3State *s, qemu_irq_handler handler,
     for (i = 0; i < s->num_cpu; i++) {
         sysbus_init_irq(sbd, &s->cpu[i].parent_vfiq);
     }
+    for (i = 0; i < s->num_cpu; i++) {
+        sysbus_init_irq(sbd, &s->cpu[i].parent_nmi);
+    }
+    for (i = 0; i < s->num_cpu; i++) {
+        sysbus_init_irq(sbd, &s->cpu[i].parent_vnmi);
+    }
 
     memory_region_init_io(&s->iomem_dist, OBJECT(s), ops, s,
                           "gicv3_dist", 0x10000);
diff --git a/include/hw/intc/arm_gic_common.h b/include/hw/intc/arm_gic_common.h
index 7080375008..97fea4102d 100644
--- a/include/hw/intc/arm_gic_common.h
+++ b/include/hw/intc/arm_gic_common.h
@@ -71,6 +71,8 @@ struct GICState {
     qemu_irq parent_fiq[GIC_NCPU];
     qemu_irq parent_virq[GIC_NCPU];
     qemu_irq parent_vfiq[GIC_NCPU];
+    qemu_irq parent_nmi[GIC_NCPU];
+    qemu_irq parent_vnmi[GIC_NCPU];
     qemu_irq maintenance_irq[GIC_NCPU];
 
     /* GICD_CTLR; for a GIC with the security extensions the NS banked version
diff --git a/include/hw/intc/arm_gicv3_common.h b/include/hw/intc/arm_gicv3_common.h
index 4e2fb518e7..7324c7d983 100644
--- a/include/hw/intc/arm_gicv3_common.h
+++ b/include/hw/intc/arm_gicv3_common.h
@@ -155,6 +155,8 @@ struct GICv3CPUState {
     qemu_irq parent_fiq;
     qemu_irq parent_virq;
     qemu_irq parent_vfiq;
+    qemu_irq parent_nmi;
+    qemu_irq parent_vnmi;
 
     /* Redistributor */
     uint32_t level;                  /* Current IRQ level */
-- 
2.34.1

[RFC PATCH v5 12/22] target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()

[Jinjie Ruan via]

According to Arm GIC section 4.6.3 Interrupt superpriority, the interrupt
with superpriority is always IRQ, never FIQ, so the NMI exception trap entry
behave like IRQ. However, VNMI can be IRQ or FIQ, FIQ can only come from
hcrx_el2.HCRX_VFNMI bit, IRQ can be raised from the GIC or come from the
hcrx_el2.HCRX_VINMI bit.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Also handle VNMI in arm_cpu_do_interrupt_aarch64().
v3:
- Remove the FIQ NMI handle.
---
 target/arm/helper.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/target/arm/helper.c b/target/arm/helper.c
index b796dbdf21..bd34b3506a 100644
--- a/target/arm/helper.c
+++ b/target/arm/helper.c
@@ -11459,12 +11459,21 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
         break;
     case EXCP_IRQ:
     case EXCP_VIRQ:
+    case EXCP_NMI:
         addr += 0x80;
         break;
     case EXCP_FIQ:
     case EXCP_VFIQ:
         addr += 0x100;
         break;
+    case EXCP_VNMI:
+        if (env->irq_line_state & CPU_INTERRUPT_VNMI ||
+            env->cp15.hcrx_el2 & HCRX_VINMI) {
+            addr += 0x80;
+        } else if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
+            addr += 0x100;
+        }
+        break;
     case EXCP_VSERR:
         addr += 0x180;
         /* Construct the SError syndrome from IDS and ISS fields. */
-- 
2.34.1

Re: [RFC PATCH v5 12/22] target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> According to Arm GIC section 4.6.3 Interrupt superpriority, the interrupt
> with superpriority is always IRQ, never FIQ, so the NMI exception trap entry
> behave like IRQ. However, VNMI can be IRQ or FIQ, FIQ can only come from
> hcrx_el2.HCRX_VFNMI bit, IRQ can be raised from the GIC or come from the
> hcrx_el2.HCRX_VINMI bit.
> 
> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
> ---
> v4:
> - Also handle VNMI in arm_cpu_do_interrupt_aarch64().
> v3:
> - Remove the FIQ NMI handle.
> ---
>   target/arm/helper.c | 9 +++++++++
>   1 file changed, 9 insertions(+)
> 
> diff --git a/target/arm/helper.c b/target/arm/helper.c
> index b796dbdf21..bd34b3506a 100644
> --- a/target/arm/helper.c
> +++ b/target/arm/helper.c
> @@ -11459,12 +11459,21 @@ static void arm_cpu_do_interrupt_aarch64(CPUState *cs)
>           break;
>       case EXCP_IRQ:
>       case EXCP_VIRQ:
> +    case EXCP_NMI:
>           addr += 0x80;
>           break;
>       case EXCP_FIQ:
>       case EXCP_VFIQ:
>           addr += 0x100;
>           break;
> +    case EXCP_VNMI:
> +        if (env->irq_line_state & CPU_INTERRUPT_VNMI ||
> +            env->cp15.hcrx_el2 & HCRX_VINMI) {
> +            addr += 0x80;
> +        } else if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
> +            addr += 0x100;
> +        }
> +        break;

By not combining VFNMI with CPU_INTERRUPT_VNMI, you don't need this complication.
Just

      case EXCP_IRQ:
      case EXCP_VIRQ:
+    case EXCP_NMI:


r~

Re: [RFC PATCH v5 12/22] target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()

[Jinjie Ruan via]

On 2024/3/1 7:09, Richard Henderson wrote:
> On 2/29/24 03:10, Jinjie Ruan via wrote:
>> According to Arm GIC section 4.6.3 Interrupt superpriority, the interrupt
>> with superpriority is always IRQ, never FIQ, so the NMI exception trap
>> entry
>> behave like IRQ. However, VNMI can be IRQ or FIQ, FIQ can only come from
>> hcrx_el2.HCRX_VFNMI bit, IRQ can be raised from the GIC or come from the
>> hcrx_el2.HCRX_VINMI bit.
>>
>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>> ---
>> v4:
>> - Also handle VNMI in arm_cpu_do_interrupt_aarch64().
>> v3:
>> - Remove the FIQ NMI handle.
>> ---
>>   target/arm/helper.c | 9 +++++++++
>>   1 file changed, 9 insertions(+)
>>
>> diff --git a/target/arm/helper.c b/target/arm/helper.c
>> index b796dbdf21..bd34b3506a 100644
>> --- a/target/arm/helper.c
>> +++ b/target/arm/helper.c
>> @@ -11459,12 +11459,21 @@ static void
>> arm_cpu_do_interrupt_aarch64(CPUState *cs)
>>           break;
>>       case EXCP_IRQ:
>>       case EXCP_VIRQ:
>> +    case EXCP_NMI:
>>           addr += 0x80;
>>           break;
>>       case EXCP_FIQ:
>>       case EXCP_VFIQ:
>>           addr += 0x100;
>>           break;
>> +    case EXCP_VNMI:
>> +        if (env->irq_line_state & CPU_INTERRUPT_VNMI ||
>> +            env->cp15.hcrx_el2 & HCRX_VINMI) {
>> +            addr += 0x80;
>> +        } else if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
>> +            addr += 0x100;
>> +        }
>> +        break;
> 
> By not combining VFNMI with CPU_INTERRUPT_VNMI, you don't need this
> complication.
> Just
> 
>      case EXCP_IRQ:
>      case EXCP_VIRQ:
> +    case EXCP_NMI:

Not understand it. both VIRQ and VFIQ will set CPU_INTERRUPT_VNMI and
cause EXCP_VNMI if they have Superpriority, the distinction jump here is
necessary.

> 
> 
> r~

Re: [RFC PATCH v5 12/22] target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()

[Richard Henderson]

On 2/29/24 17:42, Jinjie Ruan wrote:
> 
> 
> On 2024/3/1 7:09, Richard Henderson wrote:
>> On 2/29/24 03:10, Jinjie Ruan via wrote:
>>> According to Arm GIC section 4.6.3 Interrupt superpriority, the interrupt
>>> with superpriority is always IRQ, never FIQ, so the NMI exception trap
>>> entry
>>> behave like IRQ. However, VNMI can be IRQ or FIQ, FIQ can only come from
>>> hcrx_el2.HCRX_VFNMI bit, IRQ can be raised from the GIC or come from the
>>> hcrx_el2.HCRX_VINMI bit.
>>>
>>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>>> ---
>>> v4:
>>> - Also handle VNMI in arm_cpu_do_interrupt_aarch64().
>>> v3:
>>> - Remove the FIQ NMI handle.
>>> ---
>>>    target/arm/helper.c | 9 +++++++++
>>>    1 file changed, 9 insertions(+)
>>>
>>> diff --git a/target/arm/helper.c b/target/arm/helper.c
>>> index b796dbdf21..bd34b3506a 100644
>>> --- a/target/arm/helper.c
>>> +++ b/target/arm/helper.c
>>> @@ -11459,12 +11459,21 @@ static void
>>> arm_cpu_do_interrupt_aarch64(CPUState *cs)
>>>            break;
>>>        case EXCP_IRQ:
>>>        case EXCP_VIRQ:
>>> +    case EXCP_NMI:
>>>            addr += 0x80;
>>>            break;
>>>        case EXCP_FIQ:
>>>        case EXCP_VFIQ:
>>>            addr += 0x100;
>>>            break;
>>> +    case EXCP_VNMI:
>>> +        if (env->irq_line_state & CPU_INTERRUPT_VNMI ||
>>> +            env->cp15.hcrx_el2 & HCRX_VINMI) {
>>> +            addr += 0x80;
>>> +        } else if (env->cp15.hcrx_el2 & HCRX_VFNMI) {
>>> +            addr += 0x100;
>>> +        }
>>> +        break;
>>
>> By not combining VFNMI with CPU_INTERRUPT_VNMI, you don't need this
>> complication.
>> Just
>>
>>       case EXCP_IRQ:
>>       case EXCP_VIRQ:
>> +    case EXCP_NMI:
> 
> Not understand it. both VIRQ and VFIQ will set CPU_INTERRUPT_VNMI and
> cause EXCP_VNMI if they have Superpriority, the distinction jump here is
> necessary.

In my comment against patch 5, that's exactly what I said *not* to do.


r~

[RFC PATCH v5 13/22] hw/intc/arm_gicv3: Add irq superpriority information

[Jinjie Ruan via]

A SPI, PPI or SGI interrupt can have a superpriority property. So
maintain superpriority information in PendingIrq and GICR/GICD.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Acked-by: Richard Henderson <richard.henderson@linaro.org>
---
v3:
- Place this ahead of implement GICR_INMIR.
- Add Acked-by.
---
 include/hw/intc/arm_gicv3_common.h | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/include/hw/intc/arm_gicv3_common.h b/include/hw/intc/arm_gicv3_common.h
index 7324c7d983..df4380141d 100644
--- a/include/hw/intc/arm_gicv3_common.h
+++ b/include/hw/intc/arm_gicv3_common.h
@@ -146,6 +146,7 @@ typedef struct {
     int irq;
     uint8_t prio;
     int grp;
+    bool superprio;
 } PendingIrq;
 
 struct GICv3CPUState {
@@ -172,6 +173,7 @@ struct GICv3CPUState {
     uint32_t gicr_ienabler0;
     uint32_t gicr_ipendr0;
     uint32_t gicr_iactiver0;
+    uint32_t gicr_isuperprio;
     uint32_t edge_trigger; /* ICFGR0 and ICFGR1 even bits */
     uint32_t gicr_igrpmodr0;
     uint32_t gicr_nsacr;
@@ -274,6 +276,7 @@ struct GICv3State {
     GIC_DECLARE_BITMAP(active);       /* GICD_ISACTIVER */
     GIC_DECLARE_BITMAP(level);        /* Current level */
     GIC_DECLARE_BITMAP(edge_trigger); /* GICD_ICFGR even bits */
+    GIC_DECLARE_BITMAP(superprio);    /* GICD_INMIR */
     uint8_t gicd_ipriority[GICV3_MAXIRQ];
     uint64_t gicd_irouter[GICV3_MAXIRQ];
     /* Cached information: pointer to the cpu i/f for the CPUs specified
@@ -313,6 +316,7 @@ GICV3_BITMAP_ACCESSORS(pending)
 GICV3_BITMAP_ACCESSORS(active)
 GICV3_BITMAP_ACCESSORS(level)
 GICV3_BITMAP_ACCESSORS(edge_trigger)
+GICV3_BITMAP_ACCESSORS(superprio)
 
 #define TYPE_ARM_GICV3_COMMON "arm-gicv3-common"
 typedef struct ARMGICv3CommonClass ARMGICv3CommonClass;
-- 
2.34.1

[RFC PATCH v5 14/22] hw/intc/arm_gicv3_redist: Implement GICR_INMIR0

[Jinjie Ruan via]

Add GICR_INMIR0 register and support access GICR_INMIR0.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Make the GICR_INMIR0 implementation more clearer.
---
 hw/intc/arm_gicv3_redist.c | 19 +++++++++++++++++++
 hw/intc/gicv3_internal.h   |  1 +
 2 files changed, 20 insertions(+)

diff --git a/hw/intc/arm_gicv3_redist.c b/hw/intc/arm_gicv3_redist.c
index 8153525849..7a16a058b1 100644
--- a/hw/intc/arm_gicv3_redist.c
+++ b/hw/intc/arm_gicv3_redist.c
@@ -35,6 +35,15 @@ static int gicr_ns_access(GICv3CPUState *cs, int irq)
     return extract32(cs->gicr_nsacr, irq * 2, 2);
 }
 
+static void gicr_write_bitmap_reg(GICv3CPUState *cs, MemTxAttrs attrs,
+                                  uint32_t *reg, uint32_t val)
+{
+    /* Helper routine to implement writing to a "set" register */
+    val &= mask_group(cs, attrs);
+    *reg = val;
+    gicv3_redist_update(cs);
+}
+
 static void gicr_write_set_bitmap_reg(GICv3CPUState *cs, MemTxAttrs attrs,
                                       uint32_t *reg, uint32_t val)
 {
@@ -406,6 +415,10 @@ static MemTxResult gicr_readl(GICv3CPUState *cs, hwaddr offset,
         *data = value;
         return MEMTX_OK;
     }
+    case GICR_INMIR0:
+        *data = cs->gic->nmi_support ?
+                gicr_read_bitmap_reg(cs, attrs, cs->gicr_isuperprio) : 0;
+        return MEMTX_OK;
     case GICR_ICFGR0:
     case GICR_ICFGR1:
     {
@@ -555,6 +568,12 @@ static MemTxResult gicr_writel(GICv3CPUState *cs, hwaddr offset,
         gicv3_redist_update(cs);
         return MEMTX_OK;
     }
+    case GICR_INMIR0:
+        if (cs->gic->nmi_support) {
+            gicr_write_bitmap_reg(cs, attrs, &cs->gicr_isuperprio, value);
+        }
+        return MEMTX_OK;
+
     case GICR_ICFGR0:
         /* Register is all RAZ/WI or RAO/WI bits */
         return MEMTX_OK;
diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
index 29d5cdc1b6..f35b7d2f03 100644
--- a/hw/intc/gicv3_internal.h
+++ b/hw/intc/gicv3_internal.h
@@ -109,6 +109,7 @@
 #define GICR_ICFGR1           (GICR_SGI_OFFSET + 0x0C04)
 #define GICR_IGRPMODR0        (GICR_SGI_OFFSET + 0x0D00)
 #define GICR_NSACR            (GICR_SGI_OFFSET + 0x0E00)
+#define GICR_INMIR0           (GICR_SGI_OFFSET + 0x0F80)
 
 /* VLPI redistributor registers, offsets from VLPI_base */
 #define GICR_VPROPBASER       (GICR_VLPI_OFFSET + 0x70)
-- 
2.34.1

Re: [RFC PATCH v5 14/22] hw/intc/arm_gicv3_redist: Implement GICR_INMIR0

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> Add GICR_INMIR0 register and support access GICR_INMIR0.
> 
> Signed-off-by: Jinjie Ruan<ruanjinjie@huawei.com>
> ---
> v4:
> - Make the GICR_INMIR0 implementation more clearer.
> ---
>   hw/intc/arm_gicv3_redist.c | 19 +++++++++++++++++++
>   hw/intc/gicv3_internal.h   |  1 +
>   2 files changed, 20 insertions(+)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[RFC PATCH v5 15/22] hw/intc/arm_gicv3: Implement GICD_INMIR

[Jinjie Ruan via]

Add GICD_INMIR, GICD_INMIRnE register and support access GICD_INMIR0.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Make the GICD_INMIR implementation more clearer.
- Udpate the commit message.
v3:
- Add Reviewed-by.
---
 hw/intc/arm_gicv3_dist.c | 34 ++++++++++++++++++++++++++++++++++
 hw/intc/gicv3_internal.h |  2 ++
 2 files changed, 36 insertions(+)

diff --git a/hw/intc/arm_gicv3_dist.c b/hw/intc/arm_gicv3_dist.c
index 35e850685c..9739404e35 100644
--- a/hw/intc/arm_gicv3_dist.c
+++ b/hw/intc/arm_gicv3_dist.c
@@ -89,6 +89,29 @@ static int gicd_ns_access(GICv3State *s, int irq)
     return extract32(s->gicd_nsacr[irq / 16], (irq % 16) * 2, 2);
 }
 
+static void gicd_write_bitmap_reg(GICv3State *s, MemTxAttrs attrs,
+                                  uint32_t *bmp, maskfn *maskfn,
+                                  int offset, uint32_t val)
+{
+    /*
+     * Helper routine to implement writing to a "set" register
+     * (GICD_INMIR, etc).
+     * Semantics implemented here:
+     * RAZ/WI for SGIs, PPIs, unimplemented IRQs
+     * Bits corresponding to Group 0 or Secure Group 1 interrupts RAZ/WI.
+     * offset should be the offset in bytes of the register from the start
+     * of its group.
+     */
+    int irq = offset * 8;
+
+    if (irq < GIC_INTERNAL || irq >= s->num_irq) {
+        return;
+    }
+    val &= mask_group_and_nsacr(s, attrs, maskfn, irq);
+    *gic_bmp_ptr32(bmp, irq) = val;
+    gicv3_update(s, irq, 32);
+}
+
 static void gicd_write_set_bitmap_reg(GICv3State *s, MemTxAttrs attrs,
                                       uint32_t *bmp,
                                       maskfn *maskfn,
@@ -543,6 +566,11 @@ static bool gicd_readl(GICv3State *s, hwaddr offset,
         /* RAZ/WI since affinity routing is always enabled */
         *data = 0;
         return true;
+    case GICD_INMIR ... GICD_INMIR + 0x7f:
+        *data = (!s->nmi_support) ? 0 :
+                gicd_read_bitmap_reg(s, attrs, s->superprio, NULL,
+                                     offset - GICD_INMIR);
+        return true;
     case GICD_IROUTER ... GICD_IROUTER + 0x1fdf:
     {
         uint64_t r;
@@ -752,6 +780,12 @@ static bool gicd_writel(GICv3State *s, hwaddr offset,
     case GICD_SPENDSGIR ... GICD_SPENDSGIR + 0xf:
         /* RAZ/WI since affinity routing is always enabled */
         return true;
+    case GICD_INMIR ... GICD_INMIR + 0x7f:
+        if (s->nmi_support) {
+            gicd_write_bitmap_reg(s, attrs, s->superprio, NULL,
+                                  offset - GICD_INMIR, value);
+        }
+        return true;
     case GICD_IROUTER ... GICD_IROUTER + 0x1fdf:
     {
         uint64_t r;
diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
index f35b7d2f03..a1fc34597e 100644
--- a/hw/intc/gicv3_internal.h
+++ b/hw/intc/gicv3_internal.h
@@ -52,6 +52,8 @@
 #define GICD_SGIR            0x0F00
 #define GICD_CPENDSGIR       0x0F10
 #define GICD_SPENDSGIR       0x0F20
+#define GICD_INMIR           0x0F80
+#define GICD_INMIRnE         0x3B00
 #define GICD_IROUTER         0x6000
 #define GICD_IDREGS          0xFFD0
 
-- 
2.34.1

[RFC PATCH v5 16/22] hw/intc: Enable FEAT_GICv3_NMI Feature

[Jinjie Ruan via]

Added properties to enable FEAT_GICv3_NMI feature, setup distributor
and redistributor registers to indicate NMI support.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Add Reviewed-by.
---
 hw/intc/arm_gicv3_common.c         | 1 +
 hw/intc/arm_gicv3_dist.c           | 2 ++
 hw/intc/gicv3_internal.h           | 1 +
 include/hw/intc/arm_gicv3_common.h | 1 +
 4 files changed, 5 insertions(+)

diff --git a/hw/intc/arm_gicv3_common.c b/hw/intc/arm_gicv3_common.c
index c52f060026..2d2cea6858 100644
--- a/hw/intc/arm_gicv3_common.c
+++ b/hw/intc/arm_gicv3_common.c
@@ -569,6 +569,7 @@ static Property arm_gicv3_common_properties[] = {
     DEFINE_PROP_UINT32("num-irq", GICv3State, num_irq, 32),
     DEFINE_PROP_UINT32("revision", GICv3State, revision, 3),
     DEFINE_PROP_BOOL("has-lpi", GICv3State, lpi_enable, 0),
+    DEFINE_PROP_BOOL("has-nmi", GICv3State, nmi_support, 0),
     DEFINE_PROP_BOOL("has-security-extensions", GICv3State, security_extn, 0),
     /*
      * Compatibility property: force 8 bits of physical priority, even
diff --git a/hw/intc/arm_gicv3_dist.c b/hw/intc/arm_gicv3_dist.c
index 9739404e35..c4e28d209a 100644
--- a/hw/intc/arm_gicv3_dist.c
+++ b/hw/intc/arm_gicv3_dist.c
@@ -412,6 +412,7 @@ static bool gicd_readl(GICv3State *s, hwaddr offset,
          *                      by GICD_TYPER.IDbits)
          * MBIS == 0 (message-based SPIs not supported)
          * SecurityExtn == 1 if security extns supported
+         * NMI = 1 if Non-maskable interrupt property is supported
          * CPUNumber == 0 since for us ARE is always 1
          * ITLinesNumber == (((max SPI IntID + 1) / 32) - 1)
          */
@@ -425,6 +426,7 @@ static bool gicd_readl(GICv3State *s, hwaddr offset,
         bool dvis = s->revision >= 4;
 
         *data = (1 << 25) | (1 << 24) | (dvis << 18) | (sec_extn << 10) |
+            (s->nmi_support << GICD_TYPER_NMI_SHIFT) |
             (s->lpi_enable << GICD_TYPER_LPIS_SHIFT) |
             (0xf << 19) | itlinesnumber;
         return true;
diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
index a1fc34597e..8d793243f4 100644
--- a/hw/intc/gicv3_internal.h
+++ b/hw/intc/gicv3_internal.h
@@ -70,6 +70,7 @@
 #define GICD_CTLR_E1NWF             (1U << 7)
 #define GICD_CTLR_RWP               (1U << 31)
 
+#define GICD_TYPER_NMI_SHIFT           9
 #define GICD_TYPER_LPIS_SHIFT          17
 
 /* 16 bits EventId */
diff --git a/include/hw/intc/arm_gicv3_common.h b/include/hw/intc/arm_gicv3_common.h
index df4380141d..16c5fa7256 100644
--- a/include/hw/intc/arm_gicv3_common.h
+++ b/include/hw/intc/arm_gicv3_common.h
@@ -251,6 +251,7 @@ struct GICv3State {
     uint32_t num_irq;
     uint32_t revision;
     bool lpi_enable;
+    bool nmi_support;
     bool security_extn;
     bool force_8bit_prio;
     bool irq_reset_nonsecure;
-- 
2.34.1

[RFC PATCH v5 17/22] hw/intc/arm_gicv3: Add NMI handling CPU interface registers

[Jinjie Ruan via]

Add the NMIAR CPU interface registers which deal with acknowledging NMI.

When introduce NMI interrupt, there are some updates to the semantics for the
register ICC_IAR1_EL1 and ICC_HPPIR1_EL1. For ICC_IAR1_EL1 register, it
should return 1022 if the intid has super priority. And for ICC_NMIAR1_EL1
register, it should return 1023 if the intid do not have super priority.
Howerever, these are not necessary for ICC_HPPIR1_EL1 register.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Define ICC_NMIAR1_EL1 only if FEAT_GICv3_NMI is implemented.
- Check sctrl_elx.SCTLR_NMI to return 1022 for icc_iar1_read().
- Add gicv3_icc_nmiar1_read() trace event.
- Do not check icc_hppi_can_preempt() for icc_nmiar1_read().
- Add icv_nmiar1_read() and call it when EL2Enabled() and HCR_EL2.IMO == '1'
---
 hw/intc/arm_gicv3_cpuif.c | 59 +++++++++++++++++++++++++++++++++++++--
 hw/intc/gicv3_internal.h  |  1 +
 hw/intc/trace-events      |  1 +
 3 files changed, 58 insertions(+), 3 deletions(-)

diff --git a/hw/intc/arm_gicv3_cpuif.c b/hw/intc/arm_gicv3_cpuif.c
index e1a60d8c15..df82a413c6 100644
--- a/hw/intc/arm_gicv3_cpuif.c
+++ b/hw/intc/arm_gicv3_cpuif.c
@@ -795,6 +795,13 @@ static uint64_t icv_iar_read(CPUARMState *env, const ARMCPRegInfo *ri)
     return intid;
 }
 
+static uint64_t icv_nmiar1_read(CPUARMState *env, const ARMCPRegInfo *ri)
+{
+    /* todo */
+    uint64_t intid = INTID_SPURIOUS;
+    return intid;
+}
+
 static uint32_t icc_fullprio_mask(GICv3CPUState *cs)
 {
     /*
@@ -1097,7 +1104,8 @@ static uint64_t icc_hppir0_value(GICv3CPUState *cs, CPUARMState *env)
     return cs->hppi.irq;
 }
 
-static uint64_t icc_hppir1_value(GICv3CPUState *cs, CPUARMState *env)
+static uint64_t icc_hppir1_value(GICv3CPUState *cs, CPUARMState *env,
+                                 bool is_nmi, bool is_hppi)
 {
     /* Return the highest priority pending interrupt register value
      * for group 1.
@@ -1108,6 +1116,19 @@ static uint64_t icc_hppir1_value(GICv3CPUState *cs, CPUARMState *env)
         return INTID_SPURIOUS;
     }
 
+    if (!is_hppi) {
+        int el = arm_current_el(env);
+
+        if (is_nmi && (!cs->hppi.superprio)) {
+            return INTID_SPURIOUS;
+        }
+
+        if ((!is_nmi) && cs->hppi.superprio
+            && env->cp15.sctlr_el[el] & SCTLR_NMI) {
+            return INTID_NMI;
+        }
+    }
+
     /* Check whether we can return the interrupt or if we should return
      * a special identifier, as per the CheckGroup1ForSpecialIdentifiers
      * pseudocode. (We can simplify a little because for us ICC_SRE_EL1.RM
@@ -1168,7 +1189,7 @@ static uint64_t icc_iar1_read(CPUARMState *env, const ARMCPRegInfo *ri)
     if (!icc_hppi_can_preempt(cs)) {
         intid = INTID_SPURIOUS;
     } else {
-        intid = icc_hppir1_value(cs, env);
+        intid = icc_hppir1_value(cs, env, false, false);
     }
 
     if (!gicv3_intid_is_special(intid)) {
@@ -1179,6 +1200,25 @@ static uint64_t icc_iar1_read(CPUARMState *env, const ARMCPRegInfo *ri)
     return intid;
 }
 
+static uint64_t icc_nmiar1_read(CPUARMState *env, const ARMCPRegInfo *ri)
+{
+    GICv3CPUState *cs = icc_cs_from_env(env);
+    uint64_t intid;
+
+    if (icv_access(env, HCR_IMO)) {
+        return icv_nmiar1_read(env, ri);
+    }
+
+    intid = icc_hppir1_value(cs, env, true, false);
+
+    if (!gicv3_intid_is_special(intid)) {
+        icc_activate_irq(cs, intid);
+    }
+
+    trace_gicv3_icc_nmiar1_read(gicv3_redist_affid(cs), intid);
+    return intid;
+}
+
 static void icc_drop_prio(GICv3CPUState *cs, int grp)
 {
     /* Drop the priority of the currently active interrupt in
@@ -1555,7 +1595,7 @@ static uint64_t icc_hppir1_read(CPUARMState *env, const ARMCPRegInfo *ri)
         return icv_hppir_read(env, ri);
     }
 
-    value = icc_hppir1_value(cs, env);
+    value = icc_hppir1_value(cs, env, false, true);
     trace_gicv3_icc_hppir1_read(gicv3_redist_affid(cs), value);
     return value;
 }
@@ -2482,6 +2522,15 @@ static const ARMCPRegInfo gicv3_cpuif_icc_apxr23_reginfo[] = {
     },
 };
 
+static const ARMCPRegInfo gicv3_cpuif_gicv3_nmi_reginfo[] = {
+    { .name = "ICC_NMIAR1_EL1", .state = ARM_CP_STATE_BOTH,
+      .opc0 = 3, .opc1 = 0, .crn = 12, .crm = 9, .opc2 = 5,
+      .type = ARM_CP_IO | ARM_CP_NO_RAW,
+      .access = PL1_R, .accessfn = gicv3_irq_access,
+      .readfn = icc_nmiar1_read,
+    },
+};
+
 static uint64_t ich_ap_read(CPUARMState *env, const ARMCPRegInfo *ri)
 {
     GICv3CPUState *cs = icc_cs_from_env(env);
@@ -2838,6 +2887,10 @@ void gicv3_init_cpuif(GICv3State *s)
          */
         define_arm_cp_regs(cpu, gicv3_cpuif_reginfo);
 
+        if (s->nmi_support) {
+            define_arm_cp_regs(cpu, gicv3_cpuif_gicv3_nmi_reginfo);
+        }
+
         /*
          * The CPU implementation specifies the number of supported
          * bits of physical priority. For backwards compatibility
diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
index 8d793243f4..93e56b3726 100644
--- a/hw/intc/gicv3_internal.h
+++ b/hw/intc/gicv3_internal.h
@@ -511,6 +511,7 @@ FIELD(VTE, RDBASE, 42, RDBASE_PROCNUM_LENGTH)
 /* Special interrupt IDs */
 #define INTID_SECURE 1020
 #define INTID_NONSECURE 1021
+#define INTID_NMI 1022
 #define INTID_SPURIOUS 1023
 
 /* Functions internal to the emulated GICv3 */
diff --git a/hw/intc/trace-events b/hw/intc/trace-events
index 1ef29d0256..94030550d5 100644
--- a/hw/intc/trace-events
+++ b/hw/intc/trace-events
@@ -116,6 +116,7 @@ gicv3_cpuif_set_irqs(uint32_t cpuid, int fiqlevel, int irqlevel) "GICv3 CPU i/f
 gicv3_icc_generate_sgi(uint32_t cpuid, int irq, int irm, uint32_t aff, uint32_t targetlist) "GICv3 CPU i/f 0x%x generating SGI %d IRM %d target affinity 0x%xxx targetlist 0x%x"
 gicv3_icc_iar0_read(uint32_t cpu, uint64_t val) "GICv3 ICC_IAR0 read cpu 0x%x value 0x%" PRIx64
 gicv3_icc_iar1_read(uint32_t cpu, uint64_t val) "GICv3 ICC_IAR1 read cpu 0x%x value 0x%" PRIx64
+gicv3_icc_nmiar1_read(uint32_t cpu, uint64_t val) "GICv3 ICC_NMIAR1 read cpu 0x%x value 0x%" PRIx64
 gicv3_icc_eoir_write(int grp, uint32_t cpu, uint64_t val) "GICv3 ICC_EOIR%d write cpu 0x%x value 0x%" PRIx64
 gicv3_icc_hppir0_read(uint32_t cpu, uint64_t val) "GICv3 ICC_HPPIR0 read cpu 0x%x value 0x%" PRIx64
 gicv3_icc_hppir1_read(uint32_t cpu, uint64_t val) "GICv3 ICC_HPPIR1 read cpu 0x%x value 0x%" PRIx64
-- 
2.34.1

Re: [RFC PATCH v5 17/22] hw/intc/arm_gicv3: Add NMI handling CPU interface registers

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> +static uint64_t icv_nmiar1_read(CPUARMState *env, const ARMCPRegInfo *ri)
> +{
> +    /* todo */
> +    uint64_t intid = INTID_SPURIOUS;
> +    return intid;
> +}

You're going to need to fill this in.

It's like icv_iar_read() but only affects irqs with superpriority.


r~

[RFC PATCH v5 18/22] hw/intc/arm_gicv3: Implement NMI interrupt prioirty

[Jinjie Ruan via]

If GICD_CTLR_DS bit is zero and the NMI is non-secure, the NMI prioirty
is higher than 0x80, otherwise it is higher than 0x0. And save NMI
super prioirty information in hppi.superprio to deliver NMI exception.
Since both GICR and GICD can deliver NMI, it is both necessary to check
whether the pending irq is NMI in gicv3_redist_update_noirqset and
gicv3_update_noirqset. And In irqbetter(), only a non-NMI with the same
priority and a smaller interrupt number can be preempted but not NMI.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Replace is_nmi with has_superprio to not a mix NMI and superpriority.
- Update the comment in irqbetter().
- Extract gicv3_get_priority() to avoid code repeat.
---
v3:
- Add missing brace
---
 hw/intc/arm_gicv3.c | 71 ++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 63 insertions(+), 8 deletions(-)

diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
index 0b8f79a122..1d16a53b23 100644
--- a/hw/intc/arm_gicv3.c
+++ b/hw/intc/arm_gicv3.c
@@ -21,7 +21,8 @@
 #include "hw/intc/arm_gicv3.h"
 #include "gicv3_internal.h"
 
-static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
+static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio,
+                      bool has_superprio)
 {
     /* Return true if this IRQ at this priority should take
      * precedence over the current recorded highest priority
@@ -33,11 +34,24 @@ static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
     if (prio < cs->hppi.prio) {
         return true;
     }
+
+    /*
+     * Current highest prioirity pending interrupt is an IRQ without
+     * superpriority, the new IRQ with superpriority has same priority
+     * should signal to the CPU as it have the priority higher than
+     * the labelled 0x80 or 0x00.
+     */
+    if (prio == cs->hppi.prio && !cs->hppi.superprio && has_superprio) {
+        return true;
+    }
+
     /* If multiple pending interrupts have the same priority then it is an
      * IMPDEF choice which of them to signal to the CPU. We choose to
-     * signal the one with the lowest interrupt number.
+     * signal the one with the lowest interrupt number if they don't have
+     * superpriority.
      */
-    if (prio == cs->hppi.prio && irq <= cs->hppi.irq) {
+    if (prio == cs->hppi.prio && !cs->hppi.superprio &&
+        !has_superprio && irq <= cs->hppi.irq) {
         return true;
     }
     return false;
@@ -129,6 +143,35 @@ static uint32_t gicr_int_pending(GICv3CPUState *cs)
     return pend;
 }
 
+static bool gicv3_get_priority(GICv3CPUState *cs, bool is_redist,
+                               uint32_t superprio, uint8_t *prio, int irq)
+{
+    bool has_superprio = false;
+
+    if (superprio) {
+        has_superprio = true;
+
+        /* DS = 0 & Non-secure NMI */
+        if (!(cs->gic->gicd_ctlr & GICD_CTLR_DS) &&
+            ((is_redist && extract32(cs->gicr_igroupr0, irq, 1)) ||
+             (!is_redist && gicv3_gicd_group_test(cs->gic, irq)))) {
+            *prio = 0x80;
+        } else {
+            *prio = 0x0;
+        }
+    } else {
+        has_superprio = false;
+
+        if (is_redist) {
+            *prio = cs->gicr_ipriorityr[irq];
+        } else {
+            *prio = cs->gic->gicd_ipriority[irq];
+        }
+    }
+
+    return has_superprio;
+}
+
 /* Update the interrupt status after state in a redistributor
  * or CPU interface has changed, but don't tell the CPU i/f.
  */
@@ -141,6 +184,8 @@ static void gicv3_redist_update_noirqset(GICv3CPUState *cs)
     uint8_t prio;
     int i;
     uint32_t pend;
+    uint32_t superprio = 0;
+    bool has_superprio = false;
 
     /* Find out which redistributor interrupts are eligible to be
      * signaled to the CPU interface.
@@ -152,10 +197,13 @@ static void gicv3_redist_update_noirqset(GICv3CPUState *cs)
             if (!(pend & (1 << i))) {
                 continue;
             }
-            prio = cs->gicr_ipriorityr[i];
-            if (irqbetter(cs, i, prio)) {
+            superprio = extract32(cs->gicr_isuperprio, i, 1);
+            has_superprio = gicv3_get_priority(cs, true, superprio, &prio, i);
+
+            if (irqbetter(cs, i, prio, has_superprio)) {
                 cs->hppi.irq = i;
                 cs->hppi.prio = prio;
+                cs->hppi.superprio = has_superprio;
                 seenbetter = true;
             }
         }
@@ -168,7 +216,7 @@ static void gicv3_redist_update_noirqset(GICv3CPUState *cs)
     if ((cs->gicr_ctlr & GICR_CTLR_ENABLE_LPIS) && cs->gic->lpi_enable &&
         (cs->gic->gicd_ctlr & GICD_CTLR_EN_GRP1NS) &&
         (cs->hpplpi.prio != 0xff)) {
-        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio)) {
+        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio, false)) {
             cs->hppi.irq = cs->hpplpi.irq;
             cs->hppi.prio = cs->hpplpi.prio;
             cs->hppi.grp = cs->hpplpi.grp;
@@ -213,6 +261,8 @@ static void gicv3_update_noirqset(GICv3State *s, int start, int len)
     int i;
     uint8_t prio;
     uint32_t pend = 0;
+    uint32_t superprio = 0;
+    bool has_superprio = false;
 
     assert(start >= GIC_INTERNAL);
     assert(len > 0);
@@ -240,10 +290,15 @@ static void gicv3_update_noirqset(GICv3State *s, int start, int len)
              */
             continue;
         }
-        prio = s->gicd_ipriority[i];
-        if (irqbetter(cs, i, prio)) {
+
+        superprio = *gic_bmp_ptr32(s->superprio, i);
+        superprio = superprio & (1 << (i & 0x1f));
+        has_superprio = gicv3_get_priority(cs, false, superprio, &prio, i);
+
+        if (irqbetter(cs, i, prio, has_superprio)) {
             cs->hppi.irq = i;
             cs->hppi.prio = prio;
+            cs->hppi.superprio = has_superprio;
             cs->seenbetter = true;
         }
     }
-- 
2.34.1

Re: [RFC PATCH v5 18/22] hw/intc/arm_gicv3: Implement NMI interrupt prioirty

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> If GICD_CTLR_DS bit is zero and the NMI is non-secure, the NMI prioirty
> is higher than 0x80, otherwise it is higher than 0x0. And save NMI
> super prioirty information in hppi.superprio to deliver NMI exception.
> Since both GICR and GICD can deliver NMI, it is both necessary to check
> whether the pending irq is NMI in gicv3_redist_update_noirqset and
> gicv3_update_noirqset. And In irqbetter(), only a non-NMI with the same
> priority and a smaller interrupt number can be preempted but not NMI.
> 
> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
> ---
> v4:
> - Replace is_nmi with has_superprio to not a mix NMI and superpriority.
> - Update the comment in irqbetter().
> - Extract gicv3_get_priority() to avoid code repeat.
> ---
> v3:
> - Add missing brace
> ---
>   hw/intc/arm_gicv3.c | 71 ++++++++++++++++++++++++++++++++++++++++-----
>   1 file changed, 63 insertions(+), 8 deletions(-)
> 
> diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
> index 0b8f79a122..1d16a53b23 100644
> --- a/hw/intc/arm_gicv3.c
> +++ b/hw/intc/arm_gicv3.c
> @@ -21,7 +21,8 @@
>   #include "hw/intc/arm_gicv3.h"
>   #include "gicv3_internal.h"
>   
> -static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
> +static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio,
> +                      bool has_superprio)
>   {
>       /* Return true if this IRQ at this priority should take
>        * precedence over the current recorded highest priority
> @@ -33,11 +34,24 @@ static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
>       if (prio < cs->hppi.prio) {
>           return true;
>       }
> +
> +    /*
> +     * Current highest prioirity pending interrupt is an IRQ without
> +     * superpriority, the new IRQ with superpriority has same priority
> +     * should signal to the CPU as it have the priority higher than
> +     * the labelled 0x80 or 0x00.
> +     */
> +    if (prio == cs->hppi.prio && !cs->hppi.superprio && has_superprio) {
> +        return true;
> +    }
> +
>       /* If multiple pending interrupts have the same priority then it is an
>        * IMPDEF choice which of them to signal to the CPU. We choose to
> -     * signal the one with the lowest interrupt number.
> +     * signal the one with the lowest interrupt number if they don't have
> +     * superpriority.
>        */
> -    if (prio == cs->hppi.prio && irq <= cs->hppi.irq) {
> +    if (prio == cs->hppi.prio && !cs->hppi.superprio &&
> +        !has_superprio && irq <= cs->hppi.irq) {
>           return true;
>       }
>       return false;
> @@ -129,6 +143,35 @@ static uint32_t gicr_int_pending(GICv3CPUState *cs)
>       return pend;
>   }
>   
> +static bool gicv3_get_priority(GICv3CPUState *cs, bool is_redist,
> +                               uint32_t superprio, uint8_t *prio, int irq)
> +{
> +    bool has_superprio = false;
> +
> +    if (superprio) {
> +        has_superprio = true;
> +
> +        /* DS = 0 & Non-secure NMI */
> +        if (!(cs->gic->gicd_ctlr & GICD_CTLR_DS) &&
> +            ((is_redist && extract32(cs->gicr_igroupr0, irq, 1)) ||
> +             (!is_redist && gicv3_gicd_group_test(cs->gic, irq)))) {
> +            *prio = 0x80;
> +        } else {
> +            *prio = 0x0;
> +        }
> +    } else {
> +        has_superprio = false;
> +
> +        if (is_redist) {
> +            *prio = cs->gicr_ipriorityr[irq];
> +        } else {
> +            *prio = cs->gic->gicd_ipriority[irq];
> +        }
> +    }
> +
> +    return has_superprio;
> +}

Did you not like the idea to map {priority, !superpriority} into a single value?

It would eliminate the change in irqbetter(), which is a bit more complex than it needs to be.

> @@ -152,10 +197,13 @@ static void gicv3_redist_update_noirqset(GICv3CPUState *cs)
>               if (!(pend & (1 << i))) {
>                   continue;
>               }
> -            prio = cs->gicr_ipriorityr[i];
> -            if (irqbetter(cs, i, prio)) {
> +            superprio = extract32(cs->gicr_isuperprio, i, 1);
> +            has_superprio = gicv3_get_priority(cs, true, superprio, &prio, i);

It would allow moving the read of gicr_isuperprio into gicv3_get_priority(), alongside the 
read of gicr_ipriorityr.

Is there a bug here not handling is_redist for GCIR_INMI*?

> @@ -168,7 +216,7 @@ static void gicv3_redist_update_noirqset(GICv3CPUState *cs)
>       if ((cs->gicr_ctlr & GICR_CTLR_ENABLE_LPIS) && cs->gic->lpi_enable &&
>           (cs->gic->gicd_ctlr & GICD_CTLR_EN_GRP1NS) &&
>           (cs->hpplpi.prio != 0xff)) {
> -        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio)) {
> +        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio, false)) {

Always passing false here is incorrect -- again missing the redistributor nmi bit?


r~

Re: [RFC PATCH v5 18/22] hw/intc/arm_gicv3: Implement NMI interrupt prioirty

[Jinjie Ruan via]

On 2024/3/1 7:50, Richard Henderson wrote:
> On 2/29/24 03:10, Jinjie Ruan via wrote:
>> If GICD_CTLR_DS bit is zero and the NMI is non-secure, the NMI prioirty
>> is higher than 0x80, otherwise it is higher than 0x0. And save NMI
>> super prioirty information in hppi.superprio to deliver NMI exception.
>> Since both GICR and GICD can deliver NMI, it is both necessary to check
>> whether the pending irq is NMI in gicv3_redist_update_noirqset and
>> gicv3_update_noirqset. And In irqbetter(), only a non-NMI with the same
>> priority and a smaller interrupt number can be preempted but not NMI.
>>
>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>> ---
>> v4:
>> - Replace is_nmi with has_superprio to not a mix NMI and superpriority.
>> - Update the comment in irqbetter().
>> - Extract gicv3_get_priority() to avoid code repeat.
>> ---
>> v3:
>> - Add missing brace
>> ---
>>   hw/intc/arm_gicv3.c | 71 ++++++++++++++++++++++++++++++++++++++++-----
>>   1 file changed, 63 insertions(+), 8 deletions(-)
>>
>> diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
>> index 0b8f79a122..1d16a53b23 100644
>> --- a/hw/intc/arm_gicv3.c
>> +++ b/hw/intc/arm_gicv3.c
>> @@ -21,7 +21,8 @@
>>   #include "hw/intc/arm_gicv3.h"
>>   #include "gicv3_internal.h"
>>   -static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
>> +static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio,
>> +                      bool has_superprio)
>>   {
>>       /* Return true if this IRQ at this priority should take
>>        * precedence over the current recorded highest priority
>> @@ -33,11 +34,24 @@ static bool irqbetter(GICv3CPUState *cs, int irq,
>> uint8_t prio)
>>       if (prio < cs->hppi.prio) {
>>           return true;
>>       }
>> +
>> +    /*
>> +     * Current highest prioirity pending interrupt is an IRQ without
>> +     * superpriority, the new IRQ with superpriority has same priority
>> +     * should signal to the CPU as it have the priority higher than
>> +     * the labelled 0x80 or 0x00.
>> +     */
>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio && has_superprio) {
>> +        return true;
>> +    }
>> +
>>       /* If multiple pending interrupts have the same priority then it
>> is an
>>        * IMPDEF choice which of them to signal to the CPU. We choose to
>> -     * signal the one with the lowest interrupt number.
>> +     * signal the one with the lowest interrupt number if they don't
>> have
>> +     * superpriority.
>>        */
>> -    if (prio == cs->hppi.prio && irq <= cs->hppi.irq) {
>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio &&
>> +        !has_superprio && irq <= cs->hppi.irq) {
>>           return true;
>>       }
>>       return false;
>> @@ -129,6 +143,35 @@ static uint32_t gicr_int_pending(GICv3CPUState *cs)
>>       return pend;
>>   }
>>   +static bool gicv3_get_priority(GICv3CPUState *cs, bool is_redist,
>> +                               uint32_t superprio, uint8_t *prio, int
>> irq)
>> +{
>> +    bool has_superprio = false;
>> +
>> +    if (superprio) {
>> +        has_superprio = true;
>> +
>> +        /* DS = 0 & Non-secure NMI */
>> +        if (!(cs->gic->gicd_ctlr & GICD_CTLR_DS) &&
>> +            ((is_redist && extract32(cs->gicr_igroupr0, irq, 1)) ||
>> +             (!is_redist && gicv3_gicd_group_test(cs->gic, irq)))) {
>> +            *prio = 0x80;
>> +        } else {
>> +            *prio = 0x0;
>> +        }
>> +    } else {
>> +        has_superprio = false;
>> +
>> +        if (is_redist) {
>> +            *prio = cs->gicr_ipriorityr[irq];
>> +        } else {
>> +            *prio = cs->gic->gicd_ipriority[irq];
>> +        }
>> +    }
>> +
>> +    return has_superprio;
>> +}
> 
> Did you not like the idea to map {priority, !superpriority} into a
> single value?
> 
> It would eliminate the change in irqbetter(), which is a bit more
> complex than it needs to be.

I will try to change to implement this mapping scheme.

> 
>> @@ -152,10 +197,13 @@ static void
>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>               if (!(pend & (1 << i))) {
>>                   continue;
>>               }
>> -            prio = cs->gicr_ipriorityr[i];
>> -            if (irqbetter(cs, i, prio)) {
>> +            superprio = extract32(cs->gicr_isuperprio, i, 1);
>> +            has_superprio = gicv3_get_priority(cs, true, superprio,
>> &prio, i);
> 
> It would allow moving the read of gicr_isuperprio into
> gicv3_get_priority(), alongside the read of gicr_ipriorityr.
> 
> Is there a bug here not handling is_redist for GCIR_INMI*?
> 
>> @@ -168,7 +216,7 @@ static void
>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>       if ((cs->gicr_ctlr & GICR_CTLR_ENABLE_LPIS) &&
>> cs->gic->lpi_enable &&
>>           (cs->gic->gicd_ctlr & GICD_CTLR_EN_GRP1NS) &&
>>           (cs->hpplpi.prio != 0xff)) {
>> -        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio)) {
>> +        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio, false)) {
> 
> Always passing false here is incorrect -- again missing the
> redistributor nmi bit?
> 
> 
> r~

Re: [RFC PATCH v5 18/22] hw/intc/arm_gicv3: Implement NMI interrupt prioirty

[Jinjie Ruan via]

On 2024/3/4 20:18, Jinjie Ruan wrote:
> 
> 
> On 2024/3/1 7:50, Richard Henderson wrote:
>> On 2/29/24 03:10, Jinjie Ruan via wrote:
>>> If GICD_CTLR_DS bit is zero and the NMI is non-secure, the NMI prioirty
>>> is higher than 0x80, otherwise it is higher than 0x0. And save NMI
>>> super prioirty information in hppi.superprio to deliver NMI exception.
>>> Since both GICR and GICD can deliver NMI, it is both necessary to check
>>> whether the pending irq is NMI in gicv3_redist_update_noirqset and
>>> gicv3_update_noirqset. And In irqbetter(), only a non-NMI with the same
>>> priority and a smaller interrupt number can be preempted but not NMI.
>>>
>>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>>> ---
>>> v4:
>>> - Replace is_nmi with has_superprio to not a mix NMI and superpriority.
>>> - Update the comment in irqbetter().
>>> - Extract gicv3_get_priority() to avoid code repeat.
>>> ---
>>> v3:
>>> - Add missing brace
>>> ---
>>>   hw/intc/arm_gicv3.c | 71 ++++++++++++++++++++++++++++++++++++++++-----
>>>   1 file changed, 63 insertions(+), 8 deletions(-)
>>>
>>> diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
>>> index 0b8f79a122..1d16a53b23 100644
>>> --- a/hw/intc/arm_gicv3.c
>>> +++ b/hw/intc/arm_gicv3.c
>>> @@ -21,7 +21,8 @@
>>>   #include "hw/intc/arm_gicv3.h"
>>>   #include "gicv3_internal.h"
>>>   -static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
>>> +static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio,
>>> +                      bool has_superprio)
>>>   {
>>>       /* Return true if this IRQ at this priority should take
>>>        * precedence over the current recorded highest priority
>>> @@ -33,11 +34,24 @@ static bool irqbetter(GICv3CPUState *cs, int irq,
>>> uint8_t prio)
>>>       if (prio < cs->hppi.prio) {
>>>           return true;
>>>       }
>>> +
>>> +    /*
>>> +     * Current highest prioirity pending interrupt is an IRQ without
>>> +     * superpriority, the new IRQ with superpriority has same priority
>>> +     * should signal to the CPU as it have the priority higher than
>>> +     * the labelled 0x80 or 0x00.
>>> +     */
>>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio && has_superprio) {
>>> +        return true;
>>> +    }
>>> +
>>>       /* If multiple pending interrupts have the same priority then it
>>> is an
>>>        * IMPDEF choice which of them to signal to the CPU. We choose to
>>> -     * signal the one with the lowest interrupt number.
>>> +     * signal the one with the lowest interrupt number if they don't
>>> have
>>> +     * superpriority.
>>>        */
>>> -    if (prio == cs->hppi.prio && irq <= cs->hppi.irq) {
>>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio &&
>>> +        !has_superprio && irq <= cs->hppi.irq) {
>>>           return true;
>>>       }
>>>       return false;
>>> @@ -129,6 +143,35 @@ static uint32_t gicr_int_pending(GICv3CPUState *cs)
>>>       return pend;
>>>   }
>>>   +static bool gicv3_get_priority(GICv3CPUState *cs, bool is_redist,
>>> +                               uint32_t superprio, uint8_t *prio, int
>>> irq)
>>> +{
>>> +    bool has_superprio = false;
>>> +
>>> +    if (superprio) {
>>> +        has_superprio = true;
>>> +
>>> +        /* DS = 0 & Non-secure NMI */
>>> +        if (!(cs->gic->gicd_ctlr & GICD_CTLR_DS) &&
>>> +            ((is_redist && extract32(cs->gicr_igroupr0, irq, 1)) ||
>>> +             (!is_redist && gicv3_gicd_group_test(cs->gic, irq)))) {
>>> +            *prio = 0x80;
>>> +        } else {
>>> +            *prio = 0x0;
>>> +        }
>>> +    } else {
>>> +        has_superprio = false;
>>> +
>>> +        if (is_redist) {
>>> +            *prio = cs->gicr_ipriorityr[irq];
>>> +        } else {
>>> +            *prio = cs->gic->gicd_ipriority[irq];
>>> +        }
>>> +    }
>>> +
>>> +    return has_superprio;
>>> +}
>>
>> Did you not like the idea to map {priority, !superpriority} into a
>> single value?
>>
>> It would eliminate the change in irqbetter(), which is a bit more
>> complex than it needs to be.
> 
> I will try to change to implement this mapping scheme.

I try to implement this mapping scheme, but it seems more complex, more
change on a larger scale, although it would eliminate the change in
irqbetter(). The subpriority and group priority, eight-bit segmentation
is used, the same is true of the semantics of BPR. And currently, the
invalid values or initial value of all priorities in the QEMU are 0xff,
and all priority mask operations are 0xff.

Especially, what should hppi.prio save? the original priority value or
the mapped priority value. If hppi.prio save the mapped priority value,
all of the above things need to be updated. Otherwise,if hppi.prio save
the original priority value，it will be more simpler, but it also need
to convert from mapped priority to original priority when save
cs->hppi.prio and convert to mapped priority when comparing priority in
irqbetter(), it is also not more compact than existing implementation.

@Peter

> 
>>
>>> @@ -152,10 +197,13 @@ static void
>>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>>               if (!(pend & (1 << i))) {
>>>                   continue;
>>>               }
>>> -            prio = cs->gicr_ipriorityr[i];
>>> -            if (irqbetter(cs, i, prio)) {
>>> +            superprio = extract32(cs->gicr_isuperprio, i, 1);
>>> +            has_superprio = gicv3_get_priority(cs, true, superprio,
>>> &prio, i);
>>
>> It would allow moving the read of gicr_isuperprio into
>> gicv3_get_priority(), alongside the read of gicr_ipriorityr.
>>
>> Is there a bug here not handling is_redist for GCIR_INMI*?
>>
>>> @@ -168,7 +216,7 @@ static void
>>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>>       if ((cs->gicr_ctlr & GICR_CTLR_ENABLE_LPIS) &&
>>> cs->gic->lpi_enable &&
>>>           (cs->gic->gicd_ctlr & GICD_CTLR_EN_GRP1NS) &&
>>>           (cs->hpplpi.prio != 0xff)) {
>>> -        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio)) {
>>> +        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio, false)) {
>>
>> Always passing false here is incorrect -- again missing the
>> redistributor nmi bit?
>>
>>
>> r~

Re: [RFC PATCH v5 18/22] hw/intc/arm_gicv3: Implement NMI interrupt prioirty

[Jinjie Ruan via]

On 2024/3/1 7:50, Richard Henderson wrote:
> On 2/29/24 03:10, Jinjie Ruan via wrote:
>> If GICD_CTLR_DS bit is zero and the NMI is non-secure, the NMI prioirty
>> is higher than 0x80, otherwise it is higher than 0x0. And save NMI
>> super prioirty information in hppi.superprio to deliver NMI exception.
>> Since both GICR and GICD can deliver NMI, it is both necessary to check
>> whether the pending irq is NMI in gicv3_redist_update_noirqset and
>> gicv3_update_noirqset. And In irqbetter(), only a non-NMI with the same
>> priority and a smaller interrupt number can be preempted but not NMI.
>>
>> Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
>> ---
>> v4:
>> - Replace is_nmi with has_superprio to not a mix NMI and superpriority.
>> - Update the comment in irqbetter().
>> - Extract gicv3_get_priority() to avoid code repeat.
>> ---
>> v3:
>> - Add missing brace
>> ---
>>   hw/intc/arm_gicv3.c | 71 ++++++++++++++++++++++++++++++++++++++++-----
>>   1 file changed, 63 insertions(+), 8 deletions(-)
>>
>> diff --git a/hw/intc/arm_gicv3.c b/hw/intc/arm_gicv3.c
>> index 0b8f79a122..1d16a53b23 100644
>> --- a/hw/intc/arm_gicv3.c
>> +++ b/hw/intc/arm_gicv3.c
>> @@ -21,7 +21,8 @@
>>   #include "hw/intc/arm_gicv3.h"
>>   #include "gicv3_internal.h"
>>   -static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio)
>> +static bool irqbetter(GICv3CPUState *cs, int irq, uint8_t prio,
>> +                      bool has_superprio)
>>   {
>>       /* Return true if this IRQ at this priority should take
>>        * precedence over the current recorded highest priority
>> @@ -33,11 +34,24 @@ static bool irqbetter(GICv3CPUState *cs, int irq,
>> uint8_t prio)
>>       if (prio < cs->hppi.prio) {
>>           return true;
>>       }
>> +
>> +    /*
>> +     * Current highest prioirity pending interrupt is an IRQ without
>> +     * superpriority, the new IRQ with superpriority has same priority
>> +     * should signal to the CPU as it have the priority higher than
>> +     * the labelled 0x80 or 0x00.
>> +     */
>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio && has_superprio) {
>> +        return true;
>> +    }
>> +
>>       /* If multiple pending interrupts have the same priority then it
>> is an
>>        * IMPDEF choice which of them to signal to the CPU. We choose to
>> -     * signal the one with the lowest interrupt number.
>> +     * signal the one with the lowest interrupt number if they don't
>> have
>> +     * superpriority.
>>        */
>> -    if (prio == cs->hppi.prio && irq <= cs->hppi.irq) {
>> +    if (prio == cs->hppi.prio && !cs->hppi.superprio &&
>> +        !has_superprio && irq <= cs->hppi.irq) {
>>           return true;
>>       }
>>       return false;
>> @@ -129,6 +143,35 @@ static uint32_t gicr_int_pending(GICv3CPUState *cs)
>>       return pend;
>>   }
>>   +static bool gicv3_get_priority(GICv3CPUState *cs, bool is_redist,
>> +                               uint32_t superprio, uint8_t *prio, int
>> irq)
>> +{
>> +    bool has_superprio = false;
>> +
>> +    if (superprio) {
>> +        has_superprio = true;
>> +
>> +        /* DS = 0 & Non-secure NMI */
>> +        if (!(cs->gic->gicd_ctlr & GICD_CTLR_DS) &&
>> +            ((is_redist && extract32(cs->gicr_igroupr0, irq, 1)) ||
>> +             (!is_redist && gicv3_gicd_group_test(cs->gic, irq)))) {
>> +            *prio = 0x80;
>> +        } else {
>> +            *prio = 0x0;
>> +        }
>> +    } else {
>> +        has_superprio = false;
>> +
>> +        if (is_redist) {
>> +            *prio = cs->gicr_ipriorityr[irq];
>> +        } else {
>> +            *prio = cs->gic->gicd_ipriority[irq];
>> +        }
>> +    }
>> +
>> +    return has_superprio;
>> +}
> 
> Did you not like the idea to map {priority, !superpriority} into a
> single value?
> 
> It would eliminate the change in irqbetter(), which is a bit more
> complex than it needs to be.
> 
>> @@ -152,10 +197,13 @@ static void
>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>               if (!(pend & (1 << i))) {
>>                   continue;
>>               }
>> -            prio = cs->gicr_ipriorityr[i];
>> -            if (irqbetter(cs, i, prio)) {
>> +            superprio = extract32(cs->gicr_isuperprio, i, 1);
>> +            has_superprio = gicv3_get_priority(cs, true, superprio,
>> &prio, i);
> 
> It would allow moving the read of gicr_isuperprio into
> gicv3_get_priority(), alongside the read of gicr_ipriorityr.
> 
> Is there a bug here not handling is_redist for GCIR_INMI*?

There seems no bug. Since prio = cs->gicr_ipriorityr[i] or  prio =
s->gicd_ipriority[i] are separately used originally,

> 
>> @@ -168,7 +216,7 @@ static void
>> gicv3_redist_update_noirqset(GICv3CPUState *cs)
>>       if ((cs->gicr_ctlr & GICR_CTLR_ENABLE_LPIS) &&
>> cs->gic->lpi_enable &&
>>           (cs->gic->gicd_ctlr & GICD_CTLR_EN_GRP1NS) &&
>>           (cs->hpplpi.prio != 0xff)) {
>> -        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio)) {
>> +        if (irqbetter(cs, cs->hpplpi.irq, cs->hpplpi.prio, false)) {
> 
> Always passing false here is incorrect -- again missing the
> redistributor nmi bit?
> 
> 
> r~

[RFC PATCH v5 19/22] hw/intc/arm_gicv3: Report the NMI interrupt in gicv3_cpuif_update()

[Jinjie Ruan via]

In CPU Interface, if the IRQ has the superpriority property, report
NMI to the corresponding PE.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
v4:
- Swap the ordering of the IFs.
v3:
- Remove handling nmi_is_irq flag.
---
 hw/intc/arm_gicv3_cpuif.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/hw/intc/arm_gicv3_cpuif.c b/hw/intc/arm_gicv3_cpuif.c
index df82a413c6..afba98ad87 100644
--- a/hw/intc/arm_gicv3_cpuif.c
+++ b/hw/intc/arm_gicv3_cpuif.c
@@ -938,6 +938,7 @@ void gicv3_cpuif_update(GICv3CPUState *cs)
     /* Tell the CPU about its highest priority pending interrupt */
     int irqlevel = 0;
     int fiqlevel = 0;
+    int nmilevel = 0;
     ARMCPU *cpu = ARM_CPU(cs->cpu);
     CPUARMState *env = &cpu->env;
 
@@ -976,6 +977,8 @@ void gicv3_cpuif_update(GICv3CPUState *cs)
 
         if (isfiq) {
             fiqlevel = 1;
+        } else if (cs->hppi.superprio) {
+            nmilevel = 1;
         } else {
             irqlevel = 1;
         }
@@ -985,6 +988,7 @@ void gicv3_cpuif_update(GICv3CPUState *cs)
 
     qemu_set_irq(cs->parent_fiq, fiqlevel);
     qemu_set_irq(cs->parent_irq, irqlevel);
+    qemu_set_irq(cs->parent_nmi, nmilevel);
 }
 
 static uint64_t icc_pmr_read(CPUARMState *env, const ARMCPRegInfo *ri)
-- 
2.34.1

Re: [RFC PATCH v5 19/22] hw/intc/arm_gicv3: Report the NMI interrupt in gicv3_cpuif_update()

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> In CPU Interface, if the IRQ has the superpriority property, report
> NMI to the corresponding PE.
> 
> Signed-off-by: Jinjie Ruan<ruanjinjie@huawei.com>
> ---
> v4:
> - Swap the ordering of the IFs.
> v3:
> - Remove handling nmi_is_irq flag.
> ---
>   hw/intc/arm_gicv3_cpuif.c | 4 ++++
>   1 file changed, 4 insertions(+)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[RFC PATCH v5 20/22] hw/intc/arm_gicv3: Report the VNMI interrupt

[Jinjie Ruan via]

In vCPU Interface, if the vIRQ has the superpriority property, report
vNMI to the corresponding vPE.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
---
 hw/intc/arm_gicv3_cpuif.c | 14 ++++++++++++--
 hw/intc/gicv3_internal.h  |  1 +
 2 files changed, 13 insertions(+), 2 deletions(-)

diff --git a/hw/intc/arm_gicv3_cpuif.c b/hw/intc/arm_gicv3_cpuif.c
index afba98ad87..0141d259e4 100644
--- a/hw/intc/arm_gicv3_cpuif.c
+++ b/hw/intc/arm_gicv3_cpuif.c
@@ -465,6 +465,7 @@ void gicv3_cpuif_virt_irq_fiq_update(GICv3CPUState *cs)
     int idx;
     int irqlevel = 0;
     int fiqlevel = 0;
+    int nmilevel = 0;
 
     idx = hppvi_index(cs);
     trace_gicv3_cpuif_virt_update(gicv3_redist_affid(cs), idx,
@@ -482,9 +483,17 @@ void gicv3_cpuif_virt_irq_fiq_update(GICv3CPUState *cs)
         uint64_t lr = cs->ich_lr_el2[idx];
 
         if (icv_hppi_can_preempt(cs, lr)) {
-            /* Virtual interrupts are simple: G0 are always FIQ, and G1 IRQ */
+            /*
+             * Virtual interrupts are simple: G0 are always FIQ, and G1 are
+             * IRQ or NMI which depends on the ICH_LR<n>_EL2.NMI to have
+             * non-maskable property.
+             */
             if (lr & ICH_LR_EL2_GROUP) {
-                irqlevel = 1;
+                if (cs->gic->nmi_support && (lr & ICH_LR_EL2_NMI)) {
+                    nmilevel = 1;
+                } else {
+                    irqlevel = 1;
+                }
             } else {
                 fiqlevel = 1;
             }
@@ -494,6 +503,7 @@ void gicv3_cpuif_virt_irq_fiq_update(GICv3CPUState *cs)
     trace_gicv3_cpuif_virt_set_irqs(gicv3_redist_affid(cs), fiqlevel, irqlevel);
     qemu_set_irq(cs->parent_vfiq, fiqlevel);
     qemu_set_irq(cs->parent_virq, irqlevel);
+    qemu_set_irq(cs->parent_vnmi, nmilevel);
 }
 
 static void gicv3_cpuif_virt_update(GICv3CPUState *cs)
diff --git a/hw/intc/gicv3_internal.h b/hw/intc/gicv3_internal.h
index 93e56b3726..b6cb0115e7 100644
--- a/hw/intc/gicv3_internal.h
+++ b/hw/intc/gicv3_internal.h
@@ -242,6 +242,7 @@ FIELD(GICR_VPENDBASER, VALID, 63, 1)
 #define ICH_LR_EL2_PRIORITY_SHIFT 48
 #define ICH_LR_EL2_PRIORITY_LENGTH 8
 #define ICH_LR_EL2_PRIORITY_MASK (0xffULL << ICH_LR_EL2_PRIORITY_SHIFT)
+#define ICH_LR_EL2_NMI (1ULL << 59)
 #define ICH_LR_EL2_GROUP (1ULL << 60)
 #define ICH_LR_EL2_HW (1ULL << 61)
 #define ICH_LR_EL2_STATE_SHIFT 62
-- 
2.34.1

Re: [RFC PATCH v5 20/22] hw/intc/arm_gicv3: Report the VNMI interrupt

[Richard Henderson]

On 2/29/24 03:10, Jinjie Ruan via wrote:
> In vCPU Interface, if the vIRQ has the superpriority property, report
> vNMI to the corresponding vPE.
> 
> Signed-off-by: Jinjie Ruan<ruanjinjie@huawei.com>
> ---
>   hw/intc/arm_gicv3_cpuif.c | 14 ++++++++++++--
>   hw/intc/gicv3_internal.h  |  1 +
>   2 files changed, 13 insertions(+), 2 deletions(-)

Reviewed-by: Richard Henderson <richard.henderson@linaro.org>

r~

[RFC PATCH v5 21/22] target/arm: Add FEAT_NMI to max

[Jinjie Ruan via]

Enable FEAT_NMI on the 'max' CPU.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v3:
- Add Reviewed-by.
- Sorted to last.
---
 docs/system/arm/emulation.rst | 1 +
 target/arm/tcg/cpu64.c        | 1 +
 2 files changed, 2 insertions(+)

diff --git a/docs/system/arm/emulation.rst b/docs/system/arm/emulation.rst
index f67aea2d83..91baf7ad69 100644
--- a/docs/system/arm/emulation.rst
+++ b/docs/system/arm/emulation.rst
@@ -63,6 +63,7 @@ the following architecture extensions:
 - FEAT_MTE (Memory Tagging Extension)
 - FEAT_MTE2 (Memory Tagging Extension)
 - FEAT_MTE3 (MTE Asymmetric Fault Handling)
+- FEAT_NMI (Non-maskable Interrupt)
 - FEAT_NV (Nested Virtualization)
 - FEAT_NV2 (Enhanced nested virtualization support)
 - FEAT_PACIMP (Pointer authentication - IMPLEMENTATION DEFINED algorithm)
diff --git a/target/arm/tcg/cpu64.c b/target/arm/tcg/cpu64.c
index 5fba2c0f04..60f0dcd799 100644
--- a/target/arm/tcg/cpu64.c
+++ b/target/arm/tcg/cpu64.c
@@ -1175,6 +1175,7 @@ void aarch64_max_tcg_initfn(Object *obj)
     t = FIELD_DP64(t, ID_AA64PFR1, RAS_FRAC, 0);  /* FEAT_RASv1p1 + FEAT_DoubleFault */
     t = FIELD_DP64(t, ID_AA64PFR1, SME, 1);       /* FEAT_SME */
     t = FIELD_DP64(t, ID_AA64PFR1, CSV2_FRAC, 0); /* FEAT_CSV2_2 */
+    t = FIELD_DP64(t, ID_AA64PFR1, NMI, 1);       /* FEAT_NMI */
     cpu->isar.id_aa64pfr1 = t;
 
     t = cpu->isar.id_aa64mmfr0;
-- 
2.34.1

[RFC PATCH v5 22/22] hw/arm/virt: Add FEAT_GICv3_NMI feature support in virt GIC

[Jinjie Ruan via]

A PE that implements FEAT_NMI and FEAT_GICv3 also implements
FEAT_GICv3_NMI. A PE that does not implement FEAT_NMI, does not implement
FEAT_GICv3_NMI

So included support FEAT_GICv3_NMI feature as part of virt platform
GIC initialization if FEAT_NMI and FEAT_GICv3 supported.

Signed-off-by: Jinjie Ruan <ruanjinjie@huawei.com>
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
---
v4:
- Add Reviewed-by.
v3:
- Adjust to be the last after add FEAT_NMI to max.
- Check whether support FEAT_NMI and FEAT_GICv3 for FEAT_GICv3_NMI.
---
 hw/arm/virt.c | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/hw/arm/virt.c b/hw/arm/virt.c
index 2d4a187fd5..c12307ccd9 100644
--- a/hw/arm/virt.c
+++ b/hw/arm/virt.c
@@ -712,6 +712,19 @@ static void create_v2m(VirtMachineState *vms)
     vms->msi_controller = VIRT_MSI_CTRL_GICV2M;
 }
 
+/*
+ * A PE that implements FEAT_NMI and FEAT_GICv3 also implements
+ * FEAT_GICv3_NMI. A PE that does not implement FEAT_NMI, does not implement
+ * FEAT_GICv3_NMI.
+ */
+static bool gicv3_nmi_present(VirtMachineState *vms)
+{
+    ARMCPU *cpu = ARM_CPU(qemu_get_cpu(0));
+
+    return cpu_isar_feature(aa64_nmi, cpu) &&
+           (vms->gic_version != VIRT_GIC_VERSION_2);
+}
+
 static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
 {
     MachineState *ms = MACHINE(vms);
@@ -785,6 +798,11 @@ static void create_gic(VirtMachineState *vms, MemoryRegion *mem)
                               vms->virt);
         }
     }
+
+    if (gicv3_nmi_present(vms)) {
+        qdev_prop_set_bit(vms->gic, "has-nmi", true);
+    }
+
     gicbusdev = SYS_BUS_DEVICE(vms->gic);
     sysbus_realize_and_unref(gicbusdev, &error_fatal);
     sysbus_mmio_map(gicbusdev, 0, vms->memmap[VIRT_GIC_DIST].base);
-- 
2.34.1