[PATCH v2 0/2] target/riscv: Allow software access to MIP SEIP

[PATCH v2 0/2] target/riscv: Allow software access to MIP SEIP

[Alistair Francis]

From: Alistair Francis <alistair.francis@wdc.com>

The RISC-V specification states that:
  "Supervisor-level external interrupts are made pending based on the
  logical-OR of the software-writable SEIP bit and the signal from the
  external interrupt controller."

We currently only allow either the interrupt controller or software to
set the bit, which is incorrect.

This patch removes the miclaim mask when writing MIP to allow M-mode
software to inject interrupts, even with an interrupt controller.

We then also need to keep track of which source is setting MIP_SEIP. The
final value is a OR of both, so we add two bools and use that to keep
track of the current state. This way either source can change without
losing the correct value.

This fixes: https://gitlab.com/qemu-project/qemu/-/issues/904

Alistair Francis (2):
  target/riscv: cpu: Fixup indentation
  target/riscv: Allow software access to MIP SEIP

 target/riscv/cpu.h |  8 ++++++++
 target/riscv/cpu.c | 30 +++++++++++++++++++-----------
 target/riscv/csr.c |  8 ++++++--
 3 files changed, 33 insertions(+), 13 deletions(-)

-- 
2.35.1

[PATCH v2 1/2] target/riscv: cpu: Fixup indentation

[Alistair Francis]

From: Alistair Francis <alistair.francis@wdc.com>

Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
---
 target/riscv/cpu.c | 20 ++++++++++----------
 1 file changed, 10 insertions(+), 10 deletions(-)

diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index ddda4906ff..41b757995d 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -567,18 +567,18 @@ static void riscv_cpu_realize(DeviceState *dev, Error **errp)
         if (cpu->cfg.ext_i && cpu->cfg.ext_e) {
             error_setg(errp,
                        "I and E extensions are incompatible");
-                       return;
-       }
+            return;
+        }
 
         if (!cpu->cfg.ext_i && !cpu->cfg.ext_e) {
             error_setg(errp,
                        "Either I or E extension must be set");
-                       return;
-       }
+            return;
+        }
 
-       if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
-                               cpu->cfg.ext_a & cpu->cfg.ext_f &
-                               cpu->cfg.ext_d)) {
+        if (cpu->cfg.ext_g && !(cpu->cfg.ext_i & cpu->cfg.ext_m &
+                                cpu->cfg.ext_a & cpu->cfg.ext_f &
+                                cpu->cfg.ext_d)) {
             warn_report("Setting G will also set IMAFD");
             cpu->cfg.ext_i = true;
             cpu->cfg.ext_m = true;
@@ -709,11 +709,11 @@ static void riscv_cpu_set_irq(void *opaque, int irq, int level)
         case IRQ_S_EXT:
         case IRQ_VS_EXT:
         case IRQ_M_EXT:
-             if (kvm_enabled()) {
+            if (kvm_enabled()) {
                 kvm_riscv_set_irq(cpu, irq, level);
-             } else {
+            } else {
                 riscv_cpu_update_mip(cpu, 1 << irq, BOOL_TO_MASK(level));
-             }
+            }
              break;
         default:
             g_assert_not_reached();
-- 
2.35.1

[PATCH v2 2/2] target/riscv: Allow software access to MIP SEIP

[Alistair Francis]

From: Alistair Francis <alistair.francis@wdc.com>

The RISC-V specification states that:
  "Supervisor-level external interrupts are made pending based on the
  logical-OR of the software-writable SEIP bit and the signal from the
  external interrupt controller."

We currently only allow either the interrupt controller or software to
set the bit, which is incorrect.

This patch removes the miclaim mask when writing MIP to allow M-mode
software to inject interrupts, even with an interrupt controller.

We then also need to keep track of which source is setting MIP_SEIP. The
final value is a OR of both, so we add two bools and use that to keep
track of the current state. This way either source can change without
losing the correct value.

Resolves: https://gitlab.com/qemu-project/qemu/-/issues/904
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Reviewed-by: Bin Meng <bmeng.cn@gmail.com>
---
 target/riscv/cpu.h |  8 ++++++++
 target/riscv/cpu.c | 10 +++++++++-
 target/riscv/csr.c |  8 ++++++--
 3 files changed, 23 insertions(+), 3 deletions(-)

diff --git a/target/riscv/cpu.h b/target/riscv/cpu.h
index c069fe85fa..05d40f8dbd 100644
--- a/target/riscv/cpu.h
+++ b/target/riscv/cpu.h
@@ -173,6 +173,14 @@ struct CPUArchState {
     uint64_t mstatus;
 
     uint64_t mip;
+    /*
+     * MIP contains the software writable version of SEIP ORed with the
+     * external interrupt value. The MIP register is always up-to-date.
+     * To keep track of the current source, we also save booleans of the values
+     * here.
+     */
+    bool external_seip;
+    bool software_seip;
 
     uint64_t miclaim;
 
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 41b757995d..68373b769c 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -706,7 +706,6 @@ static void riscv_cpu_set_irq(void *opaque, int irq, int level)
         case IRQ_VS_TIMER:
         case IRQ_M_TIMER:
         case IRQ_U_EXT:
-        case IRQ_S_EXT:
         case IRQ_VS_EXT:
         case IRQ_M_EXT:
             if (kvm_enabled()) {
@@ -715,6 +714,15 @@ static void riscv_cpu_set_irq(void *opaque, int irq, int level)
                 riscv_cpu_update_mip(cpu, 1 << irq, BOOL_TO_MASK(level));
             }
              break;
+        case IRQ_S_EXT:
+            if (kvm_enabled()) {
+                kvm_riscv_set_irq(cpu, irq, level);
+            } else {
+                env->external_seip = level;
+                riscv_cpu_update_mip(cpu, 1 << irq,
+                                     BOOL_TO_MASK(level | env->software_seip));
+            }
+            break;
         default:
             g_assert_not_reached();
         }
diff --git a/target/riscv/csr.c b/target/riscv/csr.c
index 0606cd0ea8..48e78cf91e 100644
--- a/target/riscv/csr.c
+++ b/target/riscv/csr.c
@@ -1403,10 +1403,14 @@ static RISCVException rmw_mip64(CPURISCVState *env, int csrno,
                                 uint64_t new_val, uint64_t wr_mask)
 {
     RISCVCPU *cpu = env_archcpu(env);
-    /* Allow software control of delegable interrupts not claimed by hardware */
-    uint64_t old_mip, mask = wr_mask & delegable_ints & ~env->miclaim;
+    uint64_t old_mip, mask = wr_mask & delegable_ints;
     uint32_t gin;
 
+    if (mask & MIP_SEIP) {
+        env->software_seip = new_val & MIP_SEIP;
+    }
+    new_val |= env->external_seip << IRQ_S_EXT;
+
     if (mask) {
         old_mip = riscv_cpu_update_mip(cpu, mask, (new_val & mask));
     } else {
-- 
2.35.1

Re: [PATCH v2 2/2] target/riscv: Allow software access to MIP SEIP

[Richard Henderson]

On 3/15/22 18:59, Alistair Francis wrote:
> +    if (mask & MIP_SEIP) {
> +        env->software_seip = new_val & MIP_SEIP;
> +    }
> +    new_val |= env->external_seip << IRQ_S_EXT;

You can move the second statement inside the if as well, since you don't need bits in 
new_val set that are not in mask.

Also might be clearer as

         env->external_seip * MIP_SEIP

rather than the shift, to make it clear we're talking about the same bit.  The compiler 
will reduce the multiplication.

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


r~

Re: [PATCH v2 2/2] target/riscv: Allow software access to MIP SEIP

[Richard Henderson]

On 3/15/22 18:59, Alistair Francis wrote:
> +    if (mask & MIP_SEIP) {
> +        env->software_seip = new_val & MIP_SEIP;
> +    }
> +    new_val |= env->external_seip << IRQ_S_EXT;

You can move the second statement inside the if as well, since you don't need bits in 
new_val set that are not in mask.

Also might be clearer as

         env->external_seip * MIP_SEIP

rather than the shift, to make it clear we're talking about the same bit.  The compiler 
will reduce the multiplication.

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


r~

Re: [PATCH v2 1/2] target/riscv: cpu: Fixup indentation

[Richard Henderson]

On 3/15/22 18:59, Alistair Francis wrote:
> From: Alistair Francis<alistair.francis@wdc.com>
> 
> Signed-off-by: Alistair Francis<alistair.francis@wdc.com>
> Reviewed-by: Bin Meng<bmeng.cn@gmail.com>
> ---
>   target/riscv/cpu.c | 20 ++++++++++----------
>   1 file changed, 10 insertions(+), 10 deletions(-)

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

r~

Re: [PATCH v2 1/2] target/riscv: cpu: Fixup indentation

[Richard Henderson]

On 3/15/22 18:59, Alistair Francis wrote:
> From: Alistair Francis<alistair.francis@wdc.com>
> 
> Signed-off-by: Alistair Francis<alistair.francis@wdc.com>
> Reviewed-by: Bin Meng<bmeng.cn@gmail.com>
> ---
>   target/riscv/cpu.c | 20 ++++++++++----------
>   1 file changed, 10 insertions(+), 10 deletions(-)

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

r~