Re: [PATCH 2/2] hw/riscv/aia: Check for invalid IMSIC configs

On Mon, 2026-08-24 at 12:35 +0930, Joel Stanley wrote:
> Assert that the IMSIC layout will be valid before creating the
> devices.
> 
> aia_guests is a command line parameter but the machines that allows
> it
> to be set (currently only virt) should be passing the correct stride
> for
> a given number of guests. The assert helps catch bugs in future
> changes
> and shouldn't be triggerable by an invalid user config.
> 
> Signed-off-by: Joel Stanley <[email protected]>

Acked-by: Alistair Francis <[email protected]>

Alistair

> ---
>  hw/riscv/aia.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/hw/riscv/aia.c b/hw/riscv/aia.c
> index c33a3b32bd45..c1a5982856c5 100644
> --- a/hw/riscv/aia.c
> +++ b/hw/riscv/aia.c
> @@ -43,6 +43,10 @@ DeviceState *riscv_create_aia(bool msimode, int
> aia_guests,
>      /* The RISC-V Advanced Interrupt Architecture, Chapter 1.2.
> Limits */
>      g_assert(num_sources <= 1023);
>  
> +    /* Prevent IMSIC MMIO regions from silently overlapping */
> +    g_assert(s_imsic_stride >= IMSIC_HART_SIZE(imsic_num_bits(1 +
> aia_guests)));
> +    g_assert(m_imsic_stride >= IMSIC_HART_SIZE(0));
> +
>      if (msimode) {
>          if (!kvm_enabled()) {
>              /* Per-socket M-level IMSICs */