Re: [PATCH v1 1/1] hw/riscv: shakti_c: Mark as not user creatable

[Philippe Mathieu-Daudé]

Hi Alistair,

On 9/27/21 09:13, Alistair Francis wrote:
> From: Alistair Francis <alistair.francis@wdc.com>
> 
> Mark the shakti_c machine as not user creatable as it uses serial_hd.
> 
> Resolves: https://gitlab.com/qemu-project/qemu/-/issues/639
> Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
> ---
>  hw/riscv/shakti_c.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/hw/riscv/shakti_c.c b/hw/riscv/shakti_c.c
> index 2f084d3c8d..603992d442 100644
> --- a/hw/riscv/shakti_c.c
> +++ b/hw/riscv/shakti_c.c
> @@ -150,6 +150,8 @@ static void shakti_c_soc_class_init(ObjectClass *klass, 
> void *data)
>  {
>      DeviceClass *dc = DEVICE_CLASS(klass);
>      dc->realize = shakti_c_soc_state_realize;
> +    /* Reason: Uses serial_hds in realize function, thus can't be used twice 
> */
> +    dc->user_creatable = false;

I agree with the logical change, but not totally with the comment.
Well, the comment itself is not incorrect. But the error happens before:

ERROR:tcg/tcg.c:479:tcg_register_thread: assertion failed: (n <
tcg_max_ctxs)

The reason is TYPE_RISCV_HART_ARRAY creates CPUs in
riscv_hart_realize(), and this machine use the default CPU limit
of one CPU:

static void shakti_c_machine_class_init(ObjectClass *klass, void *data)
{
    MachineClass *mc = MACHINE_CLASS(klass);
    mc->desc = "RISC-V Board compatible with Shakti SDK";
    mc->init = shakti_c_machine_state_init;
    mc->default_cpu_type = TYPE_RISCV_CPU_SHAKTI_C;
}

The default values come from:

static void machine_class_base_init(ObjectClass *oc, void *data)
{
    MachineClass *mc = MACHINE_CLASS(oc);
    mc->max_cpus = mc->max_cpus ?: 1;
    mc->min_cpus = mc->min_cpus ?: 1;
    mc->default_cpus = mc->default_cpus ?: 1;

So the crash happens when we try to create the 2nd CPU.

>  }
>  
>  static void shakti_c_soc_instance_init(Object *obj)
>