Re: [PATCH 05/22] target/xtensa: Fix AddressSpace exposure timing

[Thomas Huth <thuth@redhat.com>]

On 06/09/2025 04.11, Akihiko Odaki wrote:
> xtensa-cpu is not hotpluggable but its instance can still be created and
> finalized when processing the device-list-properties QMP command.
> Exposing such a temporary instance to AddressSpace should be
> avoided because it leaks the instance.
> 
> Expose instances to the AddressSpace at their realization time so that
> it won't happen for the temporary instances.
> 
> Signed-off-by: Akihiko Odaki <odaki@rsg.ci.i.u-tokyo.ac.jp>
> ---
>   target/xtensa/cpu.c | 8 +++++---
>   1 file changed, 5 insertions(+), 3 deletions(-)
> 
> diff --git a/target/xtensa/cpu.c b/target/xtensa/cpu.c
> index ea9b6df3aa24178c8e6a88b02afda5db659199da..63edc3a5b2778c8379a30125481f65361655fe1c 100644
> --- a/target/xtensa/cpu.c
> +++ b/target/xtensa/cpu.c
> @@ -243,7 +243,11 @@ static void xtensa_cpu_realizefn(DeviceState *dev, Error **errp)
>       Error *local_err = NULL;
>   
>   #ifndef CONFIG_USER_ONLY
> -    xtensa_irq_init(&XTENSA_CPU(dev)->env);
> +    CPUXtensaState *env = &XTENSA_CPU(dev)->env;
> +
> +    env->address_space_er = g_malloc(sizeof(*env->address_space_er));
> +    address_space_init(env->address_space_er, env->system_er, "ER");
> +    xtensa_irq_init(env);
>   #endif
>   
>       cpu_exec_realizefn(cs, &local_err);
> @@ -268,11 +272,9 @@ static void xtensa_cpu_initfn(Object *obj)
>       env->config = xcc->config;
>   
>   #ifndef CONFIG_USER_ONLY
> -    env->address_space_er = g_malloc(sizeof(*env->address_space_er));
>       env->system_er = g_malloc(sizeof(*env->system_er));
>       memory_region_init_io(env->system_er, obj, NULL, env, "er",
>                             UINT64_C(0x100000000));
> -    address_space_init(env->address_space_er, env->system_er, "ER");
>   
>       cpu->clock = qdev_init_clock_in(DEVICE(obj), "clk-in", NULL, cpu, 0);
>       clock_set_hz(cpu->clock, env->config->clock_freq_khz * 1000);
> 

Reviewed-by: Thomas Huth <thuth@redhat.com>