On 02.06.2025 11:04, Mykola Kvach wrote:
> @@ -857,8 +860,24 @@ void arch_domain_destroy(struct domain *d)
> domain_io_free(d);
> }
>
> -void arch_domain_shutdown(struct domain *d)
> +int arch_domain_shutdown(struct domain *d)
> {
> + switch ( d->shutdown_code )
> + {
> + case SHUTDOWN_suspend:
> +#ifdef CONFIG_SYSTEM_SUSPEND
> + if ( !is_hardware_domain(d) )
> + break;
> +
> + return host_system_suspend();
> +#else
> + break;
> +#endif
> + default:
> + break;
> + }
> +
> + return 0;
> }
What's wrong with
int arch_domain_shutdown(struct domain *d)
{
switch ( d->shutdown_code )
{
#ifdef CONFIG_SYSTEM_SUSPEND
case SHUTDOWN_suspend:
if ( !is_hardware_domain(d) )
break;
return host_system_suspend();
#endif
default:
break;
}
return 0;
}
?
> @@ -1311,13 +1316,13 @@ int domain_shutdown(struct domain *d, u8 reason)
> v->paused_for_shutdown = 1;
> }
>
> - arch_domain_shutdown(d);
> + ret = arch_domain_shutdown(d);
If non-zero comes back here, is ...
> __domain_finalise_shutdown(d);
... this still appropriate to call?
> spin_unlock(&d->shutdown_lock);
>
> - return 0;
> + return ret;
> }
Most callers don't care about the return value of this function. That likely
needs to change, even if _for now_ you only alter the SHUTDOWN_suspend case
(and hence some of the callers aren't immediately impacted)?
Jan
