On 29.07.2025 10:52, Mykola Kvach wrote:
> --- a/xen/common/domain.c
> +++ b/xen/common/domain.c
> @@ -1334,16 +1334,15 @@ int domain_shutdown(struct domain *d, u8 reason)
> return 0;
> }
>
> -void domain_resume(struct domain *d)
> +#ifdef CONFIG_ARM
> +void domain_resume_nopause(struct domain *d)
> +#else
> +static void domain_resume_nopause(struct domain *d)
> +#endif
#ifndef CONFIG_ARM
static
#endif
void domain_resume_nopause(struct domain *d)
to have as little redundancy as possible.
> {
> struct vcpu *v;
>
> - /*
> - * Some code paths assume that shutdown status does not get reset under
> - * their feet (e.g., some assertions make this assumption).
> - */
> - domain_pause(d);
> -
> + domain_lock(d);
This addition of locking affects domain_resume() as well. Neither need nor
correctness are discussed in the description. If the locking was really
needed for domain_resume() as well, I suppose adding that would better be
a separate change anyway.
The addition of this locking is particularly interesting considering that
...
> spin_lock(&d->shutdown_lock);
... is what follows right after.
> --- a/xen/include/xen/sched.h
> +++ b/xen/include/xen/sched.h
> @@ -814,6 +814,9 @@ void domain_destroy(struct domain *d);
> int domain_kill(struct domain *d);
> int domain_shutdown(struct domain *d, u8 reason);
> void domain_resume(struct domain *d);
> +#ifdef CONFIG_ARM
> +void domain_resume_nopause(struct domain *d);
> +#endif
>
> int domain_soft_reset(struct domain *d, bool resuming);
>
> @@ -1010,6 +1013,10 @@ static inline struct domain *next_domain_in_cpupool(
> /* VCPU is parked. */
> #define _VPF_parked 8
> #define VPF_parked (1UL<<_VPF_parked)
> +/* VCPU is suspended. */
> +#define _VPF_suspended 9
> +#define VPF_suspended (1UL<<_VPF_suspended)
Irrespective of the style violations in pre-existing code, can you please
not add further violations, by inserting the missing blanks?
> +
>
Please also don't introduce double blank lines.
Jan
