Re: [RFC 4/6] capabilities: introduce console io as a domain capability

[Jan Beulich <jbeulich@xxxxxxxx>]

On 01.08.2023 22:20, Daniel P. Smith wrote:
> --- a/xen/include/xen/sched.h
> +++ b/xen/include/xen/sched.h
> @@ -472,8 +472,8 @@ struct domain
>  #define ROLE_HARDWARE_DOMAIN   (1U<<2)
>  #define ROLE_XENSTORE_DOMAIN   (1U<<3)
>      uint8_t          role;
> -    /* Can this guest access the Xen console? */
> -    bool             is_console;
> +#define CAP_CONSOLE_IO  (1U<<0)
> +    uint8_t          capabilities;
>      /* Is this guest being debugged by dom0? */
>      bool             debugger_attached;
>      /*
> @@ -1146,6 +1146,27 @@ static always_inline bool is_hvm_vcpu(const struct 
> vcpu *v)
>      return is_hvm_domain(v->domain);
>  }
>  
> +static always_inline bool domain_has_cap(
> +    const struct domain *d, uint8_t cap)
> +{
> +    return d->capabilities & cap;
> +}

What you implement here is effectively domain_has_any_cap(), which I
don't think is of much use. At the very least you want to assert that
cap is a power of two. But perhaps you rather want the caller to pass
in a bit number, not a mask, such that it's obvious that only one
thing can be checked at a time.

> +static always_inline bool domain_set_cap(
> +    struct domain *d, uint8_t cap)
> +{
> +    switch ( cap )
> +    {
> +    case CAP_CONSOLE_IO:
> +        d->capabilities |= cap;
> +        break;
> +    default:
> +        return false;
> +    }
> +
> +    return domain_has_cap(d, cap);
> +}

The "set" operation doesn't need to be an inline function, does it?

Jan