On 9/17/2020 10:17 AM, Andrew Cooper wrote:
> On 17/09/2020 09:12, Jan Beulich wrote:
>> On 16.09.2020 20:18, Jeff Kubascik wrote:
>>> @@ -517,27 +516,35 @@ static const struct scheduler sched_arinc653_def = {
>>> .sched_id = XEN_SCHEDULER_ARINC653,
>>> .sched_data = NULL,
>>>
>>> + .global_init = NULL,
>>> .init = a653sched_init,
>>> .deinit = a653sched_deinit,
>>>
>>> - .free_udata = a653sched_free_udata,
>>> - .alloc_udata = a653sched_alloc_udata,
>>> + .alloc_pdata = NULL,
>>> + .switch_sched = a653sched_switch_sched,
>>> + .deinit_pdata = NULL,
>>> + .free_pdata = NULL,
>>>
>>> + .alloc_domdata = NULL,
>>> + .free_domdata = NULL,
>>> +
>>> + .alloc_udata = a653sched_alloc_udata,
>>> .insert_unit = NULL,
>>> .remove_unit = NULL,
>>> + .free_udata = a653sched_free_udata,
>>>
>>> .sleep = a653sched_unit_sleep,
>>> .wake = a653sched_unit_wake,
>>> .yield = NULL,
>>> .context_saved = NULL,
>>>
>>> - .do_schedule = a653sched_do_schedule,
>>> -
>>> .pick_resource = a653sched_pick_resource,
>>> + .migrate = NULL,
>>>
>>> - .switch_sched = a653sched_switch_sched,
>>> + .do_schedule = a653sched_do_schedule,
>>>
>>> .adjust = NULL,
>>> + .adjust_affinity= NULL,
>> Adding all these not really needed NULL initializers looks to rather move
>> this scheduler away from all the others. (Oddly enough all of them
>> explicitly set .sched_data to NULL - for whatever reason.)
>
> The "= NULL" is totally redundant, because the compiler will do that for
> you.
I agree with this. This originally was intended to lay the groundwork for patch
#5, but looking at it again, was confusing and unnecessary. I'll remove the =
NULL lines.
> The last user of .sched_data was dropped by 9c95227160. Conceptually,
> it is a layering violation (it prevents different cpupools being
> properly independent), so I'd recommend just dropping the field entirely.
I'll remove .sched_data above.
-Jeff
