Hi,
I've been learning how SMP scheduler word in RTEMS and I came across the
following function definition:
--------------------------------------------------------------------------------------------
/**
* @brief Allocates the cpu for the scheduled thread.
*
* Attempts to prevent migrations but does not take into account affinity.
*
* @param context The scheduler context instance.
* @param scheduled The scheduled node that should be executed next.
* @param victim If the heir is this node's thread, no processor is
allocated.
* @param[in, out] victim_cpu The cpu to allocate.
*/
static inline void _Scheduler_SMP_Allocate_processor_lazy(
Scheduler_Context *context,
Scheduler_Node *scheduled,
Scheduler_Node *victim,
Per_CPU_Control *victim_cpu
)
{
Thread_Control *scheduled_thread = _Scheduler_Node_get_user( scheduled );
Thread_Control *victim_thread = _Scheduler_Node_get_user( victim );
Per_CPU_Control *scheduled_cpu = _Thread_Get_CPU( scheduled_thread );
Per_CPU_Control *cpu_self = _Per_CPU_Get();
Thread_Control *heir;
_Assert( _ISR_Get_level() != 0 );
if ( _Thread_Is_executing_on_a_processor( scheduled_thread ) ) {
if ( _Scheduler_SMP_Is_processor_owned_by_us( context, scheduled_cpu )
) {
heir = scheduled_cpu->heir;
_Thread_Dispatch_update_heir(
cpu_self,
scheduled_cpu,
scheduled_thread
);
} else {
/* We have to force a migration to our processor set */
heir = scheduled_thread;
}
} else {
heir = scheduled_thread;
}
if ( heir != victim_thread ) {
_Thread_Set_CPU( heir, victim_cpu );
_Thread_Dispatch_update_heir( cpu_self, victim_cpu, heir );
}
}
--------------------------------------------------------------------------------
Can someone please help me understand what this function is trying to do
and how exactly it is achieving that? I am finding it hard to understand
the meaning and use of the terms like cpu_self (why do we need cpu_self ?
Is it the CPU for scheduled or for
victim?), _Thread_Is_executing_on_a_processor( scheduled_thread ): Why
would we have the thread already executing when we need to allocate a
processor to it? Why are we allocating a processor then?
Thanks,
Richi.
_______________________________________________
devel mailing list
devel@rtems.org
http://lists.rtems.org/mailman/listinfo/devel
