Need help in understanding some of the existing code in RTEMS

[Sebastian Huber]

On 14/07/2020 11:11, Sebastian Huber wrote:

> On 14/07/2020 10:47, Richi Dubey wrote:
>
>> Can someone please help me understand how this 
>> (https://git.rtems.org/rtems/tree/cpukit/score/src/scheduleredfsmp.c#n385) 
>> if condition works. Why are we removing the ready queue Node from the 
>> chain of Affine queues when we are allocating a different processor 
>> (https://git.rtems.org/rtems/tree/cpukit/score/src/scheduleredfsmp.c#n398) 
>> to the node which was currently scheduled on the CPU of that (the 
>> ready queue mentioned in the if condition's) ready queue?
>
> |static inline void _Scheduler_EDF_SMP_Allocate_processor( 
> Scheduler_Context *context, Scheduler_Node *scheduled_base, 
> Scheduler_Node *victim_base, Per_CPU_Control *victim_cpu ) { 
> Scheduler_EDF_SMP_Context *self; Scheduler_EDF_SMP_Node *scheduled; 
> uint8_t rqi; (void) victim_base; self = _Scheduler_EDF_SMP_Get_self( 
> context ); scheduled = _Scheduler_EDF_SMP_Node_downcast( 
> scheduled_base ); rqi = scheduled->ready_queue_index; if ( rqi != 0 ) 
> { Scheduler_EDF_SMP_Ready_queue *ready_queue; Per_CPU_Control 
> *desired_cpu; ready_queue = &self->Ready[ rqi ]; if ( 
> !_Chain_Is_node_off_chain( &ready_queue->Node ) ) { 
> _Chain_Extract_unprotected( &ready_queue->Node ); 
> _Chain_Set_off_chain( &ready_queue->Node ); } Here we remove the 
> affine ready queue if it exists from the chain of affine queues since 
> now an affine thread is scheduled on a processor.  desired_cpu = 
> _Per_CPU_Get_by_index( rqi - 1 ); if ( victim_cpu != desired_cpu ) { 
> Scheduler_EDF_SMP_Node *node; This is another action. If the victim 
> CPU is not the right one for the new scheduled node, we have to make 
> room for it on the desired CPU.  node = 
> _Scheduler_EDF_SMP_Get_scheduled( self, rqi ); _Assert( 
> node->ready_queue_index == 0 ); _Scheduler_EDF_SMP_Set_scheduled( 
> self, node, victim_cpu ); _Scheduler_SMP_Allocate_processor_exact( 
> context, &node->Base.Base, NULL, victim_cpu ); victim_cpu = 
> desired_cpu; } }|

Sorry for the format.

static inline void _Scheduler_EDF_SMP_Allocate_processor(
   Scheduler_Context *context,
   Scheduler_Node    *scheduled_base,
   Scheduler_Node    *victim_base,
   Per_CPU_Control   *victim_cpu
)
{
   Scheduler_EDF_SMP_Context     *self;
   Scheduler_EDF_SMP_Node        *scheduled;
   uint8_t                        rqi;

   (void) victim_base;
   self = _Scheduler_EDF_SMP_Get_self( context );
   scheduled = _Scheduler_EDF_SMP_Node_downcast( scheduled_base );
   rqi = scheduled->ready_queue_index;

   if ( rqi != 0 ) {
     Scheduler_EDF_SMP_Ready_queue *ready_queue;
     Per_CPU_Control               *desired_cpu;

     ready_queue = &self->Ready[ rqi ];

     if ( !_Chain_Is_node_off_chain( &ready_queue->Node ) ) {
       _Chain_Extract_unprotected( &ready_queue->Node );
       _Chain_Set_off_chain( &ready_queue->Node );
     }

Here we remove the affine ready queue if it exists from the chain of 
affine queues since now an affine thread is scheduled on a processor.

     desired_cpu = _Per_CPU_Get_by_index( rqi - 1 );

     if ( victim_cpu != desired_cpu ) {
       Scheduler_EDF_SMP_Node *node;

This is another action. If the victim CPU is not the right one for the 
new scheduled node, we have to make room for it on the desired CPU.

       node = _Scheduler_EDF_SMP_Get_scheduled( self, rqi );
       _Assert( node->ready_queue_index == 0 );
       _Scheduler_EDF_SMP_Set_scheduled( self, node, victim_cpu );
       _Scheduler_SMP_Allocate_processor_exact(
         context,
         &node->Base.Base,
         NULL,
         victim_cpu
       );
       victim_cpu = desired_cpu;
     }
   }