[PATCH v4] Pre-Release Strong-APA
Richi Dubey
richidubey at gmail.com
Wed Aug 26 20:29:59 UTC 2020
---
cpukit/include/rtems/scheduler.h | 6 +-
.../include/rtems/score/schedulerstrongapa.h | 39 +-
cpukit/score/src/schedulerstrongapa.c | 901 ++++++++++++++----
3 files changed, 730 insertions(+), 216 deletions(-)
diff --git a/cpukit/include/rtems/scheduler.h b/cpukit/include/rtems/scheduler.h
index 955a83cfb4..6a05c2798a 100644
--- a/cpukit/include/rtems/scheduler.h
+++ b/cpukit/include/rtems/scheduler.h
@@ -257,16 +257,14 @@
#define RTEMS_SCHEDULER_STRONG_APA( name, prio_count ) \
static struct { \
Scheduler_strong_APA_Context Base; \
- Chain_Control Ready[ ( prio_count ) ]; \
+ Scheduler_strong_APA_CPU CPU[ CONFIGURE_MAXIMUM_PROCESSORS ]; \
} SCHEDULER_STRONG_APA_CONTEXT_NAME( name )
#define RTEMS_SCHEDULER_TABLE_STRONG_APA( name, obj_name ) \
{ \
&SCHEDULER_STRONG_APA_CONTEXT_NAME( name ).Base.Base.Base, \
SCHEDULER_STRONG_APA_ENTRY_POINTS, \
- RTEMS_ARRAY_SIZE( \
- SCHEDULER_STRONG_APA_CONTEXT_NAME( name ).Ready \
- ) - 1, \
+ SCHEDULER_STRONG_APA_MAXIMUM_PRIORITY, \
( obj_name ) \
SCHEDULER_CONTROL_IS_NON_PREEMPT_MODE_SUPPORTED( false ) \
}
diff --git a/cpukit/include/rtems/score/schedulerstrongapa.h b/cpukit/include/rtems/score/schedulerstrongapa.h
index 9354acb052..cfd79e932a 100644
--- a/cpukit/include/rtems/score/schedulerstrongapa.h
+++ b/cpukit/include/rtems/score/schedulerstrongapa.h
@@ -5,9 +5,10 @@
*
* @brief Strong APA Scheduler API
*/
-
-/* SPDX-License-Identifier: BSD-2-Clause
+
+/* SPDX-License-Identifier: BSD-2-Clause
*
+ * Copyright (C) 2020 Richi Dubey
* Copyright (c) 2013, 2018 embedded brains GmbH
*
* Redistribution and use in source and binary forms, with or without
@@ -54,14 +55,14 @@ extern "C" {
* @brief Strong APA Scheduler
*
* This is an implementation of the Strong APA scheduler defined by
- * Cerqueira et al. in Linux's Processor Affinity API, Refined:
+ * Cerqueira et al. in Linux's Processor Affinity API, Refined:
* Shifting Real-Time Tasks Towards Higher Schedulability.
*
- * The scheduled and ready nodes are accessed via the
- * Scheduler_strong_APA_Context::Ready which helps in backtracking when a
+ * The scheduled and ready nodes are accessed via the
+ * Scheduler_strong_APA_Context::Ready which helps in backtracking when a
* node which is executing on a CPU gets blocked. New node is allocated to
* the cpu by checking all the executing nodes in the affinity set of the
- * node and the subsequent nodes executing on the processors in its
+ * node and the subsequent nodes executing on the processors in its
* affinity set.
* @{
*/
@@ -74,12 +75,12 @@ typedef struct {
* @brief SMP scheduler node.
*/
Scheduler_SMP_Node Base;
-
+
/**
* @brief Chain node for Scheduler_strong_APA_Context::Ready.
*/
Chain_Node Ready_node;
-
+
/**
* @brief CPU that this node would preempt in the backtracking part of
* _Scheduler_strong_APA_Get_highest_ready and
@@ -101,23 +102,23 @@ typedef struct
{
/**
* @brief CPU in a queue.
- */
+ */
Per_CPU_Control *cpu;
-
+
/**
* @brief The node that would preempt this CPU.
- */
+ */
Scheduler_Node *preempting_node;
-
+
/**
* @brief Whether or not this cpu has been added to the queue
* (visited in BFS).
- */
+ */
bool visited;
-
+
/**
- * @brief The node currently executing on this cpu
- */
+ * @brief The node currently executing on this cpu.
+ */
Scheduler_Node *executing;
} Scheduler_strong_APA_CPU;
@@ -129,13 +130,13 @@ typedef struct {
* @brief @see Scheduler_SMP_Context.
*/
Scheduler_SMP_Context Base;
-
+
/**
* @brief Chain of all the ready and scheduled nodes present in
* the Strong APA scheduler.
*/
Chain_Control Ready;
-
+
/**
* @brief Struct with important variables for each cpu.
*/
@@ -237,7 +238,7 @@ void _Scheduler_strong_APA_Update_priority(
/**
* @brief Asks for help.
*
- * @param scheduler The scheduler control instance.
+ * @param scheduler The scheduler control instance.
* @param the_thread The thread that asks for help.
* @param node The node of @a the_thread.
*
diff --git a/cpukit/score/src/schedulerstrongapa.c b/cpukit/score/src/schedulerstrongapa.c
index 924cd86412..429bbacece 100644
--- a/cpukit/score/src/schedulerstrongapa.c
+++ b/cpukit/score/src/schedulerstrongapa.c
@@ -6,18 +6,31 @@
* @brief Strong APA Scheduler Implementation
*/
-/*
- * Copyright (c) 2013, 2016 embedded brains GmbH. All rights reserved.
+/* SPDX-License-Identifier: BSD-2-Clause
+ *
+ * Copyright (C) 2020 Richi Dubey
+ * Copyright (c) 2013, 2018 embedded brains GmbH
*
- * embedded brains GmbH
- * Dornierstr. 4
- * 82178 Puchheim
- * Germany
- * <rtems at embedded-brains.de>
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
*
- * The license and distribution terms for this file may be
- * found in the file LICENSE in this distribution or at
- * http://www.rtems.org/license/LICENSE.
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
*/
#ifdef HAVE_CONFIG_H
@@ -25,301 +38,754 @@
#endif
#include <rtems/score/schedulerstrongapa.h>
-#include <rtems/score/schedulerpriorityimpl.h>
#include <rtems/score/schedulersmpimpl.h>
+#include <rtems/score/assert.h>
-static Scheduler_strong_APA_Context *_Scheduler_strong_APA_Get_self(
- Scheduler_Context *context
-)
+static inline Scheduler_strong_APA_Context *
+_Scheduler_strong_APA_Get_context( const Scheduler_Control *scheduler )
+{
+ return (Scheduler_strong_APA_Context *) _Scheduler_Get_context( scheduler );
+}
+
+static inline Scheduler_strong_APA_Context *
+_Scheduler_strong_APA_Get_self( Scheduler_Context *context )
{
return (Scheduler_strong_APA_Context *) context;
}
-static Scheduler_strong_APA_Node *
+static inline Scheduler_strong_APA_Node *
_Scheduler_strong_APA_Node_downcast( Scheduler_Node *node )
{
return (Scheduler_strong_APA_Node *) node;
}
-static void _Scheduler_strong_APA_Move_from_scheduled_to_ready(
+static inline void _Scheduler_strong_APA_Do_update(
Scheduler_Context *context,
- Scheduler_Node *scheduled_to_ready
+ Scheduler_Node *node,
+ Priority_Control new_priority
)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_self( context );
- Scheduler_strong_APA_Node *node =
- _Scheduler_strong_APA_Node_downcast( scheduled_to_ready );
-
- _Chain_Extract_unprotected( &node->Base.Base.Node.Chain );
- _Scheduler_priority_Ready_queue_enqueue_first(
- &node->Base.Base.Node.Chain,
- &node->Ready_queue,
- &self->Bit_map
- );
+ Scheduler_SMP_Node *smp_node;
+ (void) context;
+
+ smp_node = _Scheduler_SMP_Node_downcast( node );
+ _Scheduler_SMP_Node_update_priority( smp_node, new_priority );
}
-static void _Scheduler_strong_APA_Move_from_ready_to_scheduled(
+/*
+ * Returns true if the Strong APA scheduler has ready nodes
+ * available for scheduling.
+ */
+static inline bool
+ _Scheduler_strong_APA_Has_ready( Scheduler_Context *context )
+{
+ Scheduler_strong_APA_Context *self;
+ const Chain_Node *tail;
+ Chain_Node *next;
+ Scheduler_strong_APA_Node *node;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ tail = _Chain_Immutable_tail( &self->Ready );
+ next = _Chain_First( &self->Ready );
+
+ while ( next != tail ) {
+ node = (Scheduler_strong_APA_Node *) STRONG_SCHEDULER_NODE_OF_CHAIN( next );
+
+ if (
+ _Scheduler_SMP_Node_state( &node->Base.Base ) ==
+ SCHEDULER_SMP_NODE_READY
+ ) {
+ return true;
+ }
+
+ next = _Chain_Next( next );
+ }
+
+ return false;
+}
+
+static inline void _Scheduler_strong_APA_Allocate_processor(
Scheduler_Context *context,
- Scheduler_Node *ready_to_scheduled
+ Scheduler_Node *scheduled_base,
+ Scheduler_Node *victim_base,
+ Per_CPU_Control *victim_cpu
)
{
+ Scheduler_strong_APA_Node *scheduled;
Scheduler_strong_APA_Context *self;
- Scheduler_strong_APA_Node *node;
- Priority_Control insert_priority;
+ (void) victim_base;
+
+ scheduled = _Scheduler_strong_APA_Node_downcast( scheduled_base );
self = _Scheduler_strong_APA_Get_self( context );
- node = _Scheduler_strong_APA_Node_downcast( ready_to_scheduled );
- _Scheduler_priority_Ready_queue_extract(
- &node->Base.Base.Node.Chain,
- &node->Ready_queue,
- &self->Bit_map
+ self->CPU[ _Per_CPU_Get_index( victim_cpu ) ].executing = scheduled_base;
+
+ _Scheduler_SMP_Allocate_processor_exact(
+ context,
+ &(scheduled->Base.Base),
+ NULL,
+ victim_cpu
);
- insert_priority = _Scheduler_SMP_Node_priority( &node->Base.Base );
+}
+
+/*
+ * Finds and returns the highest ready node present by accessing the
+ * _Strong_APA_Context->CPU with front and rear values.
+ */
+
+static inline Scheduler_Node * _Scheduler_strong_APA_Find_highest_ready(
+ Scheduler_strong_APA_Context *self,
+ uint32_t front,
+ uint32_t rear
+)
+{
+ Scheduler_Node *highest_ready;
+ Scheduler_strong_APA_CPU *CPU;
+ const Chain_Node *tail;
+ Chain_Node *next;
+ uint32_t index_assigned_cpu;
+ Scheduler_strong_APA_Node *node;
+ Priority_Control min_priority_num;
+ Priority_Control curr_priority;
+ Per_CPU_Control *assigned_cpu;
+ Scheduler_SMP_Node_state curr_state;
+ Per_CPU_Control *curr_CPU;
+ bool first_task;
+
+ CPU = self->CPU;
+ /*
+ * When the first task accessed has nothing to compare its priority against
+ * So, it is the task with the highest priority witnessed so far.
+ */
+ first_task = true;
+
+ //Assert rear < sizeof(Context->CPU)
+ _Assert( rear < CONFIGURE_MAXIMUM_PROCESSOR );
+
+ while( front <= rear ) {
+ curr_CPU = CPU[ front ].cpu;
+ front = front + 1;
+
+ tail = _Chain_Immutable_tail( &self->Ready );
+ next = _Chain_First( &self->Ready );
+
+ while ( next != tail ) {
+ node = (Scheduler_strong_APA_Node*) STRONG_SCHEDULER_NODE_OF_CHAIN( next );
+ //Check if the curr_CPU is in the affinity set of the node.
+ if (
+ _Processor_mask_Is_set(&node->Affinity, _Per_CPU_Get_index(curr_CPU))
+ ) {
+ curr_state = _Scheduler_SMP_Node_state( &node->Base.Base );
+
+ if ( curr_state == SCHEDULER_SMP_NODE_SCHEDULED ) {
+ assigned_cpu = _Thread_Get_CPU( node->Base.Base.user );
+ index_assigned_cpu = _Per_CPU_Get_index( assigned_cpu );
+
+ if ( CPU[ index_assigned_cpu ].visited == false ) {
+ rear = rear + 1;
+ CPU[ rear ].cpu = assigned_cpu;
+ CPU[ index_assigned_cpu ].visited = true;
+ /*
+ * The curr CPU of the queue invoked this node to add its CPU
+ * that it is executing on to the queue. So this node might get
+ * preempted because of the invoker curr_CPU and this curr_CPU
+ * is the CPU that node should preempt in case this node
+ * gets preempted.
+ */
+ node->cpu_to_preempt = curr_CPU;
+ }
+ }
+ else if ( curr_state == SCHEDULER_SMP_NODE_READY ) {
+ curr_priority = _Scheduler_Node_get_priority( &node->Base.Base );
+ curr_priority = SCHEDULER_PRIORITY_PURIFY( curr_priority );
+
+ if ( first_task == true || curr_priority < min_priority_num ) {
+ min_priority_num = curr_priority;
+ highest_ready = &node->Base.Base;
+ first_task = false;
+ /*
+ * In case curr_CPU is filter_CPU, we need to store the
+ * cpu_to_preempt value so that we go back to SMP_*
+ * function, rather than preempting the node ourselves.
+ */
+ node->cpu_to_preempt = curr_CPU;
+ }
+ }
+ }
+ next = _Chain_Next( next );
+ }
+ }
+
+ return highest_ready;
+}
+
+static inline void _Scheduler_strong_APA_Move_from_ready_to_scheduled(
+ Scheduler_Context *context,
+ Scheduler_Node *ready_to_scheduled
+)
+{
+ Priority_Control insert_priority;
+
+ insert_priority = _Scheduler_SMP_Node_priority( ready_to_scheduled );
insert_priority = SCHEDULER_PRIORITY_APPEND( insert_priority );
- _Chain_Insert_ordered_unprotected(
- &self->Base.Scheduled,
- &node->Base.Base.Node.Chain,
- &insert_priority,
- _Scheduler_SMP_Priority_less_equal
+ _Scheduler_SMP_Insert_scheduled(
+ context,
+ ready_to_scheduled,
+ insert_priority
);
}
-static void _Scheduler_strong_APA_Insert_ready(
+/*
+ * Implement the BFS Algorithm for task departure to get the highest ready task
+ * for a particular CPU, returns the highest ready Scheduler_Node
+ * Scheduler_Node filter here pointst to the victim node that is blocked
+ * resulting which this function is called.
+ */
+static inline Scheduler_Node *_Scheduler_strong_APA_Get_highest_ready(
Scheduler_Context *context,
- Scheduler_Node *node_base,
- Priority_Control insert_priority
+ Scheduler_Node *filter
)
{
Scheduler_strong_APA_Context *self;
+ Per_CPU_Control *filter_cpu;
Scheduler_strong_APA_Node *node;
+ Scheduler_Node *highest_ready;
+ Scheduler_Node *curr_node;
+ Scheduler_Node *next_node;
+ Scheduler_strong_APA_CPU *CPU;
+ uint32_t front;
+ uint32_t rear;
+ uint32_t cpu_max;
+ uint32_t cpu_index;
self = _Scheduler_strong_APA_Get_self( context );
- node = _Scheduler_strong_APA_Node_downcast( node_base );
+ //Denotes front and rear of the queue.
+ front = 0;
+ rear = -1;
- if ( SCHEDULER_PRIORITY_IS_APPEND( insert_priority ) ) {
- _Scheduler_priority_Ready_queue_enqueue(
- &node->Base.Base.Node.Chain,
- &node->Ready_queue,
- &self->Bit_map
- );
- } else {
- _Scheduler_priority_Ready_queue_enqueue_first(
- &node->Base.Base.Node.Chain,
- &node->Ready_queue,
- &self->Bit_map
- );
+ filter_cpu = _Thread_Get_CPU( filter->user );
+ CPU = self->CPU;
+ cpu_max = _SMP_Get_processor_maximum();
+
+ for ( cpu_index = 0 ; cpu_index < cpu_max ; ++cpu_index ) {
+ CPU[ cpu_index ].visited = false;
}
+
+ rear = rear + 1;
+ CPU[ rear ].cpu = filter_cpu;
+ CPU[ _Per_CPU_Get_index( filter_cpu ) ].visited = true;
+
+ highest_ready = _Scheduler_strong_APA_Find_highest_ready(
+ self,
+ front,
+ rear
+ );
+
+ if ( highest_ready != filter ) {
+ /*
+ * Backtrack on the path from
+ * filter_cpu to highest_ready, shifting along every task.
+ */
+
+ node = _Scheduler_strong_APA_Node_downcast( highest_ready );
+ /*
+ * Highest ready is not just directly reachable from the victim cpu
+ * So there is need of task shifting .
+ */
+ while( node->cpu_to_preempt != filter_cpu ){
+ curr_node = &node->Base.Base;
+ next_node = CPU[ _Per_CPU_Get_index( node->cpu_to_preempt ) ].executing;
+
+ _Scheduler_SMP_Preempt(
+ context,
+ curr_node,
+ next_node,
+ _Scheduler_strong_APA_Allocate_processor
+ );
+
+ if( curr_node == highest_ready) {
+ _Scheduler_strong_APA_Move_from_ready_to_scheduled(context, curr_node);
+ }
+
+ node = _Scheduler_strong_APA_Node_downcast( next_node );
+ }
+ /*
+ * To save the last node so that the caller SMP_* function
+ * can do the allocation
+ */
+ curr_node = &node->Base.Base;
+ highest_ready = curr_node;
+ }
+
+ return highest_ready;
}
-static void _Scheduler_strong_APA_Extract_from_ready(
+/*
+ * Checks the lowest scheduled directly reachable task
+ */
+static inline Scheduler_Node *_Scheduler_strong_APA_Get_lowest_scheduled(
Scheduler_Context *context,
- Scheduler_Node *the_thread
+ Scheduler_Node *filter_base
)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_self( context );
- Scheduler_strong_APA_Node *node =
- _Scheduler_strong_APA_Node_downcast( the_thread );
-
- _Scheduler_priority_Ready_queue_extract(
- &node->Base.Base.Node.Chain,
- &node->Ready_queue,
- &self->Bit_map
- );
+ uint32_t cpu_max;
+ uint32_t cpu_index;
+ Scheduler_Node *curr_node;
+ Scheduler_Node *lowest_scheduled;
+ Priority_Control max_priority_num;
+ Priority_Control curr_priority;
+ Scheduler_strong_APA_Node *filter_strong_node;
+ Scheduler_strong_APA_Context *self;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ lowest_scheduled = NULL; // To remove compiler warning.
+ max_priority_num = 0; // Max (Lowest) priority encountered so far.
+ filter_strong_node = _Scheduler_strong_APA_Node_downcast( filter_base );
+
+ //lowest_scheduled is NULL if affinty of a node is 0
+ _Assert( !_Processor_mask_Zero( &filter_strong_node->Affinity ) );
+ cpu_max = _SMP_Get_processor_maximum();
+
+ for ( cpu_index = 0 ; cpu_index < cpu_max ; ++cpu_index ) {
+ //Checks if the CPU is in the affinity set of filter_strong_node
+ if ( _Processor_mask_Is_set( &filter_strong_node->Affinity, cpu_index) ) {
+ Per_CPU_Control *cpu = _Per_CPU_Get_by_index( cpu_index );
+
+ if ( _Per_CPU_Is_processor_online( cpu ) ) {
+ curr_node = self->CPU[ _Per_CPU_Get_index( cpu ) ].executing;
+ curr_priority = _Scheduler_Node_get_priority( curr_node );
+ curr_priority = SCHEDULER_PRIORITY_PURIFY( curr_priority );
+
+ if ( curr_priority > max_priority_num ) {
+ lowest_scheduled = curr_node;
+ max_priority_num = curr_priority;
+ }
+ }
+ }
+ }
+
+ _Assert( lowest_scheduled != NULL );
+ return lowest_scheduled;
}
-static void _Scheduler_strong_APA_Do_update(
+static inline void _Scheduler_strong_APA_Extract_from_scheduled(
Scheduler_Context *context,
- Scheduler_Node *node_to_update,
- Priority_Control new_priority
+ Scheduler_Node *node_to_extract
)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_self( context );
- Scheduler_strong_APA_Node *node =
- _Scheduler_strong_APA_Node_downcast( node_to_update );
-
- _Scheduler_SMP_Node_update_priority( &node->Base, new_priority );
- _Scheduler_priority_Ready_queue_update(
- &node->Ready_queue,
- SCHEDULER_PRIORITY_UNMAP( new_priority ),
- &self->Bit_map,
- &self->Ready[ 0 ]
- );
+ Scheduler_strong_APA_Context *self;
+ Scheduler_strong_APA_Node *node;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ node = _Scheduler_strong_APA_Node_downcast( node_to_extract );
+
+ _Scheduler_SMP_Extract_from_scheduled( &self->Base.Base, &node->Base.Base );
+ //Not removing it from Ready since the node could go in the READY state.
}
-static Scheduler_strong_APA_Context *
-_Scheduler_strong_APA_Get_context( const Scheduler_Control *scheduler )
+static inline void _Scheduler_strong_APA_Extract_from_ready(
+ Scheduler_Context *context,
+ Scheduler_Node *node_to_extract
+)
{
- return (Scheduler_strong_APA_Context *) _Scheduler_Get_context( scheduler );
+ Scheduler_strong_APA_Context *self;
+ Scheduler_strong_APA_Node *node;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ node = _Scheduler_strong_APA_Node_downcast( node_to_extract );
+
+ _Assert( !_Chain_Is_empty(self->Ready) );
+ _Assert( !_Chain_Is_node_off_chain( &node->Ready_node ) );
+
+ _Chain_Extract_unprotected( &node->Ready_node ); //Removed from Ready
+ _Chain_Set_off_chain( &node->Ready_node );
}
-void _Scheduler_strong_APA_Initialize( const Scheduler_Control *scheduler )
+static inline void _Scheduler_strong_APA_Insert_ready(
+ Scheduler_Context *context,
+ Scheduler_Node *node_base,
+ Priority_Control insert_priority
+)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_context( scheduler );
+ Scheduler_strong_APA_Context *self;
+ Scheduler_strong_APA_Node *node;
- _Scheduler_SMP_Initialize( &self->Base );
- _Priority_bit_map_Initialize( &self->Bit_map );
- _Scheduler_priority_Ready_queue_initialize(
- &self->Ready[ 0 ],
- scheduler->maximum_priority
- );
+ self = _Scheduler_strong_APA_Get_self( context );
+ node = _Scheduler_strong_APA_Node_downcast( node_base );
+
+ if( _Chain_Is_node_off_chain( &node->Ready_node ) ) {
+ _Chain_Append_unprotected( &self->Ready, &node->Ready_node );
+ }
}
-void _Scheduler_strong_APA_Node_initialize(
- const Scheduler_Control *scheduler,
- Scheduler_Node *node,
- Thread_Control *the_thread,
- Priority_Control priority
+static inline void _Scheduler_strong_APA_Move_from_scheduled_to_ready(
+ Scheduler_Context *context,
+ Scheduler_Node *scheduled_to_ready
)
{
- Scheduler_Context *context;
- Scheduler_strong_APA_Context *self;
- Scheduler_strong_APA_Node *the_node;
+ Priority_Control insert_priority;
- the_node = _Scheduler_strong_APA_Node_downcast( node );
- _Scheduler_SMP_Node_initialize(
- scheduler,
- &the_node->Base,
- the_thread,
- priority
- );
+ _Scheduler_SMP_Extract_from_scheduled( context, scheduled_to_ready );
+ insert_priority = _Scheduler_SMP_Node_priority( scheduled_to_ready );
- context = _Scheduler_Get_context( scheduler );
- self = _Scheduler_strong_APA_Get_self( context );
- _Scheduler_priority_Ready_queue_update(
- &the_node->Ready_queue,
- SCHEDULER_PRIORITY_UNMAP( priority ),
- &self->Bit_map,
- &self->Ready[ 0 ]
+ _Scheduler_strong_APA_Insert_ready(
+ context,
+ scheduled_to_ready,
+ insert_priority
);
}
-static bool _Scheduler_strong_APA_Has_ready( Scheduler_Context *context )
+static inline Scheduler_Node* _Scheduler_strong_APA_Get_lowest_reachable(
+ Scheduler_strong_APA_Context *self,
+ uint32_t front,
+ uint32_t rear,
+ Per_CPU_Control **cpu_to_preempt
+)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_self( context );
+ Scheduler_Node *lowest_reachable;
+ Priority_Control max_priority_num;
+ uint32_t cpu_max;
+ uint32_t cpu_index;
+ Thread_Control *curr_thread;
+ Per_CPU_Control *curr_CPU;
+ Priority_Control curr_priority;
+ Scheduler_Node *curr_node;
+ Scheduler_strong_APA_Node *curr_strong_node; //Current Strong_APA_Node
+ Scheduler_strong_APA_CPU *CPU;
+
+ max_priority_num = 0; //Max (Lowest) priority encountered so far.
+ CPU = self->CPU;
+ cpu_max = _SMP_Get_processor_maximum();
+
+ while( front <= rear ) {
+ curr_CPU = CPU[ front ].cpu;
+ front = front + 1;
+
+ curr_node = CPU[ _Per_CPU_Get_index( curr_CPU ) ].executing;
+ curr_thread = curr_node->user;
+
+ curr_priority = _Scheduler_Node_get_priority( curr_node );
+ curr_priority = SCHEDULER_PRIORITY_PURIFY( curr_priority );
+
+ curr_strong_node = _Scheduler_strong_APA_Node_downcast( curr_node );
+
+ if ( curr_priority > max_priority_num ) {
+ lowest_reachable = curr_node;
+ max_priority_num = curr_priority;
+ *cpu_to_preempt = curr_CPU;
+ }
+
+ if ( !curr_thread->is_idle ) {
+ for ( cpu_index = 0 ; cpu_index < cpu_max ; ++cpu_index ) {
+ if ( _Processor_mask_Is_set( &curr_strong_node->Affinity, cpu_index ) ) {
+ //Checks if the thread_CPU is in the affinity set of the node
+ Per_CPU_Control *cpu = _Per_CPU_Get_by_index( cpu_index );
+ if ( _Per_CPU_Is_processor_online( cpu ) && CPU[ cpu_index ].visited == false ) {
+ rear = rear + 1;
+ CPU[ rear ].cpu = cpu;
+ CPU[ cpu_index ].visited = true;
+ CPU[ cpu_index ].preempting_node = curr_node;
+ }
+ }
+ }
+ }
+ }
- return !_Priority_bit_map_Is_empty( &self->Bit_map );
+ return lowest_reachable;
}
-static Scheduler_Node *_Scheduler_strong_APA_Get_highest_ready(
+static inline bool _Scheduler_strong_APA_Do_enqueue(
Scheduler_Context *context,
- Scheduler_Node *node
+ Scheduler_Node *lowest_reachable,
+ Scheduler_Node *node,
+ Priority_Control insert_priority,
+ Per_CPU_Control *cpu_to_preempt
)
{
- Scheduler_strong_APA_Context *self =
- _Scheduler_strong_APA_Get_self( context );
+ bool needs_help;
+ Priority_Control node_priority;
+ Priority_Control lowest_priority;
+ Scheduler_strong_APA_CPU *CPU;
+ Scheduler_Node *curr_node;
+ Scheduler_strong_APA_Node *curr_strong_node; //Current Strong_APA_Node
+ Per_CPU_Control *curr_CPU;
+ Scheduler_strong_APA_Context *self;
+ Scheduler_Node *next_node;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ CPU = self->CPU;
+
+ node_priority = _Scheduler_Node_get_priority( node );
+ node_priority = SCHEDULER_PRIORITY_PURIFY( node_priority );
+
+ lowest_priority = _Scheduler_Node_get_priority( lowest_reachable );
+ lowest_priority = SCHEDULER_PRIORITY_PURIFY( lowest_priority );
+
+ if( lowest_priority > node_priority ) {
+ //Backtrack on the path from
+ //_Thread_Get_CPU(lowest_reachable->user) to lowest_reachable, shifting
+ //along every task
+
+ curr_node = CPU[ _Per_CPU_Get_index(cpu_to_preempt) ].preempting_node;
+ curr_strong_node = _Scheduler_strong_APA_Node_downcast( curr_node );
+ curr_strong_node->cpu_to_preempt = cpu_to_preempt;
+
+ //Save which cpu to preempt in cpu_to_preempt value of the node
+ while ( curr_node != node ) {
+ curr_CPU = _Thread_Get_CPU( curr_node->user );
+ curr_node = CPU[ _Per_CPU_Get_index( curr_CPU ) ].preempting_node;
+ curr_strong_node = _Scheduler_strong_APA_Node_downcast( curr_node );
+ curr_strong_node->cpu_to_preempt = curr_CPU;
+ }
+
+ curr_CPU = curr_strong_node->cpu_to_preempt;
+ next_node = CPU[ _Per_CPU_Get_index( curr_CPU ) ].executing;
+
+ node_priority = _Scheduler_Node_get_priority( curr_node );
+ node_priority = SCHEDULER_PRIORITY_PURIFY( node_priority );
+
+ _Scheduler_SMP_Enqueue_to_scheduled(
+ context,
+ curr_node,
+ node_priority,
+ next_node,
+ _Scheduler_SMP_Insert_scheduled,
+ _Scheduler_strong_APA_Move_from_scheduled_to_ready,
+ _Scheduler_strong_APA_Allocate_processor
+ );
- (void) node;
+ curr_node = next_node;
+ curr_strong_node = _Scheduler_strong_APA_Node_downcast( curr_node );
- return (Scheduler_Node *) _Scheduler_priority_Ready_queue_first(
- &self->Bit_map,
- &self->Ready[ 0 ]
- );
+ while( curr_node != lowest_reachable ) {
+ curr_CPU = curr_strong_node->cpu_to_preempt;
+ next_node = CPU[ _Per_CPU_Get_index( curr_CPU ) ].executing;
+ //curr_node preempts the next_node;
+ _Scheduler_SMP_Preempt(
+ context,
+ curr_node,
+ next_node,
+ _Scheduler_strong_APA_Allocate_processor
+ );
+
+ curr_node = next_node;
+ curr_strong_node = _Scheduler_strong_APA_Node_downcast( curr_node );
+ }
+
+ _Scheduler_strong_APA_Move_from_scheduled_to_ready( context, lowest_reachable );
+
+ needs_help = false;
+ } else {
+ needs_help = true;
+ }
+
+ //Add it to Ready chain since it is now either scheduled or just ready.
+ _Scheduler_strong_APA_Insert_ready(context,node,insert_priority);
+
+ return needs_help;
}
-void _Scheduler_strong_APA_Block(
- const Scheduler_Control *scheduler,
- Thread_Control *the_thread,
- Scheduler_Node *node
+/*
+ * BFS Algorithm for task arrival
+ * Enqueue node either in the scheduled chain or in the ready chain.
+ * node is the newly arrived node and is currently not scheduled.
+ */
+static inline bool _Scheduler_strong_APA_Enqueue(
+ Scheduler_Context *context,
+ Scheduler_Node *node,
+ Priority_Control insert_priority
)
{
- Scheduler_Context *context = _Scheduler_Get_context( scheduler );
+ Scheduler_strong_APA_Context *self;
+ Scheduler_strong_APA_CPU *CPU;
+ uint32_t cpu_max;
+ uint32_t cpu_index;
+ Per_CPU_Control *cpu_to_preempt;
+ Scheduler_Node *lowest_reachable;
+ Scheduler_strong_APA_Node *strong_node;
- _Scheduler_SMP_Block(
+ //Denotes front and rear of the queue
+ uint32_t front;
+ uint32_t rear;
+
+ front = 0;
+ rear = -1;
+
+ self = _Scheduler_strong_APA_Get_self( context );
+ strong_node = _Scheduler_strong_APA_Node_downcast( node );
+ cpu_max = _SMP_Get_processor_maximum();
+ CPU = self->CPU;
+
+ for ( cpu_index = 0 ; cpu_index < cpu_max ; ++cpu_index ) {
+ CPU[ cpu_index ].visited = false;
+
+ //Checks if the thread_CPU is in the affinity set of the node
+ if ( _Processor_mask_Is_set( &strong_node->Affinity, cpu_index) ) {
+ Per_CPU_Control *cpu = _Per_CPU_Get_by_index( cpu_index );
+
+ if ( _Per_CPU_Is_processor_online( cpu ) ) {
+ rear = rear + 1;
+ CPU[ rear ].cpu = cpu;
+ CPU[ cpu_index ].visited = true;
+ CPU[ cpu_index ].preempting_node = node;
+ }
+ }
+ }
+
+ //This assert makes sure that there always exist an element in the
+ // Queue when we start the queue traversal.
+ _Assert( !_Processor_mask_Zero( &strong_node->Affinity ) );
+
+ lowest_reachable = _Scheduler_strong_APA_Get_lowest_reachable(
+ self,
+ front,
+ rear,
+ &cpu_to_preempt
+ );
+
+ return _Scheduler_strong_APA_Do_enqueue(
+ context,
+ lowest_reachable,
+ node,
+ insert_priority,
+ cpu_to_preempt
+ );
+}
+
+static inline bool _Scheduler_strong_APA_Enqueue_scheduled(
+ Scheduler_Context *context,
+ Scheduler_Node *node,
+ Priority_Control insert_priority
+)
+{
+ return _Scheduler_SMP_Enqueue_scheduled(
context,
- the_thread,
node,
- _Scheduler_SMP_Extract_from_scheduled,
+ insert_priority,
+ _Scheduler_SMP_Priority_less_equal,
_Scheduler_strong_APA_Extract_from_ready,
_Scheduler_strong_APA_Get_highest_ready,
+ _Scheduler_strong_APA_Insert_ready,
+ _Scheduler_SMP_Insert_scheduled,
_Scheduler_strong_APA_Move_from_ready_to_scheduled,
- _Scheduler_SMP_Allocate_processor_exact
+ _Scheduler_strong_APA_Allocate_processor
);
}
-static bool _Scheduler_strong_APA_Enqueue(
+static inline bool _Scheduler_strong_APA_Do_ask_for_help(
Scheduler_Context *context,
- Scheduler_Node *node,
- Priority_Control insert_priority
+ Thread_Control *the_thread,
+ Scheduler_Node *node
)
{
- return _Scheduler_SMP_Enqueue(
+ return _Scheduler_SMP_Ask_for_help(
context,
+ the_thread,
node,
- insert_priority,
_Scheduler_SMP_Priority_less_equal,
_Scheduler_strong_APA_Insert_ready,
_Scheduler_SMP_Insert_scheduled,
_Scheduler_strong_APA_Move_from_scheduled_to_ready,
- _Scheduler_SMP_Get_lowest_scheduled,
- _Scheduler_SMP_Allocate_processor_exact
+ _Scheduler_strong_APA_Get_lowest_scheduled,
+ _Scheduler_strong_APA_Allocate_processor
);
}
-static bool _Scheduler_strong_APA_Enqueue_scheduled(
+static inline void _Scheduler_strong_APA_Register_idle(
Scheduler_Context *context,
- Scheduler_Node *node,
- Priority_Control insert_priority
+ Scheduler_Node *idle_base,
+ Per_CPU_Control *cpu
)
{
- return _Scheduler_SMP_Enqueue_scheduled(
+ (void) context;
+ (void) idle_base;
+ (void) cpu;
+ /*
+ * We do not maintain a variable to access the scheduled
+ * node for a CPU. So this function does nothing.
+ */
+}
+
+static inline void _Scheduler_strong_APA_Do_set_affinity(
+ Scheduler_Context *context,
+ Scheduler_Node *node_base,
+ void *arg
+)
+{
+ Scheduler_strong_APA_Node *node;
+
+ node = _Scheduler_strong_APA_Node_downcast( node_base );
+ node->Affinity = *(( const Processor_mask *) arg);
+}
+
+void _Scheduler_strong_APA_Initialize( const Scheduler_Control *scheduler )
+{
+ Scheduler_strong_APA_Context *self =
+ _Scheduler_strong_APA_Get_context( scheduler );
+
+ _Scheduler_SMP_Initialize( &self->Base );
+ _Chain_Initialize_empty( &self->Ready );
+}
+
+void _Scheduler_strong_APA_Yield(
+ const Scheduler_Control *scheduler,
+ Thread_Control *thread,
+ Scheduler_Node *node
+)
+{
+ Scheduler_Context *context = _Scheduler_Get_context( scheduler );
+
+ _Scheduler_SMP_Yield(
context,
+ thread,
node,
- insert_priority,
- _Scheduler_SMP_Priority_less_equal,
_Scheduler_strong_APA_Extract_from_ready,
- _Scheduler_strong_APA_Get_highest_ready,
- _Scheduler_strong_APA_Insert_ready,
- _Scheduler_SMP_Insert_scheduled,
- _Scheduler_strong_APA_Move_from_ready_to_scheduled,
- _Scheduler_SMP_Allocate_processor_exact
+ _Scheduler_strong_APA_Enqueue,
+ _Scheduler_strong_APA_Enqueue_scheduled
);
}
-void _Scheduler_strong_APA_Unblock(
+void _Scheduler_strong_APA_Block(
const Scheduler_Control *scheduler,
- Thread_Control *the_thread,
+ Thread_Control *thread,
Scheduler_Node *node
)
{
Scheduler_Context *context = _Scheduler_Get_context( scheduler );
-
- _Scheduler_SMP_Unblock(
+ //The extract from ready automatically removes the node from Ready chain.
+ _Scheduler_SMP_Block(
context,
- the_thread,
+ thread,
node,
- _Scheduler_strong_APA_Do_update,
- _Scheduler_strong_APA_Enqueue
+ _Scheduler_strong_APA_Extract_from_scheduled,
+ _Scheduler_strong_APA_Extract_from_ready,
+ _Scheduler_strong_APA_Get_highest_ready,
+ _Scheduler_strong_APA_Move_from_ready_to_scheduled,
+ _Scheduler_strong_APA_Allocate_processor
);
}
-static bool _Scheduler_strong_APA_Do_ask_for_help(
- Scheduler_Context *context,
- Thread_Control *the_thread,
- Scheduler_Node *node
+void _Scheduler_strong_APA_Unblock(
+ const Scheduler_Control *scheduler,
+ Thread_Control *thread,
+ Scheduler_Node *node
)
{
- return _Scheduler_SMP_Ask_for_help(
+ Scheduler_Context *context = _Scheduler_Get_context( scheduler );
+
+ _Scheduler_SMP_Unblock(
context,
- the_thread,
+ thread,
node,
- _Scheduler_SMP_Priority_less_equal,
- _Scheduler_strong_APA_Insert_ready,
- _Scheduler_SMP_Insert_scheduled,
- _Scheduler_strong_APA_Move_from_scheduled_to_ready,
- _Scheduler_SMP_Get_lowest_scheduled,
- _Scheduler_SMP_Allocate_processor_lazy
+ _Scheduler_strong_APA_Do_update,
+ _Scheduler_strong_APA_Enqueue
);
}
void _Scheduler_strong_APA_Update_priority(
const Scheduler_Control *scheduler,
- Thread_Control *the_thread,
+ Thread_Control *thread,
Scheduler_Node *node
)
{
@@ -327,7 +793,7 @@ void _Scheduler_strong_APA_Update_priority(
_Scheduler_SMP_Update_priority(
context,
- the_thread,
+ thread,
node,
_Scheduler_strong_APA_Extract_from_ready,
_Scheduler_strong_APA_Do_update,
@@ -345,7 +811,11 @@ bool _Scheduler_strong_APA_Ask_for_help(
{
Scheduler_Context *context = _Scheduler_Get_context( scheduler );
- return _Scheduler_strong_APA_Do_ask_for_help( context, the_thread, node );
+ return _Scheduler_strong_APA_Do_ask_for_help(
+ context,
+ the_thread,
+ node
+ );
}
void _Scheduler_strong_APA_Reconsider_help_request(
@@ -381,7 +851,7 @@ void _Scheduler_strong_APA_Withdraw_node(
_Scheduler_strong_APA_Extract_from_ready,
_Scheduler_strong_APA_Get_highest_ready,
_Scheduler_strong_APA_Move_from_ready_to_scheduled,
- _Scheduler_SMP_Allocate_processor_lazy
+ _Scheduler_strong_APA_Allocate_processor
);
}
@@ -416,20 +886,65 @@ Thread_Control *_Scheduler_strong_APA_Remove_processor(
);
}
-void _Scheduler_strong_APA_Yield(
+void _Scheduler_strong_APA_Node_initialize(
const Scheduler_Control *scheduler,
+ Scheduler_Node *node,
Thread_Control *the_thread,
- Scheduler_Node *node
+ Priority_Control priority
)
{
- Scheduler_Context *context = _Scheduler_Get_context( scheduler );
+ Scheduler_SMP_Node *smp_node;
+ Scheduler_strong_APA_Node *strong_node;
- _Scheduler_SMP_Yield(
- context,
- the_thread,
- node,
- _Scheduler_strong_APA_Extract_from_ready,
- _Scheduler_strong_APA_Enqueue,
- _Scheduler_strong_APA_Enqueue_scheduled
+ smp_node = _Scheduler_SMP_Node_downcast( node );
+ strong_node = _Scheduler_strong_APA_Node_downcast( node );
+
+ _Scheduler_SMP_Node_initialize( scheduler, smp_node, the_thread, priority );
+
+ _Processor_mask_Assign(
+ &strong_node->Affinity,
+ _SMP_Get_online_processors()
);
}
+
+bool _Scheduler_strong_APA_Set_affinity(
+ const Scheduler_Control *scheduler,
+ Thread_Control *thread,
+ Scheduler_Node *node_base,
+ const Processor_mask *affinity
+)
+{
+ Scheduler_Context *context;
+ Scheduler_strong_APA_Node *node;
+ Processor_mask local_affinity;
+
+ context = _Scheduler_Get_context( scheduler );
+ _Processor_mask_And( &local_affinity, &context->Processors, affinity );
+
+ if ( _Processor_mask_Is_zero( &local_affinity ) ) {
+ return false;
+ }
+
+ node = _Scheduler_strong_APA_Node_downcast( node_base );
+
+ if ( _Processor_mask_Is_equal( &node->Affinity, affinity ) )
+ return true; //Nothing to do. Return true.
+
+ _Processor_mask_Assign( &node->Affinity, &local_affinity );
+
+ _Scheduler_SMP_Set_affinity(
+ context,
+ thread,
+ node_base,
+ &local_affinity,
+ _Scheduler_strong_APA_Do_set_affinity,
+ _Scheduler_strong_APA_Extract_from_ready,
+ _Scheduler_strong_APA_Get_highest_ready,
+ _Scheduler_strong_APA_Move_from_ready_to_scheduled,
+ _Scheduler_strong_APA_Enqueue,
+ _Scheduler_strong_APA_Allocate_processor
+ );
+
+ return true;
+}
+
--
2.17.1
More information about the devel
mailing list