score: Fix CPU time used by executing threads

The CPU time used of a thread was previously maintained per-processor
mostly during _Thread_Dispatch().  However, on SMP configurations the
actual processor of a thread is difficult to figure out since thread
dispatching is a highly asynchronous process (e.g. via inter-processor
interrupts).  Only the intended processor of a thread is known to the
scheduler easily.  Do the CPU usage accounting during thread heir
updates in the context of the scheduler operations.  Provide the
function _Thread_Get_CPU_time_used() to get the CPU usage of a thread
using proper locks to get a consistent value.

Close #2627.

7 files changed, 99 insertions, 69 deletions

diff --git a/cpukit/score/Makefile.am b/cpukit/score/Makefile.am
index af78f15645..b6824ad446 100644
--- a/cpukit/score/Makefile.am
+++ b/cpukit/score/Makefile.am
@@ -305,6 +305,7 @@ libscore_a_SOURCES += src/thread.c src/threadchangepriority.c \
 libscore_a_SOURCES += src/threadentryadaptoridle.c
 libscore_a_SOURCES += src/threadentryadaptornumeric.c
 libscore_a_SOURCES += src/threadentryadaptorpointer.c
+libscore_a_SOURCES += src/threadgetcputimeused.c
 libscore_a_SOURCES += src/threadglobalconstruction.c
 libscore_a_SOURCES += src/threadtimeout.c
 libscore_a_SOURCES += src/threadyield.c
diff --git a/cpukit/score/include/rtems/score/percpu.h b/cpukit/score/include/rtems/score/percpu.h
index 22017889b3..95c93ddf3a 100644
--- a/cpukit/score/include/rtems/score/percpu.h
+++ b/cpukit/score/include/rtems/score/percpu.h
@@ -333,8 +333,17 @@ typedef struct Per_CPU_Control {
    */
   volatile bool dispatch_necessary;
 
-  /** This is the time of the last context switch on this CPU. */
-  Timestamp_Control time_of_last_context_switch;
+  /**
+   * @brief The CPU usage timestamp contains the time point of the last heir
+   * thread change or last CPU usage update of the executing thread of this
+   * processor.
+   *
+   * Protected by the scheduler lock.
+   *
+   * @see _Scheduler_Update_heir(), _Thread_Dispatch_update_heir() and
+   * _Thread_Get_CPU_time_used().
+   */
+  Timestamp_Control cpu_usage_timestamp;
 
   /**
    * @brief Watchdog state for this processor.
@@ -681,8 +690,6 @@ bool _Per_CPU_State_wait_for_non_initial_state(
   _Per_CPU_Get()->interrupt_stack_high
 #define _Thread_Dispatch_necessary \
   _Per_CPU_Get()->dispatch_necessary
-#define _Thread_Time_of_last_context_switch \
-  _Per_CPU_Get()->time_of_last_context_switch
 
 /**
  * @brief Returns the thread control block of the executing thread.
diff --git a/cpukit/score/include/rtems/score/schedulerimpl.h b/cpukit/score/include/rtems/score/schedulerimpl.h
index d50c36a7b9..a41b7ea62f 100644
--- a/cpukit/score/include/rtems/score/schedulerimpl.h
+++ b/cpukit/score/include/rtems/score/schedulerimpl.h
@@ -627,19 +627,6 @@ bool _Scheduler_Set_affinity(
 
 #endif /* defined(__RTEMS_HAVE_SYS_CPUSET_H__) */
 
-RTEMS_INLINE_ROUTINE void _Scheduler_Update_heir(
-  Thread_Control *new_heir,
-  bool            force_dispatch
-)
-{
-  Thread_Control *heir = _Thread_Heir;
-
-  if ( heir != new_heir && ( heir->is_preemptible || force_dispatch ) ) {
-    _Thread_Heir = new_heir;
-    _Thread_Dispatch_necessary = true;
-  }
-}
-
 RTEMS_INLINE_ROUTINE void _Scheduler_Generic_block(
   const Scheduler_Control *scheduler,
   Thread_Control          *the_thread,
@@ -1356,6 +1343,25 @@ RTEMS_INLINE_ROUTINE bool _Scheduler_Ask_blocked_node_for_help(
 
 ISR_LOCK_DECLARE( extern, _Scheduler_Lock )
 
+RTEMS_INLINE_ROUTINE void _Scheduler_Update_heir(
+  Thread_Control *new_heir,
+  bool            force_dispatch
+)
+{
+  Thread_Control *heir = _Thread_Heir;
+
+  if ( heir != new_heir && ( heir->is_preemptible || force_dispatch ) ) {
+#if defined(RTEMS_SMP)
+    /* We need this state only for _Thread_Get_CPU_time_used() */
+    _Scheduler_Thread_change_state( heir, THREAD_SCHEDULER_BLOCKED );
+    _Scheduler_Thread_change_state( new_heir, THREAD_SCHEDULER_SCHEDULED );
+#endif
+    _Thread_Update_CPU_time_used( heir, _Thread_Get_CPU( heir ) );
+    _Thread_Heir = new_heir;
+    _Thread_Dispatch_necessary = true;
+  }
+}
+
 /**
  * @brief Acquires the scheduler instance of the thread.
  *
diff --git a/cpukit/score/include/rtems/score/threadimpl.h b/cpukit/score/include/rtems/score/threadimpl.h
index ec9851dba0..3f8df856d5 100644
--- a/cpukit/score/include/rtems/score/threadimpl.h
+++ b/cpukit/score/include/rtems/score/threadimpl.h
@@ -578,36 +578,6 @@ RTEMS_INLINE_ROUTINE bool _Thread_Is_executing_on_a_processor(
 #endif
 
 /**
- * @brief Returns @a true and sets time_of_context_switch to the
- * time of the last context switch when the thread is currently executing
- * in the system, otherwise @a false.
- */
-RTEMS_INLINE_ROUTINE bool _Thread_Get_time_of_last_context_switch(
-  Thread_Control    *the_thread,
-  Timestamp_Control *time_of_context_switch
-)
-{
-  bool retval = false;
-
-  _Thread_Disable_dispatch();
-  #ifndef RTEMS_SMP
-    if ( _Thread_Executing->Object.id == the_thread->Object.id ) {
-      *time_of_context_switch = _Thread_Time_of_last_context_switch;
-      retval = true;
-    }
-  #else
-    if ( _Thread_Is_executing_on_a_processor( the_thread ) ) {
-      *time_of_context_switch =
-        _Thread_Get_CPU( the_thread )->time_of_last_context_switch;
-      retval = true;
-    }
-  #endif
-  _Thread_Enable_dispatch();
-  return retval;
-}
-
-
-/**
  * This function returns true if the_thread is the heir
  * thread, and false otherwise.
  */
@@ -803,6 +773,20 @@ RTEMS_INLINE_ROUTINE Thread_Control *_Thread_Get_heir_and_make_it_executing(
   return heir;
 }
 
+RTEMS_INLINE_ROUTINE void _Thread_Update_CPU_time_used(
+  Thread_Control  *the_thread,
+  Per_CPU_Control *cpu
+)
+{
+  Timestamp_Control last;
+  Timestamp_Control ran;
+
+  last = cpu->cpu_usage_timestamp;
+  _TOD_Get_uptime( &cpu->cpu_usage_timestamp );
+  _Timestamp_Subtract( &last, &cpu->cpu_usage_timestamp, &ran );
+  _Timestamp_Add_to( &the_thread->cpu_time_used, &ran );
+}
+
 #if defined( RTEMS_SMP )
 RTEMS_INLINE_ROUTINE void _Thread_Dispatch_update_heir(
   Per_CPU_Control *cpu_self,
@@ -810,6 +794,8 @@ RTEMS_INLINE_ROUTINE void _Thread_Dispatch_update_heir(
   Thread_Control  *heir
 )
 {
+  _Thread_Update_CPU_time_used( cpu_for_heir->heir, cpu_for_heir );
+
   cpu_for_heir->heir = heir;
 
   if ( cpu_for_heir == cpu_self ) {
@@ -820,23 +806,10 @@ RTEMS_INLINE_ROUTINE void _Thread_Dispatch_update_heir(
 }
 #endif
 
-RTEMS_INLINE_ROUTINE void _Thread_Update_cpu_time_used(
-  Thread_Control *executing,
-  Timestamp_Control *time_of_last_context_switch
-)
-{
-  Timestamp_Control uptime;
-  Timestamp_Control ran;
-
-  _TOD_Get_uptime( &uptime );
-  _Timestamp_Subtract(
-    time_of_last_context_switch,
-    &uptime,
-    &ran
-  );
-  *time_of_last_context_switch = uptime;
-  _Timestamp_Add_to( &executing->cpu_time_used, &ran );
-}
+void _Thread_Get_CPU_time_used(
+  Thread_Control    *the_thread,
+  Timestamp_Control *cpu_time_used
+);
 
 RTEMS_INLINE_ROUTINE void _Thread_Action_control_initialize(
   Thread_Action_control *action_control
diff --git a/cpukit/score/src/schedulersmpstartidle.c b/cpukit/score/src/schedulersmpstartidle.c
index de125d3323..0ffb628550 100644
--- a/cpukit/score/src/schedulersmpstartidle.c
+++ b/cpukit/score/src/schedulersmpstartidle.c
@@ -22,6 +22,7 @@ void _Scheduler_SMP_Start_idle(
   Scheduler_SMP_Context *self = _Scheduler_SMP_Get_self( context );
   Scheduler_SMP_Node *node = _Scheduler_SMP_Thread_get_node( thread );
 
+  _Scheduler_Thread_change_state( thread, THREAD_SCHEDULER_SCHEDULED );
   node->state = SCHEDULER_SMP_NODE_SCHEDULED;
 
   _Thread_Set_CPU( thread, cpu );
diff --git a/cpukit/score/src/threaddispatch.c b/cpukit/score/src/threaddispatch.c
index 1b36c93fbb..ce5d118a22 100644
--- a/cpukit/score/src/threaddispatch.c
+++ b/cpukit/score/src/threaddispatch.c
@@ -108,11 +108,6 @@ void _Thread_Do_dispatch( Per_CPU_Control *cpu_self, ISR_Level level )
     _ISR_Enable( level );
 #endif
 
-    _Thread_Update_cpu_time_used(
-      executing,
-      &cpu_self->time_of_last_context_switch
-    );
-
     _User_extensions_Thread_switch( executing, heir );
     _Thread_Save_fp( executing );
     _Context_Switch( &executing->Registers, &heir->Registers );
diff --git a/cpukit/score/src/threadgetcputimeused.c b/cpukit/score/src/threadgetcputimeused.c
new file mode 100644
index 0000000000..6bfe8ea216
--- /dev/null
+++ b/cpukit/score/src/threadgetcputimeused.c
@@ -0,0 +1,47 @@
+/*
+ * Copyright (c) 2016 embedded brains GmbH.  All rights reserved.
+ *
+ *  embedded brains GmbH
+ *  Dornierstr. 4
+ *  82178 Puchheim
+ *  Germany
+ *  <rtems@embedded-brains.de>
+ *
+ * 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.
+ */
+
+#if HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <rtems/score/threadimpl.h>
+#include <rtems/score/schedulerimpl.h>
+
+static bool _Thread_Is_scheduled( const Thread_Control *the_thread )
+{
+#if defined(RTEMS_SMP)
+  return the_thread->Scheduler.state == THREAD_SCHEDULER_SCHEDULED;
+#else
+  return _Thread_Is_executing( the_thread );
+#endif
+}
+
+void _Thread_Get_CPU_time_used(
+  Thread_Control    *the_thread,
+  Timestamp_Control *cpu_time_used
+)
+{
+  ISR_lock_Context lock_context;
+
+  _Scheduler_Acquire( the_thread, &lock_context );
+
+  if ( _Thread_Is_scheduled( the_thread ) ) {
+    _Thread_Update_CPU_time_used( the_thread, _Thread_Get_CPU( the_thread ) );
+  }
+
+  *cpu_time_used = the_thread->cpu_time_used;
+
+  _Scheduler_Release( the_thread, &lock_context );
+}