sched/task: It is not appropriate for logic in task_exit() to call the new version of this_task(). sched/irq: Remove redundant fetch of CPU index; configs/sabre-6qguad: update README.

configs/sabre-6quad/README.txt

@@ -127,21 +127,28 @@ Status
   deferred setting g_cpu_irqlock().  That latter setting is now deferred
   until sched_resume_scheduler() runs.  These commits were made:
 
-    commit de34b4523fc33c6f2f20619349af8fa081a3bfcd
-      sched/ and arch/arm/src/armv7-a:  Replace a few more occurrences
-      of this_task() with current_task(cpu) in an effort to get the i.MX6
-      working in SMP mode again.  It does not yet work, sadly.
+     commit 0ba78530164814360eb09ed9805137b934c6f03b
+      sched/irq: Fix a infinite recursion problem that a recent change
+      introduced into the i.MX6 SMP implementation.
 
     commit 8aa15385060bf705bbca2c22a5682128740e55a8
       arch/arm/src/armv7-a:  Found some additional places were the new
       this_task() function cannot be called in the i.MX6 SMP configuration.
 
-     commit 0ba78530164814360eb09ed9805137b934c6f03b
-      sched/irq: Fix a infinite recursion problem that a recent change
-      introduced into the i.MX6 SMP implementation.
+    commit de34b4523fc33c6f2f20619349af8fa081a3bfcd
+      sched/ and arch/arm/src/armv7-a:  Replace a few more occurrences
+      of this_task() with current_task(cpu) in an effort to get the i.MX6
+      working in SMP mode again.  It does not yet work, sadly.
 
-  With these changes, basic SMP functionality is restored.  Insufficient
-  stress testing has been done to prove that the solution is stable, however.
+    commit cce21bef3292a40dcd97b6176ea016e2b559de8b
+      sched/sched: sched_lock() and sched_unlock().. back out some changes
+      I made recently.  The seemed correct but apparently not.  Also
+      reorder to logic so that g_global_lockcount is incremented for the very
+      minimum amount of time.
+
+  With these changes, basic SMP functionality is restored and there are no
+  known issues.  Insufficient stress testing has been done to prove that the
+  solution is stable, however.
 
 Platform Features
 =================

sched/irq/irq_csection.c

@@ -342,7 +342,6 @@ try_again:
                * the spinlock.
                */
 
-              cpu = this_cpu();
               DEBUGASSERT((g_cpu_irqset & (1 << cpu)) == 0);
 
               if (!irq_waitlock(cpu))

sched/task/task_exit.c

@@ -1,7 +1,8 @@
 /****************************************************************************
  * sched/task/task_exit.c
  *
- *   Copyright (C) 2008-2009, 2012-2014, 2016 Gregory Nutt. All rights reserved.
+ *   Copyright (C) 2008-2009, 2012-2014, 2016, 2018 Gregory Nutt. All
+ *     rights reserved.
  *   Author: Gregory Nutt <gnutt@nuttx.org>
  *
  * Redistribution and use in source and binary forms, with or without
@@ -82,9 +83,24 @@
 
 int task_exit(void)
 {
-  FAR struct tcb_s *dtcb = this_task();
+  FAR struct tcb_s *dtcb;
   FAR struct tcb_s *rtcb;
   int ret;
+#ifdef CONFIG_SMP
+  int cpu;
+
+  /* Get the current CPU.  By assumption, we are within a critical section
+   * and, hence, the CPU index will remain stable.
+   *
+   * Avoid using this_task() because it may assume a state that is not
+   * appropriate for an exiting task.
+   */
+
+  cpu  = this_cpu();
+  dtcb = current_task(cpu);
+#else
+  dtcb = this_task();
+#endif
 
   /* Remove the TCB of the current task from the ready-to-run list.  A context
    * switch will definitely be necessary -- that must be done by the
@@ -95,7 +111,14 @@ int task_exit(void)
    */
 
   (void)sched_removereadytorun(dtcb);
+
+  /* Get the new task at the head of the ready to run list */
+
+#ifdef CONFIG_SMP
+  rtcb = current_task(cpu);
+#else
   rtcb = this_task();
+#endif
 
 #ifdef CONFIG_SMP
   /* Because clearing the global IRQ control in sched_removereadytorun()