---

yaml
---
r: 323917
b: refs/heads/master
c: edde96e
h: refs/heads/master
i:
  323915: 9011be0
v: v3

[refs]

@@ -1,2 +1,2 @@
 ---
-refs/heads/master: 532b1858c5241bedfff5ab863d7cf012e8b81a6b
+refs/heads/master: edde96eafc91a510f404e7b82cfc0ecb608505ee

trunk/kernel/sched/core.c

@@ -3367,6 +3367,40 @@ pick_next_task(struct rq *rq)
 
 /*
  * __schedule() is the main scheduler function.
+ *
+ * The main means of driving the scheduler and thus entering this function are:
+ *
+ *   1. Explicit blocking: mutex, semaphore, waitqueue, etc.
+ *
+ *   2. TIF_NEED_RESCHED flag is checked on interrupt and userspace return
+ *      paths. For example, see arch/x86/entry_64.S.
+ *
+ *      To drive preemption between tasks, the scheduler sets the flag in timer
+ *      interrupt handler scheduler_tick().
+ *
+ *   3. Wakeups don't really cause entry into schedule(). They add a
+ *      task to the run-queue and that's it.
+ *
+ *      Now, if the new task added to the run-queue preempts the current
+ *      task, then the wakeup sets TIF_NEED_RESCHED and schedule() gets
+ *      called on the nearest possible occasion:
+ *
+ *       - If the kernel is preemptible (CONFIG_PREEMPT=y):
+ *
+ *         - in syscall or exception context, at the next outmost
+ *           preempt_enable(). (this might be as soon as the wake_up()'s
+ *           spin_unlock()!)
+ *
+ *         - in IRQ context, return from interrupt-handler to
+ *           preemptible context
+ *
+ *       - If the kernel is not preemptible (CONFIG_PREEMPT is not set)
+ *         then at the next:
+ *
+ *          - cond_resched() call
+ *          - explicit schedule() call
+ *          - return from syscall or exception to user-space
+ *          - return from interrupt-handler to user-space
  */
 static void __sched __schedule(void)
 {