[PATCH v2 01/11] sched: Improve try_invoke_on_locked_down_task()

[Peter Zijlstra]

Clarify and tighten try_invoke_on_locked_down_task().

Basically the function calls @func under task_rq_lock(), except it
avoids taking rq->lock when possible.

This makes calling @func unconditional (the function will get renamed
in a later patch to remove the try).

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 kernel/sched/core.c |   63 ++++++++++++++++++++++++++++++++--------------------
 1 file changed, 39 insertions(+), 24 deletions(-)

--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4111,41 +4111,56 @@ try_to_wake_up(struct task_struct *p, un
  * @func: Function to invoke.
  * @arg: Argument to function.
  *
- * If the specified task can be quickly locked into a definite state
- * (either sleeping or on a given runqueue), arrange to keep it in that
- * state while invoking @func(@arg).  This function can use ->on_rq and
- * task_curr() to work out what the state is, if required.  Given that
- * @func can be invoked with a runqueue lock held, it had better be quite
- * lightweight.
+ * Fix the task in it's current state by avoiding wakeups and or rq operations
+ * and call @func(@arg) on it.  This function can use ->on_rq and task_curr()
+ * to work out what the state is, if required.  Given that @func can be invoked
+ * with a runqueue lock held, it had better be quite lightweight.
  *
  * Returns:
- *	@false if the task slipped out from under the locks.
- *	@true if the task was locked onto a runqueue or is sleeping.
- *		However, @func can override this by returning @false.
+ *   Whatever @func returns
  */
 bool try_invoke_on_locked_down_task(struct task_struct *p, bool (*func)(struct task_struct *t, void *arg), void *arg)
 {
+	struct rq *rq = NULL;
+	unsigned int state;
 	struct rq_flags rf;
 	bool ret = false;
-	struct rq *rq;
 
 	raw_spin_lock_irqsave(&p->pi_lock, rf.flags);
-	if (p->on_rq) {
+
+	state = READ_ONCE(p->__state);
+
+	/*
+	 * Ensure we load p->on_rq after p->__state, otherwise it would be
+	 * possible to, falsely, observe p->on_rq == 0.
+	 *
+	 * See try_to_wake_up() for a longer comment.
+	 */
+	smp_rmb();
+
+	/*
+	 * Since pi->lock blocks try_to_wake_up(), we don't need rq->lock when
+	 * the task is blocked. Make sure to check @state since ttwu() can drop
+	 * locks at the end, see ttwu_queue_wakelist().
+	 */
+	if (state == TASK_RUNNING || state == TASK_WAKING || p->on_rq)
 		rq = __task_rq_lock(p, &rf);
-		if (task_rq(p) == rq)
-			ret = func(p, arg);
+
+	/*
+	 * At this point the task is pinned; either:
+	 *  - blocked and we're holding off wakeups	 (pi->lock)
+	 *  - woken, and we're holding off enqueue	 (rq->lock)
+	 *  - queued, and we're holding off schedule	 (rq->lock)
+	 *  - running, and we're holding off de-schedule (rq->lock)
+	 *
+	 * The called function (@func) can use: task_curr(), p->on_rq and
+	 * p->__state to differentiate between these states.
+	 */
+	ret = func(p, arg);
+
+	if (rq)
 		rq_unlock(rq, &rf);
-	} else {
-		switch (READ_ONCE(p->__state)) {
-		case TASK_RUNNING:
-		case TASK_WAKING:
-			break;
-		default:
-			smp_rmb(); // See smp_rmb() comment in try_to_wake_up().
-			if (!p->on_rq)
-				ret = func(p, arg);
-		}
-	}
+
 	raw_spin_unlock_irqrestore(&p->pi_lock, rf.flags);
 	return ret;
 }