Re: 4.2-rc5 rcu stalls.

[Sasha Levin <sasha.levin@oracle.com>]

[-- Attachment #1: Type: text/plain, Size: 223 bytes --]

On 08/03/2015 06:03 PM, Paul E. McKenney wrote:
>> > Ugh, that doesn't revert cleanly.  Got something handy ?
> I do not, but perhaps either Sasha or Frederic do.

I've attached a revert courtesy of Peter.


Thanks,
Sasha


[-- Attachment #2: revert.patch --]
[-- Type: text/x-patch, Size: 3961 bytes --]

 include/linux/preempt.h | 12 ------------
 kernel/sched/core.c     | 34 +++++++++++++++++++---------------
 2 files changed, 19 insertions(+), 27 deletions(-)

diff --git a/include/linux/preempt.h b/include/linux/preempt.h
index 84991f185173..3a93d4cdcce9 100644
--- a/include/linux/preempt.h
+++ b/include/linux/preempt.h
@@ -137,18 +137,6 @@ extern void preempt_count_sub(int val);
 #define preempt_count_inc() preempt_count_add(1)
 #define preempt_count_dec() preempt_count_sub(1)
 
-#define preempt_active_enter() \
-do { \
-	preempt_count_add(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET); \
-	barrier(); \
-} while (0)
-
-#define preempt_active_exit() \
-do { \
-	barrier(); \
-	preempt_count_sub(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET); \
-} while (0)
-
 #ifdef CONFIG_PREEMPT_COUNT
 
 #define preempt_disable() \
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index 78b4bad10081..bd378bd21a0e 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -2983,7 +2983,9 @@ pick_next_task(struct rq *rq, struct task_struct *prev)
  *          - return from syscall or exception to user-space
  *          - return from interrupt-handler to user-space
  *
- * WARNING: must be called with preemption disabled!
+ * WARNING: all callers must re-check need_resched() afterward and reschedule
+ * accordingly in case an event triggered the need for rescheduling (such as
+ * an interrupt waking up a task) while preemption was disabled in __schedule().
  */
 static void __sched __schedule(void)
 {
@@ -2992,6 +2994,7 @@ static void __sched __schedule(void)
 	struct rq *rq;
 	int cpu;
 
+	preempt_disable();
 	cpu = smp_processor_id();
 	rq = cpu_rq(cpu);
 	rcu_note_context_switch();
@@ -3058,6 +3061,8 @@ static void __sched __schedule(void)
 	}
 
 	balance_callback(rq);
+
+	sched_preempt_enable_no_resched();
 }
 
 static inline void sched_submit_work(struct task_struct *tsk)
@@ -3078,9 +3083,7 @@ asmlinkage __visible void __sched schedule(void)
 
 	sched_submit_work(tsk);
 	do {
-		preempt_disable();
 		__schedule();
-		sched_preempt_enable_no_resched();
 	} while (need_resched());
 }
 EXPORT_SYMBOL(schedule);
@@ -3119,14 +3122,15 @@ void __sched schedule_preempt_disabled(void)
 static void __sched notrace preempt_schedule_common(void)
 {
 	do {
-		preempt_active_enter();
+		__preempt_count_add(PREEMPT_ACTIVE);
 		__schedule();
-		preempt_active_exit();
+		__preempt_count_sub(PREEMPT_ACTIVE);
 
 		/*
 		 * Check again in case we missed a preemption opportunity
 		 * between schedule and now.
 		 */
+		barrier();
 	} while (need_resched());
 }
 
@@ -3172,13 +3176,7 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
 		return;
 
 	do {
-		/*
-		 * Use raw __prempt_count() ops that don't call function.
-		 * We can't call functions before disabling preemption which
-		 * disarm preemption tracing recursions.
-		 */
-		__preempt_count_add(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET);
-		barrier();
+		__preempt_count_add(PREEMPT_ACTIVE);
 		/*
 		 * Needs preempt disabled in case user_exit() is traced
 		 * and the tracer calls preempt_enable_notrace() causing
@@ -3188,8 +3186,8 @@ asmlinkage __visible void __sched notrace preempt_schedule_notrace(void)
 		__schedule();
 		exception_exit(prev_ctx);
 
+		__preempt_count_sub(PREEMPT_ACTIVE);
 		barrier();
-		__preempt_count_sub(PREEMPT_ACTIVE + PREEMPT_DISABLE_OFFSET);
 	} while (need_resched());
 }
 EXPORT_SYMBOL_GPL(preempt_schedule_notrace);
@@ -3212,11 +3210,17 @@ asmlinkage __visible void __sched preempt_schedule_irq(void)
 	prev_state = exception_enter();
 
 	do {
-		preempt_active_enter();
+		__preempt_count_add(PREEMPT_ACTIVE);
 		local_irq_enable();
 		__schedule();
 		local_irq_disable();
-		preempt_active_exit();
+		__preempt_count_sub(PREEMPT_ACTIVE);
+
+		/*
+		 * Check again in case we missed a preemption opportunity
+		 * between schedule and now.
+		 */
+		barrier();
 	} while (need_resched());
 
 	exception_exit(prev_state);