aboutsummaryrefslogtreecommitdiffstats
path: root/kernel
diff options
context:
space:
mode:
authorHiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>2008-03-10 14:01:20 -0400
committerIngo Molnar <mingo@elte.hu>2008-03-14 22:02:49 -0400
commit0e1f34833bd9170ccc93ab759e48e695917fa48f (patch)
tree2ca4b89dd755f2d82dbd34466a8cd6d8b5b4a0af /kernel
parent4faa8496650f9417189dacce8f933e8ec61dc032 (diff)
sched: fix race in schedule()
Fix a hard to trigger crash seen in the -rt kernel that also affects the vanilla scheduler. There is a race condition between schedule() and some dequeue/enqueue functions; rt_mutex_setprio(), __setscheduler() and sched_move_task(). When scheduling to idle, idle_balance() is called to pull tasks from other busy processor. It might drop the rq lock. It means that those 3 functions encounter on_rq=0 and running=1. The current task should be put when running. Here is a possible scenario: CPU0 CPU1 | schedule() | ->deactivate_task() | ->idle_balance() | -->load_balance_newidle() rt_mutex_setprio() | | --->double_lock_balance() *get lock *rel lock * on_rq=0, ruuning=1 | * sched_class is changed | *rel lock *get lock : | : ->put_prev_task_rt() ->pick_next_task_fair() => panic The current process of CPU1(P1) is scheduling. Deactivated P1, and the scheduler looks for another process on other CPU's runqueue because CPU1 will be idle. idle_balance(), load_balance_newidle() and double_lock_balance() are called and double_lock_balance() could drop the rq lock. On the other hand, CPU0 is trying to boost the priority of P1. The result of boosting only P1's prio and sched_class are changed to RT. The sched entities of P1 and P1's group are never put. It makes cfs_rq invalid, because the cfs_rq has curr and no leaf, but pick_next_task_fair() is called, then the kernel panics. Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com> Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl> Signed-off-by: Ingo Molnar <mingo@elte.hu>
Diffstat (limited to 'kernel')
-rw-r--r--kernel/sched.c38
1 files changed, 16 insertions, 22 deletions
diff --git a/kernel/sched.c b/kernel/sched.c
index 1cb53fb1fe3d..9df9ba73cb7a 100644
--- a/kernel/sched.c
+++ b/kernel/sched.c
@@ -4268,11 +4268,10 @@ void rt_mutex_setprio(struct task_struct *p, int prio)
4268 oldprio = p->prio; 4268 oldprio = p->prio;
4269 on_rq = p->se.on_rq; 4269 on_rq = p->se.on_rq;
4270 running = task_current(rq, p); 4270 running = task_current(rq, p);
4271 if (on_rq) { 4271 if (on_rq)
4272 dequeue_task(rq, p, 0); 4272 dequeue_task(rq, p, 0);
4273 if (running) 4273 if (running)
4274 p->sched_class->put_prev_task(rq, p); 4274 p->sched_class->put_prev_task(rq, p);
4275 }
4276 4275
4277 if (rt_prio(prio)) 4276 if (rt_prio(prio))
4278 p->sched_class = &rt_sched_class; 4277 p->sched_class = &rt_sched_class;
@@ -4281,10 +4280,9 @@ void rt_mutex_setprio(struct task_struct *p, int prio)
4281 4280
4282 p->prio = prio; 4281 p->prio = prio;
4283 4282
4283 if (running)
4284 p->sched_class->set_curr_task(rq);
4284 if (on_rq) { 4285 if (on_rq) {
4285 if (running)
4286 p->sched_class->set_curr_task(rq);
4287
4288 enqueue_task(rq, p, 0); 4286 enqueue_task(rq, p, 0);
4289 4287
4290 check_class_changed(rq, p, prev_class, oldprio, running); 4288 check_class_changed(rq, p, prev_class, oldprio, running);
@@ -4581,19 +4579,17 @@ recheck:
4581 update_rq_clock(rq); 4579 update_rq_clock(rq);
4582 on_rq = p->se.on_rq; 4580 on_rq = p->se.on_rq;
4583 running = task_current(rq, p); 4581 running = task_current(rq, p);
4584 if (on_rq) { 4582 if (on_rq)
4585 deactivate_task(rq, p, 0); 4583 deactivate_task(rq, p, 0);
4586 if (running) 4584 if (running)
4587 p->sched_class->put_prev_task(rq, p); 4585 p->sched_class->put_prev_task(rq, p);
4588 }
4589 4586
4590 oldprio = p->prio; 4587 oldprio = p->prio;
4591 __setscheduler(rq, p, policy, param->sched_priority); 4588 __setscheduler(rq, p, policy, param->sched_priority);
4592 4589
4590 if (running)
4591 p->sched_class->set_curr_task(rq);
4593 if (on_rq) { 4592 if (on_rq) {
4594 if (running)
4595 p->sched_class->set_curr_task(rq);
4596
4597 activate_task(rq, p, 0); 4593 activate_task(rq, p, 0);
4598 4594
4599 check_class_changed(rq, p, prev_class, oldprio, running); 4595 check_class_changed(rq, p, prev_class, oldprio, running);
@@ -7618,11 +7614,10 @@ void sched_move_task(struct task_struct *tsk)
7618 running = task_current(rq, tsk); 7614 running = task_current(rq, tsk);
7619 on_rq = tsk->se.on_rq; 7615 on_rq = tsk->se.on_rq;
7620 7616
7621 if (on_rq) { 7617 if (on_rq)
7622 dequeue_task(rq, tsk, 0); 7618 dequeue_task(rq, tsk, 0);
7623 if (unlikely(running)) 7619 if (unlikely(running))
7624 tsk->sched_class->put_prev_task(rq, tsk); 7620 tsk->sched_class->put_prev_task(rq, tsk);
7625 }
7626 7621
7627 set_task_rq(tsk, task_cpu(tsk)); 7622 set_task_rq(tsk, task_cpu(tsk));
7628 7623
@@ -7631,11 +7626,10 @@ void sched_move_task(struct task_struct *tsk)
7631 tsk->sched_class->moved_group(tsk); 7626 tsk->sched_class->moved_group(tsk);
7632#endif 7627#endif
7633 7628
7634 if (on_rq) { 7629 if (unlikely(running))
7635 if (unlikely(running)) 7630 tsk->sched_class->set_curr_task(rq);
7636 tsk->sched_class->set_curr_task(rq); 7631 if (on_rq)
7637 enqueue_task(rq, tsk, 0); 7632 enqueue_task(rq, tsk, 0);
7638 }
7639 7633
7640 task_rq_unlock(rq, &flags); 7634 task_rq_unlock(rq, &flags);
7641} 7635}