From 0abf17bc7790dd0467ed0e38522242f23c5da1c4 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 12 Jun 2018 10:34:51 +0200 Subject: sched/swait: Switch to full exclusive mode Linus noted that swait basically implements exclusive mode -- because swake_up() only wakes a single waiter. And because of that it should take care to properly deal with the interruptible case. In short, the problem is that swake_up() can race with a signal. In this this case it is possible the swake_up() 'wakes' the waiter that is already on the way out because it just got a signal and the wakeup gets lost. The normal wait code is very careful and avoids this situation, make sure we do too. Copy the exact exclusive semantics from wait. Suggested-by: Linus Torvalds Signed-off-by: Peter Zijlstra (Intel) Signed-off-by: Thomas Gleixner Acked-by: Linus Torvalds Cc: bigeasy@linutronix.de Cc: oleg@redhat.com Cc: paulmck@linux.vnet.ibm.com Cc: pbonzini@redhat.com Link: https://lkml.kernel.org/r/20180612083909.209762413@infradead.org --- kernel/sched/swait.c | 22 +++++++++++++++++----- 1 file changed, 17 insertions(+), 5 deletions(-) (limited to 'kernel/sched/swait.c') diff --git a/kernel/sched/swait.c b/kernel/sched/swait.c index e68bb1398b05..66890de93ee5 100644 --- a/kernel/sched/swait.c +++ b/kernel/sched/swait.c @@ -73,7 +73,7 @@ static void __prepare_to_swait(struct swait_queue_head *q, struct swait_queue *w { wait->task = current; if (list_empty(&wait->task_list)) - list_add(&wait->task_list, &q->task_list); + list_add_tail(&wait->task_list, &q->task_list); } void prepare_to_swait(struct swait_queue_head *q, struct swait_queue *wait, int state) @@ -89,12 +89,24 @@ EXPORT_SYMBOL(prepare_to_swait); long prepare_to_swait_event(struct swait_queue_head *q, struct swait_queue *wait, int state) { - if (signal_pending_state(state, current)) - return -ERESTARTSYS; + unsigned long flags; + long ret = 0; - prepare_to_swait(q, wait, state); + raw_spin_lock_irqsave(&q->lock, flags); + if (unlikely(signal_pending_state(state, current))) { + /* + * See prepare_to_wait_event(). TL;DR, subsequent swake_up() + * must not see us. + */ + list_del_init(&wait->task_list); + ret = -ERESTARTSYS; + } else { + __prepare_to_swait(q, wait); + set_current_state(state); + } + raw_spin_unlock_irqrestore(&q->lock, flags); - return 0; + return ret; } EXPORT_SYMBOL(prepare_to_swait_event); -- cgit v1.2.3