[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <5538A0DF.50401@siemens.com>
Date: Thu, 23 Apr 2015 09:35:59 +0200
From: Jan Kiszka <jan.kiszka@...mens.com>
To: Sebastian Andrzej Siewior <bigeasy@...utronix.de>
CC: RT <linux-rt-users@...r.kernel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
Steven Rostedt <rostedt@...dmis.org>,
Mike Galbraith <umgwanakikbuti@...il.com>
Subject: [PATCH v2 RT 3.18] irq_work: Provide a soft-irq based queue
Instead of turning all irq_work requests into lazy ones on -rt, just
move their execution from hard into soft-irq context.
This resolves deadlocks of ftrace which will queue work from arbitrary
contexts, including those that have locks held that are needed for
raising a soft-irq.
Signed-off-by: Jan Kiszka <jan.kiszka@...mens.com>
---
Changes in v2:
- fix execution of raised list (discovered by Mike Galbraith)
- added comment of irq_work_run (derived from Mike's suggestion)
kernel/irq_work.c | 34 +++++++++++++++++-----------------
1 file changed, 17 insertions(+), 17 deletions(-)
diff --git a/kernel/irq_work.c b/kernel/irq_work.c
index 9dda38a..171dfac 100644
--- a/kernel/irq_work.c
+++ b/kernel/irq_work.c
@@ -85,12 +85,9 @@ bool irq_work_queue_on(struct irq_work *work, int cpu)
raise_irqwork = llist_add(&work->llnode,
&per_cpu(hirq_work_list, cpu));
else
- raise_irqwork = llist_add(&work->llnode,
- &per_cpu(lazy_list, cpu));
-#else
+#endif
raise_irqwork = llist_add(&work->llnode,
&per_cpu(raised_list, cpu));
-#endif
if (raise_irqwork)
arch_send_call_function_single_ipi(cpu);
@@ -114,21 +111,20 @@ bool irq_work_queue(struct irq_work *work)
if (work->flags & IRQ_WORK_HARD_IRQ) {
if (llist_add(&work->llnode, this_cpu_ptr(&hirq_work_list)))
arch_irq_work_raise();
- } else {
+ } else
+#endif
+ if (work->flags & IRQ_WORK_LAZY) {
if (llist_add(&work->llnode, this_cpu_ptr(&lazy_list)) &&
tick_nohz_tick_stopped())
+#ifdef CONFIG_PREEMPT_RT_FULL
raise_softirq(TIMER_SOFTIRQ);
- }
#else
- if (work->flags & IRQ_WORK_LAZY) {
- if (llist_add(&work->llnode, this_cpu_ptr(&lazy_list)) &&
- tick_nohz_tick_stopped())
arch_irq_work_raise();
+#endif
} else {
if (llist_add(&work->llnode, this_cpu_ptr(&raised_list)))
arch_irq_work_raise();
}
-#endif
preempt_enable();
@@ -202,6 +198,13 @@ void irq_work_run(void)
{
#ifdef CONFIG_PREEMPT_RT_FULL
irq_work_run_list(this_cpu_ptr(&hirq_work_list));
+ /*
+ * NOTE: we raise softirq via IPI for safety (caller may hold locks
+ * that raise_softirq needs) and execute in irq_work_tick() to move
+ * the overhead from hard to soft irq context.
+ */
+ if (!llist_empty(this_cpu_ptr(&raised_list)))
+ raise_softirq(TIMER_SOFTIRQ);
#else
irq_work_run_list(this_cpu_ptr(&raised_list));
irq_work_run_list(this_cpu_ptr(&lazy_list));
@@ -211,15 +214,12 @@ EXPORT_SYMBOL_GPL(irq_work_run);
void irq_work_tick(void)
{
-#ifdef CONFIG_PREEMPT_RT_FULL
- irq_work_run_list(this_cpu_ptr(&lazy_list));
-#else
- struct llist_head *raised = &__get_cpu_var(raised_list);
+ struct llist_head *raised = this_cpu_ptr(&raised_list);
- if (!llist_empty(raised) && !arch_irq_work_has_interrupt())
+ if (!llist_empty(raised) && (!arch_irq_work_has_interrupt() ||
+ IS_ENABLED(CONFIG_PREEMPT_RT_FULL)))
irq_work_run_list(raised);
- irq_work_run_list(&__get_cpu_var(lazy_list));
-#endif
+ irq_work_run_list(this_cpu_ptr(&lazy_list));
}
/*
--
2.1.4
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists