Subject: rcu: Make ksoftirqd do RCU quiescent states
From: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Date: Wed, 5 Oct 2011 11:45:18 -0700

Implementing RCU-bh in terms of RCU-preempt makes the system vulnerable
to network-based denial-of-service attacks.  This patch therefore
makes __do_softirq() invoke rcu_bh_qs(), but only when __do_softirq()
is running in ksoftirqd context.  A wrapper layer in interposed so that
other calls to __do_softirq() avoid invoking rcu_bh_qs().  The underlying
function __do_softirq_common() does the actual work.

The reason that rcu_bh_qs() is bad in these non-ksoftirqd contexts is
that there might be a local_bh_enable() inside an RCU-preempt read-side
critical section.  This local_bh_enable() can invoke __do_softirq()
directly, so if __do_softirq() were to invoke rcu_bh_qs() (which just
calls rcu_preempt_qs() in the PREEMPT_RT_FULL case), there would be
an illegal RCU-preempt quiescent state in the middle of an RCU-preempt
read-side critical section.  Therefore, quiescent states can only happen
in cases where __do_softirq() is invoked directly from ksoftirqd.

Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Link: http://lkml.kernel.org/r/20111005184518.GA21601@linux.vnet.ibm.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

---
 include/linux/rcupdate.h |    6 ------
 kernel/rcutree.c         |    9 ++++++++-
 kernel/rcutree_plugin.h  |    8 +++++++-
 kernel/softirq.c         |   20 +++++++++++++-------
 4 files changed, 28 insertions(+), 15 deletions(-)

--- a/include/linux/rcupdate.h
+++ b/include/linux/rcupdate.h
@@ -228,13 +228,7 @@ static inline int rcu_preempt_depth(void
 
 /* Internal to kernel */
 extern void rcu_sched_qs(int cpu);
-
-#ifndef CONFIG_PREEMPT_RT_FULL
 extern void rcu_bh_qs(int cpu);
-#else
-static inline void rcu_bh_qs(int cpu) { }
-#endif
-
 extern void rcu_check_callbacks(int cpu, int user);
 struct notifier_block;
 extern void rcu_idle_enter(void);
--- a/kernel/rcutree.c
+++ b/kernel/rcutree.c
@@ -182,7 +182,14 @@ void rcu_sched_qs(int cpu)
 	rdp->passed_quiesce = 1;
 }
 
-#ifndef CONFIG_PREEMPT_RT_FULL
+#ifdef CONFIG_PREEMPT_RT_FULL
+static void rcu_preempt_qs(int cpu);
+
+void rcu_bh_qs(int cpu)
+{
+	rcu_preempt_qs(cpu);
+}
+#else
 void rcu_bh_qs(int cpu)
 {
 	struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);
--- a/kernel/rcutree_plugin.h
+++ b/kernel/rcutree_plugin.h
@@ -1530,7 +1530,7 @@ static void __cpuinit rcu_prepare_kthrea
 
 #endif /* #else #ifdef CONFIG_RCU_BOOST */
 
-#if !defined(CONFIG_RCU_FAST_NO_HZ)
+#if !defined(CONFIG_RCU_FAST_NO_HZ) || defined(CONFIG_PREEMPT_RT_FULL)
 
 /*
  * Check to see if any future RCU-related work will need to be done
@@ -1546,6 +1546,9 @@ int rcu_needs_cpu(int cpu, unsigned long
 	*delta_jiffies = ULONG_MAX;
 	return rcu_cpu_has_callbacks(cpu, NULL);
 }
+#endif	/* !defined(CONFIG_RCU_FAST_NO_HZ) || defined(CONFIG_PREEMPT_RT_FULL) */
+
+#if !defined(CONFIG_RCU_FAST_NO_HZ)
 
 /*
  * Because we do not have RCU_FAST_NO_HZ, don't bother cleaning up
@@ -1637,6 +1640,8 @@ static bool rcu_try_advance_all_cbs(void
 	return cbs_ready;
 }
 
+#ifndef CONFIG_PREEMPT_RT_FULL
+
 /*
  * Allow the CPU to enter dyntick-idle mode unless it has callbacks ready
  * to invoke.  If the CPU has callbacks, try to advance them.  Tell the
@@ -1675,6 +1680,7 @@ int rcu_needs_cpu(int cpu, unsigned long
 	}
 	return 0;
 }
+#endif	/* #ifndef CONFIG_PREEMPT_RT_FULL */
 
 /*
  * Prepare a CPU for idle from an RCU perspective.  The first major task
--- a/kernel/softirq.c
+++ b/kernel/softirq.c
@@ -142,7 +142,7 @@ static void wakeup_softirqd(void)
 		wake_up_process(tsk);
 }
 
-static void handle_pending_softirqs(u32 pending, int cpu)
+static void handle_pending_softirqs(u32 pending, int cpu, int need_rcu_bh_qs)
 {
 	struct softirq_action *h = softirq_vec;
 	unsigned int prev_count = preempt_count();
@@ -165,7 +165,8 @@ static void handle_pending_softirqs(u32
 			       prev_count, (unsigned int) preempt_count());
 			preempt_count() = prev_count;
 		}
-		rcu_bh_qs(cpu);
+		if (need_rcu_bh_qs)
+			rcu_bh_qs(cpu);
 	}
 	local_irq_disable();
 }
@@ -331,7 +332,7 @@ restart:
 	/* Reset the pending bitmask before enabling irqs */
 	set_softirq_pending(0);
 
-	handle_pending_softirqs(pending, cpu);
+	handle_pending_softirqs(pending, cpu, 1);
 
 	pending = local_softirq_pending();
 	if (pending) {
@@ -384,7 +385,12 @@ static void ksoftirqd_clr_sched_params(u
 static DEFINE_LOCAL_IRQ_LOCK(local_softirq_lock);
 static DEFINE_PER_CPU(struct task_struct *, local_softirq_runner);
 
-static void __do_softirq(void);
+static void __do_softirq_common(int need_rcu_bh_qs);
+
+void __do_softirq(void)
+{
+	__do_softirq_common(0);
+}
 
 void __init softirq_early_init(void)
 {
@@ -455,7 +461,7 @@ EXPORT_SYMBOL(in_serving_softirq);
  * Called with bh and local interrupts disabled. For full RT cpu must
  * be pinned.
  */
-static void __do_softirq(void)
+static void __do_softirq_common(int need_rcu_bh_qs)
 {
 	u32 pending = local_softirq_pending();
 	int cpu = smp_processor_id();
@@ -469,7 +475,7 @@ static void __do_softirq(void)
 
 	lockdep_softirq_enter();
 
-	handle_pending_softirqs(pending, cpu);
+	handle_pending_softirqs(pending, cpu, need_rcu_bh_qs);
 
 	pending = local_softirq_pending();
 	if (pending)
@@ -508,7 +514,7 @@ static int __thread_do_softirq(int cpu)
 	 * schedule!
 	 */
 	if (local_softirq_pending())
-		__do_softirq();
+		__do_softirq_common(cpu >= 0);
 	local_unlock(local_softirq_lock);
 	unpin_current_cpu();
 	preempt_disable();