*/
#include <linux/delay.h>
-#include <linux/stop_machine.h>
+
+#define RCU_KTHREAD_PRIO 1
+
+#ifdef CONFIG_RCU_BOOST
+#define RCU_BOOST_PRIO CONFIG_RCU_BOOST_PRIO
+#else
+#define RCU_BOOST_PRIO RCU_KTHREAD_PRIO
+#endif
/*
* Check the RCU kernel configuration parameters and print informative
printk(KERN_INFO "\tRCU torture testing starts during boot.\n");
#endif
#if defined(CONFIG_TREE_PREEMPT_RCU) && !defined(CONFIG_RCU_CPU_STALL_VERBOSE)
- printk(KERN_INFO "\tVerbose stalled-CPUs detection is disabled.\n");
+ printk(KERN_INFO "\tDump stacks of tasks blocking RCU-preempt GP.\n");
+#endif
+#if defined(CONFIG_RCU_CPU_STALL_INFO)
+ printk(KERN_INFO "\tAdditional per-CPU info printed with stalls.\n");
#endif
#if NUM_RCU_LVL_4 != 0
printk(KERN_INFO "\tExperimental four-level hierarchy is enabled.\n");
{
struct rcu_data *rdp = &per_cpu(rcu_preempt_data, cpu);
- rdp->passed_quiesc_completed = rdp->gpnum - 1;
+ rdp->passed_quiesce_gpnum = rdp->gpnum;
barrier();
- rdp->passed_quiesc = 1;
+ if (rdp->passed_quiesce == 0)
+ trace_rcu_grace_period("rcu_preempt", rdp->gpnum, "cpuqs");
+ rdp->passed_quiesce = 1;
current->rcu_read_unlock_special &= ~RCU_READ_UNLOCK_NEED_QS;
}
if (rnp->qsmask & rdp->grpmask)
rnp->gp_tasks = &t->rcu_node_entry;
}
+ trace_rcu_preempt_task(rdp->rsp->name,
+ t->pid,
+ (rnp->qsmask & rdp->grpmask)
+ ? rnp->gpnum
+ : rnp->gpnum + 1);
raw_spin_unlock_irqrestore(&rnp->lock, flags);
} else if (t->rcu_read_lock_nesting < 0 &&
t->rcu_read_unlock_special) {
{
int empty;
int empty_exp;
+ int empty_exp_now;
unsigned long flags;
struct list_head *np;
+#ifdef CONFIG_RCU_BOOST
+ struct rt_mutex *rbmp = NULL;
+#endif /* #ifdef CONFIG_RCU_BOOST */
struct rcu_node *rnp;
int special;
smp_mb(); /* ensure expedited fastpath sees end of RCU c-s. */
np = rcu_next_node_entry(t, rnp);
list_del_init(&t->rcu_node_entry);
+ t->rcu_blocked_node = NULL;
+ trace_rcu_unlock_preempted_task("rcu_preempt",
+ rnp->gpnum, t->pid);
if (&t->rcu_node_entry == rnp->gp_tasks)
rnp->gp_tasks = np;
if (&t->rcu_node_entry == rnp->exp_tasks)
#ifdef CONFIG_RCU_BOOST
if (&t->rcu_node_entry == rnp->boost_tasks)
rnp->boost_tasks = np;
- /* Snapshot and clear ->rcu_boosted with rcu_node lock held. */
- if (t->rcu_boosted) {
- special |= RCU_READ_UNLOCK_BOOSTED;
- t->rcu_boosted = 0;
+ /* Snapshot/clear ->rcu_boost_mutex with rcu_node lock held. */
+ if (t->rcu_boost_mutex) {
+ rbmp = t->rcu_boost_mutex;
+ t->rcu_boost_mutex = NULL;
}
#endif /* #ifdef CONFIG_RCU_BOOST */
- t->rcu_blocked_node = NULL;
/*
* If this was the last task on the current list, and if
* we aren't waiting on any CPUs, report the quiescent state.
- * Note that rcu_report_unblock_qs_rnp() releases rnp->lock.
+ * Note that rcu_report_unblock_qs_rnp() releases rnp->lock,
+ * so we must take a snapshot of the expedited state.
*/
- if (empty)
- raw_spin_unlock_irqrestore(&rnp->lock, flags);
- else
+ empty_exp_now = !rcu_preempted_readers_exp(rnp);
+ if (!empty && !rcu_preempt_blocked_readers_cgp(rnp)) {
+ trace_rcu_quiescent_state_report("preempt_rcu",
+ rnp->gpnum,
+ 0, rnp->qsmask,
+ rnp->level,
+ rnp->grplo,
+ rnp->grphi,
+ !!rnp->gp_tasks);
rcu_report_unblock_qs_rnp(rnp, flags);
+ } else
+ raw_spin_unlock_irqrestore(&rnp->lock, flags);
#ifdef CONFIG_RCU_BOOST
/* Unboost if we were boosted. */
- if (special & RCU_READ_UNLOCK_BOOSTED) {
- rt_mutex_unlock(t->rcu_boost_mutex);
- t->rcu_boost_mutex = NULL;
- }
+ if (rbmp)
+ rt_mutex_unlock(rbmp);
#endif /* #ifdef CONFIG_RCU_BOOST */
/*
* If this was the last task on the expedited lists,
* then we need to report up the rcu_node hierarchy.
*/
- if (!empty_exp && !rcu_preempted_readers_exp(rnp))
- rcu_report_exp_rnp(&rcu_preempt_state, rnp);
+ if (!empty_exp && empty_exp_now)
+ rcu_report_exp_rnp(&rcu_preempt_state, rnp, true);
} else {
local_irq_restore(flags);
}
{
struct task_struct *t = current;
- barrier(); /* needed if we ever invoke rcu_read_unlock in rcutree.c */
if (t->rcu_read_lock_nesting != 1)
--t->rcu_read_lock_nesting;
else {
+ barrier(); /* critical section before exit code. */
t->rcu_read_lock_nesting = INT_MIN;
barrier(); /* assign before ->rcu_read_unlock_special load */
if (unlikely(ACCESS_ONCE(t->rcu_read_unlock_special)))
#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_VERBOSE */
+#ifdef CONFIG_RCU_CPU_STALL_INFO
+
+static void rcu_print_task_stall_begin(struct rcu_node *rnp)
+{
+ printk(KERN_ERR "\tTasks blocked on level-%d rcu_node (CPUs %d-%d):",
+ rnp->level, rnp->grplo, rnp->grphi);
+}
+
+static void rcu_print_task_stall_end(void)
+{
+ printk(KERN_CONT "\n");
+}
+
+#else /* #ifdef CONFIG_RCU_CPU_STALL_INFO */
+
+static void rcu_print_task_stall_begin(struct rcu_node *rnp)
+{
+}
+
+static void rcu_print_task_stall_end(void)
+{
+}
+
+#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_INFO */
+
/*
* Scan the current list of tasks blocked within RCU read-side critical
* sections, printing out the tid of each.
*/
-static void rcu_print_task_stall(struct rcu_node *rnp)
+static int rcu_print_task_stall(struct rcu_node *rnp)
{
struct task_struct *t;
+ int ndetected = 0;
if (!rcu_preempt_blocked_readers_cgp(rnp))
- return;
+ return 0;
+ rcu_print_task_stall_begin(rnp);
t = list_entry(rnp->gp_tasks,
struct task_struct, rcu_node_entry);
- list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry)
- printk(" P%d", t->pid);
+ list_for_each_entry_continue(t, &rnp->blkd_tasks, rcu_node_entry) {
+ printk(KERN_CONT " P%d", t->pid);
+ ndetected++;
+ }
+ rcu_print_task_stall_end();
+ return ndetected;
}
/*
* absolutely necessary, but this is a good performance/complexity
* tradeoff.
*/
- if (rcu_preempt_blocked_readers_cgp(rnp))
+ if (rcu_preempt_blocked_readers_cgp(rnp) && rnp->qsmask == 0)
retval |= RCU_OFL_TASKS_NORM_GP;
if (rcu_preempted_readers_exp(rnp))
retval |= RCU_OFL_TASKS_EXP_GP;
return retval;
}
+#endif /* #ifdef CONFIG_HOTPLUG_CPU */
+
/*
* Do CPU-offline processing for preemptible RCU.
*/
-static void rcu_preempt_offline_cpu(int cpu)
+static void rcu_preempt_cleanup_dead_cpu(int cpu)
{
- __rcu_offline_cpu(cpu, &rcu_preempt_state);
+ rcu_cleanup_dead_cpu(cpu, &rcu_preempt_state);
}
-#endif /* #ifdef CONFIG_HOTPLUG_CPU */
-
/*
* Check for a quiescent state from the current CPU. When a task blocks,
* the task is recorded in the corresponding CPU's rcu_node structure,
*/
void call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
{
- __call_rcu(head, func, &rcu_preempt_state);
+ __call_rcu(head, func, &rcu_preempt_state, 0);
}
EXPORT_SYMBOL_GPL(call_rcu);
+/*
+ * Queue an RCU callback for lazy invocation after a grace period.
+ * This will likely be later named something like "call_rcu_lazy()",
+ * but this change will require some way of tagging the lazy RCU
+ * callbacks in the list of pending callbacks. Until then, this
+ * function may only be called from __kfree_rcu().
+ */
+void kfree_call_rcu(struct rcu_head *head,
+ void (*func)(struct rcu_head *rcu))
+{
+ __call_rcu(head, func, &rcu_preempt_state, 1);
+}
+EXPORT_SYMBOL_GPL(kfree_call_rcu);
+
/**
* synchronize_rcu - wait until a grace period has elapsed.
*
*/
void synchronize_rcu(void)
{
+ rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&
+ !lock_is_held(&rcu_lock_map) &&
+ !lock_is_held(&rcu_sched_lock_map),
+ "Illegal synchronize_rcu() in RCU read-side critical section");
if (!rcu_scheduler_active)
return;
wait_rcu_gp(call_rcu);
* recursively up the tree. (Calm down, calm down, we do the recursion
* iteratively!)
*
+ * Most callers will set the "wake" flag, but the task initiating the
+ * expedited grace period need not wake itself.
+ *
* Caller must hold sync_rcu_preempt_exp_mutex.
*/
-static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp)
+static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
+ bool wake)
{
unsigned long flags;
unsigned long mask;
}
if (rnp->parent == NULL) {
raw_spin_unlock_irqrestore(&rnp->lock, flags);
- wake_up(&sync_rcu_preempt_exp_wq);
+ if (wake)
+ wake_up(&sync_rcu_preempt_exp_wq);
break;
}
mask = rnp->grpmask;
must_wait = 1;
}
if (!must_wait)
- rcu_report_exp_rnp(rsp, rnp);
+ rcu_report_exp_rnp(rsp, rnp, false); /* Don't wake self. */
}
-/*
- * Wait for an rcu-preempt grace period, but expedite it. The basic idea
- * is to invoke synchronize_sched_expedited() to push all the tasks to
- * the ->blkd_tasks lists and wait for this list to drain.
+/**
+ * synchronize_rcu_expedited - Brute-force RCU grace period
+ *
+ * Wait for an RCU-preempt grace period, but expedite it. The basic
+ * idea is to invoke synchronize_sched_expedited() to push all the tasks to
+ * the ->blkd_tasks lists and wait for this list to drain. This consumes
+ * significant time on all CPUs and is unfriendly to real-time workloads,
+ * so is thus not recommended for any sort of common-case code.
+ * In fact, if you are using synchronize_rcu_expedited() in a loop,
+ * please restructure your code to batch your updates, and then Use a
+ * single synchronize_rcu() instead.
+ *
+ * Note that it is illegal to call this function while holding any lock
+ * that is acquired by a CPU-hotplug notifier. And yes, it is also illegal
+ * to call this function from a CPU-hotplug notifier. Failing to observe
+ * these restriction will result in deadlock.
*/
void synchronize_rcu_expedited(void)
{
}
/*
- * Does preemptible RCU need the CPU to stay out of dynticks mode?
+ * Does preemptible RCU have callbacks on this CPU?
*/
-static int rcu_preempt_needs_cpu(int cpu)
+static int rcu_preempt_cpu_has_callbacks(int cpu)
{
return !!per_cpu(rcu_preempt_data, cpu).nxtlist;
}
}
/*
- * Move preemptible RCU's callbacks from dying CPU to other online CPU.
+ * Move preemptible RCU's callbacks from dying CPU to other online CPU
+ * and record a quiescent state.
*/
-static void rcu_preempt_send_cbs_to_online(void)
+static void rcu_preempt_cleanup_dying_cpu(void)
{
- rcu_send_cbs_to_online(&rcu_preempt_state);
+ rcu_cleanup_dying_cpu(&rcu_preempt_state);
}
/*
* Because preemptible RCU does not exist, we never have to check for
* tasks blocked within RCU read-side critical sections.
*/
-static void rcu_print_task_stall(struct rcu_node *rnp)
+static int rcu_print_task_stall(struct rcu_node *rnp)
{
+ return 0;
}
/*
return 0;
}
+#endif /* #ifdef CONFIG_HOTPLUG_CPU */
+
/*
* Because preemptible RCU does not exist, it never needs CPU-offline
* processing.
*/
-static void rcu_preempt_offline_cpu(int cpu)
+static void rcu_preempt_cleanup_dead_cpu(int cpu)
{
}
-#endif /* #ifdef CONFIG_HOTPLUG_CPU */
-
/*
* Because preemptible RCU does not exist, it never has any callbacks
* to check.
}
/*
+ * Queue an RCU callback for lazy invocation after a grace period.
+ * This will likely be later named something like "call_rcu_lazy()",
+ * but this change will require some way of tagging the lazy RCU
+ * callbacks in the list of pending callbacks. Until then, this
+ * function may only be called from __kfree_rcu().
+ *
+ * Because there is no preemptible RCU, we use RCU-sched instead.
+ */
+void kfree_call_rcu(struct rcu_head *head,
+ void (*func)(struct rcu_head *rcu))
+{
+ __call_rcu(head, func, &rcu_sched_state, 1);
+}
+EXPORT_SYMBOL_GPL(kfree_call_rcu);
+
+/*
* Wait for an rcu-preempt grace period, but make it happen quickly.
* But because preemptible RCU does not exist, map to rcu-sched.
*/
* report on tasks preempted in RCU read-side critical sections during
* expedited RCU grace periods.
*/
-static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp)
+static void rcu_report_exp_rnp(struct rcu_state *rsp, struct rcu_node *rnp,
+ bool wake)
{
- return;
}
#endif /* #ifdef CONFIG_HOTPLUG_CPU */
}
/*
- * Because preemptible RCU does not exist, it never needs any CPU.
+ * Because preemptible RCU does not exist, it never has callbacks
*/
-static int rcu_preempt_needs_cpu(int cpu)
+static int rcu_preempt_cpu_has_callbacks(int cpu)
{
return 0;
}
}
/*
- * Because there is no preemptible RCU, there are no callbacks to move.
+ * Because there is no preemptible RCU, there is no cleanup to do.
*/
-static void rcu_preempt_send_cbs_to_online(void)
+static void rcu_preempt_cleanup_dying_cpu(void)
{
}
t = container_of(tb, struct task_struct, rcu_node_entry);
rt_mutex_init_proxy_locked(&mtx, t);
t->rcu_boost_mutex = &mtx;
- t->rcu_boosted = 1;
raw_spin_unlock_irqrestore(&rnp->lock, flags);
rt_mutex_lock(&mtx); /* Side effect: boosts task t's priority. */
rt_mutex_unlock(&mtx); /* Keep lockdep happy. */
- return rnp->exp_tasks != NULL || rnp->boost_tasks != NULL;
+ return ACCESS_ONCE(rnp->exp_tasks) != NULL ||
+ ACCESS_ONCE(rnp->boost_tasks) != NULL;
}
/*
int spincnt = 0;
int more2boost;
+ trace_rcu_utilization("Start boost kthread@init");
for (;;) {
rnp->boost_kthread_status = RCU_KTHREAD_WAITING;
+ trace_rcu_utilization("End boost kthread@rcu_wait");
rcu_wait(rnp->boost_tasks || rnp->exp_tasks);
+ trace_rcu_utilization("Start boost kthread@rcu_wait");
rnp->boost_kthread_status = RCU_KTHREAD_RUNNING;
more2boost = rcu_boost(rnp);
if (more2boost)
else
spincnt = 0;
if (spincnt > 10) {
+ trace_rcu_utilization("End boost kthread@rcu_yield");
rcu_yield(rcu_boost_kthread_timer, (unsigned long)rnp);
+ trace_rcu_utilization("Start boost kthread@rcu_yield");
spincnt = 0;
}
}
/* NOTREACHED */
+ trace_rcu_utilization("End boost kthread@notreached");
return 0;
}
}
/*
+ * Is the current CPU running the RCU-callbacks kthread?
+ * Caller must have preemption disabled.
+ */
+static bool rcu_is_callbacks_kthread(void)
+{
+ return __get_cpu_var(rcu_cpu_kthread_task) == current;
+}
+
+/*
* Set the affinity of the boost kthread. The CPU-hotplug locks are
* held, so no one should be messing with the existence of the boost
* kthread.
if (rnp->boost_kthread_task != NULL)
return 0;
t = kthread_create(rcu_boost_kthread, (void *)rnp,
- "rcub%d", rnp_index);
+ "rcub/%d", rnp_index);
if (IS_ERR(t))
return PTR_ERR(t);
raw_spin_lock_irqsave(&rnp->lock, flags);
rnp->boost_kthread_task = t;
raw_spin_unlock_irqrestore(&rnp->lock, flags);
- sp.sched_priority = RCU_KTHREAD_PRIO;
+ sp.sched_priority = RCU_BOOST_PRIO;
sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
wake_up_process(t); /* get to TASK_INTERRUPTIBLE quickly. */
return 0;
{
struct sched_param sp;
struct timer_list yield_timer;
+ int prio = current->rt_priority;
setup_timer_on_stack(&yield_timer, f, arg);
mod_timer(&yield_timer, jiffies + 2);
sched_setscheduler_nocheck(current, SCHED_NORMAL, &sp);
set_user_nice(current, 19);
schedule();
- sp.sched_priority = RCU_KTHREAD_PRIO;
+ set_user_nice(current, 0);
+ sp.sched_priority = prio;
sched_setscheduler_nocheck(current, SCHED_FIFO, &sp);
del_timer(&yield_timer);
}
/*
* Per-CPU kernel thread that invokes RCU callbacks. This replaces the
- * earlier RCU softirq.
+ * RCU softirq used in flavors and configurations of RCU that do not
+ * support RCU priority boosting.
*/
static int rcu_cpu_kthread(void *arg)
{
char work;
char *workp = &per_cpu(rcu_cpu_has_work, cpu);
+ trace_rcu_utilization("Start CPU kthread@init");
for (;;) {
*statusp = RCU_KTHREAD_WAITING;
+ trace_rcu_utilization("End CPU kthread@rcu_wait");
rcu_wait(*workp != 0 || kthread_should_stop());
+ trace_rcu_utilization("Start CPU kthread@rcu_wait");
local_bh_disable();
if (rcu_cpu_kthread_should_stop(cpu)) {
local_bh_enable();
spincnt = 0;
if (spincnt > 10) {
*statusp = RCU_KTHREAD_YIELDING;
+ trace_rcu_utilization("End CPU kthread@rcu_yield");
rcu_yield(rcu_cpu_kthread_timer, (unsigned long)cpu);
+ trace_rcu_utilization("Start CPU kthread@rcu_yield");
spincnt = 0;
}
}
*statusp = RCU_KTHREAD_STOPPED;
+ trace_rcu_utilization("End CPU kthread@term");
return 0;
}
t = kthread_create_on_node(rcu_cpu_kthread,
(void *)(long)cpu,
cpu_to_node(cpu),
- "rcuc%d", cpu);
+ "rcuc/%d", cpu);
if (IS_ERR(t))
return PTR_ERR(t);
if (cpu_online(cpu))
return 0;
if (rnp->node_kthread_task == NULL) {
t = kthread_create(rcu_node_kthread, (void *)rnp,
- "rcun%d", rnp_index);
+ "rcun/%d", rnp_index);
if (IS_ERR(t))
return PTR_ERR(t);
raw_spin_lock_irqsave(&rnp->lock, flags);
WARN_ON_ONCE(1);
}
+static bool rcu_is_callbacks_kthread(void)
+{
+ return false;
+}
+
static void rcu_preempt_boost_start_gp(struct rcu_node *rnp)
{
}
#endif /* #else #ifdef CONFIG_RCU_BOOST */
-#ifndef CONFIG_SMP
+#if !defined(CONFIG_RCU_FAST_NO_HZ)
-void synchronize_sched_expedited(void)
+/*
+ * Check to see if any future RCU-related work will need to be done
+ * by the current CPU, even if none need be done immediately, returning
+ * 1 if so. This function is part of the RCU implementation; it is -not-
+ * an exported member of the RCU API.
+ *
+ * Because we not have RCU_FAST_NO_HZ, just check whether this CPU needs
+ * any flavor of RCU.
+ */
+int rcu_needs_cpu(int cpu)
{
- cond_resched();
+ return rcu_cpu_has_callbacks(cpu);
}
-EXPORT_SYMBOL_GPL(synchronize_sched_expedited);
-#else /* #ifndef CONFIG_SMP */
+/*
+ * Because we do not have RCU_FAST_NO_HZ, don't bother initializing for it.
+ */
+static void rcu_prepare_for_idle_init(int cpu)
+{
+}
-static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0);
-static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0);
+/*
+ * Because we do not have RCU_FAST_NO_HZ, don't bother cleaning up
+ * after it.
+ */
+static void rcu_cleanup_after_idle(int cpu)
+{
+}
-static int synchronize_sched_expedited_cpu_stop(void *data)
+/*
+ * Do the idle-entry grace-period work, which, because CONFIG_RCU_FAST_NO_HZ=n,
+ * is nothing.
+ */
+static void rcu_prepare_for_idle(int cpu)
{
- /*
- * There must be a full memory barrier on each affected CPU
- * between the time that try_stop_cpus() is called and the
- * time that it returns.
- *
- * In the current initial implementation of cpu_stop, the
- * above condition is already met when the control reaches
- * this point and the following smp_mb() is not strictly
- * necessary. Do smp_mb() anyway for documentation and
- * robustness against future implementation changes.
- */
- smp_mb(); /* See above comment block. */
- return 0;
}
+#else /* #if !defined(CONFIG_RCU_FAST_NO_HZ) */
+
/*
- * Wait for an rcu-sched grace period to elapse, but use "big hammer"
- * approach to force grace period to end quickly. This consumes
- * significant time on all CPUs, and is thus not recommended for
- * any sort of common-case code.
+ * This code is invoked when a CPU goes idle, at which point we want
+ * to have the CPU do everything required for RCU so that it can enter
+ * the energy-efficient dyntick-idle mode. This is handled by a
+ * state machine implemented by rcu_prepare_for_idle() below.
*
- * Note that it is illegal to call this function while holding any
- * lock that is acquired by a CPU-hotplug notifier. Failing to
- * observe this restriction will result in deadlock.
+ * The following three proprocessor symbols control this state machine:
*
- * This implementation can be thought of as an application of ticket
- * locking to RCU, with sync_sched_expedited_started and
- * sync_sched_expedited_done taking on the roles of the halves
- * of the ticket-lock word. Each task atomically increments
- * sync_sched_expedited_started upon entry, snapshotting the old value,
- * then attempts to stop all the CPUs. If this succeeds, then each
- * CPU will have executed a context switch, resulting in an RCU-sched
- * grace period. We are then done, so we use atomic_cmpxchg() to
- * update sync_sched_expedited_done to match our snapshot -- but
- * only if someone else has not already advanced past our snapshot.
+ * RCU_IDLE_FLUSHES gives the maximum number of times that we will attempt
+ * to satisfy RCU. Beyond this point, it is better to incur a periodic
+ * scheduling-clock interrupt than to loop through the state machine
+ * at full power.
+ * RCU_IDLE_OPT_FLUSHES gives the number of RCU_IDLE_FLUSHES that are
+ * optional if RCU does not need anything immediately from this
+ * CPU, even if this CPU still has RCU callbacks queued. The first
+ * times through the state machine are mandatory: we need to give
+ * the state machine a chance to communicate a quiescent state
+ * to the RCU core.
+ * RCU_IDLE_GP_DELAY gives the number of jiffies that a CPU is permitted
+ * to sleep in dyntick-idle mode with RCU callbacks pending. This
+ * is sized to be roughly one RCU grace period. Those energy-efficiency
+ * benchmarkers who might otherwise be tempted to set this to a large
+ * number, be warned: Setting RCU_IDLE_GP_DELAY too high can hang your
+ * system. And if you are -that- concerned about energy efficiency,
+ * just power the system down and be done with it!
+ * RCU_IDLE_LAZY_GP_DELAY gives the number of jiffies that a CPU is
+ * permitted to sleep in dyntick-idle mode with only lazy RCU
+ * callbacks pending. Setting this too high can OOM your system.
*
- * On the other hand, if try_stop_cpus() fails, we check the value
- * of sync_sched_expedited_done. If it has advanced past our
- * initial snapshot, then someone else must have forced a grace period
- * some time after we took our snapshot. In this case, our work is
- * done for us, and we can simply return. Otherwise, we try again,
- * but keep our initial snapshot for purposes of checking for someone
- * doing our work for us.
- *
- * If we fail too many times in a row, we fall back to synchronize_sched().
+ * The values below work well in practice. If future workloads require
+ * adjustment, they can be converted into kernel config parameters, though
+ * making the state machine smarter might be a better option.
*/
-void synchronize_sched_expedited(void)
-{
- int firstsnap, s, snap, trycount = 0;
+#define RCU_IDLE_FLUSHES 5 /* Number of dyntick-idle tries. */
+#define RCU_IDLE_OPT_FLUSHES 3 /* Optional dyntick-idle tries. */
+#define RCU_IDLE_GP_DELAY 6 /* Roughly one grace period. */
+#define RCU_IDLE_LAZY_GP_DELAY (6 * HZ) /* Roughly six seconds. */
- /* Note that atomic_inc_return() implies full memory barrier. */
- firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started);
- get_online_cpus();
-
- /*
- * Each pass through the following loop attempts to force a
- * context switch on each CPU.
- */
- while (try_stop_cpus(cpu_online_mask,
- synchronize_sched_expedited_cpu_stop,
- NULL) == -EAGAIN) {
- put_online_cpus();
+static DEFINE_PER_CPU(int, rcu_dyntick_drain);
+static DEFINE_PER_CPU(unsigned long, rcu_dyntick_holdoff);
+static DEFINE_PER_CPU(struct hrtimer, rcu_idle_gp_timer);
+static ktime_t rcu_idle_gp_wait; /* If some non-lazy callbacks. */
+static ktime_t rcu_idle_lazy_gp_wait; /* If only lazy callbacks. */
- /* No joy, try again later. Or just synchronize_sched(). */
- if (trycount++ < 10)
- udelay(trycount * num_online_cpus());
- else {
- synchronize_sched();
- return;
- }
+/*
+ * Allow the CPU to enter dyntick-idle mode if either: (1) There are no
+ * callbacks on this CPU, (2) this CPU has not yet attempted to enter
+ * dyntick-idle mode, or (3) this CPU is in the process of attempting to
+ * enter dyntick-idle mode. Otherwise, if we have recently tried and failed
+ * to enter dyntick-idle mode, we refuse to try to enter it. After all,
+ * it is better to incur scheduling-clock interrupts than to spin
+ * continuously for the same time duration!
+ */
+int rcu_needs_cpu(int cpu)
+{
+ /* If no callbacks, RCU doesn't need the CPU. */
+ if (!rcu_cpu_has_callbacks(cpu))
+ return 0;
+ /* Otherwise, RCU needs the CPU only if it recently tried and failed. */
+ return per_cpu(rcu_dyntick_holdoff, cpu) == jiffies;
+}
- /* Check to see if someone else did our work for us. */
- s = atomic_read(&sync_sched_expedited_done);
- if (UINT_CMP_GE((unsigned)s, (unsigned)firstsnap)) {
- smp_mb(); /* ensure test happens before caller kfree */
- return;
- }
+/*
+ * Does the specified flavor of RCU have non-lazy callbacks pending on
+ * the specified CPU? Both RCU flavor and CPU are specified by the
+ * rcu_data structure.
+ */
+static bool __rcu_cpu_has_nonlazy_callbacks(struct rcu_data *rdp)
+{
+ return rdp->qlen != rdp->qlen_lazy;
+}
- /*
- * Refetching sync_sched_expedited_started allows later
- * callers to piggyback on our grace period. We subtract
- * 1 to get the same token that the last incrementer got.
- * We retry after they started, so our grace period works
- * for them, and they started after our first try, so their
- * grace period works for us.
- */
- get_online_cpus();
- snap = atomic_read(&sync_sched_expedited_started) - 1;
- smp_mb(); /* ensure read is before try_stop_cpus(). */
- }
+#ifdef CONFIG_TREE_PREEMPT_RCU
- /*
- * Everyone up to our most recent fetch is covered by our grace
- * period. Update the counter, but only if our work is still
- * relevant -- which it won't be if someone who started later
- * than we did beat us to the punch.
- */
- do {
- s = atomic_read(&sync_sched_expedited_done);
- if (UINT_CMP_GE((unsigned)s, (unsigned)snap)) {
- smp_mb(); /* ensure test happens before caller kfree */
- break;
- }
- } while (atomic_cmpxchg(&sync_sched_expedited_done, s, snap) != s);
+/*
+ * Are there non-lazy RCU-preempt callbacks? (There cannot be if there
+ * is no RCU-preempt in the kernel.)
+ */
+static bool rcu_preempt_cpu_has_nonlazy_callbacks(int cpu)
+{
+ struct rcu_data *rdp = &per_cpu(rcu_preempt_data, cpu);
- put_online_cpus();
+ return __rcu_cpu_has_nonlazy_callbacks(rdp);
}
-EXPORT_SYMBOL_GPL(synchronize_sched_expedited);
-#endif /* #else #ifndef CONFIG_SMP */
+#else /* #ifdef CONFIG_TREE_PREEMPT_RCU */
-#if !defined(CONFIG_RCU_FAST_NO_HZ)
+static bool rcu_preempt_cpu_has_nonlazy_callbacks(int cpu)
+{
+ return 0;
+}
+
+#endif /* else #ifdef CONFIG_TREE_PREEMPT_RCU */
/*
- * Check to see if any future RCU-related work will need to be done
- * by the current CPU, even if none need be done immediately, returning
- * 1 if so. This function is part of the RCU implementation; it is -not-
- * an exported member of the RCU API.
- *
- * Because we have preemptible RCU, just check whether this CPU needs
- * any flavor of RCU. Do not chew up lots of CPU cycles with preemption
- * disabled in a most-likely vain attempt to cause RCU not to need this CPU.
+ * Does any flavor of RCU have non-lazy callbacks on the specified CPU?
*/
-int rcu_needs_cpu(int cpu)
+static bool rcu_cpu_has_nonlazy_callbacks(int cpu)
{
- return rcu_needs_cpu_quick_check(cpu);
+ return __rcu_cpu_has_nonlazy_callbacks(&per_cpu(rcu_sched_data, cpu)) ||
+ __rcu_cpu_has_nonlazy_callbacks(&per_cpu(rcu_bh_data, cpu)) ||
+ rcu_preempt_cpu_has_nonlazy_callbacks(cpu);
}
/*
- * Check to see if we need to continue a callback-flush operations to
- * allow the last CPU to enter dyntick-idle mode. But fast dyntick-idle
- * entry is not configured, so we never do need to.
+ * Timer handler used to force CPU to start pushing its remaining RCU
+ * callbacks in the case where it entered dyntick-idle mode with callbacks
+ * pending. The hander doesn't really need to do anything because the
+ * real work is done upon re-entry to idle, or by the next scheduling-clock
+ * interrupt should idle not be re-entered.
*/
-static void rcu_needs_cpu_flush(void)
+static enum hrtimer_restart rcu_idle_gp_timer_func(struct hrtimer *hrtp)
{
+ trace_rcu_prep_idle("Timer");
+ return HRTIMER_NORESTART;
}
-#else /* #if !defined(CONFIG_RCU_FAST_NO_HZ) */
+/*
+ * Initialize the timer used to pull CPUs out of dyntick-idle mode.
+ */
+static void rcu_prepare_for_idle_init(int cpu)
+{
+ static int firsttime = 1;
+ struct hrtimer *hrtp = &per_cpu(rcu_idle_gp_timer, cpu);
-#define RCU_NEEDS_CPU_FLUSHES 5
-static DEFINE_PER_CPU(int, rcu_dyntick_drain);
-static DEFINE_PER_CPU(unsigned long, rcu_dyntick_holdoff);
+ hrtimer_init(hrtp, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
+ hrtp->function = rcu_idle_gp_timer_func;
+ if (firsttime) {
+ unsigned int upj = jiffies_to_usecs(RCU_IDLE_GP_DELAY);
+
+ rcu_idle_gp_wait = ns_to_ktime(upj * (u64)1000);
+ upj = jiffies_to_usecs(RCU_IDLE_LAZY_GP_DELAY);
+ rcu_idle_lazy_gp_wait = ns_to_ktime(upj * (u64)1000);
+ firsttime = 0;
+ }
+}
/*
- * Check to see if any future RCU-related work will need to be done
- * by the current CPU, even if none need be done immediately, returning
- * 1 if so. This function is part of the RCU implementation; it is -not-
- * an exported member of the RCU API.
+ * Clean up for exit from idle. Because we are exiting from idle, there
+ * is no longer any point to rcu_idle_gp_timer, so cancel it. This will
+ * do nothing if this timer is not active, so just cancel it unconditionally.
+ */
+static void rcu_cleanup_after_idle(int cpu)
+{
+ hrtimer_cancel(&per_cpu(rcu_idle_gp_timer, cpu));
+}
+
+/*
+ * Check to see if any RCU-related work can be done by the current CPU,
+ * and if so, schedule a softirq to get it done. This function is part
+ * of the RCU implementation; it is -not- an exported member of the RCU API.
*
- * Because we are not supporting preemptible RCU, attempt to accelerate
- * any current grace periods so that RCU no longer needs this CPU, but
- * only if all other CPUs are already in dynticks-idle mode. This will
- * allow the CPU cores to be powered down immediately, as opposed to after
- * waiting many milliseconds for grace periods to elapse.
+ * The idea is for the current CPU to clear out all work required by the
+ * RCU core for the current grace period, so that this CPU can be permitted
+ * to enter dyntick-idle mode. In some cases, it will need to be awakened
+ * at the end of the grace period by whatever CPU ends the grace period.
+ * This allows CPUs to go dyntick-idle more quickly, and to reduce the
+ * number of wakeups by a modest integer factor.
*
* Because it is not legal to invoke rcu_process_callbacks() with irqs
* disabled, we do one pass of force_quiescent_state(), then do a
* invoke_rcu_core() to cause rcu_process_callbacks() to be invoked
* later. The per-cpu rcu_dyntick_drain variable controls the sequencing.
+ *
+ * The caller must have disabled interrupts.
*/
-int rcu_needs_cpu(int cpu)
+static void rcu_prepare_for_idle(int cpu)
{
- int c = 0;
- int snap;
- int thatcpu;
-
- /* Check for being in the holdoff period. */
- if (per_cpu(rcu_dyntick_holdoff, cpu) == jiffies)
- return rcu_needs_cpu_quick_check(cpu);
-
- /* Don't bother unless we are the last non-dyntick-idle CPU. */
- for_each_online_cpu(thatcpu) {
- if (thatcpu == cpu)
- continue;
- snap = atomic_add_return(0, &per_cpu(rcu_dynticks,
- thatcpu).dynticks);
- smp_mb(); /* Order sampling of snap with end of grace period. */
- if ((snap & 0x1) != 0) {
- per_cpu(rcu_dyntick_drain, cpu) = 0;
- per_cpu(rcu_dyntick_holdoff, cpu) = jiffies - 1;
- return rcu_needs_cpu_quick_check(cpu);
- }
+ /*
+ * If there are no callbacks on this CPU, enter dyntick-idle mode.
+ * Also reset state to avoid prejudicing later attempts.
+ */
+ if (!rcu_cpu_has_callbacks(cpu)) {
+ per_cpu(rcu_dyntick_holdoff, cpu) = jiffies - 1;
+ per_cpu(rcu_dyntick_drain, cpu) = 0;
+ trace_rcu_prep_idle("No callbacks");
+ return;
+ }
+
+ /*
+ * If in holdoff mode, just return. We will presumably have
+ * refrained from disabling the scheduling-clock tick.
+ */
+ if (per_cpu(rcu_dyntick_holdoff, cpu) == jiffies) {
+ trace_rcu_prep_idle("In holdoff");
+ return;
}
/* Check and update the rcu_dyntick_drain sequencing. */
if (per_cpu(rcu_dyntick_drain, cpu) <= 0) {
/* First time through, initialize the counter. */
- per_cpu(rcu_dyntick_drain, cpu) = RCU_NEEDS_CPU_FLUSHES;
+ per_cpu(rcu_dyntick_drain, cpu) = RCU_IDLE_FLUSHES;
+ } else if (per_cpu(rcu_dyntick_drain, cpu) <= RCU_IDLE_OPT_FLUSHES &&
+ !rcu_pending(cpu) &&
+ !local_softirq_pending()) {
+ /* Can we go dyntick-idle despite still having callbacks? */
+ trace_rcu_prep_idle("Dyntick with callbacks");
+ per_cpu(rcu_dyntick_drain, cpu) = 0;
+ per_cpu(rcu_dyntick_holdoff, cpu) = jiffies;
+ if (rcu_cpu_has_nonlazy_callbacks(cpu))
+ hrtimer_start(&per_cpu(rcu_idle_gp_timer, cpu),
+ rcu_idle_gp_wait, HRTIMER_MODE_REL);
+ else
+ hrtimer_start(&per_cpu(rcu_idle_gp_timer, cpu),
+ rcu_idle_lazy_gp_wait, HRTIMER_MODE_REL);
+ return; /* Nothing more to do immediately. */
} else if (--per_cpu(rcu_dyntick_drain, cpu) <= 0) {
/* We have hit the limit, so time to give up. */
per_cpu(rcu_dyntick_holdoff, cpu) = jiffies;
- return rcu_needs_cpu_quick_check(cpu);
+ trace_rcu_prep_idle("Begin holdoff");
+ invoke_rcu_core(); /* Force the CPU out of dyntick-idle. */
+ return;
}
- /* Do one step pushing remaining RCU callbacks through. */
+ /*
+ * Do one step of pushing the remaining RCU callbacks through
+ * the RCU core state machine.
+ */
+#ifdef CONFIG_TREE_PREEMPT_RCU
+ if (per_cpu(rcu_preempt_data, cpu).nxtlist) {
+ rcu_preempt_qs(cpu);
+ force_quiescent_state(&rcu_preempt_state, 0);
+ }
+#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
if (per_cpu(rcu_sched_data, cpu).nxtlist) {
rcu_sched_qs(cpu);
force_quiescent_state(&rcu_sched_state, 0);
- c = c || per_cpu(rcu_sched_data, cpu).nxtlist;
}
if (per_cpu(rcu_bh_data, cpu).nxtlist) {
rcu_bh_qs(cpu);
force_quiescent_state(&rcu_bh_state, 0);
- c = c || per_cpu(rcu_bh_data, cpu).nxtlist;
}
- /* If RCU callbacks are still pending, RCU still needs this CPU. */
- if (c)
+ /*
+ * If RCU callbacks are still pending, RCU still needs this CPU.
+ * So try forcing the callbacks through the grace period.
+ */
+ if (rcu_cpu_has_callbacks(cpu)) {
+ trace_rcu_prep_idle("More callbacks");
invoke_rcu_core();
- return c;
+ } else
+ trace_rcu_prep_idle("Callbacks drained");
+}
+
+#endif /* #else #if !defined(CONFIG_RCU_FAST_NO_HZ) */
+
+#ifdef CONFIG_RCU_CPU_STALL_INFO
+
+#ifdef CONFIG_RCU_FAST_NO_HZ
+
+static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
+{
+ struct hrtimer *hrtp = &per_cpu(rcu_idle_gp_timer, cpu);
+
+ sprintf(cp, "drain=%d %c timer=%lld",
+ per_cpu(rcu_dyntick_drain, cpu),
+ per_cpu(rcu_dyntick_holdoff, cpu) == jiffies ? 'H' : '.',
+ hrtimer_active(hrtp)
+ ? ktime_to_us(hrtimer_get_remaining(hrtp))
+ : -1);
+}
+
+#else /* #ifdef CONFIG_RCU_FAST_NO_HZ */
+
+static void print_cpu_stall_fast_no_hz(char *cp, int cpu)
+{
+}
+
+#endif /* #else #ifdef CONFIG_RCU_FAST_NO_HZ */
+
+/* Initiate the stall-info list. */
+static void print_cpu_stall_info_begin(void)
+{
+ printk(KERN_CONT "\n");
}
/*
- * Check to see if we need to continue a callback-flush operations to
- * allow the last CPU to enter dyntick-idle mode.
+ * Print out diagnostic information for the specified stalled CPU.
+ *
+ * If the specified CPU is aware of the current RCU grace period
+ * (flavor specified by rsp), then print the number of scheduling
+ * clock interrupts the CPU has taken during the time that it has
+ * been aware. Otherwise, print the number of RCU grace periods
+ * that this CPU is ignorant of, for example, "1" if the CPU was
+ * aware of the previous grace period.
+ *
+ * Also print out idle and (if CONFIG_RCU_FAST_NO_HZ) idle-entry info.
*/
-static void rcu_needs_cpu_flush(void)
+static void print_cpu_stall_info(struct rcu_state *rsp, int cpu)
{
- int cpu = smp_processor_id();
- unsigned long flags;
+ char fast_no_hz[72];
+ struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
+ struct rcu_dynticks *rdtp = rdp->dynticks;
+ char *ticks_title;
+ unsigned long ticks_value;
- if (per_cpu(rcu_dyntick_drain, cpu) <= 0)
- return;
- local_irq_save(flags);
- (void)rcu_needs_cpu(cpu);
- local_irq_restore(flags);
+ if (rsp->gpnum == rdp->gpnum) {
+ ticks_title = "ticks this GP";
+ ticks_value = rdp->ticks_this_gp;
+ } else {
+ ticks_title = "GPs behind";
+ ticks_value = rsp->gpnum - rdp->gpnum;
+ }
+ print_cpu_stall_fast_no_hz(fast_no_hz, cpu);
+ printk(KERN_ERR "\t%d: (%lu %s) idle=%03x/%llx/%d %s\n",
+ cpu, ticks_value, ticks_title,
+ atomic_read(&rdtp->dynticks) & 0xfff,
+ rdtp->dynticks_nesting, rdtp->dynticks_nmi_nesting,
+ fast_no_hz);
}
-#endif /* #else #if !defined(CONFIG_RCU_FAST_NO_HZ) */
+/* Terminate the stall-info list. */
+static void print_cpu_stall_info_end(void)
+{
+ printk(KERN_ERR "\t");
+}
+
+/* Zero ->ticks_this_gp for all flavors of RCU. */
+static void zero_cpu_stall_ticks(struct rcu_data *rdp)
+{
+ rdp->ticks_this_gp = 0;
+}
+
+/* Increment ->ticks_this_gp for all flavors of RCU. */
+static void increment_cpu_stall_ticks(void)
+{
+ __get_cpu_var(rcu_sched_data).ticks_this_gp++;
+ __get_cpu_var(rcu_bh_data).ticks_this_gp++;
+#ifdef CONFIG_TREE_PREEMPT_RCU
+ __get_cpu_var(rcu_preempt_data).ticks_this_gp++;
+#endif /* #ifdef CONFIG_TREE_PREEMPT_RCU */
+}
+
+#else /* #ifdef CONFIG_RCU_CPU_STALL_INFO */
+
+static void print_cpu_stall_info_begin(void)
+{
+ printk(KERN_CONT " {");
+}
+
+static void print_cpu_stall_info(struct rcu_state *rsp, int cpu)
+{
+ printk(KERN_CONT " %d", cpu);
+}
+
+static void print_cpu_stall_info_end(void)
+{
+ printk(KERN_CONT "} ");
+}
+
+static void zero_cpu_stall_ticks(struct rcu_data *rdp)
+{
+}
+
+static void increment_cpu_stall_ticks(void)
+{
+}
+
+#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_INFO */