The rcu_momentary_dyntick_idle() function scans the RCU flavors, checking
that one of them still needs a quiescent state before doing an expensive
atomic operation on the ->dynticks counter. However, this check reduces
overhead only after a rare race condition, and increases complexity. This
commit therefore removes the scan and the mechanism enabling the scan.
Signed-off-by: Paul E. McKenney <paul...@linux.vnet.ibm.com>
---
.../Design/Data-Structures/Data-Structures.html | 4 +-
kernel/rcu/tree.c | 62 +++++-----------------
kernel/rcu/tree.h | 3 +-
3 files changed, 15 insertions(+), 54 deletions(-)
diff --git a/Documentation/RCU/Design/Data-Structures/Data-Structures.html
b/Documentation/RCU/Design/Data-Structures/Data-Structures.html
index 3d0311657533..e4bf20a68fa3 100644
--- a/Documentation/RCU/Design/Data-Structures/Data-Structures.html
+++ b/Documentation/RCU/Design/Data-Structures/Data-Structures.html
@@ -1104,7 +1104,7 @@ Its fields are as follows:
1 int dynticks_nesting;
2 int dynticks_nmi_nesting;
3 atomic_t dynticks;
- 4 int rcu_sched_qs_mask;
+ 4 bool rcu_need_heavy_qs;
5 unsigned long rcu_qs_ctr;
</pre>
@@ -1124,7 +1124,7 @@ CPU's transitions to and from dyntick-idle mode, so that
this counter
has an even value when the CPU is in dyntick-idle mode and an odd
value otherwise.
-</p><p>The <tt>->rcu_sched_qs_mask</tt> field is used
+</p><p>The <tt>->rcu_need_heavy_qs</tt> field is used
to record the fact that the RCU core code would really like to
see a quiescent state from the corresponding CPU, so much so that
it is willing to call for heavy-weight dyntick-counter operations.
diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
index be40ebb4f0bf..e0c75488f30d 100644
--- a/kernel/rcu/tree.c
+++ b/kernel/rcu/tree.c
@@ -443,44 +443,14 @@ bool rcu_eqs_special_set(int cpu)
* memory barriers to let the RCU core know about it, regardless of what
* this CPU might (or might not) do in the near future.
*
- * We inform the RCU core by emulating a zero-duration dyntick-idle
- * period, which we in turn do by incrementing the ->dynticks counter
- * by two.
+ * We inform the RCU core by emulating a zero-duration dyntick-idle period.
*
* The caller must have disabled interrupts.
*/
static void rcu_momentary_dyntick_idle(void)
{
- struct rcu_data *rdp;
- int resched_mask;
- struct rcu_state *rsp;
-
- /*
- * Yes, we can lose flag-setting operations. This is OK, because
- * the flag will be set again after some delay.
- */
- resched_mask = raw_cpu_read(rcu_dynticks.rcu_sched_qs_mask);
- raw_cpu_write(rcu_dynticks.rcu_sched_qs_mask, 0);
-
- /* Find the flavor that needs a quiescent state. */
- for_each_rcu_flavor(rsp) {
- rdp = raw_cpu_ptr(rsp->rda);
- if (!(resched_mask & rsp->flavor_mask))
- continue;
- smp_mb(); /* rcu_sched_qs_mask before cond_resched_completed. */
- if (READ_ONCE(rdp->mynode->completed) !=
- READ_ONCE(rdp->cond_resched_completed))
- continue;
-
- /*
- * Pretend to be momentarily idle for the quiescent state.
- * This allows the grace-period kthread to record the
- * quiescent state, with no need for this CPU to do anything
- * further.
- */
- rcu_dynticks_momentary_idle();
- break;
- }
+ raw_cpu_write(rcu_dynticks.rcu_need_heavy_qs, false);
+ rcu_dynticks_momentary_idle();
}
/*
@@ -496,7 +466,7 @@ void rcu_note_context_switch(void)
trace_rcu_utilization(TPS("Start context switch"));
rcu_sched_qs();
rcu_preempt_note_context_switch();
- if (unlikely(raw_cpu_read(rcu_dynticks.rcu_sched_qs_mask)))
+ if (unlikely(raw_cpu_read(rcu_dynticks.rcu_need_heavy_qs)))
rcu_momentary_dyntick_idle();
for_each_rcu_flavor(rsp)
do_nocb_deferred_wakeup(this_cpu_ptr(rsp->rda));
@@ -524,7 +494,7 @@ void rcu_all_qs(void)
struct rcu_state *rsp;
barrier(); /* Avoid RCU read-side critical sections leaking down. */
- if (unlikely(raw_cpu_read(rcu_dynticks.rcu_sched_qs_mask))) {
+ if (unlikely(raw_cpu_read(rcu_dynticks.rcu_need_heavy_qs))) {
local_irq_save(flags);
rcu_momentary_dyntick_idle();
local_irq_restore(flags);
@@ -1285,7 +1255,7 @@ static int rcu_implicit_dynticks_qs(struct rcu_data *rdp,
bool *isidle, unsigned long *maxj)
{
unsigned long jtsq;
- int *rcrmp;
+ bool *rnhqp;
unsigned long rjtsc;
struct rcu_node *rnp;
@@ -1342,7 +1312,7 @@ static int rcu_implicit_dynticks_qs(struct rcu_data *rdp,
* in-kernel CPU-bound tasks cannot advance grace periods.
* So if the grace period is old enough, make the CPU pay attention.
* Note that the unsynchronized assignments to the per-CPU
- * rcu_sched_qs_mask variable are safe. Yes, setting of
+ * rcu_need_heavy_qs variable are safe. Yes, setting of
* bits can be lost, but they will be set again on the next
* force-quiescent-state pass. So lost bit sets do not result
* in incorrect behavior, merely in a grace period lasting
@@ -1356,16 +1326,11 @@ static int rcu_implicit_dynticks_qs(struct rcu_data
*rdp,
* is set too high, we override with half of the RCU CPU stall
* warning delay.
*/
- rcrmp = &per_cpu(rcu_dynticks.rcu_sched_qs_mask, rdp->cpu);
- if (time_after(jiffies, rdp->rsp->gp_start + jtsq) ||
- time_after(jiffies, rdp->rsp->jiffies_resched)) {
- if (!(READ_ONCE(*rcrmp) & rdp->rsp->flavor_mask)) {
- WRITE_ONCE(rdp->cond_resched_completed,
- READ_ONCE(rdp->mynode->completed));
- smp_mb(); /* ->cond_resched_completed before *rcrmp. */
- WRITE_ONCE(*rcrmp,
- READ_ONCE(*rcrmp) + rdp->rsp->flavor_mask);
- }
+ rnhqp = &per_cpu(rcu_dynticks.rcu_need_heavy_qs, rdp->cpu);
+ if (!READ_ONCE(*rnhqp) &&
+ (time_after(jiffies, rdp->rsp->gp_start + jtsq) ||
+ time_after(jiffies, rdp->rsp->jiffies_resched))) {
+ WRITE_ONCE(*rnhqp, true);
rdp->rsp->jiffies_resched += 5; /* Re-enable beating. */
}
@@ -4179,7 +4144,6 @@ static void __init rcu_init_one(struct rcu_state *rsp)
static const char * const fqs[] = RCU_FQS_NAME_INIT;
static struct lock_class_key rcu_node_class[RCU_NUM_LVLS];
static struct lock_class_key rcu_fqs_class[RCU_NUM_LVLS];
- static u8 fl_mask = 0x1;
int levelcnt[RCU_NUM_LVLS]; /* # nodes in each level. */
int levelspread[RCU_NUM_LVLS]; /* kids/node in each level. */
@@ -4201,8 +4165,6 @@ static void __init rcu_init_one(struct rcu_state *rsp)
for (i = 1; i < rcu_num_lvls; i++)
rsp->level[i] = rsp->level[i - 1] + levelcnt[i - 1];
rcu_init_levelspread(levelspread, levelcnt);
- rsp->flavor_mask = fl_mask;
- fl_mask <<= 1;
/* Initialize the elements themselves, starting from the leaves. */
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index 76e4467bc765..b212cd0f22c7 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -113,7 +113,7 @@ struct rcu_dynticks {
/* Process level is worth LLONG_MAX/2. */
int dynticks_nmi_nesting; /* Track NMI nesting level. */
atomic_t dynticks; /* Even value for idle, else odd. */
- int rcu_sched_qs_mask; /* GP old, need heavy quiescent state. */
+ bool rcu_need_heavy_qs; /* GP old, need heavy quiescent state. */
unsigned long rcu_qs_ctr; /* Light universal quiescent state ctr. */
#ifdef CONFIG_NO_HZ_FULL_SYSIDLE
long long dynticks_idle_nesting;
@@ -484,7 +484,6 @@ struct rcu_state {
struct rcu_node *level[RCU_NUM_LVLS + 1];
/* Hierarchy levels (+1 to */
/* shut bogus gcc warning) */
- u8 flavor_mask; /* bit in flavor mask. */
struct rcu_data __percpu *rda; /* pointer of percu rcu_data. */
call_rcu_func_t call; /* call_rcu() flavor. */
int ncpus; /* # CPUs seen so far. */
--
2.5.2
