Skip to content

Commit 483d5bf

Browse files
Frederic Weisbeckerpaulmckrcu
Frederic Weisbecker
authored and
paulmckrcu
committed
rcu/nocb: Use kthread parking instead of ad-hoc implementation
Upon NOCB deoffloading, the rcuo kthread must be forced to sleep until the corresponding rdp is ever offloaded again. The deoffloader clears the SEGCBLIST_OFFLOADED flag, wakes up the rcuo kthread which then notices that change and clears in turn its SEGCBLIST_KTHREAD_CB flag before going to sleep, until it ever sees the SEGCBLIST_OFFLOADED flag again, should a re-offloading happen. Upon NOCB offloading, the rcuo kthread must be forced to wake up and handle callbacks until the corresponding rdp is ever deoffloaded again. The offloader sets the SEGCBLIST_OFFLOADED flag, wakes up the rcuo kthread which then notices that change and sets in turn its SEGCBLIST_KTHREAD_CB flag before going to check callbacks, until it ever sees the SEGCBLIST_OFFLOADED flag cleared again, should a de-offloading happen again. This is all a crude ad-hoc and error-prone kthread (un-)parking re-implementation. Consolidate the behaviour with the appropriate API instead. [ paulmck: Apply Qiang Zhang feedback provided in Link: below. ] Link: https://lore.kernel.org/all/[email protected]/ Signed-off-by: Frederic Weisbecker <[email protected]> Signed-off-by: Paul E. McKenney <[email protected]>
1 parent aa97b9a commit 483d5bf

File tree

3 files changed

+72
-128
lines changed

3 files changed

+72
-128
lines changed

‎include/linux/rcu_segcblist.h

Lines changed: 36 additions & 45 deletions
Original file line numberDiff line numberDiff line change
@@ -84,31 +84,31 @@ struct rcu_cblist {
8484
* ----------------------------------------------------------------------------
8585
* |
8686
* v
87-
* -----------------------------------
88-
* | |
89-
* v v
90-
* --------------------------------------- ----------------------------------|
91-
* | SEGCBLIST_RCU_CORE | | | SEGCBLIST_RCU_CORE | |
92-
* | SEGCBLIST_LOCKING | | | SEGCBLIST_LOCKING | |
93-
* | SEGCBLIST_OFFLOADED | | | SEGCBLIST_OFFLOADED | |
94-
* | SEGCBLIST_KTHREAD_CB | | SEGCBLIST_KTHREAD_GP |
95-
* | | | |
96-
* | | | |
97-
* | CB kthread woke up and | | GP kthread woke up and |
98-
* | acknowledged SEGCBLIST_OFFLOADED. | | acknowledged SEGCBLIST_OFFLOADED|
99-
* | Processes callbacks concurrently | | |
100-
* | with rcu_core(), holding | | |
101-
* | nocb_lock. | | |
102-
* --------------------------------------- -----------------------------------
103-
* | |
104-
* -----------------------------------
87+
* ----------------------------------------------------------------------------
88+
* | SEGCBLIST_RCU_CORE | SEGCBLIST_LOCKING | SEGCBLIST_OFFLOADED |
89+
* | + unparked CB kthread |
90+
* | |
91+
* | CB kthread got unparked and processes callbacks concurrently with |
92+
* | rcu_core(), holding nocb_lock. |
93+
* ---------------------------------------------------------------------------
94+
* |
95+
* v
96+
* ---------------------------------------------------------------------------|
97+
* | SEGCBLIST_RCU_CORE | |
98+
* | SEGCBLIST_LOCKING | |
99+
* | SEGCBLIST_OFFLOADED | |
100+
* | SEGCBLIST_KTHREAD_GP |
101+
* | + unparked CB kthread |
102+
* | |
103+
* | GP kthread woke up and acknowledged nocb_lock. |
104+
* ---------------------------------------- -----------------------------------
105105
* |
106106
* v
107107
* |--------------------------------------------------------------------------|
108-
* | SEGCBLIST_LOCKING | |
109-
* | SEGCBLIST_OFFLOADED | |
108+
* | SEGCBLIST_LOCKING | |
109+
* | SEGCBLIST_OFFLOADED | |
110110
* | SEGCBLIST_KTHREAD_GP | |
111-
* | SEGCBLIST_KTHREAD_CB |
111+
* | + unparked CB kthread |
112112
* | |
113113
* | Kthreads handle callbacks holding nocb_lock, local rcu_core() stops |
114114
* | handling callbacks. Enable bypass queueing. |
@@ -124,8 +124,8 @@ struct rcu_cblist {
124124
* |--------------------------------------------------------------------------|
125125
* | SEGCBLIST_LOCKING | |
126126
* | SEGCBLIST_OFFLOADED | |
127-
* | SEGCBLIST_KTHREAD_CB | |
128127
* | SEGCBLIST_KTHREAD_GP |
128+
* | + unparked CB kthread |
129129
* | |
130130
* | CB/GP kthreads handle callbacks holding nocb_lock, local rcu_core() |
131131
* | ignores callbacks. Bypass enqueue is enabled. |
@@ -136,8 +136,8 @@ struct rcu_cblist {
136136
* | SEGCBLIST_RCU_CORE | |
137137
* | SEGCBLIST_LOCKING | |
138138
* | SEGCBLIST_OFFLOADED | |
139-
* | SEGCBLIST_KTHREAD_CB | |
140139
* | SEGCBLIST_KTHREAD_GP |
140+
* | + unparked CB kthread |
141141
* | |
142142
* | CB/GP kthreads handle callbacks holding nocb_lock, local rcu_core() |
143143
* | handles callbacks concurrently. Bypass enqueue is disabled. |
@@ -149,40 +149,31 @@ struct rcu_cblist {
149149
* |--------------------------------------------------------------------------|
150150
* | SEGCBLIST_RCU_CORE | |
151151
* | SEGCBLIST_LOCKING | |
152-
* | SEGCBLIST_KTHREAD_CB | |
153152
* | SEGCBLIST_KTHREAD_GP |
153+
* | + unparked CB kthread |
154154
* | |
155155
* | CB/GP kthreads and local rcu_core() handle callbacks concurrently |
156-
* | holding nocb_lock. Wake up CB and GP kthreads if necessary. |
156+
* | holding nocb_lock. Wake up GP kthread if necessary. |
157157
* ----------------------------------------------------------------------------
158158
* |
159159
* v
160-
* -----------------------------------
161-
* | |
162-
* v v
163-
* ---------------------------------------------------------------------------|
164-
* | | |
165-
* | SEGCBLIST_RCU_CORE | | SEGCBLIST_RCU_CORE | |
166-
* | SEGCBLIST_LOCKING | | SEGCBLIST_LOCKING | |
167-
* | SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP |
168-
* | | |
169-
* | GP kthread woke up and | CB kthread woke up and |
170-
* | acknowledged the fact that | acknowledged the fact that |
171-
* | SEGCBLIST_OFFLOADED got cleared. | SEGCBLIST_OFFLOADED got cleared. |
172-
* | | The CB kthread goes to sleep |
173-
* | The callbacks from the target CPU | until it ever gets re-offloaded. |
174-
* | will be ignored from the GP kthread | |
175-
* | loop. | |
160+
* |--------------------------------------------------------------------------|
161+
* | SEGCBLIST_RCU_CORE | |
162+
* | SEGCBLIST_LOCKING | |
163+
* | + unparked CB kthread |
164+
* | |
165+
* | GP kthread woke up and acknowledged the fact that SEGCBLIST_OFFLOADED |
166+
* | got cleared. The callbacks from the target CPU will be ignored from the|
167+
* | GP kthread loop. |
176168
* ----------------------------------------------------------------------------
177-
* | |
178-
* -----------------------------------
179169
* |
180170
* v
181171
* ----------------------------------------------------------------------------
182172
* | SEGCBLIST_RCU_CORE | SEGCBLIST_LOCKING |
173+
* | + parked CB kthread |
183174
* | |
184-
* | Callbacks processed by rcu_core() from softirqs or local |
185-
* | rcuc kthread, while holding nocb_lock. |
175+
* | CB kthread is parked. Callbacks processed by rcu_core() from softirqs or |
176+
* | local rcuc kthread, while holding nocb_lock. |
186177
* ----------------------------------------------------------------------------
187178
* |
188179
* v

‎kernel/rcu/tree_nocb.h

Lines changed: 34 additions & 81 deletions
Original file line numberDiff line numberDiff line change
@@ -635,8 +635,7 @@ static void call_rcu_nocb(struct rcu_data *rdp, struct rcu_head *head,
635635
}
636636
}
637637

638-
static int nocb_gp_toggle_rdp(struct rcu_data *rdp,
639-
bool *wake_state)
638+
static int nocb_gp_toggle_rdp(struct rcu_data *rdp)
640639
{
641640
struct rcu_segcblist *cblist = &rdp->cblist;
642641
unsigned long flags;
@@ -650,8 +649,6 @@ static int nocb_gp_toggle_rdp(struct rcu_data *rdp,
650649
* We will handle this rdp until it ever gets de-offloaded.
651650
*/
652651
rcu_segcblist_set_flags(cblist, SEGCBLIST_KTHREAD_GP);
653-
if (rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB))
654-
*wake_state = true;
655652
ret = 1;
656653
} else if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_OFFLOADED) &&
657654
rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP)) {
@@ -660,8 +657,6 @@ static int nocb_gp_toggle_rdp(struct rcu_data *rdp,
660657
* We will ignore this rdp until it ever gets re-offloaded.
661658
*/
662659
rcu_segcblist_clear_flags(cblist, SEGCBLIST_KTHREAD_GP);
663-
if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB))
664-
*wake_state = true;
665660
ret = 0;
666661
} else {
667662
WARN_ON_ONCE(1);
@@ -877,16 +872,15 @@ static void nocb_gp_wait(struct rcu_data *my_rdp)
877872
}
878873

879874
if (rdp_toggling) {
880-
bool wake_state = false;
881875
int ret;
882876

883-
ret = nocb_gp_toggle_rdp(rdp_toggling, &wake_state);
877+
ret = nocb_gp_toggle_rdp(rdp_toggling);
884878
if (ret == 1)
885879
list_add_tail(&rdp_toggling->nocb_entry_rdp, &my_rdp->nocb_head_rdp);
886880
else if (ret == 0)
887881
list_del(&rdp_toggling->nocb_entry_rdp);
888-
if (wake_state)
889-
swake_up_one(&rdp_toggling->nocb_state_wq);
882+
883+
swake_up_one(&rdp_toggling->nocb_state_wq);
890884
}
891885

892886
my_rdp->nocb_gp_seq = -1;
@@ -913,16 +907,9 @@ static int rcu_nocb_gp_kthread(void *arg)
913907
return 0;
914908
}
915909

916-
static inline bool nocb_cb_can_run(struct rcu_data *rdp)
917-
{
918-
u8 flags = SEGCBLIST_OFFLOADED | SEGCBLIST_KTHREAD_CB;
919-
920-
return rcu_segcblist_test_flags(&rdp->cblist, flags);
921-
}
922-
923910
static inline bool nocb_cb_wait_cond(struct rcu_data *rdp)
924911
{
925-
return nocb_cb_can_run(rdp) && !READ_ONCE(rdp->nocb_cb_sleep);
912+
return !READ_ONCE(rdp->nocb_cb_sleep) || kthread_should_park();
926913
}
927914

928915
/*
@@ -934,21 +921,19 @@ static void nocb_cb_wait(struct rcu_data *rdp)
934921
struct rcu_segcblist *cblist = &rdp->cblist;
935922
unsigned long cur_gp_seq;
936923
unsigned long flags;
937-
bool needwake_state = false;
938924
bool needwake_gp = false;
939-
bool can_sleep = true;
940925
struct rcu_node *rnp = rdp->mynode;
941926

942-
do {
943-
swait_event_interruptible_exclusive(rdp->nocb_cb_wq,
944-
nocb_cb_wait_cond(rdp));
945-
946-
if (READ_ONCE(rdp->nocb_cb_sleep)) {
947-
WARN_ON(signal_pending(current));
948-
trace_rcu_nocb_wake(rcu_state.name, rdp->cpu, TPS("WokeEmpty"));
949-
}
950-
} while (!nocb_cb_can_run(rdp));
927+
swait_event_interruptible_exclusive(rdp->nocb_cb_wq,
928+
nocb_cb_wait_cond(rdp));
929+
if (kthread_should_park()) {
930+
kthread_parkme();
931+
} else if (READ_ONCE(rdp->nocb_cb_sleep)) {
932+
WARN_ON(signal_pending(current));
933+
trace_rcu_nocb_wake(rcu_state.name, rdp->cpu, TPS("WokeEmpty"));
934+
}
951935

936+
WARN_ON_ONCE(!rcu_rdp_is_offloaded(rdp));
952937

953938
local_irq_save(flags);
954939
rcu_momentary_dyntick_idle();
@@ -971,37 +956,16 @@ static void nocb_cb_wait(struct rcu_data *rdp)
971956
raw_spin_unlock_rcu_node(rnp); /* irqs remain disabled. */
972957
}
973958

974-
if (rcu_segcblist_test_flags(cblist, SEGCBLIST_OFFLOADED)) {
975-
if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB)) {
976-
rcu_segcblist_set_flags(cblist, SEGCBLIST_KTHREAD_CB);
977-
if (rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP))
978-
needwake_state = true;
979-
}
980-
if (rcu_segcblist_ready_cbs(cblist))
981-
can_sleep = false;
959+
if (!rcu_segcblist_ready_cbs(cblist)) {
960+
WRITE_ONCE(rdp->nocb_cb_sleep, true);
961+
trace_rcu_nocb_wake(rcu_state.name, rdp->cpu, TPS("CBSleep"));
982962
} else {
983-
/*
984-
* De-offloading. Clear our flag and notify the de-offload worker.
985-
* We won't touch the callbacks and keep sleeping until we ever
986-
* get re-offloaded.
987-
*/
988-
WARN_ON_ONCE(!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB));
989-
rcu_segcblist_clear_flags(cblist, SEGCBLIST_KTHREAD_CB);
990-
if (!rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP))
991-
needwake_state = true;
963+
WRITE_ONCE(rdp->nocb_cb_sleep, false);
992964
}
993965

994-
WRITE_ONCE(rdp->nocb_cb_sleep, can_sleep);
995-
996-
if (rdp->nocb_cb_sleep)
997-
trace_rcu_nocb_wake(rcu_state.name, rdp->cpu, TPS("CBSleep"));
998-
999966
rcu_nocb_unlock_irqrestore(rdp, flags);
1000967
if (needwake_gp)
1001968
rcu_gp_kthread_wake();
1002-
1003-
if (needwake_state)
1004-
swake_up_one(&rdp->nocb_state_wq);
1005969
}
1006970

1007971
/*
@@ -1094,17 +1058,8 @@ static int rdp_offload_toggle(struct rcu_data *rdp,
10941058
bool wake_gp = false;
10951059

10961060
rcu_segcblist_offload(cblist, offload);
1097-
1098-
if (rdp->nocb_cb_sleep)
1099-
rdp->nocb_cb_sleep = false;
11001061
rcu_nocb_unlock_irqrestore(rdp, flags);
11011062

1102-
/*
1103-
* Ignore former value of nocb_cb_sleep and force wake up as it could
1104-
* have been spuriously set to false already.
1105-
*/
1106-
swake_up_one(&rdp->nocb_cb_wq);
1107-
11081063
raw_spin_lock_irqsave(&rdp_gp->nocb_gp_lock, flags);
11091064
// Queue this rdp for add/del to/from the list to iterate on rcuog
11101065
WRITE_ONCE(rdp_gp->nocb_toggling_rdp, rdp);
@@ -1161,28 +1116,19 @@ static long rcu_nocb_rdp_deoffload(void *arg)
11611116
if (wake_gp)
11621117
wake_up_process(rdp_gp->nocb_gp_kthread);
11631118

1164-
/*
1165-
* If rcuo[p] kthread spawn failed, directly remove SEGCBLIST_KTHREAD_CB.
1166-
* Just wait SEGCBLIST_KTHREAD_GP to be cleared by rcuog.
1167-
*/
1168-
if (!rdp->nocb_cb_kthread) {
1169-
rcu_nocb_lock_irqsave(rdp, flags);
1170-
rcu_segcblist_clear_flags(&rdp->cblist, SEGCBLIST_KTHREAD_CB);
1171-
rcu_nocb_unlock_irqrestore(rdp, flags);
1172-
}
1173-
11741119
swait_event_exclusive(rdp->nocb_state_wq,
1175-
!rcu_segcblist_test_flags(cblist,
1176-
SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP));
1120+
!rcu_segcblist_test_flags(cblist,
1121+
SEGCBLIST_KTHREAD_GP));
1122+
if (rdp->nocb_cb_kthread)
1123+
kthread_park(rdp->nocb_cb_kthread);
11771124
} else {
11781125
/*
11791126
* No kthread to clear the flags for us or remove the rdp from the nocb list
11801127
* to iterate. Do it here instead. Locking doesn't look stricly necessary
11811128
* but we stick to paranoia in this rare path.
11821129
*/
11831130
rcu_nocb_lock_irqsave(rdp, flags);
1184-
rcu_segcblist_clear_flags(&rdp->cblist,
1185-
SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP);
1131+
rcu_segcblist_clear_flags(&rdp->cblist, SEGCBLIST_KTHREAD_GP);
11861132
rcu_nocb_unlock_irqrestore(rdp, flags);
11871133

11881134
list_del(&rdp->nocb_entry_rdp);
@@ -1282,8 +1228,10 @@ static long rcu_nocb_rdp_offload(void *arg)
12821228
wake_gp = rdp_offload_toggle(rdp, true, flags);
12831229
if (wake_gp)
12841230
wake_up_process(rdp_gp->nocb_gp_kthread);
1231+
1232+
kthread_unpark(rdp->nocb_cb_kthread);
1233+
12851234
swait_event_exclusive(rdp->nocb_state_wq,
1286-
rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_CB) &&
12871235
rcu_segcblist_test_flags(cblist, SEGCBLIST_KTHREAD_GP));
12881236

12891237
/*
@@ -1468,7 +1416,7 @@ void __init rcu_init_nohz(void)
14681416
if (rcu_segcblist_empty(&rdp->cblist))
14691417
rcu_segcblist_init(&rdp->cblist);
14701418
rcu_segcblist_offload(&rdp->cblist, true);
1471-
rcu_segcblist_set_flags(&rdp->cblist, SEGCBLIST_KTHREAD_CB | SEGCBLIST_KTHREAD_GP);
1419+
rcu_segcblist_set_flags(&rdp->cblist, SEGCBLIST_KTHREAD_GP);
14721420
rcu_segcblist_clear_flags(&rdp->cblist, SEGCBLIST_RCU_CORE);
14731421
}
14741422
rcu_organize_nocb_kthreads();
@@ -1526,11 +1474,16 @@ static void rcu_spawn_cpu_nocb_kthread(int cpu)
15261474
mutex_unlock(&rdp_gp->nocb_gp_kthread_mutex);
15271475

15281476
/* Spawn the kthread for this CPU. */
1529-
t = kthread_run(rcu_nocb_cb_kthread, rdp,
1530-
"rcuo%c/%d", rcu_state.abbr, cpu);
1477+
t = kthread_create(rcu_nocb_cb_kthread, rdp,
1478+
"rcuo%c/%d", rcu_state.abbr, cpu);
15311479
if (WARN_ONCE(IS_ERR(t), "%s: Could not start rcuo CB kthread, OOM is now expected behavior\n", __func__))
15321480
goto end;
15331481

1482+
if (rcu_rdp_is_offloaded(rdp))
1483+
wake_up_process(t);
1484+
else
1485+
kthread_park(t);
1486+
15341487
if (IS_ENABLED(CONFIG_RCU_NOCB_CPU_CB_BOOST) && kthread_prio)
15351488
sched_setscheduler_nocheck(t, SCHED_FIFO, &sp);
15361489

‎kernel/rcu/tree_plugin.h

Lines changed: 2 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -28,8 +28,8 @@ static bool rcu_rdp_is_offloaded(struct rcu_data *rdp)
2828
!(lockdep_is_held(&rcu_state.barrier_mutex) ||
2929
(IS_ENABLED(CONFIG_HOTPLUG_CPU) && lockdep_is_cpus_held()) ||
3030
rcu_lockdep_is_held_nocb(rdp) ||
31-
(rdp == this_cpu_ptr(&rcu_data) &&
32-
!(IS_ENABLED(CONFIG_PREEMPT_COUNT) && preemptible())) ||
31+
(!(IS_ENABLED(CONFIG_PREEMPT_COUNT) && preemptible()) &&
32+
rdp == this_cpu_ptr(&rcu_data)) ||
3333
rcu_current_is_nocb_kthread(rdp)),
3434
"Unsafe read of RCU_NOCB offloaded state"
3535
);

0 commit comments

Comments
 (0)