1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
|
From: Peter Zijlstra <peterz@infradead.org>
Date: Fri, 23 Oct 2020 12:12:01 +0200
Subject: [PATCH 03/19] sched/hotplug: Ensure only per-cpu kthreads run during
hotplug
In preparation for migrate_disable(), make sure only per-cpu kthreads
are allowed to run on !active CPUs.
This is ran (as one of the very first steps) from the cpu-hotplug
task which is a per-cpu kthread and completion of the hotplug
operation only requires such tasks.
This constraint enables the migrate_disable() implementation to wait
for completion of all migrate_disable regions on this CPU at hotplug
time without fear of any new ones starting.
This replaces the unlikely(rq->balance_callbacks) test at the tail of
context_switch with an unlikely(rq->balance_work), the fast path is
not affected.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
---
kernel/sched/core.c | 114 ++++++++++++++++++++++++++++++++++++++++++++++++++-
kernel/sched/sched.h | 7 ++-
2 files changed, 118 insertions(+), 3 deletions(-)
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3514,8 +3514,10 @@ static inline struct callback_head *spli
struct callback_head *head = rq->balance_callback;
lockdep_assert_held(&rq->lock);
- if (head)
+ if (head) {
rq->balance_callback = NULL;
+ rq->balance_flags &= ~BALANCE_WORK;
+ }
return head;
}
@@ -3536,6 +3538,21 @@ static inline void balance_callbacks(str
}
}
+static void balance_push(struct rq *rq);
+
+static inline void balance_switch(struct rq *rq)
+{
+ if (likely(!rq->balance_flags))
+ return;
+
+ if (rq->balance_flags & BALANCE_PUSH) {
+ balance_push(rq);
+ return;
+ }
+
+ __balance_callbacks(rq);
+}
+
#else
static inline void __balance_callbacks(struct rq *rq)
@@ -3551,6 +3568,10 @@ static inline void balance_callbacks(str
{
}
+static inline void balance_switch(struct rq *rq)
+{
+}
+
#endif
static inline void
@@ -3578,7 +3599,7 @@ static inline void finish_lock_switch(st
* prev into current:
*/
spin_acquire(&rq->lock.dep_map, 0, 0, _THIS_IP_);
- __balance_callbacks(rq);
+ balance_switch(rq);
raw_spin_unlock_irq(&rq->lock);
}
@@ -6834,6 +6855,90 @@ static void migrate_tasks(struct rq *dea
rq->stop = stop;
}
+
+static int __balance_push_cpu_stop(void *arg)
+{
+ struct task_struct *p = arg;
+ struct rq *rq = this_rq();
+ struct rq_flags rf;
+ int cpu;
+
+ raw_spin_lock_irq(&p->pi_lock);
+ rq_lock(rq, &rf);
+
+ update_rq_clock(rq);
+
+ if (task_rq(p) == rq && task_on_rq_queued(p)) {
+ cpu = select_fallback_rq(rq->cpu, p);
+ rq = __migrate_task(rq, &rf, p, cpu);
+ }
+
+ rq_unlock(rq, &rf);
+ raw_spin_unlock_irq(&p->pi_lock);
+
+ put_task_struct(p);
+
+ return 0;
+}
+
+static DEFINE_PER_CPU(struct cpu_stop_work, push_work);
+
+/*
+ * Ensure we only run per-cpu kthreads once the CPU goes !active.
+ */
+static void balance_push(struct rq *rq)
+{
+ struct task_struct *push_task = rq->curr;
+
+ lockdep_assert_held(&rq->lock);
+ SCHED_WARN_ON(rq->cpu != smp_processor_id());
+
+ /*
+ * Both the cpu-hotplug and stop task are in this case and are
+ * required to complete the hotplug process.
+ */
+ if (is_per_cpu_kthread(push_task))
+ return;
+
+ get_task_struct(push_task);
+ /*
+ * Temporarily drop rq->lock such that we can wake-up the stop task.
+ * Both preemption and IRQs are still disabled.
+ */
+ raw_spin_unlock(&rq->lock);
+ stop_one_cpu_nowait(rq->cpu, __balance_push_cpu_stop, push_task,
+ this_cpu_ptr(&push_work));
+ /*
+ * At this point need_resched() is true and we'll take the loop in
+ * schedule(). The next pick is obviously going to be the stop task
+ * which is_per_cpu_kthread() and will push this task away.
+ */
+ raw_spin_lock(&rq->lock);
+}
+
+static void balance_push_set(int cpu, bool on)
+{
+ struct rq *rq = cpu_rq(cpu);
+ struct rq_flags rf;
+
+ rq_lock_irqsave(rq, &rf);
+ if (on)
+ rq->balance_flags |= BALANCE_PUSH;
+ else
+ rq->balance_flags &= ~BALANCE_PUSH;
+ rq_unlock_irqrestore(rq, &rf);
+}
+
+#else
+
+static inline void balance_push(struct rq *rq)
+{
+}
+
+static inline void balance_push_set(int cpu, bool on)
+{
+}
+
#endif /* CONFIG_HOTPLUG_CPU */
void set_rq_online(struct rq *rq)
@@ -6919,6 +7024,8 @@ int sched_cpu_activate(unsigned int cpu)
struct rq *rq = cpu_rq(cpu);
struct rq_flags rf;
+ balance_push_set(cpu, false);
+
#ifdef CONFIG_SCHED_SMT
/*
* When going up, increment the number of cores with SMT present.
@@ -6966,6 +7073,8 @@ int sched_cpu_deactivate(unsigned int cp
*/
synchronize_rcu();
+ balance_push_set(cpu, true);
+
#ifdef CONFIG_SCHED_SMT
/*
* When going down, decrement the number of cores with SMT present.
@@ -6979,6 +7088,7 @@ int sched_cpu_deactivate(unsigned int cp
ret = cpuset_cpu_inactive(cpu);
if (ret) {
+ balance_push_set(cpu, false);
set_cpu_active(cpu, true);
return ret;
}
--- a/kernel/sched/sched.h
+++ b/kernel/sched/sched.h
@@ -967,6 +967,7 @@ struct rq {
unsigned long cpu_capacity_orig;
struct callback_head *balance_callback;
+ unsigned char balance_flags;
unsigned char nohz_idle_balance;
unsigned char idle_balance;
@@ -1379,6 +1380,9 @@ init_numa_balancing(unsigned long clone_
#ifdef CONFIG_SMP
+#define BALANCE_WORK 0x01
+#define BALANCE_PUSH 0x02
+
static inline void
queue_balance_callback(struct rq *rq,
struct callback_head *head,
@@ -1386,12 +1390,13 @@ queue_balance_callback(struct rq *rq,
{
lockdep_assert_held(&rq->lock);
- if (unlikely(head->next))
+ if (unlikely(head->next || (rq->balance_flags & BALANCE_PUSH)))
return;
head->func = (void (*)(struct callback_head *))func;
head->next = rq->balance_callback;
rq->balance_callback = head;
+ rq->balance_flags |= BALANCE_WORK;
}
#define rcu_dereference_check_sched_domain(p) \
|
