Doug / smarc-fsl-linux-kernel

1

/*

1

/*

2

* Read-Copy Update mechanism for mutual exclusion

2

* Read-Copy Update mechanism for mutual exclusion

3

*

3

*

4

* This program is free software; you can redistribute it and/or modify

4

* This program is free software; you can redistribute it and/or modify

5

* it under the terms of the GNU General Public License as published by

5

* it under the terms of the GNU General Public License as published by

6

* the Free Software Foundation; either version 2 of the License, or

6

* the Free Software Foundation; either version 2 of the License, or

7

* (at your option) any later version.

7

* (at your option) any later version.

8

*

8

*

9

* This program is distributed in the hope that it will be useful,

9

* This program is distributed in the hope that it will be useful,

10

* but WITHOUT ANY WARRANTY; without even the implied warranty of

10

* but WITHOUT ANY WARRANTY; without even the implied warranty of

11

* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

11

* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the

12

* GNU General Public License for more details.

12

* GNU General Public License for more details.

13

*

13

*

14

* You should have received a copy of the GNU General Public License

14

* You should have received a copy of the GNU General Public License

15

* along with this program; if not, write to the Free Software

15

* along with this program; if not, write to the Free Software

16

* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

16

* Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.

17

*

17

*

18

* Copyright IBM Corporation, 2008

18

* Copyright IBM Corporation, 2008

19

*

19

*

20

* Authors: Dipankar Sarma <dipankar@in.ibm.com>

20

* Authors: Dipankar Sarma <dipankar@in.ibm.com>

21

* Manfred Spraul <manfred@colorfullife.com>

21

* Manfred Spraul <manfred@colorfullife.com>

22

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> Hierarchical version

22

* Paul E. McKenney <paulmck@linux.vnet.ibm.com> Hierarchical version

23

*

23

*

24

* Based on the original work by Paul McKenney <paulmck@us.ibm.com>

24

* Based on the original work by Paul McKenney <paulmck@us.ibm.com>

25

* and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.

25

* and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.

26

*

26

*

27

* For detailed explanation of Read-Copy Update mechanism see -

27

* For detailed explanation of Read-Copy Update mechanism see -

28

* Documentation/RCU

28

* Documentation/RCU

29

*/

29

*/

30

#include <linux/types.h>

30

#include <linux/types.h>

31

#include <linux/kernel.h>

31

#include <linux/kernel.h>

32

#include <linux/init.h>

32

#include <linux/init.h>

33

#include <linux/spinlock.h>

33

#include <linux/spinlock.h>

34

#include <linux/smp.h>

34

#include <linux/smp.h>

35

#include <linux/rcupdate.h>

35

#include <linux/rcupdate.h>

36

#include <linux/interrupt.h>

36

#include <linux/interrupt.h>

37

#include <linux/sched.h>

37

#include <linux/sched.h>

38

#include <linux/nmi.h>

38

#include <linux/nmi.h>

39

#include <linux/atomic.h>

39

#include <linux/atomic.h>

40

#include <linux/bitops.h>

40

#include <linux/bitops.h>

41

#include <linux/export.h>

41

#include <linux/export.h>

42

#include <linux/completion.h>

42

#include <linux/completion.h>

43

#include <linux/moduleparam.h>

43

#include <linux/moduleparam.h>

44

#include <linux/percpu.h>

44

#include <linux/percpu.h>

45

#include <linux/notifier.h>

45

#include <linux/notifier.h>

46

#include <linux/cpu.h>

46

#include <linux/cpu.h>

47

#include <linux/mutex.h>

47

#include <linux/mutex.h>

48

#include <linux/time.h>

48

#include <linux/time.h>

49

#include <linux/kernel_stat.h>

49

#include <linux/kernel_stat.h>

50

#include <linux/wait.h>

50

#include <linux/wait.h>

51

#include <linux/kthread.h>

51

#include <linux/kthread.h>

52

#include <linux/prefetch.h>

52

#include <linux/prefetch.h>

53

#include <linux/delay.h>

53

#include <linux/delay.h>

54

#include <linux/stop_machine.h>

54

#include <linux/stop_machine.h>

55

56

#include "rcutree.h"

56

#include "rcutree.h"

57

#include <trace/events/rcu.h>

57

#include <trace/events/rcu.h>

58

59

#include "rcu.h"

59

#include "rcu.h"

60

61

/* Data structures. */

61

/* Data structures. */

62

63

static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];

63

static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];

64

65

#define RCU_STATE_INITIALIZER(structname) { \

65

#define RCU_STATE_INITIALIZER(structname) { \

66

.level = { &structname##_state.node[0] }, \

66

.level = { &structname##_state.node[0] }, \

67

.levelcnt = { \

67

.levelcnt = { \

68

NUM_RCU_LVL_0, /* root of hierarchy. */ \

68

NUM_RCU_LVL_0, /* root of hierarchy. */ \

69

NUM_RCU_LVL_1, \

69

NUM_RCU_LVL_1, \

70

NUM_RCU_LVL_2, \

70

NUM_RCU_LVL_2, \

71

NUM_RCU_LVL_3, \

71

NUM_RCU_LVL_3, \

72

NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \

72

NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \

73

}, \

73

}, \

74

.fqs_state = RCU_GP_IDLE, \

74

.fqs_state = RCU_GP_IDLE, \

75

.gpnum = -300, \

75

.gpnum = -300, \

76

.completed = -300, \

76

.completed = -300, \

77

.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.onofflock), \

77

.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.onofflock), \

78

.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.fqslock), \

78

.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.fqslock), \

79

.n_force_qs = 0, \

79

.n_force_qs = 0, \

80

.n_force_qs_ngp = 0, \

80

.n_force_qs_ngp = 0, \

81

.name = #structname, \

81

.name = #structname, \

82

}

82

}

83

84

struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched);

84

struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched);

85

DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);

85

DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);

86

87

struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh);

87

struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh);

88

DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);

88

DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);

89

90

static struct rcu_state *rcu_state;

90

static struct rcu_state *rcu_state;

91

92

/*

92

/*

93

* The rcu_scheduler_active variable transitions from zero to one just

93

* The rcu_scheduler_active variable transitions from zero to one just

94

* before the first task is spawned. So when this variable is zero, RCU

94

* before the first task is spawned. So when this variable is zero, RCU

95

* can assume that there is but one task, allowing RCU to (for example)

95

* can assume that there is but one task, allowing RCU to (for example)

96

* optimized synchronize_sched() to a simple barrier(). When this variable

96

* optimized synchronize_sched() to a simple barrier(). When this variable

97

* is one, RCU must actually do all the hard work required to detect real

97

* is one, RCU must actually do all the hard work required to detect real

98

* grace periods. This variable is also used to suppress boot-time false

98

* grace periods. This variable is also used to suppress boot-time false

99

* positives from lockdep-RCU error checking.

99

* positives from lockdep-RCU error checking.

100

*/

100

*/

101

int rcu_scheduler_active __read_mostly;

101

int rcu_scheduler_active __read_mostly;

102

EXPORT_SYMBOL_GPL(rcu_scheduler_active);

102

EXPORT_SYMBOL_GPL(rcu_scheduler_active);

103

104

/*

104

/*

105

* The rcu_scheduler_fully_active variable transitions from zero to one

105

* The rcu_scheduler_fully_active variable transitions from zero to one

106

* during the early_initcall() processing, which is after the scheduler

106

* during the early_initcall() processing, which is after the scheduler

107

* is capable of creating new tasks. So RCU processing (for example,

107

* is capable of creating new tasks. So RCU processing (for example,

108

* creating tasks for RCU priority boosting) must be delayed until after

108

* creating tasks for RCU priority boosting) must be delayed until after

109

* rcu_scheduler_fully_active transitions from zero to one. We also

109

* rcu_scheduler_fully_active transitions from zero to one. We also

110

* currently delay invocation of any RCU callbacks until after this point.

110

* currently delay invocation of any RCU callbacks until after this point.

111

*

111

*

112

* It might later prove better for people registering RCU callbacks during

112

* It might later prove better for people registering RCU callbacks during

113

* early boot to take responsibility for these callbacks, but one step at

113

* early boot to take responsibility for these callbacks, but one step at

114

* a time.

114

* a time.

115

*/

115

*/

116

static int rcu_scheduler_fully_active __read_mostly;

116

static int rcu_scheduler_fully_active __read_mostly;

117

118

#ifdef CONFIG_RCU_BOOST

118

#ifdef CONFIG_RCU_BOOST

119

120

/*

120

/*

121

* Control variables for per-CPU and per-rcu_node kthreads. These

121

* Control variables for per-CPU and per-rcu_node kthreads. These

122

* handle all flavors of RCU.

122

* handle all flavors of RCU.

123

*/

123

*/

124

static DEFINE_PER_CPU(struct task_struct *, rcu_cpu_kthread_task);

124

static DEFINE_PER_CPU(struct task_struct *, rcu_cpu_kthread_task);

125

DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);

125

DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);

126

DEFINE_PER_CPU(int, rcu_cpu_kthread_cpu);

126

DEFINE_PER_CPU(int, rcu_cpu_kthread_cpu);

127

DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);

127

DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);

128

DEFINE_PER_CPU(char, rcu_cpu_has_work);

128

DEFINE_PER_CPU(char, rcu_cpu_has_work);

129

130

#endif /* #ifdef CONFIG_RCU_BOOST */

130

#endif /* #ifdef CONFIG_RCU_BOOST */

131

132

static void rcu_node_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);

132

static void rcu_node_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);

133

static void invoke_rcu_core(void);

133

static void invoke_rcu_core(void);

134

static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);

134

static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);

135

136

/*

136

/*

137

* Track the rcutorture test sequence number and the update version

137

* Track the rcutorture test sequence number and the update version

138

* number within a given test. The rcutorture_testseq is incremented

138

* number within a given test. The rcutorture_testseq is incremented

139

* on every rcutorture module load and unload, so has an odd value

139

* on every rcutorture module load and unload, so has an odd value

140

* when a test is running. The rcutorture_vernum is set to zero

140

* when a test is running. The rcutorture_vernum is set to zero

141

* when rcutorture starts and is incremented on each rcutorture update.

141

* when rcutorture starts and is incremented on each rcutorture update.

142

* These variables enable correlating rcutorture output with the

142

* These variables enable correlating rcutorture output with the

143

* RCU tracing information.

143

* RCU tracing information.

144

*/

144

*/

145

unsigned long rcutorture_testseq;

145

unsigned long rcutorture_testseq;

146

unsigned long rcutorture_vernum;

146

unsigned long rcutorture_vernum;

147

148

/*

148

/*

149

* Return true if an RCU grace period is in progress. The ACCESS_ONCE()s

149

* Return true if an RCU grace period is in progress. The ACCESS_ONCE()s

150

* permit this function to be invoked without holding the root rcu_node

150

* permit this function to be invoked without holding the root rcu_node

151

* structure's ->lock, but of course results can be subject to change.

151

* structure's ->lock, but of course results can be subject to change.

152

*/

152

*/

153

static int rcu_gp_in_progress(struct rcu_state *rsp)

153

static int rcu_gp_in_progress(struct rcu_state *rsp)

154

{

154

{

155

return ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum);

155

return ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum);

156

}

156

}

157

158

/*

158

/*

159

* Note a quiescent state. Because we do not need to know

159

* Note a quiescent state. Because we do not need to know

160

* how many quiescent states passed, just if there was at least

160

* how many quiescent states passed, just if there was at least

161

* one since the start of the grace period, this just sets a flag.

161

* one since the start of the grace period, this just sets a flag.

162

* The caller must have disabled preemption.

162

* The caller must have disabled preemption.

163

*/

163

*/

164

void rcu_sched_qs(int cpu)

164

void rcu_sched_qs(int cpu)

165

{

165

{

166

struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);

166

struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);

167

168

rdp->passed_quiesce_gpnum = rdp->gpnum;

168

rdp->passed_quiesce_gpnum = rdp->gpnum;

169

barrier();

169

barrier();

170

if (rdp->passed_quiesce == 0)

170

if (rdp->passed_quiesce == 0)

171

trace_rcu_grace_period("rcu_sched", rdp->gpnum, "cpuqs");

171

trace_rcu_grace_period("rcu_sched", rdp->gpnum, "cpuqs");

172

rdp->passed_quiesce = 1;

172

rdp->passed_quiesce = 1;

173

}

173

}

174

175

void rcu_bh_qs(int cpu)

175

void rcu_bh_qs(int cpu)

176

{

176

{

177

struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);

177

struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);

178

179

rdp->passed_quiesce_gpnum = rdp->gpnum;

179

rdp->passed_quiesce_gpnum = rdp->gpnum;

180

barrier();

180

barrier();

181

if (rdp->passed_quiesce == 0)

181

if (rdp->passed_quiesce == 0)

182

trace_rcu_grace_period("rcu_bh", rdp->gpnum, "cpuqs");

182

trace_rcu_grace_period("rcu_bh", rdp->gpnum, "cpuqs");

183

rdp->passed_quiesce = 1;

183

rdp->passed_quiesce = 1;

184

}

184

}

185

186

/*

186

/*

187

* Note a context switch. This is a quiescent state for RCU-sched,

187

* Note a context switch. This is a quiescent state for RCU-sched,

188

* and requires special handling for preemptible RCU.

188

* and requires special handling for preemptible RCU.

189

* The caller must have disabled preemption.

189

* The caller must have disabled preemption.

190

*/

190

*/

191

void rcu_note_context_switch(int cpu)

191

void rcu_note_context_switch(int cpu)

192

{

192

{

193

trace_rcu_utilization("Start context switch");

193

trace_rcu_utilization("Start context switch");

194

rcu_sched_qs(cpu);

194

rcu_sched_qs(cpu);

195

rcu_preempt_note_context_switch(cpu);

195

rcu_preempt_note_context_switch(cpu);

196

trace_rcu_utilization("End context switch");

196

trace_rcu_utilization("End context switch");

197

}

197

}

198

EXPORT_SYMBOL_GPL(rcu_note_context_switch);

198

EXPORT_SYMBOL_GPL(rcu_note_context_switch);

199

200

DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {

200

DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {

201

.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,

201

.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,

202

.dynticks = ATOMIC_INIT(1),

202

.dynticks = ATOMIC_INIT(1),

203

};

203

};

204

205

static int blimit = 10; /* Maximum callbacks per rcu_do_batch. */

205

static int blimit = 10; /* Maximum callbacks per rcu_do_batch. */

206

static int qhimark = 10000; /* If this many pending, ignore blimit. */

206

static int qhimark = 10000; /* If this many pending, ignore blimit. */

207

static int qlowmark = 100; /* Once only this many pending, use blimit. */

207

static int qlowmark = 100; /* Once only this many pending, use blimit. */

208

209

module_param(blimit, int, 0);

209

module_param(blimit, int, 0);

210

module_param(qhimark, int, 0);

210

module_param(qhimark, int, 0);

211

module_param(qlowmark, int, 0);

211

module_param(qlowmark, int, 0);

212

213

int rcu_cpu_stall_suppress __read_mostly; /* 1 = suppress stall warnings. */

213

int rcu_cpu_stall_suppress __read_mostly; /* 1 = suppress stall warnings. */

214

int rcu_cpu_stall_timeout __read_mostly = CONFIG_RCU_CPU_STALL_TIMEOUT;

214

int rcu_cpu_stall_timeout __read_mostly = CONFIG_RCU_CPU_STALL_TIMEOUT;

215

216

module_param(rcu_cpu_stall_suppress, int, 0644);

216

module_param(rcu_cpu_stall_suppress, int, 0644);

217

module_param(rcu_cpu_stall_timeout, int, 0644);

217

module_param(rcu_cpu_stall_timeout, int, 0644);

218

219

static void force_quiescent_state(struct rcu_state *rsp, int relaxed);

219

static void force_quiescent_state(struct rcu_state *rsp, int relaxed);

220

static int rcu_pending(int cpu);

220

static int rcu_pending(int cpu);

221

222

/*

222

/*

223

* Return the number of RCU-sched batches processed thus far for debug & stats.

223

* Return the number of RCU-sched batches processed thus far for debug & stats.

224

*/

224

*/

225

long rcu_batches_completed_sched(void)

225

long rcu_batches_completed_sched(void)

226

{

226

{

227

return rcu_sched_state.completed;

227

return rcu_sched_state.completed;

228

}

228

}

229

EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);

229

EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);

230

231

/*

231

/*

232

* Return the number of RCU BH batches processed thus far for debug & stats.

232

* Return the number of RCU BH batches processed thus far for debug & stats.

233

*/

233

*/

234

long rcu_batches_completed_bh(void)

234

long rcu_batches_completed_bh(void)

235

{

235

{

236

return rcu_bh_state.completed;

236

return rcu_bh_state.completed;

237

}

237

}

238

EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

238

EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

239

240

/*

240

/*

241

* Force a quiescent state for RCU BH.

241

* Force a quiescent state for RCU BH.

242

*/

242

*/

243

void rcu_bh_force_quiescent_state(void)

243

void rcu_bh_force_quiescent_state(void)

244

{

244

{

245

force_quiescent_state(&rcu_bh_state, 0);

245

force_quiescent_state(&rcu_bh_state, 0);

246

}

246

}

247

EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

247

EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

248

249

/*

249

/*

250

* Record the number of times rcutorture tests have been initiated and

250

* Record the number of times rcutorture tests have been initiated and

251

* terminated. This information allows the debugfs tracing stats to be

251

* terminated. This information allows the debugfs tracing stats to be

252

* correlated to the rcutorture messages, even when the rcutorture module

252

* correlated to the rcutorture messages, even when the rcutorture module

253

* is being repeatedly loaded and unloaded. In other words, we cannot

253

* is being repeatedly loaded and unloaded. In other words, we cannot

254

* store this state in rcutorture itself.

254

* store this state in rcutorture itself.

255

*/

255

*/

256

void rcutorture_record_test_transition(void)

256

void rcutorture_record_test_transition(void)

257

{

257

{

258

rcutorture_testseq++;

258

rcutorture_testseq++;

259

rcutorture_vernum = 0;

259

rcutorture_vernum = 0;

260

}

260

}

261

EXPORT_SYMBOL_GPL(rcutorture_record_test_transition);

261

EXPORT_SYMBOL_GPL(rcutorture_record_test_transition);

262

263

/*

263

/*

264

* Record the number of writer passes through the current rcutorture test.

264

* Record the number of writer passes through the current rcutorture test.

265

* This is also used to correlate debugfs tracing stats with the rcutorture

265

* This is also used to correlate debugfs tracing stats with the rcutorture

266

* messages.

266

* messages.

267

*/

267

*/

268

void rcutorture_record_progress(unsigned long vernum)

268

void rcutorture_record_progress(unsigned long vernum)

269

{

269

{

270

rcutorture_vernum++;

270

rcutorture_vernum++;

271

}

271

}

272

EXPORT_SYMBOL_GPL(rcutorture_record_progress);

272

EXPORT_SYMBOL_GPL(rcutorture_record_progress);

273

274

/*

274

/*

275

* Force a quiescent state for RCU-sched.

275

* Force a quiescent state for RCU-sched.

276

*/

276

*/

277

void rcu_sched_force_quiescent_state(void)

277

void rcu_sched_force_quiescent_state(void)

278

{

278

{

279

force_quiescent_state(&rcu_sched_state, 0);

279

force_quiescent_state(&rcu_sched_state, 0);

280

}

280

}

281

EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

281

EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

282

283

/*

283

/*

284

* Does the CPU have callbacks ready to be invoked?

284

* Does the CPU have callbacks ready to be invoked?

285

*/

285

*/

286

static int

286

static int

287

cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)

287

cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)

288

{

288

{

289

return &rdp->nxtlist != rdp->nxttail[RCU_DONE_TAIL];

289

return &rdp->nxtlist != rdp->nxttail[RCU_DONE_TAIL];

290

}

290

}

291

292

/*

292

/*

293

* Does the current CPU require a yet-as-unscheduled grace period?

293

* Does the current CPU require a yet-as-unscheduled grace period?

294

*/

294

*/

295

static int

295

static int

296

cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)

296

cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)

297

{

297

{

298

return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);

298

return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);

299

}

299

}

300

301

/*

301

/*

302

* Return the root node of the specified rcu_state structure.

302

* Return the root node of the specified rcu_state structure.

303

*/

303

*/

304

static struct rcu_node *rcu_get_root(struct rcu_state *rsp)

304

static struct rcu_node *rcu_get_root(struct rcu_state *rsp)

305

{

305

{

306

return &rsp->node[0];

306

return &rsp->node[0];

307

}

307

}

308

309

/*

309

/*

310

* If the specified CPU is offline, tell the caller that it is in

310

* If the specified CPU is offline, tell the caller that it is in

311

* a quiescent state. Otherwise, whack it with a reschedule IPI.

311

* a quiescent state. Otherwise, whack it with a reschedule IPI.

312

* Grace periods can end up waiting on an offline CPU when that

312

* Grace periods can end up waiting on an offline CPU when that

313

* CPU is in the process of coming online -- it will be added to the

313

* CPU is in the process of coming online -- it will be added to the

314

* rcu_node bitmasks before it actually makes it online. The same thing

314

* rcu_node bitmasks before it actually makes it online. The same thing

315

* can happen while a CPU is in the process of coming online. Because this

315

* can happen while a CPU is in the process of coming online. Because this

316

* race is quite rare, we check for it after detecting that the grace

316

* race is quite rare, we check for it after detecting that the grace

317

* period has been delayed rather than checking each and every CPU

317

* period has been delayed rather than checking each and every CPU

318

* each and every time we start a new grace period.

318

* each and every time we start a new grace period.

319

*/

319

*/

320

static int rcu_implicit_offline_qs(struct rcu_data *rdp)

320

static int rcu_implicit_offline_qs(struct rcu_data *rdp)

321

{

321

{

322

/*

322

/*

323

* If the CPU is offline for more than a jiffy, it is in a quiescent

323

* If the CPU is offline for more than a jiffy, it is in a quiescent

324

* state. We can trust its state not to change because interrupts

324

* state. We can trust its state not to change because interrupts

325

* are disabled. The reason for the jiffy's worth of slack is to

325

* are disabled. The reason for the jiffy's worth of slack is to

326

* handle CPUs initializing on the way up and finding their way

326

* handle CPUs initializing on the way up and finding their way

327

* to the idle loop on the way down.

327

* to the idle loop on the way down.

328

*/

328

*/

329

if (cpu_is_offline(rdp->cpu) &&

329

if (cpu_is_offline(rdp->cpu) &&

330

ULONG_CMP_LT(rdp->rsp->gp_start + 2, jiffies)) {

330

ULONG_CMP_LT(rdp->rsp->gp_start + 2, jiffies)) {

331

trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "ofl");

331

trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "ofl");

332

rdp->offline_fqs++;

332

rdp->offline_fqs++;

333

return 1;

333

return 1;

334

}

334

}

335

return 0;

335

return 0;

336

}

336

}

337

338

/*

338

/*

339

* rcu_idle_enter_common - inform RCU that current CPU is moving towards idle

339

* rcu_idle_enter_common - inform RCU that current CPU is moving towards idle

340

*

340

*

341

* If the new value of the ->dynticks_nesting counter now is zero,

341

* If the new value of the ->dynticks_nesting counter now is zero,

342

* we really have entered idle, and must do the appropriate accounting.

342

* we really have entered idle, and must do the appropriate accounting.

343

* The caller must have disabled interrupts.

343

* The caller must have disabled interrupts.

344

*/

344

*/

345

static void rcu_idle_enter_common(struct rcu_dynticks *rdtp, long long oldval)

345

static void rcu_idle_enter_common(struct rcu_dynticks *rdtp, long long oldval)

346

{

346

{

347

trace_rcu_dyntick("Start", oldval, 0);

347

trace_rcu_dyntick("Start", oldval, 0);

348

if (!is_idle_task(current)) {

348

if (!is_idle_task(current)) {

349

struct task_struct *idle = idle_task(smp_processor_id());

349

struct task_struct *idle = idle_task(smp_processor_id());

350

351

trace_rcu_dyntick("Error on entry: not idle task", oldval, 0);

351

trace_rcu_dyntick("Error on entry: not idle task", oldval, 0);

352

ftrace_dump(DUMP_ALL);

352

ftrace_dump(DUMP_ALL);

353

WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",

353

WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",

354

current->pid, current->comm,

354

current->pid, current->comm,

355

idle->pid, idle->comm); /* must be idle task! */

355

idle->pid, idle->comm); /* must be idle task! */

356

}

356

}

357

rcu_prepare_for_idle(smp_processor_id());

357

rcu_prepare_for_idle(smp_processor_id());

358

/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */

358

/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */

359

smp_mb__before_atomic_inc(); /* See above. */

359

smp_mb__before_atomic_inc(); /* See above. */

360

atomic_inc(&rdtp->dynticks);

360

atomic_inc(&rdtp->dynticks);

361

smp_mb__after_atomic_inc(); /* Force ordering with next sojourn. */

361

smp_mb__after_atomic_inc(); /* Force ordering with next sojourn. */

362

WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);

362

WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);

363

364

/*

364

/*

365

* The idle task is not permitted to enter the idle loop while

365

* The idle task is not permitted to enter the idle loop while

366

* in an RCU read-side critical section.

366

* in an RCU read-side critical section.

367

*/

367

*/

368

rcu_lockdep_assert(!lock_is_held(&rcu_lock_map),

368

rcu_lockdep_assert(!lock_is_held(&rcu_lock_map),

369

"Illegal idle entry in RCU read-side critical section.");

369

"Illegal idle entry in RCU read-side critical section.");

370

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map),

370

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map),

371

"Illegal idle entry in RCU-bh read-side critical section.");

371

"Illegal idle entry in RCU-bh read-side critical section.");

372

rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),

372

rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),

373

"Illegal idle entry in RCU-sched read-side critical section.");

373

"Illegal idle entry in RCU-sched read-side critical section.");

374

}

374

}

375

376

/**

376

/**

377

* rcu_idle_enter - inform RCU that current CPU is entering idle

377

* rcu_idle_enter - inform RCU that current CPU is entering idle

378

*

378

*

379

* Enter idle mode, in other words, -leave- the mode in which RCU

379

* Enter idle mode, in other words, -leave- the mode in which RCU

380

* read-side critical sections can occur. (Though RCU read-side

380

* read-side critical sections can occur. (Though RCU read-side

381

* critical sections can occur in irq handlers in idle, a possibility

381

* critical sections can occur in irq handlers in idle, a possibility

382

* handled by irq_enter() and irq_exit().)

382

* handled by irq_enter() and irq_exit().)

383

*

383

*

384

* We crowbar the ->dynticks_nesting field to zero to allow for

384

* We crowbar the ->dynticks_nesting field to zero to allow for

385

* the possibility of usermode upcalls having messed up our count

385

* the possibility of usermode upcalls having messed up our count

386

* of interrupt nesting level during the prior busy period.

386

* of interrupt nesting level during the prior busy period.

387

*/

387

*/

388

void rcu_idle_enter(void)

388

void rcu_idle_enter(void)

389

{

389

{

390

unsigned long flags;

390

unsigned long flags;

391

long long oldval;

391

long long oldval;

392

struct rcu_dynticks *rdtp;

392

struct rcu_dynticks *rdtp;

393

394

local_irq_save(flags);

394

local_irq_save(flags);

395

rdtp = &__get_cpu_var(rcu_dynticks);

395

rdtp = &__get_cpu_var(rcu_dynticks);

396

oldval = rdtp->dynticks_nesting;

396

oldval = rdtp->dynticks_nesting;

397

WARN_ON_ONCE((oldval & DYNTICK_TASK_NEST_MASK) == 0);

397

WARN_ON_ONCE((oldval & DYNTICK_TASK_NEST_MASK) == 0);

398

if ((oldval & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)

398

if ((oldval & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)

399

rdtp->dynticks_nesting = 0;

399

rdtp->dynticks_nesting = 0;

400

else

400

else

401

rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;

401

rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;

402

rcu_idle_enter_common(rdtp, oldval);

402

rcu_idle_enter_common(rdtp, oldval);

403

local_irq_restore(flags);

403

local_irq_restore(flags);

404

}

404

}

405

EXPORT_SYMBOL_GPL(rcu_idle_enter);

405

EXPORT_SYMBOL_GPL(rcu_idle_enter);

406

407

/**

407

/**

408

* rcu_irq_exit - inform RCU that current CPU is exiting irq towards idle

408

* rcu_irq_exit - inform RCU that current CPU is exiting irq towards idle

409

*

409

*

410

* Exit from an interrupt handler, which might possibly result in entering

410

* Exit from an interrupt handler, which might possibly result in entering

411

* idle mode, in other words, leaving the mode in which read-side critical

411

* idle mode, in other words, leaving the mode in which read-side critical

412

* sections can occur.

412

* sections can occur.

413

*

413

*

414

* This code assumes that the idle loop never does anything that might

414

* This code assumes that the idle loop never does anything that might

415

* result in unbalanced calls to irq_enter() and irq_exit(). If your

415

* result in unbalanced calls to irq_enter() and irq_exit(). If your

416

* architecture violates this assumption, RCU will give you what you

416

* architecture violates this assumption, RCU will give you what you

417

* deserve, good and hard. But very infrequently and irreproducibly.

417

* deserve, good and hard. But very infrequently and irreproducibly.

418

*

418

*

419

* Use things like work queues to work around this limitation.

419

* Use things like work queues to work around this limitation.

420

*

420

*

421

* You have been warned.

421

* You have been warned.

422

*/

422

*/

423

void rcu_irq_exit(void)

423

void rcu_irq_exit(void)

424

{

424

{

425

unsigned long flags;

425

unsigned long flags;

426

long long oldval;

426

long long oldval;

427

struct rcu_dynticks *rdtp;

427

struct rcu_dynticks *rdtp;

428

429

local_irq_save(flags);

429

local_irq_save(flags);

430

rdtp = &__get_cpu_var(rcu_dynticks);

430

rdtp = &__get_cpu_var(rcu_dynticks);

431

oldval = rdtp->dynticks_nesting;

431

oldval = rdtp->dynticks_nesting;

432

rdtp->dynticks_nesting--;

432

rdtp->dynticks_nesting--;

433

WARN_ON_ONCE(rdtp->dynticks_nesting < 0);

433

WARN_ON_ONCE(rdtp->dynticks_nesting < 0);

434

if (rdtp->dynticks_nesting)

434

if (rdtp->dynticks_nesting)

435

trace_rcu_dyntick("--=", oldval, rdtp->dynticks_nesting);

435

trace_rcu_dyntick("--=", oldval, rdtp->dynticks_nesting);

436

else

436

else

437

rcu_idle_enter_common(rdtp, oldval);

437

rcu_idle_enter_common(rdtp, oldval);

438

local_irq_restore(flags);

438

local_irq_restore(flags);

439

}

439

}

440

441

/*

441

/*

442

* rcu_idle_exit_common - inform RCU that current CPU is moving away from idle

442

* rcu_idle_exit_common - inform RCU that current CPU is moving away from idle

443

*

443

*

444

* If the new value of the ->dynticks_nesting counter was previously zero,

444

* If the new value of the ->dynticks_nesting counter was previously zero,

445

* we really have exited idle, and must do the appropriate accounting.

445

* we really have exited idle, and must do the appropriate accounting.

446

* The caller must have disabled interrupts.

446

* The caller must have disabled interrupts.

447

*/

447

*/

448

static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)

448

static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)

449

{

449

{

450

smp_mb__before_atomic_inc(); /* Force ordering w/previous sojourn. */

450

smp_mb__before_atomic_inc(); /* Force ordering w/previous sojourn. */

451

atomic_inc(&rdtp->dynticks);

451

atomic_inc(&rdtp->dynticks);

452

/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */

452

/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */

453

smp_mb__after_atomic_inc(); /* See above. */

453

smp_mb__after_atomic_inc(); /* See above. */

454

WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));

454

WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));

455

rcu_cleanup_after_idle(smp_processor_id());

455

rcu_cleanup_after_idle(smp_processor_id());

456

trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);

456

trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);

457

if (!is_idle_task(current)) {

457

if (!is_idle_task(current)) {

458

struct task_struct *idle = idle_task(smp_processor_id());

458

struct task_struct *idle = idle_task(smp_processor_id());

459

460

trace_rcu_dyntick("Error on exit: not idle task",

460

trace_rcu_dyntick("Error on exit: not idle task",

461

oldval, rdtp->dynticks_nesting);

461

oldval, rdtp->dynticks_nesting);

462

ftrace_dump(DUMP_ALL);

462

ftrace_dump(DUMP_ALL);

463

WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",

463

WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",

464

current->pid, current->comm,

464

current->pid, current->comm,

465

idle->pid, idle->comm); /* must be idle task! */

465

idle->pid, idle->comm); /* must be idle task! */

466

}

466

}

467

}

467

}

468

469

/**

469

/**

470

* rcu_idle_exit - inform RCU that current CPU is leaving idle

470

* rcu_idle_exit - inform RCU that current CPU is leaving idle

471

*

471

*

472

* Exit idle mode, in other words, -enter- the mode in which RCU

472

* Exit idle mode, in other words, -enter- the mode in which RCU

473

* read-side critical sections can occur.

473

* read-side critical sections can occur.

474

*

474

*

475

* We crowbar the ->dynticks_nesting field to DYNTICK_TASK_NEST to

475

* We crowbar the ->dynticks_nesting field to DYNTICK_TASK_NEST to

476

* allow for the possibility of usermode upcalls messing up our count

476

* allow for the possibility of usermode upcalls messing up our count

477

* of interrupt nesting level during the busy period that is just

477

* of interrupt nesting level during the busy period that is just

478

* now starting.

478

* now starting.

479

*/

479

*/

480

void rcu_idle_exit(void)

480

void rcu_idle_exit(void)

481

{

481

{

482

unsigned long flags;

482

unsigned long flags;

483

struct rcu_dynticks *rdtp;

483

struct rcu_dynticks *rdtp;

484

long long oldval;

484

long long oldval;

485

486

local_irq_save(flags);

486

local_irq_save(flags);

487

rdtp = &__get_cpu_var(rcu_dynticks);

487

rdtp = &__get_cpu_var(rcu_dynticks);

488

oldval = rdtp->dynticks_nesting;

488

oldval = rdtp->dynticks_nesting;

489

WARN_ON_ONCE(oldval < 0);

489

WARN_ON_ONCE(oldval < 0);

490

if (oldval & DYNTICK_TASK_NEST_MASK)

490

if (oldval & DYNTICK_TASK_NEST_MASK)

491

rdtp->dynticks_nesting += DYNTICK_TASK_NEST_VALUE;

491

rdtp->dynticks_nesting += DYNTICK_TASK_NEST_VALUE;

492

else

492

else

493

rdtp->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;

493

rdtp->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;

494

rcu_idle_exit_common(rdtp, oldval);

494

rcu_idle_exit_common(rdtp, oldval);

495

local_irq_restore(flags);

495

local_irq_restore(flags);

496

}

496

}

497

EXPORT_SYMBOL_GPL(rcu_idle_exit);

497

EXPORT_SYMBOL_GPL(rcu_idle_exit);

498

499

/**

499

/**

500

* rcu_irq_enter - inform RCU that current CPU is entering irq away from idle

500

* rcu_irq_enter - inform RCU that current CPU is entering irq away from idle

501

*

501

*

502

* Enter an interrupt handler, which might possibly result in exiting

502

* Enter an interrupt handler, which might possibly result in exiting

503

* idle mode, in other words, entering the mode in which read-side critical

503

* idle mode, in other words, entering the mode in which read-side critical

504

* sections can occur.

504

* sections can occur.

505

*

505

*

506

* Note that the Linux kernel is fully capable of entering an interrupt

506

* Note that the Linux kernel is fully capable of entering an interrupt

507

* handler that it never exits, for example when doing upcalls to

507

* handler that it never exits, for example when doing upcalls to

508

* user mode! This code assumes that the idle loop never does upcalls to

508

* user mode! This code assumes that the idle loop never does upcalls to

509

* user mode. If your architecture does do upcalls from the idle loop (or

509

* user mode. If your architecture does do upcalls from the idle loop (or

510

* does anything else that results in unbalanced calls to the irq_enter()

510

* does anything else that results in unbalanced calls to the irq_enter()

511

* and irq_exit() functions), RCU will give you what you deserve, good

511

* and irq_exit() functions), RCU will give you what you deserve, good

512

* and hard. But very infrequently and irreproducibly.

512

* and hard. But very infrequently and irreproducibly.

513

*

513

*

514

* Use things like work queues to work around this limitation.

514

* Use things like work queues to work around this limitation.

515

*

515

*

516

* You have been warned.

516

* You have been warned.

517

*/

517

*/

518

void rcu_irq_enter(void)

518

void rcu_irq_enter(void)

519

{

519

{

520

unsigned long flags;

520

unsigned long flags;

521

struct rcu_dynticks *rdtp;

521

struct rcu_dynticks *rdtp;

522

long long oldval;

522

long long oldval;

523

524

local_irq_save(flags);

524

local_irq_save(flags);

525

rdtp = &__get_cpu_var(rcu_dynticks);

525

rdtp = &__get_cpu_var(rcu_dynticks);

526

oldval = rdtp->dynticks_nesting;

526

oldval = rdtp->dynticks_nesting;

527

rdtp->dynticks_nesting++;

527

rdtp->dynticks_nesting++;

528

WARN_ON_ONCE(rdtp->dynticks_nesting == 0);

528

WARN_ON_ONCE(rdtp->dynticks_nesting == 0);

529

if (oldval)

529

if (oldval)

530

trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);

530

trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);

531

else

531

else

532

rcu_idle_exit_common(rdtp, oldval);

532

rcu_idle_exit_common(rdtp, oldval);

533

local_irq_restore(flags);

533

local_irq_restore(flags);

534

}

534

}

535

536

/**

536

/**

537

* rcu_nmi_enter - inform RCU of entry to NMI context

537

* rcu_nmi_enter - inform RCU of entry to NMI context

538

*

538

*

539

* If the CPU was idle with dynamic ticks active, and there is no

539

* If the CPU was idle with dynamic ticks active, and there is no

540

* irq handler running, this updates rdtp->dynticks_nmi to let the

540

* irq handler running, this updates rdtp->dynticks_nmi to let the

541

* RCU grace-period handling know that the CPU is active.

541

* RCU grace-period handling know that the CPU is active.

542

*/

542

*/

543

void rcu_nmi_enter(void)

543

void rcu_nmi_enter(void)

544

{

544

{

545

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

545

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

546

547

if (rdtp->dynticks_nmi_nesting == 0 &&

547

if (rdtp->dynticks_nmi_nesting == 0 &&

548

(atomic_read(&rdtp->dynticks) & 0x1))

548

(atomic_read(&rdtp->dynticks) & 0x1))

549

return;

549

return;

550

rdtp->dynticks_nmi_nesting++;

550

rdtp->dynticks_nmi_nesting++;

551

smp_mb__before_atomic_inc(); /* Force delay from prior write. */

551

smp_mb__before_atomic_inc(); /* Force delay from prior write. */

552

atomic_inc(&rdtp->dynticks);

552

atomic_inc(&rdtp->dynticks);

553

/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */

553

/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */

554

smp_mb__after_atomic_inc(); /* See above. */

554

smp_mb__after_atomic_inc(); /* See above. */

555

WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));

555

WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));

556

}

556

}

557

558

/**

558

/**

559

* rcu_nmi_exit - inform RCU of exit from NMI context

559

* rcu_nmi_exit - inform RCU of exit from NMI context

560

*

560

*

561

* If the CPU was idle with dynamic ticks active, and there is no

561

* If the CPU was idle with dynamic ticks active, and there is no

562

* irq handler running, this updates rdtp->dynticks_nmi to let the

562

* irq handler running, this updates rdtp->dynticks_nmi to let the

563

* RCU grace-period handling know that the CPU is no longer active.

563

* RCU grace-period handling know that the CPU is no longer active.

564

*/

564

*/

565

void rcu_nmi_exit(void)

565

void rcu_nmi_exit(void)

566

{

566

{

567

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

567

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

568

569

if (rdtp->dynticks_nmi_nesting == 0 ||

569

if (rdtp->dynticks_nmi_nesting == 0 ||

570

--rdtp->dynticks_nmi_nesting != 0)

570

--rdtp->dynticks_nmi_nesting != 0)

571

return;

571

return;

572

/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */

572

/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */

573

smp_mb__before_atomic_inc(); /* See above. */

573

smp_mb__before_atomic_inc(); /* See above. */

574

atomic_inc(&rdtp->dynticks);

574

atomic_inc(&rdtp->dynticks);

575

smp_mb__after_atomic_inc(); /* Force delay to next write. */

575

smp_mb__after_atomic_inc(); /* Force delay to next write. */

576

WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);

576

WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);

577

}

577

}

578

579

#ifdef CONFIG_PROVE_RCU

579

#ifdef CONFIG_PROVE_RCU

580

581

/**

581

/**

582

* rcu_is_cpu_idle - see if RCU thinks that the current CPU is idle

582

* rcu_is_cpu_idle - see if RCU thinks that the current CPU is idle

583

*

583

*

584

* If the current CPU is in its idle loop and is neither in an interrupt

584

* If the current CPU is in its idle loop and is neither in an interrupt

585

* or NMI handler, return true.

585

* or NMI handler, return true.

586

*/

586

*/

587

int rcu_is_cpu_idle(void)

587

int rcu_is_cpu_idle(void)

588

{

588

{

589

int ret;

589

int ret;

590

591

preempt_disable();

591

preempt_disable();

592

ret = (atomic_read(&__get_cpu_var(rcu_dynticks).dynticks) & 0x1) == 0;

592

ret = (atomic_read(&__get_cpu_var(rcu_dynticks).dynticks) & 0x1) == 0;

593

preempt_enable();

593

preempt_enable();

594

return ret;

594

return ret;

595

}

595

}

596

EXPORT_SYMBOL(rcu_is_cpu_idle);

596

EXPORT_SYMBOL(rcu_is_cpu_idle);

597

598

#ifdef CONFIG_HOTPLUG_CPU

598

#ifdef CONFIG_HOTPLUG_CPU

599

600

/*

600

/*

601

* Is the current CPU online? Disable preemption to avoid false positives

601

* Is the current CPU online? Disable preemption to avoid false positives

602

* that could otherwise happen due to the current CPU number being sampled,

602

* that could otherwise happen due to the current CPU number being sampled,

603

* this task being preempted, its old CPU being taken offline, resuming

603

* this task being preempted, its old CPU being taken offline, resuming

604

* on some other CPU, then determining that its old CPU is now offline.

604

* on some other CPU, then determining that its old CPU is now offline.

605

* It is OK to use RCU on an offline processor during initial boot, hence

605

* It is OK to use RCU on an offline processor during initial boot, hence

606

* the check for rcu_scheduler_fully_active. Note also that it is OK

606

* the check for rcu_scheduler_fully_active. Note also that it is OK

607

* for a CPU coming online to use RCU for one jiffy prior to marking itself

607

* for a CPU coming online to use RCU for one jiffy prior to marking itself

608

* online in the cpu_online_mask. Similarly, it is OK for a CPU going

608

* online in the cpu_online_mask. Similarly, it is OK for a CPU going

609

* offline to continue to use RCU for one jiffy after marking itself

609

* offline to continue to use RCU for one jiffy after marking itself

610

* offline in the cpu_online_mask. This leniency is necessary given the

610

* offline in the cpu_online_mask. This leniency is necessary given the

611

* non-atomic nature of the online and offline processing, for example,

611

* non-atomic nature of the online and offline processing, for example,

612

* the fact that a CPU enters the scheduler after completing the CPU_DYING

612

* the fact that a CPU enters the scheduler after completing the CPU_DYING

613

* notifiers.

613

* notifiers.

614

*

614

*

615

* This is also why RCU internally marks CPUs online during the

615

* This is also why RCU internally marks CPUs online during the

616

* CPU_UP_PREPARE phase and offline during the CPU_DEAD phase.

616

* CPU_UP_PREPARE phase and offline during the CPU_DEAD phase.

617

*

617

*

618

* Disable checking if in an NMI handler because we cannot safely report

618

* Disable checking if in an NMI handler because we cannot safely report

619

* errors from NMI handlers anyway.

619

* errors from NMI handlers anyway.

620

*/

620

*/

621

bool rcu_lockdep_current_cpu_online(void)

621

bool rcu_lockdep_current_cpu_online(void)

622

{

622

{

623

struct rcu_data *rdp;

623

struct rcu_data *rdp;

624

struct rcu_node *rnp;

624

struct rcu_node *rnp;

625

bool ret;

625

bool ret;

626

627

if (in_nmi())

627

if (in_nmi())

628

return 1;

628

return 1;

629

preempt_disable();

629

preempt_disable();

630

rdp = &__get_cpu_var(rcu_sched_data);

630

rdp = &__get_cpu_var(rcu_sched_data);

631

rnp = rdp->mynode;

631

rnp = rdp->mynode;

632

ret = (rdp->grpmask & rnp->qsmaskinit) ||

632

ret = (rdp->grpmask & rnp->qsmaskinit) ||

633

!rcu_scheduler_fully_active;

633

!rcu_scheduler_fully_active;

634

preempt_enable();

634

preempt_enable();

635

return ret;

635

return ret;

636

}

636

}

637

EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);

637

EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);

638

639

#endif /* #ifdef CONFIG_HOTPLUG_CPU */

639

#endif /* #ifdef CONFIG_HOTPLUG_CPU */

640

641

#endif /* #ifdef CONFIG_PROVE_RCU */

641

#endif /* #ifdef CONFIG_PROVE_RCU */

642

643

/**

643

/**

644

* rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle

644

* rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle

645

*

645

*

646

* If the current CPU is idle or running at a first-level (not nested)

646

* If the current CPU is idle or running at a first-level (not nested)

647

* interrupt from idle, return true. The caller must have at least

647

* interrupt from idle, return true. The caller must have at least

648

* disabled preemption.

648

* disabled preemption.

649

*/

649

*/

650

int rcu_is_cpu_rrupt_from_idle(void)

650

int rcu_is_cpu_rrupt_from_idle(void)

651

{

651

{

652

return __get_cpu_var(rcu_dynticks).dynticks_nesting <= 1;

652

return __get_cpu_var(rcu_dynticks).dynticks_nesting <= 1;

653

}

653

}

654

655

/*

655

/*

656

* Snapshot the specified CPU's dynticks counter so that we can later

656

* Snapshot the specified CPU's dynticks counter so that we can later

657

* credit them with an implicit quiescent state. Return 1 if this CPU

657

* credit them with an implicit quiescent state. Return 1 if this CPU

658

* is in dynticks idle mode, which is an extended quiescent state.

658

* is in dynticks idle mode, which is an extended quiescent state.

659

*/

659

*/

660

static int dyntick_save_progress_counter(struct rcu_data *rdp)

660

static int dyntick_save_progress_counter(struct rcu_data *rdp)

661

{

661

{

662

rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);

662

rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);

663

return (rdp->dynticks_snap & 0x1) == 0;

663

return (rdp->dynticks_snap & 0x1) == 0;

664

}

664

}

665

666

/*

666

/*

667

* Return true if the specified CPU has passed through a quiescent

667

* Return true if the specified CPU has passed through a quiescent

668

* state by virtue of being in or having passed through an dynticks

668

* state by virtue of being in or having passed through an dynticks

669

* idle state since the last call to dyntick_save_progress_counter()

669

* idle state since the last call to dyntick_save_progress_counter()

670

* for this same CPU.

670

* for this same CPU.

671

*/

671

*/

672

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

672

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

673

{

673

{

674

unsigned int curr;

674

unsigned int curr;

675

unsigned int snap;

675

unsigned int snap;

676

677

curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);

677

curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);

678

snap = (unsigned int)rdp->dynticks_snap;

678

snap = (unsigned int)rdp->dynticks_snap;

679

680

/*

680

/*

681

* If the CPU passed through or entered a dynticks idle phase with

681

* If the CPU passed through or entered a dynticks idle phase with

682

* no active irq/NMI handlers, then we can safely pretend that the CPU

682

* no active irq/NMI handlers, then we can safely pretend that the CPU

683

* already acknowledged the request to pass through a quiescent

683

* already acknowledged the request to pass through a quiescent

684

* state. Either way, that CPU cannot possibly be in an RCU

684

* state. Either way, that CPU cannot possibly be in an RCU

685

* read-side critical section that started before the beginning

685

* read-side critical section that started before the beginning

686

* of the current RCU grace period.

686

* of the current RCU grace period.

687

*/

687

*/

688

if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {

688

if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {

689

trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "dti");

689

trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "dti");

690

rdp->dynticks_fqs++;

690

rdp->dynticks_fqs++;

691

return 1;

691

return 1;

692

}

692

}

693

694

/* Go check for the CPU being offline. */

694

/* Go check for the CPU being offline. */

695

return rcu_implicit_offline_qs(rdp);

695

return rcu_implicit_offline_qs(rdp);

696

}

696

}

697

698

static int jiffies_till_stall_check(void)

698

static int jiffies_till_stall_check(void)

699

{

699

{

700

int till_stall_check = ACCESS_ONCE(rcu_cpu_stall_timeout);

700

int till_stall_check = ACCESS_ONCE(rcu_cpu_stall_timeout);

701

702

/*

702

/*

703

* Limit check must be consistent with the Kconfig limits

703

* Limit check must be consistent with the Kconfig limits

704

* for CONFIG_RCU_CPU_STALL_TIMEOUT.

704

* for CONFIG_RCU_CPU_STALL_TIMEOUT.

705

*/

705

*/

706

if (till_stall_check < 3) {

706

if (till_stall_check < 3) {

707

ACCESS_ONCE(rcu_cpu_stall_timeout) = 3;

707

ACCESS_ONCE(rcu_cpu_stall_timeout) = 3;

708

till_stall_check = 3;

708

till_stall_check = 3;

709

} else if (till_stall_check > 300) {

709

} else if (till_stall_check > 300) {

710

ACCESS_ONCE(rcu_cpu_stall_timeout) = 300;

710

ACCESS_ONCE(rcu_cpu_stall_timeout) = 300;

711

till_stall_check = 300;

711

till_stall_check = 300;

712

}

712

}

713

return till_stall_check * HZ + RCU_STALL_DELAY_DELTA;

713

return till_stall_check * HZ + RCU_STALL_DELAY_DELTA;

714

}

714

}

715

716

static void record_gp_stall_check_time(struct rcu_state *rsp)

716

static void record_gp_stall_check_time(struct rcu_state *rsp)

717

{

717

{

718

rsp->gp_start = jiffies;

718

rsp->gp_start = jiffies;

719

rsp->jiffies_stall = jiffies + jiffies_till_stall_check();

719

rsp->jiffies_stall = jiffies + jiffies_till_stall_check();

720

}

720

}

721

722

static void print_other_cpu_stall(struct rcu_state *rsp)

722

static void print_other_cpu_stall(struct rcu_state *rsp)

723

{

723

{

724

int cpu;

724

int cpu;

725

long delta;

725

long delta;

726

unsigned long flags;

726

unsigned long flags;

727

int ndetected;

727

int ndetected;

728

struct rcu_node *rnp = rcu_get_root(rsp);

728

struct rcu_node *rnp = rcu_get_root(rsp);

729

730

/* Only let one CPU complain about others per time interval. */

730

/* Only let one CPU complain about others per time interval. */

731

732

raw_spin_lock_irqsave(&rnp->lock, flags);

732

raw_spin_lock_irqsave(&rnp->lock, flags);

733

delta = jiffies - rsp->jiffies_stall;

733

delta = jiffies - rsp->jiffies_stall;

734

if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {

734

if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {

735

raw_spin_unlock_irqrestore(&rnp->lock, flags);

735

raw_spin_unlock_irqrestore(&rnp->lock, flags);

736

return;

736

return;

737

}

737

}

738

rsp->jiffies_stall = jiffies + 3 * jiffies_till_stall_check() + 3;

738

rsp->jiffies_stall = jiffies + 3 * jiffies_till_stall_check() + 3;

739

raw_spin_unlock_irqrestore(&rnp->lock, flags);

739

raw_spin_unlock_irqrestore(&rnp->lock, flags);

740

741

/*

741

/*

742

* OK, time to rat on our buddy...

742

* OK, time to rat on our buddy...

743

* See Documentation/RCU/stallwarn.txt for info on how to debug

743

* See Documentation/RCU/stallwarn.txt for info on how to debug

744

* RCU CPU stall warnings.

744

* RCU CPU stall warnings.

745

*/

745

*/

746

printk(KERN_ERR "INFO: %s detected stalls on CPUs/tasks:",

746

printk(KERN_ERR "INFO: %s detected stalls on CPUs/tasks:",

747

rsp->name);

747

rsp->name);

748

print_cpu_stall_info_begin();

748

print_cpu_stall_info_begin();

749

rcu_for_each_leaf_node(rsp, rnp) {

749

rcu_for_each_leaf_node(rsp, rnp) {

750

raw_spin_lock_irqsave(&rnp->lock, flags);

750

raw_spin_lock_irqsave(&rnp->lock, flags);

751

ndetected += rcu_print_task_stall(rnp);

751

ndetected += rcu_print_task_stall(rnp);

752

raw_spin_unlock_irqrestore(&rnp->lock, flags);

752

raw_spin_unlock_irqrestore(&rnp->lock, flags);

753

if (rnp->qsmask == 0)

753

if (rnp->qsmask == 0)

754

continue;

754

continue;

755

for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)

755

for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)

756

if (rnp->qsmask & (1UL << cpu)) {

756

if (rnp->qsmask & (1UL << cpu)) {

757

print_cpu_stall_info(rsp, rnp->grplo + cpu);

757

print_cpu_stall_info(rsp, rnp->grplo + cpu);

758

ndetected++;

758

ndetected++;

759

}

759

}

760

}

760

}

761

762

/*

762

/*

763

* Now rat on any tasks that got kicked up to the root rcu_node

763

* Now rat on any tasks that got kicked up to the root rcu_node

764

* due to CPU offlining.

764

* due to CPU offlining.

765

*/

765

*/

766

rnp = rcu_get_root(rsp);

766

rnp = rcu_get_root(rsp);

767

raw_spin_lock_irqsave(&rnp->lock, flags);

767

raw_spin_lock_irqsave(&rnp->lock, flags);

768

ndetected = rcu_print_task_stall(rnp);

768

ndetected = rcu_print_task_stall(rnp);

769

raw_spin_unlock_irqrestore(&rnp->lock, flags);

769

raw_spin_unlock_irqrestore(&rnp->lock, flags);

770

771

print_cpu_stall_info_end();

771

print_cpu_stall_info_end();

772

printk(KERN_CONT "(detected by %d, t=%ld jiffies)\n",

772

printk(KERN_CONT "(detected by %d, t=%ld jiffies)\n",

773

smp_processor_id(), (long)(jiffies - rsp->gp_start));

773

smp_processor_id(), (long)(jiffies - rsp->gp_start));

774

if (ndetected == 0)

774

if (ndetected == 0)

775

printk(KERN_ERR "INFO: Stall ended before state dump start\n");

775

printk(KERN_ERR "INFO: Stall ended before state dump start\n");

776

else if (!trigger_all_cpu_backtrace())

776

else if (!trigger_all_cpu_backtrace())

777

dump_stack();

777

dump_stack();

778

779

/* If so configured, complain about tasks blocking the grace period. */

779

/* If so configured, complain about tasks blocking the grace period. */

780

781

rcu_print_detail_task_stall(rsp);

781

rcu_print_detail_task_stall(rsp);

782

783

force_quiescent_state(rsp, 0); /* Kick them all. */

783

force_quiescent_state(rsp, 0); /* Kick them all. */

784

}

784

}

785

786

static void print_cpu_stall(struct rcu_state *rsp)

786

static void print_cpu_stall(struct rcu_state *rsp)

787

{

787

{

788

unsigned long flags;

788

unsigned long flags;

789

struct rcu_node *rnp = rcu_get_root(rsp);

789

struct rcu_node *rnp = rcu_get_root(rsp);

790

791

/*

791

/*

792

* OK, time to rat on ourselves...

792

* OK, time to rat on ourselves...

793

* See Documentation/RCU/stallwarn.txt for info on how to debug

793

* See Documentation/RCU/stallwarn.txt for info on how to debug

794

* RCU CPU stall warnings.

794

* RCU CPU stall warnings.

795

*/

795

*/

796

printk(KERN_ERR "INFO: %s self-detected stall on CPU", rsp->name);

796

printk(KERN_ERR "INFO: %s self-detected stall on CPU", rsp->name);

797

print_cpu_stall_info_begin();

797

print_cpu_stall_info_begin();

798

print_cpu_stall_info(rsp, smp_processor_id());

798

print_cpu_stall_info(rsp, smp_processor_id());

799

print_cpu_stall_info_end();

799

print_cpu_stall_info_end();

800

printk(KERN_CONT " (t=%lu jiffies)\n", jiffies - rsp->gp_start);

800

printk(KERN_CONT " (t=%lu jiffies)\n", jiffies - rsp->gp_start);

801

if (!trigger_all_cpu_backtrace())

801

if (!trigger_all_cpu_backtrace())

802

dump_stack();

802

dump_stack();

803

804

raw_spin_lock_irqsave(&rnp->lock, flags);

804

raw_spin_lock_irqsave(&rnp->lock, flags);

805

if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))

805

if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))

806

rsp->jiffies_stall = jiffies +

806

rsp->jiffies_stall = jiffies +

807

3 * jiffies_till_stall_check() + 3;

807

3 * jiffies_till_stall_check() + 3;

808

raw_spin_unlock_irqrestore(&rnp->lock, flags);

808

raw_spin_unlock_irqrestore(&rnp->lock, flags);

809

810

set_need_resched(); /* kick ourselves to get things going. */

810

set_need_resched(); /* kick ourselves to get things going. */

811

}

811

}

812

813

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)

813

static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)

814

{

814

{

815

unsigned long j;

815

unsigned long j;

816

unsigned long js;

816

unsigned long js;

817

struct rcu_node *rnp;

817

struct rcu_node *rnp;

818

819

if (rcu_cpu_stall_suppress)

819

if (rcu_cpu_stall_suppress)

820

return;

820

return;

821

j = ACCESS_ONCE(jiffies);

821

j = ACCESS_ONCE(jiffies);

822

js = ACCESS_ONCE(rsp->jiffies_stall);

822

js = ACCESS_ONCE(rsp->jiffies_stall);

823

rnp = rdp->mynode;

823

rnp = rdp->mynode;

824

if ((ACCESS_ONCE(rnp->qsmask) & rdp->grpmask) && ULONG_CMP_GE(j, js)) {

824

if ((ACCESS_ONCE(rnp->qsmask) & rdp->grpmask) && ULONG_CMP_GE(j, js)) {

825

826

/* We haven't checked in, so go dump stack. */

826

/* We haven't checked in, so go dump stack. */

827

print_cpu_stall(rsp);

827

print_cpu_stall(rsp);

828

829

} else if (rcu_gp_in_progress(rsp) &&

829

} else if (rcu_gp_in_progress(rsp) &&

830

ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {

830

ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {

831

832

/* They had a few time units to dump stack, so complain. */

832

/* They had a few time units to dump stack, so complain. */

833

print_other_cpu_stall(rsp);

833

print_other_cpu_stall(rsp);

834

}

834

}

835

}

835

}

836

837

static int rcu_panic(struct notifier_block *this, unsigned long ev, void *ptr)

837

static int rcu_panic(struct notifier_block *this, unsigned long ev, void *ptr)

838

{

838

{

839

rcu_cpu_stall_suppress = 1;

839

rcu_cpu_stall_suppress = 1;

840

return NOTIFY_DONE;

840

return NOTIFY_DONE;

841

}

841

}

842

843

/**

843

/**

844

* rcu_cpu_stall_reset - prevent further stall warnings in current grace period

844

* rcu_cpu_stall_reset - prevent further stall warnings in current grace period

845

*

845

*

846

* Set the stall-warning timeout way off into the future, thus preventing

846

* Set the stall-warning timeout way off into the future, thus preventing

847

* any RCU CPU stall-warning messages from appearing in the current set of

847

* any RCU CPU stall-warning messages from appearing in the current set of

848

* RCU grace periods.

848

* RCU grace periods.

849

*

849

*

850

* The caller must disable hard irqs.

850

* The caller must disable hard irqs.

851

*/

851

*/

852

void rcu_cpu_stall_reset(void)

852

void rcu_cpu_stall_reset(void)

853

{

853

{

854

rcu_sched_state.jiffies_stall = jiffies + ULONG_MAX / 2;

854

rcu_sched_state.jiffies_stall = jiffies + ULONG_MAX / 2;

855

rcu_bh_state.jiffies_stall = jiffies + ULONG_MAX / 2;

855

rcu_bh_state.jiffies_stall = jiffies + ULONG_MAX / 2;

856

rcu_preempt_stall_reset();

856

rcu_preempt_stall_reset();

857

}

857

}

858

859

static struct notifier_block rcu_panic_block = {

859

static struct notifier_block rcu_panic_block = {

860

.notifier_call = rcu_panic,

860

.notifier_call = rcu_panic,

861

};

861

};

862

863

static void __init check_cpu_stall_init(void)

863

static void __init check_cpu_stall_init(void)

864

{

864

{

865

atomic_notifier_chain_register(&panic_notifier_list, &rcu_panic_block);

865

atomic_notifier_chain_register(&panic_notifier_list, &rcu_panic_block);

866

}

866

}

867

868

/*

868

/*

869

* Update CPU-local rcu_data state to record the newly noticed grace period.

869

* Update CPU-local rcu_data state to record the newly noticed grace period.

870

* This is used both when we started the grace period and when we notice

870

* This is used both when we started the grace period and when we notice

871

* that someone else started the grace period. The caller must hold the

871

* that someone else started the grace period. The caller must hold the

872

* ->lock of the leaf rcu_node structure corresponding to the current CPU,

872

* ->lock of the leaf rcu_node structure corresponding to the current CPU,

873

* and must have irqs disabled.

873

* and must have irqs disabled.

874

*/

874

*/

875

static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

875

static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

876

{

876

{

877

if (rdp->gpnum != rnp->gpnum) {

877

if (rdp->gpnum != rnp->gpnum) {

878

/*

878

/*

879

* If the current grace period is waiting for this CPU,

879

* If the current grace period is waiting for this CPU,

880

* set up to detect a quiescent state, otherwise don't

880

* set up to detect a quiescent state, otherwise don't

881

* go looking for one.

881

* go looking for one.

882

*/

882

*/

883

rdp->gpnum = rnp->gpnum;

883

rdp->gpnum = rnp->gpnum;

884

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpustart");

884

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpustart");

885

if (rnp->qsmask & rdp->grpmask) {

885

if (rnp->qsmask & rdp->grpmask) {

886

rdp->qs_pending = 1;

886

rdp->qs_pending = 1;

887

rdp->passed_quiesce = 0;

887

rdp->passed_quiesce = 0;

888

} else

888

} else

889

rdp->qs_pending = 0;

889

rdp->qs_pending = 0;

890

zero_cpu_stall_ticks(rdp);

890

zero_cpu_stall_ticks(rdp);

891

}

891

}

892

}

892

}

893

894

static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)

894

static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)

895

{

895

{

896

unsigned long flags;

896

unsigned long flags;

897

struct rcu_node *rnp;

897

struct rcu_node *rnp;

898

899

local_irq_save(flags);

899

local_irq_save(flags);

900

rnp = rdp->mynode;

900

rnp = rdp->mynode;

901

if (rdp->gpnum == ACCESS_ONCE(rnp->gpnum) || /* outside lock. */

901

if (rdp->gpnum == ACCESS_ONCE(rnp->gpnum) || /* outside lock. */

902

!raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */

902

!raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */

903

local_irq_restore(flags);

903

local_irq_restore(flags);

904

return;

904

return;

905

}

905

}

906

__note_new_gpnum(rsp, rnp, rdp);

906

__note_new_gpnum(rsp, rnp, rdp);

907

raw_spin_unlock_irqrestore(&rnp->lock, flags);

907

raw_spin_unlock_irqrestore(&rnp->lock, flags);

908

}

908

}

909

910

/*

910

/*

911

* Did someone else start a new RCU grace period start since we last

911

* Did someone else start a new RCU grace period start since we last

912

* checked? Update local state appropriately if so. Must be called

912

* checked? Update local state appropriately if so. Must be called

913

* on the CPU corresponding to rdp.

913

* on the CPU corresponding to rdp.

914

*/

914

*/

915

static int

915

static int

916

check_for_new_grace_period(struct rcu_state *rsp, struct rcu_data *rdp)

916

check_for_new_grace_period(struct rcu_state *rsp, struct rcu_data *rdp)

917

{

917

{

918

unsigned long flags;

918

unsigned long flags;

919

int ret = 0;

919

int ret = 0;

920

921

local_irq_save(flags);

921

local_irq_save(flags);

922

if (rdp->gpnum != rsp->gpnum) {

922

if (rdp->gpnum != rsp->gpnum) {

923

note_new_gpnum(rsp, rdp);

923

note_new_gpnum(rsp, rdp);

924

ret = 1;

924

ret = 1;

925

}

925

}

926

local_irq_restore(flags);

926

local_irq_restore(flags);

927

return ret;

927

return ret;

928

}

928

}

929

930

/*

930

/*

931

* Advance this CPU's callbacks, but only if the current grace period

931

* Advance this CPU's callbacks, but only if the current grace period

932

* has ended. This may be called only from the CPU to whom the rdp

932

* has ended. This may be called only from the CPU to whom the rdp

933

* belongs. In addition, the corresponding leaf rcu_node structure's

933

* belongs. In addition, the corresponding leaf rcu_node structure's

934

* ->lock must be held by the caller, with irqs disabled.

934

* ->lock must be held by the caller, with irqs disabled.

935

*/

935

*/

936

static void

936

static void

937

__rcu_process_gp_end(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

937

__rcu_process_gp_end(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

938

{

938

{

939

/* Did another grace period end? */

939

/* Did another grace period end? */

940

if (rdp->completed != rnp->completed) {

940

if (rdp->completed != rnp->completed) {

941

942

/* Advance callbacks. No harm if list empty. */

942

/* Advance callbacks. No harm if list empty. */

943

rdp->nxttail[RCU_DONE_TAIL] = rdp->nxttail[RCU_WAIT_TAIL];

943

rdp->nxttail[RCU_DONE_TAIL] = rdp->nxttail[RCU_WAIT_TAIL];

944

rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_READY_TAIL];

944

rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_READY_TAIL];

945

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

945

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

946

947

/* Remember that we saw this grace-period completion. */

947

/* Remember that we saw this grace-period completion. */

948

rdp->completed = rnp->completed;

948

rdp->completed = rnp->completed;

949

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuend");

949

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuend");

950

951

/*

951

/*

952

* If we were in an extended quiescent state, we may have

952

* If we were in an extended quiescent state, we may have

953

* missed some grace periods that others CPUs handled on

953

* missed some grace periods that others CPUs handled on

954

* our behalf. Catch up with this state to avoid noting

954

* our behalf. Catch up with this state to avoid noting

955

* spurious new grace periods. If another grace period

955

* spurious new grace periods. If another grace period

956

* has started, then rnp->gpnum will have advanced, so

956

* has started, then rnp->gpnum will have advanced, so

957

* we will detect this later on.

957

* we will detect this later on.

958

*/

958

*/

959

if (ULONG_CMP_LT(rdp->gpnum, rdp->completed))

959

if (ULONG_CMP_LT(rdp->gpnum, rdp->completed))

960

rdp->gpnum = rdp->completed;

960

rdp->gpnum = rdp->completed;

961

962

/*

962

/*

963

* If RCU does not need a quiescent state from this CPU,

963

* If RCU does not need a quiescent state from this CPU,

964

* then make sure that this CPU doesn't go looking for one.

964

* then make sure that this CPU doesn't go looking for one.

965

*/

965

*/

966

if ((rnp->qsmask & rdp->grpmask) == 0)

966

if ((rnp->qsmask & rdp->grpmask) == 0)

967

rdp->qs_pending = 0;

967

rdp->qs_pending = 0;

968

}

968

}

969

}

969

}

970

971

/*

971

/*

972

* Advance this CPU's callbacks, but only if the current grace period

972

* Advance this CPU's callbacks, but only if the current grace period

973

* has ended. This may be called only from the CPU to whom the rdp

973

* has ended. This may be called only from the CPU to whom the rdp

974

* belongs.

974

* belongs.

975

*/

975

*/

976

static void

976

static void

977

rcu_process_gp_end(struct rcu_state *rsp, struct rcu_data *rdp)

977

rcu_process_gp_end(struct rcu_state *rsp, struct rcu_data *rdp)

978

{

978

{

979

unsigned long flags;

979

unsigned long flags;

980

struct rcu_node *rnp;

980

struct rcu_node *rnp;

981

982

local_irq_save(flags);

982

local_irq_save(flags);

983

rnp = rdp->mynode;

983

rnp = rdp->mynode;

984

if (rdp->completed == ACCESS_ONCE(rnp->completed) || /* outside lock. */

984

if (rdp->completed == ACCESS_ONCE(rnp->completed) || /* outside lock. */

985

!raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */

985

!raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */

986

local_irq_restore(flags);

986

local_irq_restore(flags);

987

return;

987

return;

988

}

988

}

989

__rcu_process_gp_end(rsp, rnp, rdp);

989

__rcu_process_gp_end(rsp, rnp, rdp);

990

raw_spin_unlock_irqrestore(&rnp->lock, flags);

990

raw_spin_unlock_irqrestore(&rnp->lock, flags);

991

}

991

}

992

993

/*

993

/*

994

* Do per-CPU grace-period initialization for running CPU. The caller

994

* Do per-CPU grace-period initialization for running CPU. The caller

995

* must hold the lock of the leaf rcu_node structure corresponding to

995

* must hold the lock of the leaf rcu_node structure corresponding to

996

* this CPU.

996

* this CPU.

997

*/

997

*/

998

static void

998

static void

999

rcu_start_gp_per_cpu(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

999

rcu_start_gp_per_cpu(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)

1000

{

1000

{

1001

/* Prior grace period ended, so advance callbacks for current CPU. */

1001

/* Prior grace period ended, so advance callbacks for current CPU. */

1002

__rcu_process_gp_end(rsp, rnp, rdp);

1002

__rcu_process_gp_end(rsp, rnp, rdp);

1003

1004

/*

1004

/*

1005

* Because this CPU just now started the new grace period, we know

1005

* Because this CPU just now started the new grace period, we know

1006

* that all of its callbacks will be covered by this upcoming grace

1006

* that all of its callbacks will be covered by this upcoming grace

1007

* period, even the ones that were registered arbitrarily recently.

1007

* period, even the ones that were registered arbitrarily recently.

1008

* Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.

1008

* Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.

1009

*

1009

*

1010

* Other CPUs cannot be sure exactly when the grace period started.

1010

* Other CPUs cannot be sure exactly when the grace period started.

1011

* Therefore, their recently registered callbacks must pass through

1011

* Therefore, their recently registered callbacks must pass through

1012

* an additional RCU_NEXT_READY stage, so that they will be handled

1012

* an additional RCU_NEXT_READY stage, so that they will be handled

1013

* by the next RCU grace period.

1013

* by the next RCU grace period.

1014

*/

1014

*/

1015

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1015

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1016

rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1016

rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1017

1018

/* Set state so that this CPU will detect the next quiescent state. */

1018

/* Set state so that this CPU will detect the next quiescent state. */

1019

__note_new_gpnum(rsp, rnp, rdp);

1019

__note_new_gpnum(rsp, rnp, rdp);

1020

}

1020

}

1021

1022

/*

1022

/*

1023

* Start a new RCU grace period if warranted, re-initializing the hierarchy

1023

* Start a new RCU grace period if warranted, re-initializing the hierarchy

1024

* in preparation for detecting the next grace period. The caller must hold

1024

* in preparation for detecting the next grace period. The caller must hold

1025

* the root node's ->lock, which is released before return. Hard irqs must

1025

* the root node's ->lock, which is released before return. Hard irqs must

1026

* be disabled.

1026

* be disabled.

1027

*

1027

*

1028

* Note that it is legal for a dying CPU (which is marked as offline) to

1028

* Note that it is legal for a dying CPU (which is marked as offline) to

1029

* invoke this function. This can happen when the dying CPU reports its

1029

* invoke this function. This can happen when the dying CPU reports its

1030

* quiescent state.

1030

* quiescent state.

1031

*/

1031

*/

1032

static void

1032

static void

1033

rcu_start_gp(struct rcu_state *rsp, unsigned long flags)

1033

rcu_start_gp(struct rcu_state *rsp, unsigned long flags)

1034

__releases(rcu_get_root(rsp)->lock)

1034

__releases(rcu_get_root(rsp)->lock)

1035

{

1035

{

1036

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1036

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1037

struct rcu_node *rnp = rcu_get_root(rsp);

1037

struct rcu_node *rnp = rcu_get_root(rsp);

1038

1039

if (!rcu_scheduler_fully_active ||

1039

if (!rcu_scheduler_fully_active ||

1040

!cpu_needs_another_gp(rsp, rdp)) {

1040

!cpu_needs_another_gp(rsp, rdp)) {

1041

/*

1041

/*

1042

* Either the scheduler hasn't yet spawned the first

1042

* Either the scheduler hasn't yet spawned the first

1043

* non-idle task or this CPU does not need another

1043

* non-idle task or this CPU does not need another

1044

* grace period. Either way, don't start a new grace

1044

* grace period. Either way, don't start a new grace

1045

* period.

1045

* period.

1046

*/

1046

*/

1047

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1047

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1048

return;

1048

return;

1049

}

1049

}

1050

1051

if (rsp->fqs_active) {

1051

if (rsp->fqs_active) {

1052

/*

1052

/*

1053

* This CPU needs a grace period, but force_quiescent_state()

1053

* This CPU needs a grace period, but force_quiescent_state()

1054

* is running. Tell it to start one on this CPU's behalf.

1054

* is running. Tell it to start one on this CPU's behalf.

1055

*/

1055

*/

1056

rsp->fqs_need_gp = 1;

1056

rsp->fqs_need_gp = 1;

1057

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1057

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1058

return;

1058

return;

1059

}

1059

}

1060

1061

/* Advance to a new grace period and initialize state. */

1061

/* Advance to a new grace period and initialize state. */

1062

rsp->gpnum++;

1062

rsp->gpnum++;

1063

trace_rcu_grace_period(rsp->name, rsp->gpnum, "start");

1063

trace_rcu_grace_period(rsp->name, rsp->gpnum, "start");

1064

WARN_ON_ONCE(rsp->fqs_state == RCU_GP_INIT);

1064

WARN_ON_ONCE(rsp->fqs_state == RCU_GP_INIT);

1065

rsp->fqs_state = RCU_GP_INIT; /* Hold off force_quiescent_state. */

1065

rsp->fqs_state = RCU_GP_INIT; /* Hold off force_quiescent_state. */

1066

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1066

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1067

record_gp_stall_check_time(rsp);

1067

record_gp_stall_check_time(rsp);

1068

raw_spin_unlock(&rnp->lock); /* leave irqs disabled. */

1068

raw_spin_unlock(&rnp->lock); /* leave irqs disabled. */

1069

1070

/* Exclude any concurrent CPU-hotplug operations. */

1070

/* Exclude any concurrent CPU-hotplug operations. */

1071

raw_spin_lock(&rsp->onofflock); /* irqs already disabled. */

1071

raw_spin_lock(&rsp->onofflock); /* irqs already disabled. */

1072

1073

/*

1073

/*

1074

* Set the quiescent-state-needed bits in all the rcu_node

1074

* Set the quiescent-state-needed bits in all the rcu_node

1075

* structures for all currently online CPUs in breadth-first

1075

* structures for all currently online CPUs in breadth-first

1076

* order, starting from the root rcu_node structure. This

1076

* order, starting from the root rcu_node structure. This

1077

* operation relies on the layout of the hierarchy within the

1077

* operation relies on the layout of the hierarchy within the

1078

* rsp->node[] array. Note that other CPUs will access only

1078

* rsp->node[] array. Note that other CPUs will access only

1079

* the leaves of the hierarchy, which still indicate that no

1079

* the leaves of the hierarchy, which still indicate that no

1080

* grace period is in progress, at least until the corresponding

1080

* grace period is in progress, at least until the corresponding

1081

* leaf node has been initialized. In addition, we have excluded

1081

* leaf node has been initialized. In addition, we have excluded

1082

* CPU-hotplug operations.

1082

* CPU-hotplug operations.

1083

*

1083

*

1084

* Note that the grace period cannot complete until we finish

1084

* Note that the grace period cannot complete until we finish

1085

* the initialization process, as there will be at least one

1085

* the initialization process, as there will be at least one

1086

* qsmask bit set in the root node until that time, namely the

1086

* qsmask bit set in the root node until that time, namely the

1087

* one corresponding to this CPU, due to the fact that we have

1087

* one corresponding to this CPU, due to the fact that we have

1088

* irqs disabled.

1088

* irqs disabled.

1089

*/

1089

*/

1090

rcu_for_each_node_breadth_first(rsp, rnp) {

1090

rcu_for_each_node_breadth_first(rsp, rnp) {

1091

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1091

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1092

rcu_preempt_check_blocked_tasks(rnp);

1092

rcu_preempt_check_blocked_tasks(rnp);

1093

rnp->qsmask = rnp->qsmaskinit;

1093

rnp->qsmask = rnp->qsmaskinit;

1094

rnp->gpnum = rsp->gpnum;

1094

rnp->gpnum = rsp->gpnum;

1095

rnp->completed = rsp->completed;

1095

rnp->completed = rsp->completed;

1096

if (rnp == rdp->mynode)

1096

if (rnp == rdp->mynode)

1097

rcu_start_gp_per_cpu(rsp, rnp, rdp);

1097

rcu_start_gp_per_cpu(rsp, rnp, rdp);

1098

rcu_preempt_boost_start_gp(rnp);

1098

rcu_preempt_boost_start_gp(rnp);

1099

trace_rcu_grace_period_init(rsp->name, rnp->gpnum,

1099

trace_rcu_grace_period_init(rsp->name, rnp->gpnum,

1100

rnp->level, rnp->grplo,

1100

rnp->level, rnp->grplo,

1101

rnp->grphi, rnp->qsmask);

1101

rnp->grphi, rnp->qsmask);

1102

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1102

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1103

}

1103

}

1104

1105

rnp = rcu_get_root(rsp);

1105

rnp = rcu_get_root(rsp);

1106

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1106

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1107

rsp->fqs_state = RCU_SIGNAL_INIT; /* force_quiescent_state now OK. */

1107

rsp->fqs_state = RCU_SIGNAL_INIT; /* force_quiescent_state now OK. */

1108

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1108

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1109

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

1109

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

1110

}

1110

}

1111

1112

/*

1112

/*

1113

* Report a full set of quiescent states to the specified rcu_state

1113

* Report a full set of quiescent states to the specified rcu_state

1114

* data structure. This involves cleaning up after the prior grace

1114

* data structure. This involves cleaning up after the prior grace

1115

* period and letting rcu_start_gp() start up the next grace period

1115

* period and letting rcu_start_gp() start up the next grace period

1116

* if one is needed. Note that the caller must hold rnp->lock, as

1116

* if one is needed. Note that the caller must hold rnp->lock, as

1117

* required by rcu_start_gp(), which will release it.

1117

* required by rcu_start_gp(), which will release it.

1118

*/

1118

*/

1119

static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)

1119

static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)

1120

__releases(rcu_get_root(rsp)->lock)

1120

__releases(rcu_get_root(rsp)->lock)

1121

{

1121

{

1122

unsigned long gp_duration;

1122

unsigned long gp_duration;

1123

struct rcu_node *rnp = rcu_get_root(rsp);

1123

struct rcu_node *rnp = rcu_get_root(rsp);

1124

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1124

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1125

1126

WARN_ON_ONCE(!rcu_gp_in_progress(rsp));

1126

WARN_ON_ONCE(!rcu_gp_in_progress(rsp));

1127

1128

/*

1128

/*

1129

* Ensure that all grace-period and pre-grace-period activity

1129

* Ensure that all grace-period and pre-grace-period activity

1130

* is seen before the assignment to rsp->completed.

1130

* is seen before the assignment to rsp->completed.

1131

*/

1131

*/

1132

smp_mb(); /* See above block comment. */

1132

smp_mb(); /* See above block comment. */

1133

gp_duration = jiffies - rsp->gp_start;

1133

gp_duration = jiffies - rsp->gp_start;

1134

if (gp_duration > rsp->gp_max)

1134

if (gp_duration > rsp->gp_max)

1135

rsp->gp_max = gp_duration;

1135

rsp->gp_max = gp_duration;

1136

1137

/*

1137

/*

1138

* We know the grace period is complete, but to everyone else

1138

* We know the grace period is complete, but to everyone else

1139

* it appears to still be ongoing. But it is also the case

1139

* it appears to still be ongoing. But it is also the case

1140

* that to everyone else it looks like there is nothing that

1140

* that to everyone else it looks like there is nothing that

1141

* they can do to advance the grace period. It is therefore

1141

* they can do to advance the grace period. It is therefore

1142

* safe for us to drop the lock in order to mark the grace

1142

* safe for us to drop the lock in order to mark the grace

1143

* period as completed in all of the rcu_node structures.

1143

* period as completed in all of the rcu_node structures.

1144

*

1144

*

1145

* But if this CPU needs another grace period, it will take

1145

* But if this CPU needs another grace period, it will take

1146

* care of this while initializing the next grace period.

1146

* care of this while initializing the next grace period.

1147

* We use RCU_WAIT_TAIL instead of the usual RCU_DONE_TAIL

1147

* We use RCU_WAIT_TAIL instead of the usual RCU_DONE_TAIL

1148

* because the callbacks have not yet been advanced: Those

1148

* because the callbacks have not yet been advanced: Those

1149

* callbacks are waiting on the grace period that just now

1149

* callbacks are waiting on the grace period that just now

1150

* completed.

1150

* completed.

1151

*/

1151

*/

1152

if (*rdp->nxttail[RCU_WAIT_TAIL] == NULL) {

1152

if (*rdp->nxttail[RCU_WAIT_TAIL] == NULL) {

1153

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1153

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1154

1155

/*

1155

/*

1156

* Propagate new ->completed value to rcu_node structures

1156

* Propagate new ->completed value to rcu_node structures

1157

* so that other CPUs don't have to wait until the start

1157

* so that other CPUs don't have to wait until the start

1158

* of the next grace period to process their callbacks.

1158

* of the next grace period to process their callbacks.

1159

*/

1159

*/

1160

rcu_for_each_node_breadth_first(rsp, rnp) {

1160

rcu_for_each_node_breadth_first(rsp, rnp) {

1161

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1161

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1162

rnp->completed = rsp->gpnum;

1162

rnp->completed = rsp->gpnum;

1163

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1163

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1164

}

1164

}

1165

rnp = rcu_get_root(rsp);

1165

rnp = rcu_get_root(rsp);

1166

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1166

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1167

}

1167

}

1168

1169

rsp->completed = rsp->gpnum; /* Declare the grace period complete. */

1169

rsp->completed = rsp->gpnum; /* Declare the grace period complete. */

1170

trace_rcu_grace_period(rsp->name, rsp->completed, "end");

1170

trace_rcu_grace_period(rsp->name, rsp->completed, "end");

1171

rsp->fqs_state = RCU_GP_IDLE;

1171

rsp->fqs_state = RCU_GP_IDLE;

1172

rcu_start_gp(rsp, flags); /* releases root node's rnp->lock. */

1172

rcu_start_gp(rsp, flags); /* releases root node's rnp->lock. */

1173

}

1173

}

1174

1175

/*

1175

/*

1176

* Similar to rcu_report_qs_rdp(), for which it is a helper function.

1176

* Similar to rcu_report_qs_rdp(), for which it is a helper function.

1177

* Allows quiescent states for a group of CPUs to be reported at one go

1177

* Allows quiescent states for a group of CPUs to be reported at one go

1178

* to the specified rcu_node structure, though all the CPUs in the group

1178

* to the specified rcu_node structure, though all the CPUs in the group

1179

* must be represented by the same rcu_node structure (which need not be

1179

* must be represented by the same rcu_node structure (which need not be

1180

* a leaf rcu_node structure, though it often will be). That structure's

1180

* a leaf rcu_node structure, though it often will be). That structure's

1181

* lock must be held upon entry, and it is released before return.

1181

* lock must be held upon entry, and it is released before return.

1182

*/

1182

*/

1183

static void

1183

static void

1184

rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,

1184

rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,

1185

struct rcu_node *rnp, unsigned long flags)

1185

struct rcu_node *rnp, unsigned long flags)

1186

__releases(rnp->lock)

1186

__releases(rnp->lock)

1187

{

1187

{

1188

struct rcu_node *rnp_c;

1188

struct rcu_node *rnp_c;

1189

1190

/* Walk up the rcu_node hierarchy. */

1190

/* Walk up the rcu_node hierarchy. */

1191

for (;;) {

1191

for (;;) {

1192

if (!(rnp->qsmask & mask)) {

1192

if (!(rnp->qsmask & mask)) {

1193

1194

/* Our bit has already been cleared, so done. */

1194

/* Our bit has already been cleared, so done. */

1195

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1195

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1196

return;

1196

return;

1197

}

1197

}

1198

rnp->qsmask &= ~mask;

1198

rnp->qsmask &= ~mask;

1199

trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,

1199

trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,

1200

mask, rnp->qsmask, rnp->level,

1200

mask, rnp->qsmask, rnp->level,

1201

rnp->grplo, rnp->grphi,

1201

rnp->grplo, rnp->grphi,

1202

!!rnp->gp_tasks);

1202

!!rnp->gp_tasks);

1203

if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {

1203

if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {

1204

1205

/* Other bits still set at this level, so done. */

1205

/* Other bits still set at this level, so done. */

1206

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1206

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1207

return;

1207

return;

1208

}

1208

}

1209

mask = rnp->grpmask;

1209

mask = rnp->grpmask;

1210

if (rnp->parent == NULL) {

1210

if (rnp->parent == NULL) {

1211

1212

/* No more levels. Exit loop holding root lock. */

1212

/* No more levels. Exit loop holding root lock. */

1213

1214

break;

1214

break;

1215

}

1215

}

1216

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1216

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1217

rnp_c = rnp;

1217

rnp_c = rnp;

1218

rnp = rnp->parent;

1218

rnp = rnp->parent;

1219

raw_spin_lock_irqsave(&rnp->lock, flags);

1219

raw_spin_lock_irqsave(&rnp->lock, flags);

1220

WARN_ON_ONCE(rnp_c->qsmask);

1220

WARN_ON_ONCE(rnp_c->qsmask);

1221

}

1221

}

1222

1223

/*

1223

/*

1224

* Get here if we are the last CPU to pass through a quiescent

1224

* Get here if we are the last CPU to pass through a quiescent

1225

* state for this grace period. Invoke rcu_report_qs_rsp()

1225

* state for this grace period. Invoke rcu_report_qs_rsp()

1226

* to clean up and start the next grace period if one is needed.

1226

* to clean up and start the next grace period if one is needed.

1227

*/

1227

*/

1228

rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */

1228

rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */

1229

}

1229

}

1230

1231

/*

1231

/*

1232

* Record a quiescent state for the specified CPU to that CPU's rcu_data

1232

* Record a quiescent state for the specified CPU to that CPU's rcu_data

1233

* structure. This must be either called from the specified CPU, or

1233

* structure. This must be either called from the specified CPU, or

1234

* called when the specified CPU is known to be offline (and when it is

1234

* called when the specified CPU is known to be offline (and when it is

1235

* also known that no other CPU is concurrently trying to help the offline

1235

* also known that no other CPU is concurrently trying to help the offline

1236

* CPU). The lastcomp argument is used to make sure we are still in the

1236

* CPU). The lastcomp argument is used to make sure we are still in the

1237

* grace period of interest. We don't want to end the current grace period

1237

* grace period of interest. We don't want to end the current grace period

1238

* based on quiescent states detected in an earlier grace period!

1238

* based on quiescent states detected in an earlier grace period!

1239

*/

1239

*/

1240

static void

1240

static void

1241

rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp, long lastgp)

1241

rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp, long lastgp)

1242

{

1242

{

1243

unsigned long flags;

1243

unsigned long flags;

1244

unsigned long mask;

1244

unsigned long mask;

1245

struct rcu_node *rnp;

1245

struct rcu_node *rnp;

1246

1247

rnp = rdp->mynode;

1247

rnp = rdp->mynode;

1248

raw_spin_lock_irqsave(&rnp->lock, flags);

1248

raw_spin_lock_irqsave(&rnp->lock, flags);

1249

if (lastgp != rnp->gpnum || rnp->completed == rnp->gpnum) {

1249

if (lastgp != rnp->gpnum || rnp->completed == rnp->gpnum) {

1250

1251

/*

1251

/*

1252

* The grace period in which this quiescent state was

1252

* The grace period in which this quiescent state was

1253

* recorded has ended, so don't report it upwards.

1253

* recorded has ended, so don't report it upwards.

1254

* We will instead need a new quiescent state that lies

1254

* We will instead need a new quiescent state that lies

1255

* within the current grace period.

1255

* within the current grace period.

1256

*/

1256

*/

1257

rdp->passed_quiesce = 0; /* need qs for new gp. */

1257

rdp->passed_quiesce = 0; /* need qs for new gp. */

1258

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1258

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1259

return;

1259

return;

1260

}

1260

}

1261

mask = rdp->grpmask;

1261

mask = rdp->grpmask;

1262

if ((rnp->qsmask & mask) == 0) {

1262

if ((rnp->qsmask & mask) == 0) {

1263

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1263

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1264

} else {

1264

} else {

1265

rdp->qs_pending = 0;

1265

rdp->qs_pending = 0;

1266

1267

/*

1267

/*

1268

* This GP can't end until cpu checks in, so all of our

1268

* This GP can't end until cpu checks in, so all of our

1269

* callbacks can be processed during the next GP.

1269

* callbacks can be processed during the next GP.

1270

*/

1270

*/

1271

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1271

rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];

1272

1273

rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */

1273

rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */

1274

}

1274

}

1275

}

1275

}

1276

1277

/*

1277

/*

1278

* Check to see if there is a new grace period of which this CPU

1278

* Check to see if there is a new grace period of which this CPU

1279

* is not yet aware, and if so, set up local rcu_data state for it.

1279

* is not yet aware, and if so, set up local rcu_data state for it.

1280

* Otherwise, see if this CPU has just passed through its first

1280

* Otherwise, see if this CPU has just passed through its first

1281

* quiescent state for this grace period, and record that fact if so.

1281

* quiescent state for this grace period, and record that fact if so.

1282

*/

1282

*/

1283

static void

1283

static void

1284

rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp)

1284

rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp)

1285

{

1285

{

1286

/* If there is now a new grace period, record and return. */

1286

/* If there is now a new grace period, record and return. */

1287

if (check_for_new_grace_period(rsp, rdp))

1287

if (check_for_new_grace_period(rsp, rdp))

1288

return;

1288

return;

1289

1290

/*

1290

/*

1291

* Does this CPU still need to do its part for current grace period?

1291

* Does this CPU still need to do its part for current grace period?

1292

* If no, return and let the other CPUs do their part as well.

1292

* If no, return and let the other CPUs do their part as well.

1293

*/

1293

*/

1294

if (!rdp->qs_pending)

1294

if (!rdp->qs_pending)

1295

return;

1295

return;

1296

1297

/*

1297

/*

1298

* Was there a quiescent state since the beginning of the grace

1298

* Was there a quiescent state since the beginning of the grace

1299

* period? If no, then exit and wait for the next call.

1299

* period? If no, then exit and wait for the next call.

1300

*/

1300

*/

1301

if (!rdp->passed_quiesce)

1301

if (!rdp->passed_quiesce)

1302

return;

1302

return;

1303

1304

/*

1304

/*

1305

* Tell RCU we are done (but rcu_report_qs_rdp() will be the

1305

* Tell RCU we are done (but rcu_report_qs_rdp() will be the

1306

* judge of that).

1306

* judge of that).

1307

*/

1307

*/

1308

rcu_report_qs_rdp(rdp->cpu, rsp, rdp, rdp->passed_quiesce_gpnum);

1308

rcu_report_qs_rdp(rdp->cpu, rsp, rdp, rdp->passed_quiesce_gpnum);

1309

}

1309

}

1310

1311

#ifdef CONFIG_HOTPLUG_CPU

1311

#ifdef CONFIG_HOTPLUG_CPU

1312

1313

/*

1313

/*

1314

* Move a dying CPU's RCU callbacks to online CPU's callback list.

1314

* Move a dying CPU's RCU callbacks to online CPU's callback list.

1315

* Also record a quiescent state for this CPU for the current grace period.

1315

* Also record a quiescent state for this CPU for the current grace period.

1316

* Synchronization and interrupt disabling are not required because

1316

* Synchronization and interrupt disabling are not required because

1317

* this function executes in stop_machine() context. Therefore, cleanup

1317

* this function executes in stop_machine() context. Therefore, cleanup

1318

* operations that might block must be done later from the CPU_DEAD

1318

* operations that might block must be done later from the CPU_DEAD

1319

* notifier.

1319

* notifier.

1320

*

1320

*

1321

* Note that the outgoing CPU's bit has already been cleared in the

1321

* Note that the outgoing CPU's bit has already been cleared in the

1322

* cpu_online_mask. This allows us to randomly pick a callback

1322

* cpu_online_mask. This allows us to randomly pick a callback

1323

* destination from the bits set in that mask.

1323

* destination from the bits set in that mask.

1324

*/

1324

*/

1325

static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)

1325

static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)

1326

{

1326

{

1327

int i;

1327

int i;

1328

unsigned long mask;

1328

unsigned long mask;

1329

int receive_cpu = cpumask_any(cpu_online_mask);

1329

int receive_cpu = cpumask_any(cpu_online_mask);

1330

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1330

struct rcu_data *rdp = this_cpu_ptr(rsp->rda);

1331

struct rcu_data *receive_rdp = per_cpu_ptr(rsp->rda, receive_cpu);

1331

struct rcu_data *receive_rdp = per_cpu_ptr(rsp->rda, receive_cpu);

1332

RCU_TRACE(struct rcu_node *rnp = rdp->mynode); /* For dying CPU. */

1332

RCU_TRACE(struct rcu_node *rnp = rdp->mynode); /* For dying CPU. */

1333

1334

/* First, adjust the counts. */

1334

/* First, adjust the counts. */

1335

if (rdp->nxtlist != NULL) {

1335

if (rdp->nxtlist != NULL) {

1336

receive_rdp->qlen_lazy += rdp->qlen_lazy;

1336

receive_rdp->qlen_lazy += rdp->qlen_lazy;

1337

receive_rdp->qlen += rdp->qlen;

1337

receive_rdp->qlen += rdp->qlen;

1338

rdp->qlen_lazy = 0;

1338

rdp->qlen_lazy = 0;

1339

rdp->qlen = 0;

1339

rdp->qlen = 0;

1340

}

1340

}

1341

1342

/*

1342

/*

1343

* Next, move ready-to-invoke callbacks to be invoked on some

1343

* Next, move ready-to-invoke callbacks to be invoked on some

1344

* other CPU. These will not be required to pass through another

1344

* other CPU. These will not be required to pass through another

1345

* grace period: They are done, regardless of CPU.

1345

* grace period: They are done, regardless of CPU.

1346

*/

1346

*/

1347

if (rdp->nxtlist != NULL &&

1347

if (rdp->nxtlist != NULL &&

1348

rdp->nxttail[RCU_DONE_TAIL] != &rdp->nxtlist) {

1348

rdp->nxttail[RCU_DONE_TAIL] != &rdp->nxtlist) {

1349

struct rcu_head *oldhead;

1349

struct rcu_head *oldhead;

1350

struct rcu_head **oldtail;

1350

struct rcu_head **oldtail;

1351

struct rcu_head **newtail;

1351

struct rcu_head **newtail;

1352

1353

oldhead = rdp->nxtlist;

1353

oldhead = rdp->nxtlist;

1354

oldtail = receive_rdp->nxttail[RCU_DONE_TAIL];

1354

oldtail = receive_rdp->nxttail[RCU_DONE_TAIL];

1355

rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];

1355

rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];

1356

*rdp->nxttail[RCU_DONE_TAIL] = *oldtail;

1356

*rdp->nxttail[RCU_DONE_TAIL] = *oldtail;

1357

*receive_rdp->nxttail[RCU_DONE_TAIL] = oldhead;

1357

*receive_rdp->nxttail[RCU_DONE_TAIL] = oldhead;

1358

newtail = rdp->nxttail[RCU_DONE_TAIL];

1358

newtail = rdp->nxttail[RCU_DONE_TAIL];

1359

for (i = RCU_DONE_TAIL; i < RCU_NEXT_SIZE; i++) {

1359

for (i = RCU_DONE_TAIL; i < RCU_NEXT_SIZE; i++) {

1360

if (receive_rdp->nxttail[i] == oldtail)

1360

if (receive_rdp->nxttail[i] == oldtail)

1361

receive_rdp->nxttail[i] = newtail;

1361

receive_rdp->nxttail[i] = newtail;

1362

if (rdp->nxttail[i] == newtail)

1362

if (rdp->nxttail[i] == newtail)

1363

rdp->nxttail[i] = &rdp->nxtlist;

1363

rdp->nxttail[i] = &rdp->nxtlist;

1364

}

1364

}

1365

}

1365

}

1366

1367

/*

1367

/*

1368

* Finally, put the rest of the callbacks at the end of the list.

1368

* Finally, put the rest of the callbacks at the end of the list.

1369

* The ones that made it partway through get to start over: We

1369

* The ones that made it partway through get to start over: We

1370

* cannot assume that grace periods are synchronized across CPUs.

1370

* cannot assume that grace periods are synchronized across CPUs.

1371

* (We could splice RCU_WAIT_TAIL into RCU_NEXT_READY_TAIL, but

1371

* (We could splice RCU_WAIT_TAIL into RCU_NEXT_READY_TAIL, but

1372

* this does not seem compelling. Not yet, anyway.)

1372

* this does not seem compelling. Not yet, anyway.)

1373

*/

1373

*/

1374

if (rdp->nxtlist != NULL) {

1374

if (rdp->nxtlist != NULL) {

1375

*receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist;

1375

*receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist;

1376

receive_rdp->nxttail[RCU_NEXT_TAIL] =

1376

receive_rdp->nxttail[RCU_NEXT_TAIL] =

1377

rdp->nxttail[RCU_NEXT_TAIL];

1377

rdp->nxttail[RCU_NEXT_TAIL];

1378

receive_rdp->n_cbs_adopted += rdp->qlen;

1378

receive_rdp->n_cbs_adopted += rdp->qlen;

1379

rdp->n_cbs_orphaned += rdp->qlen;

1379

rdp->n_cbs_orphaned += rdp->qlen;

1380

1381

rdp->nxtlist = NULL;

1381

rdp->nxtlist = NULL;

1382

for (i = 0; i < RCU_NEXT_SIZE; i++)

1382

for (i = 0; i < RCU_NEXT_SIZE; i++)

1383

rdp->nxttail[i] = &rdp->nxtlist;

1383

rdp->nxttail[i] = &rdp->nxtlist;

1384

}

1384

}

1385

1386

/*

1386

/*

1387

* Record a quiescent state for the dying CPU. This is safe

1387

* Record a quiescent state for the dying CPU. This is safe

1388

* only because we have already cleared out the callbacks.

1388

* only because we have already cleared out the callbacks.

1389

* (Otherwise, the RCU core might try to schedule the invocation

1389

* (Otherwise, the RCU core might try to schedule the invocation

1390

* of callbacks on this now-offline CPU, which would be bad.)

1390

* of callbacks on this now-offline CPU, which would be bad.)

1391

*/

1391

*/

1392

mask = rdp->grpmask; /* rnp->grplo is constant. */

1392

mask = rdp->grpmask; /* rnp->grplo is constant. */

1393

trace_rcu_grace_period(rsp->name,

1393

trace_rcu_grace_period(rsp->name,

1394

rnp->gpnum + 1 - !!(rnp->qsmask & mask),

1394

rnp->gpnum + 1 - !!(rnp->qsmask & mask),

1395

"cpuofl");

1395

"cpuofl");

1396

rcu_report_qs_rdp(smp_processor_id(), rsp, rdp, rsp->gpnum);

1396

rcu_report_qs_rdp(smp_processor_id(), rsp, rdp, rsp->gpnum);

1397

/* Note that rcu_report_qs_rdp() might call trace_rcu_grace_period(). */

1397

/* Note that rcu_report_qs_rdp() might call trace_rcu_grace_period(). */

1398

}

1398

}

1399

1400

/*

1400

/*

1401

* The CPU has been completely removed, and some other CPU is reporting

1401

* The CPU has been completely removed, and some other CPU is reporting

1402

* this fact from process context. Do the remainder of the cleanup.

1402

* this fact from process context. Do the remainder of the cleanup.

1403

* There can only be one CPU hotplug operation at a time, so no other

1403

* There can only be one CPU hotplug operation at a time, so no other

1404

* CPU can be attempting to update rcu_cpu_kthread_task.

1404

* CPU can be attempting to update rcu_cpu_kthread_task.

1405

*/

1405

*/

1406

static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)

1406

static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)

1407

{

1407

{

1408

unsigned long flags;

1408

unsigned long flags;

1409

unsigned long mask;

1409

unsigned long mask;

1410

int need_report = 0;

1410

int need_report = 0;

1411

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

1411

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

1412

struct rcu_node *rnp = rdp->mynode; /* Outgoing CPU's rnp. */

1412

struct rcu_node *rnp = rdp->mynode; /* Outgoing CPU's rnp. */

1413

1414

/* Adjust any no-longer-needed kthreads. */

1414

/* Adjust any no-longer-needed kthreads. */

1415

rcu_stop_cpu_kthread(cpu);

1415

rcu_stop_cpu_kthread(cpu);

1416

rcu_node_kthread_setaffinity(rnp, -1);

1416

rcu_node_kthread_setaffinity(rnp, -1);

1417

1418

/* Remove the dying CPU from the bitmasks in the rcu_node hierarchy. */

1418

/* Remove the dying CPU from the bitmasks in the rcu_node hierarchy. */

1419

1420

/* Exclude any attempts to start a new grace period. */

1420

/* Exclude any attempts to start a new grace period. */

1421

raw_spin_lock_irqsave(&rsp->onofflock, flags);

1421

raw_spin_lock_irqsave(&rsp->onofflock, flags);

1422

1423

/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */

1423

/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */

1424

mask = rdp->grpmask; /* rnp->grplo is constant. */

1424

mask = rdp->grpmask; /* rnp->grplo is constant. */

1425

do {

1425

do {

1426

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1426

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

1427

rnp->qsmaskinit &= ~mask;

1427

rnp->qsmaskinit &= ~mask;

1428

if (rnp->qsmaskinit != 0) {

1428

if (rnp->qsmaskinit != 0) {

1429

if (rnp != rdp->mynode)

1429

if (rnp != rdp->mynode)

1430

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1430

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1431

break;

1431

break;

1432

}

1432

}

1433

if (rnp == rdp->mynode)

1433

if (rnp == rdp->mynode)

1434

need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);

1434

need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);

1435

else

1435

else

1436

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1436

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

1437

mask = rnp->grpmask;

1437

mask = rnp->grpmask;

1438

rnp = rnp->parent;

1438

rnp = rnp->parent;

1439

} while (rnp != NULL);

1439

} while (rnp != NULL);

1440

1441

/*

1441

/*

1442

* We still hold the leaf rcu_node structure lock here, and

1442

* We still hold the leaf rcu_node structure lock here, and

1443

* irqs are still disabled. The reason for this subterfuge is

1443

* irqs are still disabled. The reason for this subterfuge is

1444

* because invoking rcu_report_unblock_qs_rnp() with ->onofflock

1444

* because invoking rcu_report_unblock_qs_rnp() with ->onofflock

1445

* held leads to deadlock.

1445

* held leads to deadlock.

1446

*/

1446

*/

1447

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

1447

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

1448

rnp = rdp->mynode;

1448

rnp = rdp->mynode;

1449

if (need_report & RCU_OFL_TASKS_NORM_GP)

1449

if (need_report & RCU_OFL_TASKS_NORM_GP)

1450

rcu_report_unblock_qs_rnp(rnp, flags);

1450

rcu_report_unblock_qs_rnp(rnp, flags);

1451

else

1451

else

1452

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1452

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1453

if (need_report & RCU_OFL_TASKS_EXP_GP)

1453

if (need_report & RCU_OFL_TASKS_EXP_GP)

1454

rcu_report_exp_rnp(rsp, rnp, true);

1454

rcu_report_exp_rnp(rsp, rnp, true);

1455

}

1455

}

1456

1457

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1457

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1458

1459

static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)

1459

static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)

1460

{

1460

{

1461

}

1461

}

1462

1463

static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)

1463

static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)

1464

{

1464

{

1465

}

1465

}

1466

1467

#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */

1467

#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */

1468

1469

/*

1469

/*

1470

* Invoke any RCU callbacks that have made it to the end of their grace

1470

* Invoke any RCU callbacks that have made it to the end of their grace

1471

* period. Thottle as specified by rdp->blimit.

1471

* period. Thottle as specified by rdp->blimit.

1472

*/

1472

*/

1473

static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)

1473

static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)

1474

{

1474

{

1475

unsigned long flags;

1475

unsigned long flags;

1476

struct rcu_head *next, *list, **tail;

1476

struct rcu_head *next, *list, **tail;

1477

int bl, count, count_lazy;

1477

int bl, count, count_lazy;

1478

1479

/* If no callbacks are ready, just return.*/

1479

/* If no callbacks are ready, just return.*/

1480

if (!cpu_has_callbacks_ready_to_invoke(rdp)) {

1480

if (!cpu_has_callbacks_ready_to_invoke(rdp)) {

1481

trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, 0);

1481

trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, 0);

1482

trace_rcu_batch_end(rsp->name, 0, !!ACCESS_ONCE(rdp->nxtlist),

1482

trace_rcu_batch_end(rsp->name, 0, !!ACCESS_ONCE(rdp->nxtlist),

1483

need_resched(), is_idle_task(current),

1483

need_resched(), is_idle_task(current),

1484

rcu_is_callbacks_kthread());

1484

rcu_is_callbacks_kthread());

1485

return;

1485

return;

1486

}

1486

}

1487

1488

/*

1488

/*

1489

* Extract the list of ready callbacks, disabling to prevent

1489

* Extract the list of ready callbacks, disabling to prevent

1490

* races with call_rcu() from interrupt handlers.

1490

* races with call_rcu() from interrupt handlers.

1491

*/

1491

*/

1492

local_irq_save(flags);

1492

local_irq_save(flags);

1493

WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));

1493

WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));

1494

bl = rdp->blimit;

1494

bl = rdp->blimit;

1495

trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, bl);

1495

trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, bl);

1496

list = rdp->nxtlist;

1496

list = rdp->nxtlist;

1497

rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];

1497

rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];

1498

*rdp->nxttail[RCU_DONE_TAIL] = NULL;

1498

*rdp->nxttail[RCU_DONE_TAIL] = NULL;

1499

tail = rdp->nxttail[RCU_DONE_TAIL];

1499

tail = rdp->nxttail[RCU_DONE_TAIL];

1500

for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)

1500

for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)

1501

if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])

1501

if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])

1502

rdp->nxttail[count] = &rdp->nxtlist;

1502

rdp->nxttail[count] = &rdp->nxtlist;

1503

local_irq_restore(flags);

1503

local_irq_restore(flags);

1504

1505

/* Invoke callbacks. */

1505

/* Invoke callbacks. */

1506

count = count_lazy = 0;

1506

count = count_lazy = 0;

1507

while (list) {

1507

while (list) {

1508

next = list->next;

1508

next = list->next;

1509

prefetch(next);

1509

prefetch(next);

1510

debug_rcu_head_unqueue(list);

1510

debug_rcu_head_unqueue(list);

1511

if (__rcu_reclaim(rsp->name, list))

1511

if (__rcu_reclaim(rsp->name, list))

1512

count_lazy++;

1512

count_lazy++;

1513

list = next;

1513

list = next;

1514

/* Stop only if limit reached and CPU has something to do. */

1514

/* Stop only if limit reached and CPU has something to do. */

1515

if (++count >= bl &&

1515

if (++count >= bl &&

1516

(need_resched() ||

1516

(need_resched() ||

1517

(!is_idle_task(current) && !rcu_is_callbacks_kthread())))

1517

(!is_idle_task(current) && !rcu_is_callbacks_kthread())))

1518

break;

1518

break;

1519

}

1519

}

1520

1521

local_irq_save(flags);

1521

local_irq_save(flags);

1522

trace_rcu_batch_end(rsp->name, count, !!list, need_resched(),

1522

trace_rcu_batch_end(rsp->name, count, !!list, need_resched(),

1523

is_idle_task(current),

1523

is_idle_task(current),

1524

rcu_is_callbacks_kthread());

1524

rcu_is_callbacks_kthread());

1525

1526

/* Update count, and requeue any remaining callbacks. */

1526

/* Update count, and requeue any remaining callbacks. */

1527

rdp->qlen_lazy -= count_lazy;

1527

rdp->qlen_lazy -= count_lazy;

1528

rdp->qlen -= count;

1528

rdp->qlen -= count;

1529

rdp->n_cbs_invoked += count;

1529

rdp->n_cbs_invoked += count;

1530

if (list != NULL) {

1530

if (list != NULL) {

1531

*tail = rdp->nxtlist;

1531

*tail = rdp->nxtlist;

1532

rdp->nxtlist = list;

1532

rdp->nxtlist = list;

1533

for (count = 0; count < RCU_NEXT_SIZE; count++)

1533

for (count = 0; count < RCU_NEXT_SIZE; count++)

1534

if (&rdp->nxtlist == rdp->nxttail[count])

1534

if (&rdp->nxtlist == rdp->nxttail[count])

1535

rdp->nxttail[count] = tail;

1535

rdp->nxttail[count] = tail;

1536

else

1536

else

1537

break;

1537

break;

1538

}

1538

}

1539

1540

/* Reinstate batch limit if we have worked down the excess. */

1540

/* Reinstate batch limit if we have worked down the excess. */

1541

if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)

1541

if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)

1542

rdp->blimit = blimit;

1542

rdp->blimit = blimit;

1543

1544

/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */

1544

/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */

1545

if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {

1545

if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {

1546

rdp->qlen_last_fqs_check = 0;

1546

rdp->qlen_last_fqs_check = 0;

1547

rdp->n_force_qs_snap = rsp->n_force_qs;

1547

rdp->n_force_qs_snap = rsp->n_force_qs;

1548

} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)

1548

} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)

1549

rdp->qlen_last_fqs_check = rdp->qlen;

1549

rdp->qlen_last_fqs_check = rdp->qlen;

1550

1551

local_irq_restore(flags);

1551

local_irq_restore(flags);

1552

1553

/* Re-invoke RCU core processing if there are callbacks remaining. */

1553

/* Re-invoke RCU core processing if there are callbacks remaining. */

1554

if (cpu_has_callbacks_ready_to_invoke(rdp))

1554

if (cpu_has_callbacks_ready_to_invoke(rdp))

1555

invoke_rcu_core();

1555

invoke_rcu_core();

1556

}

1556

}

1557

1558

/*

1558

/*

1559

* Check to see if this CPU is in a non-context-switch quiescent state

1559

* Check to see if this CPU is in a non-context-switch quiescent state

1560

* (user mode or idle loop for rcu, non-softirq execution for rcu_bh).

1560

* (user mode or idle loop for rcu, non-softirq execution for rcu_bh).

1561

* Also schedule RCU core processing.

1561

* Also schedule RCU core processing.

1562

*

1562

*

1563

* This function must be called from hardirq context. It is normally

1563

* This function must be called from hardirq context. It is normally

1564

* invoked from the scheduling-clock interrupt. If rcu_pending returns

1564

* invoked from the scheduling-clock interrupt. If rcu_pending returns

1565

* false, there is no point in invoking rcu_check_callbacks().

1565

* false, there is no point in invoking rcu_check_callbacks().

1566

*/

1566

*/

1567

void rcu_check_callbacks(int cpu, int user)

1567

void rcu_check_callbacks(int cpu, int user)

1568

{

1568

{

1569

trace_rcu_utilization("Start scheduler-tick");

1569

trace_rcu_utilization("Start scheduler-tick");

1570

increment_cpu_stall_ticks();

1570

increment_cpu_stall_ticks();

1571

if (user || rcu_is_cpu_rrupt_from_idle()) {

1571

if (user || rcu_is_cpu_rrupt_from_idle()) {

1572

1573

/*

1573

/*

1574

* Get here if this CPU took its interrupt from user

1574

* Get here if this CPU took its interrupt from user

1575

* mode or from the idle loop, and if this is not a

1575

* mode or from the idle loop, and if this is not a

1576

* nested interrupt. In this case, the CPU is in

1576

* nested interrupt. In this case, the CPU is in

1577

* a quiescent state, so note it.

1577

* a quiescent state, so note it.

1578

*

1578

*

1579

* No memory barrier is required here because both

1579

* No memory barrier is required here because both

1580

* rcu_sched_qs() and rcu_bh_qs() reference only CPU-local

1580

* rcu_sched_qs() and rcu_bh_qs() reference only CPU-local

1581

* variables that other CPUs neither access nor modify,

1581

* variables that other CPUs neither access nor modify,

1582

* at least not while the corresponding CPU is online.

1582

* at least not while the corresponding CPU is online.

1583

*/

1583

*/

1584

1585

rcu_sched_qs(cpu);

1585

rcu_sched_qs(cpu);

1586

rcu_bh_qs(cpu);

1586

rcu_bh_qs(cpu);

1587

1588

} else if (!in_softirq()) {

1588

} else if (!in_softirq()) {

1589

1590

/*

1590

/*

1591

* Get here if this CPU did not take its interrupt from

1591

* Get here if this CPU did not take its interrupt from

1592

* softirq, in other words, if it is not interrupting

1592

* softirq, in other words, if it is not interrupting

1593

* a rcu_bh read-side critical section. This is an _bh

1593

* a rcu_bh read-side critical section. This is an _bh

1594

* critical section, so note it.

1594

* critical section, so note it.

1595

*/

1595

*/

1596

1597

rcu_bh_qs(cpu);

1597

rcu_bh_qs(cpu);

1598

}

1598

}

1599

rcu_preempt_check_callbacks(cpu);

1599

rcu_preempt_check_callbacks(cpu);

1600

if (rcu_pending(cpu))

1600

if (rcu_pending(cpu))

1601

invoke_rcu_core();

1601

invoke_rcu_core();

1602

trace_rcu_utilization("End scheduler-tick");

1602

trace_rcu_utilization("End scheduler-tick");

1603

}

1603

}

1604

1605

/*

1605

/*

1606

* Scan the leaf rcu_node structures, processing dyntick state for any that

1606

* Scan the leaf rcu_node structures, processing dyntick state for any that

1607

* have not yet encountered a quiescent state, using the function specified.

1607

* have not yet encountered a quiescent state, using the function specified.

1608

* Also initiate boosting for any threads blocked on the root rcu_node.

1608

* Also initiate boosting for any threads blocked on the root rcu_node.

1609

*

1609

*

1610

* The caller must have suppressed start of new grace periods.

1610

* The caller must have suppressed start of new grace periods.

1611

*/

1611

*/

1612

static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))

1612

static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))

1613

{

1613

{

1614

unsigned long bit;

1614

unsigned long bit;

1615

int cpu;

1615

int cpu;

1616

unsigned long flags;

1616

unsigned long flags;

1617

unsigned long mask;

1617

unsigned long mask;

1618

struct rcu_node *rnp;

1618

struct rcu_node *rnp;

1619

1620

rcu_for_each_leaf_node(rsp, rnp) {

1620

rcu_for_each_leaf_node(rsp, rnp) {

1621

mask = 0;

1621

mask = 0;

1622

raw_spin_lock_irqsave(&rnp->lock, flags);

1622

raw_spin_lock_irqsave(&rnp->lock, flags);

1623

if (!rcu_gp_in_progress(rsp)) {

1623

if (!rcu_gp_in_progress(rsp)) {

1624

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1624

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1625

return;

1625

return;

1626

}

1626

}

1627

if (rnp->qsmask == 0) {

1627

if (rnp->qsmask == 0) {

1628

rcu_initiate_boost(rnp, flags); /* releases rnp->lock */

1628

rcu_initiate_boost(rnp, flags); /* releases rnp->lock */

1629

continue;

1629

continue;

1630

}

1630

}

1631

cpu = rnp->grplo;

1631

cpu = rnp->grplo;

1632

bit = 1;

1632

bit = 1;

1633

for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {

1633

for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {

1634

if ((rnp->qsmask & bit) != 0 &&

1634

if ((rnp->qsmask & bit) != 0 &&

1635

f(per_cpu_ptr(rsp->rda, cpu)))

1635

f(per_cpu_ptr(rsp->rda, cpu)))

1636

mask |= bit;

1636

mask |= bit;

1637

}

1637

}

1638

if (mask != 0) {

1638

if (mask != 0) {

1639

1640

/* rcu_report_qs_rnp() releases rnp->lock. */

1640

/* rcu_report_qs_rnp() releases rnp->lock. */

1641

rcu_report_qs_rnp(mask, rsp, rnp, flags);

1641

rcu_report_qs_rnp(mask, rsp, rnp, flags);

1642

continue;

1642

continue;

1643

}

1643

}

1644

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1644

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1645

}

1645

}

1646

rnp = rcu_get_root(rsp);

1646

rnp = rcu_get_root(rsp);

1647

if (rnp->qsmask == 0) {

1647

if (rnp->qsmask == 0) {

1648

raw_spin_lock_irqsave(&rnp->lock, flags);

1648

raw_spin_lock_irqsave(&rnp->lock, flags);

1649

rcu_initiate_boost(rnp, flags); /* releases rnp->lock. */

1649

rcu_initiate_boost(rnp, flags); /* releases rnp->lock. */

1650

}

1650

}

1651

}

1651

}

1652

1653

/*

1653

/*

1654

* Force quiescent states on reluctant CPUs, and also detect which

1654

* Force quiescent states on reluctant CPUs, and also detect which

1655

* CPUs are in dyntick-idle mode.

1655

* CPUs are in dyntick-idle mode.

1656

*/

1656

*/

1657

static void force_quiescent_state(struct rcu_state *rsp, int relaxed)

1657

static void force_quiescent_state(struct rcu_state *rsp, int relaxed)

1658

{

1658

{

1659

unsigned long flags;

1659

unsigned long flags;

1660

struct rcu_node *rnp = rcu_get_root(rsp);

1660

struct rcu_node *rnp = rcu_get_root(rsp);

1661

1662

trace_rcu_utilization("Start fqs");

1662

trace_rcu_utilization("Start fqs");

1663

if (!rcu_gp_in_progress(rsp)) {

1663

if (!rcu_gp_in_progress(rsp)) {

1664

trace_rcu_utilization("End fqs");

1664

trace_rcu_utilization("End fqs");

1665

return; /* No grace period in progress, nothing to force. */

1665

return; /* No grace period in progress, nothing to force. */

1666

}

1666

}

1667

if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {

1667

if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {

1668

rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */

1668

rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */

1669

trace_rcu_utilization("End fqs");

1669

trace_rcu_utilization("End fqs");

1670

return; /* Someone else is already on the job. */

1670

return; /* Someone else is already on the job. */

1671

}

1671

}

1672

if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))

1672

if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))

1673

goto unlock_fqs_ret; /* no emergency and done recently. */

1673

goto unlock_fqs_ret; /* no emergency and done recently. */

1674

rsp->n_force_qs++;

1674

rsp->n_force_qs++;

1675

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1675

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1676

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1676

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1677

if(!rcu_gp_in_progress(rsp)) {

1677

if(!rcu_gp_in_progress(rsp)) {

1678

rsp->n_force_qs_ngp++;

1678

rsp->n_force_qs_ngp++;

1679

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1679

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1680

goto unlock_fqs_ret; /* no GP in progress, time updated. */

1680

goto unlock_fqs_ret; /* no GP in progress, time updated. */

1681

}

1681

}

1682

rsp->fqs_active = 1;

1682

rsp->fqs_active = 1;

1683

switch (rsp->fqs_state) {

1683

switch (rsp->fqs_state) {

1684

case RCU_GP_IDLE:

1684

case RCU_GP_IDLE:

1685

case RCU_GP_INIT:

1685

case RCU_GP_INIT:

1686

1687

break; /* grace period idle or initializing, ignore. */

1687

break; /* grace period idle or initializing, ignore. */

1688

1689

case RCU_SAVE_DYNTICK:

1689

case RCU_SAVE_DYNTICK:

1690

if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)

1690

if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)

1691

break; /* So gcc recognizes the dead code. */

1691

break; /* So gcc recognizes the dead code. */

1692

1693

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1693

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1694

1695

/* Record dyntick-idle state. */

1695

/* Record dyntick-idle state. */

1696

force_qs_rnp(rsp, dyntick_save_progress_counter);

1696

force_qs_rnp(rsp, dyntick_save_progress_counter);

1697

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1697

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1698

if (rcu_gp_in_progress(rsp))

1698

if (rcu_gp_in_progress(rsp))

1699

rsp->fqs_state = RCU_FORCE_QS;

1699

rsp->fqs_state = RCU_FORCE_QS;

1700

break;

1700

break;

1701

1702

case RCU_FORCE_QS:

1702

case RCU_FORCE_QS:

1703

1704

/* Check dyntick-idle state, send IPI to laggarts. */

1704

/* Check dyntick-idle state, send IPI to laggarts. */

1705

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1705

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1706

force_qs_rnp(rsp, rcu_implicit_dynticks_qs);

1706

force_qs_rnp(rsp, rcu_implicit_dynticks_qs);

1707

1708

/* Leave state in case more forcing is required. */

1708

/* Leave state in case more forcing is required. */

1709

1710

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1710

raw_spin_lock(&rnp->lock); /* irqs already disabled */

1711

break;

1711

break;

1712

}

1712

}

1713

rsp->fqs_active = 0;

1713

rsp->fqs_active = 0;

1714

if (rsp->fqs_need_gp) {

1714

if (rsp->fqs_need_gp) {

1715

raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */

1715

raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */

1716

rsp->fqs_need_gp = 0;

1716

rsp->fqs_need_gp = 0;

1717

rcu_start_gp(rsp, flags); /* releases rnp->lock */

1717

rcu_start_gp(rsp, flags); /* releases rnp->lock */

1718

trace_rcu_utilization("End fqs");

1718

trace_rcu_utilization("End fqs");

1719

return;

1719

return;

1720

}

1720

}

1721

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1721

raw_spin_unlock(&rnp->lock); /* irqs remain disabled */

1722

unlock_fqs_ret:

1722

unlock_fqs_ret:

1723

raw_spin_unlock_irqrestore(&rsp->fqslock, flags);

1723

raw_spin_unlock_irqrestore(&rsp->fqslock, flags);

1724

trace_rcu_utilization("End fqs");

1724

trace_rcu_utilization("End fqs");

1725

}

1725

}

1726

1727

/*

1727

/*

1728

* This does the RCU core processing work for the specified rcu_state

1728

* This does the RCU core processing work for the specified rcu_state

1729

* and rcu_data structures. This may be called only from the CPU to

1729

* and rcu_data structures. This may be called only from the CPU to

1730

* whom the rdp belongs.

1730

* whom the rdp belongs.

1731

*/

1731

*/

1732

static void

1732

static void

1733

__rcu_process_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)

1733

__rcu_process_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)

1734

{

1734

{

1735

unsigned long flags;

1735

unsigned long flags;

1736

1737

WARN_ON_ONCE(rdp->beenonline == 0);

1737

WARN_ON_ONCE(rdp->beenonline == 0);

1738

1739

/*

1739

/*

1740

* If an RCU GP has gone long enough, go check for dyntick

1740

* If an RCU GP has gone long enough, go check for dyntick

1741

* idle CPUs and, if needed, send resched IPIs.

1741

* idle CPUs and, if needed, send resched IPIs.

1742

*/

1742

*/

1743

if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1743

if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1744

force_quiescent_state(rsp, 1);

1744

force_quiescent_state(rsp, 1);

1745

1746

/*

1746

/*

1747

* Advance callbacks in response to end of earlier grace

1747

* Advance callbacks in response to end of earlier grace

1748

* period that some other CPU ended.

1748

* period that some other CPU ended.

1749

*/

1749

*/

1750

rcu_process_gp_end(rsp, rdp);

1750

rcu_process_gp_end(rsp, rdp);

1751

1752

/* Update RCU state based on any recent quiescent states. */

1752

/* Update RCU state based on any recent quiescent states. */

1753

rcu_check_quiescent_state(rsp, rdp);

1753

rcu_check_quiescent_state(rsp, rdp);

1754

1755

/* Does this CPU require a not-yet-started grace period? */

1755

/* Does this CPU require a not-yet-started grace period? */

1756

if (cpu_needs_another_gp(rsp, rdp)) {

1756

if (cpu_needs_another_gp(rsp, rdp)) {

1757

raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);

1757

raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);

1758

rcu_start_gp(rsp, flags); /* releases above lock */

1758

rcu_start_gp(rsp, flags); /* releases above lock */

1759

}

1759

}

1760

1761

/* If there are callbacks ready, invoke them. */

1761

/* If there are callbacks ready, invoke them. */

1762

if (cpu_has_callbacks_ready_to_invoke(rdp))

1762

if (cpu_has_callbacks_ready_to_invoke(rdp))

1763

invoke_rcu_callbacks(rsp, rdp);

1763

invoke_rcu_callbacks(rsp, rdp);

1764

}

1764

}

1765

1766

/*

1766

/*

1767

* Do RCU core processing for the current CPU.

1767

* Do RCU core processing for the current CPU.

1768

*/

1768

*/

1769

static void rcu_process_callbacks(struct softirq_action *unused)

1769

static void rcu_process_callbacks(struct softirq_action *unused)

1770

{

1770

{

1771

trace_rcu_utilization("Start RCU core");

1771

trace_rcu_utilization("Start RCU core");

1772

__rcu_process_callbacks(&rcu_sched_state,

1772

__rcu_process_callbacks(&rcu_sched_state,

1773

&__get_cpu_var(rcu_sched_data));

1773

&__get_cpu_var(rcu_sched_data));

1774

__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));

1774

__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));

1775

rcu_preempt_process_callbacks();

1775

rcu_preempt_process_callbacks();

1776

trace_rcu_utilization("End RCU core");

1776

trace_rcu_utilization("End RCU core");

1777

}

1777

}

1778

1779

/*

1779

/*

1780

* Schedule RCU callback invocation. If the specified type of RCU

1780

* Schedule RCU callback invocation. If the specified type of RCU

1781

* does not support RCU priority boosting, just do a direct call,

1781

* does not support RCU priority boosting, just do a direct call,

1782

* otherwise wake up the per-CPU kernel kthread. Note that because we

1782

* otherwise wake up the per-CPU kernel kthread. Note that because we

1783

* are running on the current CPU with interrupts disabled, the

1783

* are running on the current CPU with interrupts disabled, the

1784

* rcu_cpu_kthread_task cannot disappear out from under us.

1784

* rcu_cpu_kthread_task cannot disappear out from under us.

1785

*/

1785

*/

1786

static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)

1786

static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)

1787

{

1787

{

1788

if (unlikely(!ACCESS_ONCE(rcu_scheduler_fully_active)))

1788

if (unlikely(!ACCESS_ONCE(rcu_scheduler_fully_active)))

1789

return;

1789

return;

1790

if (likely(!rsp->boost)) {

1790

if (likely(!rsp->boost)) {

1791

rcu_do_batch(rsp, rdp);

1791

rcu_do_batch(rsp, rdp);

1792

return;

1792

return;

1793

}

1793

}

1794

invoke_rcu_callbacks_kthread();

1794

invoke_rcu_callbacks_kthread();

1795

}

1795

}

1796

1797

static void invoke_rcu_core(void)

1797

static void invoke_rcu_core(void)

1798

{

1798

{

1799

raise_softirq(RCU_SOFTIRQ);

1799

raise_softirq(RCU_SOFTIRQ);

1800

}

1800

}

1801

1802

static void

1802

static void

1803

__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),

1803

__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),

1804

struct rcu_state *rsp, bool lazy)

1804

struct rcu_state *rsp, bool lazy)

1805

{

1805

{

1806

unsigned long flags;

1806

unsigned long flags;

1807

struct rcu_data *rdp;

1807

struct rcu_data *rdp;

1808

1809

WARN_ON_ONCE((unsigned long)head & 0x3); /* Misaligned rcu_head! */

1809

WARN_ON_ONCE((unsigned long)head & 0x3); /* Misaligned rcu_head! */

1810

debug_rcu_head_queue(head);

1810

debug_rcu_head_queue(head);

1811

head->func = func;

1811

head->func = func;

1812

head->next = NULL;

1812

head->next = NULL;

1813

1814

smp_mb(); /* Ensure RCU update seen before callback registry. */

1814

smp_mb(); /* Ensure RCU update seen before callback registry. */

1815

1816

/*

1816

/*

1817

* Opportunistically note grace-period endings and beginnings.

1817

* Opportunistically note grace-period endings and beginnings.

1818

* Note that we might see a beginning right after we see an

1818

* Note that we might see a beginning right after we see an

1819

* end, but never vice versa, since this CPU has to pass through

1819

* end, but never vice versa, since this CPU has to pass through

1820

* a quiescent state betweentimes.

1820

* a quiescent state betweentimes.

1821

*/

1821

*/

1822

local_irq_save(flags);

1822

local_irq_save(flags);

1823

WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));

1823

WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));

1824

rdp = this_cpu_ptr(rsp->rda);

1824

rdp = this_cpu_ptr(rsp->rda);

1825

1826

/* Add the callback to our list. */

1826

/* Add the callback to our list. */

1827

*rdp->nxttail[RCU_NEXT_TAIL] = head;

1827

*rdp->nxttail[RCU_NEXT_TAIL] = head;

1828

rdp->nxttail[RCU_NEXT_TAIL] = &head->next;

1828

rdp->nxttail[RCU_NEXT_TAIL] = &head->next;

1829

rdp->qlen++;

1829

rdp->qlen++;

1830

if (lazy)

1830

if (lazy)

1831

rdp->qlen_lazy++;

1831

rdp->qlen_lazy++;

1832

1833

if (__is_kfree_rcu_offset((unsigned long)func))

1833

if (__is_kfree_rcu_offset((unsigned long)func))

1834

trace_rcu_kfree_callback(rsp->name, head, (unsigned long)func,

1834

trace_rcu_kfree_callback(rsp->name, head, (unsigned long)func,

1835

rdp->qlen_lazy, rdp->qlen);

1835

rdp->qlen_lazy, rdp->qlen);

1836

else

1836

else

1837

trace_rcu_callback(rsp->name, head, rdp->qlen_lazy, rdp->qlen);

1837

trace_rcu_callback(rsp->name, head, rdp->qlen_lazy, rdp->qlen);

1838

1839

/* If interrupts were disabled, don't dive into RCU core. */

1839

/* If interrupts were disabled, don't dive into RCU core. */

1840

if (irqs_disabled_flags(flags)) {

1840

if (irqs_disabled_flags(flags)) {

1841

local_irq_restore(flags);

1841

local_irq_restore(flags);

1842

return;

1842

return;

1843

}

1843

}

1844

1845

/*

1845

/*

1846

* Force the grace period if too many callbacks or too long waiting.

1846

* Force the grace period if too many callbacks or too long waiting.

1847

* Enforce hysteresis, and don't invoke force_quiescent_state()

1847

* Enforce hysteresis, and don't invoke force_quiescent_state()

1848

* if some other CPU has recently done so. Also, don't bother

1848

* if some other CPU has recently done so. Also, don't bother

1849

* invoking force_quiescent_state() if the newly enqueued callback

1849

* invoking force_quiescent_state() if the newly enqueued callback

1850

* is the only one waiting for a grace period to complete.

1850

* is the only one waiting for a grace period to complete.

1851

*/

1851

*/

1852

if (unlikely(rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {

1852

if (unlikely(rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {

1853

1854

/* Are we ignoring a completed grace period? */

1854

/* Are we ignoring a completed grace period? */

1855

rcu_process_gp_end(rsp, rdp);

1855

rcu_process_gp_end(rsp, rdp);

1856

check_for_new_grace_period(rsp, rdp);

1856

check_for_new_grace_period(rsp, rdp);

1857

1858

/* Start a new grace period if one not already started. */

1858

/* Start a new grace period if one not already started. */

1859

if (!rcu_gp_in_progress(rsp)) {

1859

if (!rcu_gp_in_progress(rsp)) {

1860

unsigned long nestflag;

1860

unsigned long nestflag;

1861

struct rcu_node *rnp_root = rcu_get_root(rsp);

1861

struct rcu_node *rnp_root = rcu_get_root(rsp);

1862

1863

raw_spin_lock_irqsave(&rnp_root->lock, nestflag);

1863

raw_spin_lock_irqsave(&rnp_root->lock, nestflag);

1864

rcu_start_gp(rsp, nestflag); /* rlses rnp_root->lock */

1864

rcu_start_gp(rsp, nestflag); /* rlses rnp_root->lock */

1865

} else {

1865

} else {

1866

/* Give the grace period a kick. */

1866

/* Give the grace period a kick. */

1867

rdp->blimit = LONG_MAX;

1867

rdp->blimit = LONG_MAX;

1868

if (rsp->n_force_qs == rdp->n_force_qs_snap &&

1868

if (rsp->n_force_qs == rdp->n_force_qs_snap &&

1869

*rdp->nxttail[RCU_DONE_TAIL] != head)

1869

*rdp->nxttail[RCU_DONE_TAIL] != head)

1870

force_quiescent_state(rsp, 0);

1870

force_quiescent_state(rsp, 0);

1871

rdp->n_force_qs_snap = rsp->n_force_qs;

1871

rdp->n_force_qs_snap = rsp->n_force_qs;

1872

rdp->qlen_last_fqs_check = rdp->qlen;

1872

rdp->qlen_last_fqs_check = rdp->qlen;

1873

}

1873

}

1874

} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1874

} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1875

force_quiescent_state(rsp, 1);

1875

force_quiescent_state(rsp, 1);

1876

local_irq_restore(flags);

1876

local_irq_restore(flags);

1877

}

1877

}

1878

1879

/*

1879

/*

1880

* Queue an RCU-sched callback for invocation after a grace period.

1880

* Queue an RCU-sched callback for invocation after a grace period.

1881

*/

1881

*/

1882

void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1882

void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1883

{

1883

{

1884

__call_rcu(head, func, &rcu_sched_state, 0);

1884

__call_rcu(head, func, &rcu_sched_state, 0);

1885

}

1885

}

1886

EXPORT_SYMBOL_GPL(call_rcu_sched);

1886

EXPORT_SYMBOL_GPL(call_rcu_sched);

1887

1888

/*

1888

/*

1889

* Queue an RCU callback for invocation after a quicker grace period.

1889

* Queue an RCU callback for invocation after a quicker grace period.

1890

*/

1890

*/

1891

void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1891

void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1892

{

1892

{

1893

__call_rcu(head, func, &rcu_bh_state, 0);

1893

__call_rcu(head, func, &rcu_bh_state, 0);

1894

}

1894

}

1895

EXPORT_SYMBOL_GPL(call_rcu_bh);

1895

EXPORT_SYMBOL_GPL(call_rcu_bh);

1896

1897

/*

1898

* Because a context switch is a grace period for RCU-sched and RCU-bh,

1899

* any blocking grace-period wait automatically implies a grace period

1900

* if there is only one CPU online at any point time during execution

1901

* of either synchronize_sched() or synchronize_rcu_bh(). It is OK to

1902

* occasionally incorrectly indicate that there are multiple CPUs online

1903

* when there was in fact only one the whole time, as this just adds

1904

* some overhead: RCU still operates correctly.

1905

*

1906

* Of course, sampling num_online_cpus() with preemption enabled can

1907

* give erroneous results if there are concurrent CPU-hotplug operations.

1908

* For example, given a demonic sequence of preemptions in num_online_cpus()

1909

* and CPU-hotplug operations, there could be two or more CPUs online at

1910

* all times, but num_online_cpus() might well return one (or even zero).

1911

*

1912

* However, all such demonic sequences require at least one CPU-offline

1913

* operation. Furthermore, rcu_blocking_is_gp() giving the wrong answer

1914

* is only a problem if there is an RCU read-side critical section executing

1915

* throughout. But RCU-sched and RCU-bh read-side critical sections

1916

* disable either preemption or bh, which prevents a CPU from going offline.

1917

* Therefore, the only way that rcu_blocking_is_gp() can incorrectly return

1918

* that there is only one CPU when in fact there was more than one throughout

1919

* is when there were no RCU readers in the system. If there are no

1920

* RCU readers, the grace period by definition can be of zero length,

1921

* regardless of the number of online CPUs.

1922

*/

1923

static inline int rcu_blocking_is_gp(void)

1924

{

1925

might_sleep(); /* Check for RCU read-side critical section. */

1926

return num_online_cpus() <= 1;

1927

}

1928

1897

/**

1929

/**

1898

* synchronize_sched - wait until an rcu-sched grace period has elapsed.

1930

* synchronize_sched - wait until an rcu-sched grace period has elapsed.

1899

*

1931

*

1900

* Control will return to the caller some time after a full rcu-sched

1932

* Control will return to the caller some time after a full rcu-sched

1901

* grace period has elapsed, in other words after all currently executing

1933

* grace period has elapsed, in other words after all currently executing

1902

* rcu-sched read-side critical sections have completed. These read-side

1934

* rcu-sched read-side critical sections have completed. These read-side

1903

* critical sections are delimited by rcu_read_lock_sched() and

1935

* critical sections are delimited by rcu_read_lock_sched() and

1904

* rcu_read_unlock_sched(), and may be nested. Note that preempt_disable(),

1936

* rcu_read_unlock_sched(), and may be nested. Note that preempt_disable(),

1905

* local_irq_disable(), and so on may be used in place of

1937

* local_irq_disable(), and so on may be used in place of

1906

* rcu_read_lock_sched().

1938

* rcu_read_lock_sched().

1907

*

1939

*

1908

* This means that all preempt_disable code sequences, including NMI and

1940

* This means that all preempt_disable code sequences, including NMI and

1909

* hardware-interrupt handlers, in progress on entry will have completed

1941

* hardware-interrupt handlers, in progress on entry will have completed

1910

* before this primitive returns. However, this does not guarantee that

1942

* before this primitive returns. However, this does not guarantee that

1911

* softirq handlers will have completed, since in some kernels, these

1943

* softirq handlers will have completed, since in some kernels, these

1912

* handlers can run in process context, and can block.

1944

* handlers can run in process context, and can block.

1913

*

1945

*

1914

* This primitive provides the guarantees made by the (now removed)

1946

* This primitive provides the guarantees made by the (now removed)

1915

* synchronize_kernel() API. In contrast, synchronize_rcu() only

1947

* synchronize_kernel() API. In contrast, synchronize_rcu() only

1916

* guarantees that rcu_read_lock() sections will have completed.

1948

* guarantees that rcu_read_lock() sections will have completed.

1917

* In "classic RCU", these two guarantees happen to be one and

1949

* In "classic RCU", these two guarantees happen to be one and

1918

* the same, but can differ in realtime RCU implementations.

1950

* the same, but can differ in realtime RCU implementations.

1919

*/

1951

*/

1920

void synchronize_sched(void)

1952

void synchronize_sched(void)

1921

{

1953

{

1922

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&

1954

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&

1923

!lock_is_held(&rcu_lock_map) &&

1955

!lock_is_held(&rcu_lock_map) &&

1924

!lock_is_held(&rcu_sched_lock_map),

1956

!lock_is_held(&rcu_sched_lock_map),

1925

"Illegal synchronize_sched() in RCU-sched read-side critical section");

1957

"Illegal synchronize_sched() in RCU-sched read-side critical section");

1926

if (rcu_blocking_is_gp())

1958

if (rcu_blocking_is_gp())

1927

return;

1959

return;

1928

wait_rcu_gp(call_rcu_sched);

1960

wait_rcu_gp(call_rcu_sched);

1929

}

1961

}

1930

EXPORT_SYMBOL_GPL(synchronize_sched);

1962

EXPORT_SYMBOL_GPL(synchronize_sched);

1931

1963

1932

/**

1964

/**

1933

* synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.

1965

* synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.

1934

*

1966

*

1935

* Control will return to the caller some time after a full rcu_bh grace

1967

* Control will return to the caller some time after a full rcu_bh grace

1936

* period has elapsed, in other words after all currently executing rcu_bh

1968

* period has elapsed, in other words after all currently executing rcu_bh

1937

* read-side critical sections have completed. RCU read-side critical

1969

* read-side critical sections have completed. RCU read-side critical

1938

* sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),

1970

* sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),

1939

* and may be nested.

1971

* and may be nested.

1940

*/

1972

*/

1941

void synchronize_rcu_bh(void)

1973

void synchronize_rcu_bh(void)

1942

{

1974

{

1943

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&

1975

rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map) &&

1944

!lock_is_held(&rcu_lock_map) &&

1976

!lock_is_held(&rcu_lock_map) &&

1945

!lock_is_held(&rcu_sched_lock_map),

1977

!lock_is_held(&rcu_sched_lock_map),

1946

"Illegal synchronize_rcu_bh() in RCU-bh read-side critical section");

1978

"Illegal synchronize_rcu_bh() in RCU-bh read-side critical section");

1947

if (rcu_blocking_is_gp())

1979

if (rcu_blocking_is_gp())

1948

return;

1980

return;

1949

wait_rcu_gp(call_rcu_bh);

1981

wait_rcu_gp(call_rcu_bh);

1950

}

1982

}

1951

EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

1983

EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

1952

1984

1953

static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0);

1985

static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0);

1954

static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0);

1986

static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0);

1955

1987

1956

static int synchronize_sched_expedited_cpu_stop(void *data)

1988

static int synchronize_sched_expedited_cpu_stop(void *data)

1957

{

1989

{

1958

/*

1990

/*

1959

* There must be a full memory barrier on each affected CPU

1991

* There must be a full memory barrier on each affected CPU

1960

* between the time that try_stop_cpus() is called and the

1992

* between the time that try_stop_cpus() is called and the

1961

* time that it returns.

1993

* time that it returns.

1962

*

1994

*

1963

* In the current initial implementation of cpu_stop, the

1995

* In the current initial implementation of cpu_stop, the

1964

* above condition is already met when the control reaches

1996

* above condition is already met when the control reaches

1965

* this point and the following smp_mb() is not strictly

1997

* this point and the following smp_mb() is not strictly

1966

* necessary. Do smp_mb() anyway for documentation and

1998

* necessary. Do smp_mb() anyway for documentation and

1967

* robustness against future implementation changes.

1999

* robustness against future implementation changes.

1968

*/

2000

*/

1969

smp_mb(); /* See above comment block. */

2001

smp_mb(); /* See above comment block. */

1970

return 0;

2002

return 0;

1971

}

2003

}

1972

2004

1973

/**

2005

/**

1974

* synchronize_sched_expedited - Brute-force RCU-sched grace period

2006

* synchronize_sched_expedited - Brute-force RCU-sched grace period

1975

*

2007

*

1976

* Wait for an RCU-sched grace period to elapse, but use a "big hammer"

2008

* Wait for an RCU-sched grace period to elapse, but use a "big hammer"

1977

* approach to force the grace period to end quickly. This consumes

2009

* approach to force the grace period to end quickly. This consumes

1978

* significant time on all CPUs and is unfriendly to real-time workloads,

2010

* significant time on all CPUs and is unfriendly to real-time workloads,

1979

* so is thus not recommended for any sort of common-case code. In fact,

2011

* so is thus not recommended for any sort of common-case code. In fact,

1980

* if you are using synchronize_sched_expedited() in a loop, please

2012

* if you are using synchronize_sched_expedited() in a loop, please

1981

* restructure your code to batch your updates, and then use a single

2013

* restructure your code to batch your updates, and then use a single

1982

* synchronize_sched() instead.

2014

* synchronize_sched() instead.

1983

*

2015

*

1984

* Note that it is illegal to call this function while holding any lock

2016

* Note that it is illegal to call this function while holding any lock

1985

* that is acquired by a CPU-hotplug notifier. And yes, it is also illegal

2017

* that is acquired by a CPU-hotplug notifier. And yes, it is also illegal

1986

* to call this function from a CPU-hotplug notifier. Failing to observe

2018

* to call this function from a CPU-hotplug notifier. Failing to observe

1987

* these restriction will result in deadlock.

2019

* these restriction will result in deadlock.

1988

*

2020

*

1989

* This implementation can be thought of as an application of ticket

2021

* This implementation can be thought of as an application of ticket

1990

* locking to RCU, with sync_sched_expedited_started and

2022

* locking to RCU, with sync_sched_expedited_started and

1991

* sync_sched_expedited_done taking on the roles of the halves

2023

* sync_sched_expedited_done taking on the roles of the halves

1992

* of the ticket-lock word. Each task atomically increments

2024

* of the ticket-lock word. Each task atomically increments

1993

* sync_sched_expedited_started upon entry, snapshotting the old value,

2025

* sync_sched_expedited_started upon entry, snapshotting the old value,

1994

* then attempts to stop all the CPUs. If this succeeds, then each

2026

* then attempts to stop all the CPUs. If this succeeds, then each

1995

* CPU will have executed a context switch, resulting in an RCU-sched

2027

* CPU will have executed a context switch, resulting in an RCU-sched

1996

* grace period. We are then done, so we use atomic_cmpxchg() to

2028

* grace period. We are then done, so we use atomic_cmpxchg() to

1997

* update sync_sched_expedited_done to match our snapshot -- but

2029

* update sync_sched_expedited_done to match our snapshot -- but

1998

* only if someone else has not already advanced past our snapshot.

2030

* only if someone else has not already advanced past our snapshot.

1999

*

2031

*

2000

* On the other hand, if try_stop_cpus() fails, we check the value

2032

* On the other hand, if try_stop_cpus() fails, we check the value

2001

* of sync_sched_expedited_done. If it has advanced past our

2033

* of sync_sched_expedited_done. If it has advanced past our

2002

* initial snapshot, then someone else must have forced a grace period

2034

* initial snapshot, then someone else must have forced a grace period

2003

* some time after we took our snapshot. In this case, our work is

2035

* some time after we took our snapshot. In this case, our work is

2004

* done for us, and we can simply return. Otherwise, we try again,

2036

* done for us, and we can simply return. Otherwise, we try again,

2005

* but keep our initial snapshot for purposes of checking for someone

2037

* but keep our initial snapshot for purposes of checking for someone

2006

* doing our work for us.

2038

* doing our work for us.

2007

*

2039

*

2008

* If we fail too many times in a row, we fall back to synchronize_sched().

2040

* If we fail too many times in a row, we fall back to synchronize_sched().

2009

*/

2041

*/

2010

void synchronize_sched_expedited(void)

2042

void synchronize_sched_expedited(void)

2011

{

2043

{

2012

int firstsnap, s, snap, trycount = 0;

2044

int firstsnap, s, snap, trycount = 0;

2013

2045

2014

/* Note that atomic_inc_return() implies full memory barrier. */

2046

/* Note that atomic_inc_return() implies full memory barrier. */

2015

firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started);

2047

firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started);

2016

get_online_cpus();

2048

get_online_cpus();

2017

WARN_ON_ONCE(cpu_is_offline(raw_smp_processor_id()));

2049

WARN_ON_ONCE(cpu_is_offline(raw_smp_processor_id()));

2018

2050

2019

/*

2051

/*

2020

* Each pass through the following loop attempts to force a

2052

* Each pass through the following loop attempts to force a

2021

* context switch on each CPU.

2053

* context switch on each CPU.

2022

*/

2054

*/

2023

while (try_stop_cpus(cpu_online_mask,

2055

while (try_stop_cpus(cpu_online_mask,

2024

synchronize_sched_expedited_cpu_stop,

2056

synchronize_sched_expedited_cpu_stop,

2025

NULL) == -EAGAIN) {

2057

NULL) == -EAGAIN) {

2026

put_online_cpus();

2058

put_online_cpus();

2027

2059

2028

/* No joy, try again later. Or just synchronize_sched(). */

2060

/* No joy, try again later. Or just synchronize_sched(). */

2029

if (trycount++ < 10)

2061

if (trycount++ < 10)

2030

udelay(trycount * num_online_cpus());

2062

udelay(trycount * num_online_cpus());

2031

else {

2063

else {

2032

synchronize_sched();

2064

synchronize_sched();

2033

return;

2065

return;

2034

}

2066

}

2035

2067

2036

/* Check to see if someone else did our work for us. */

2068

/* Check to see if someone else did our work for us. */

2037

s = atomic_read(&sync_sched_expedited_done);

2069

s = atomic_read(&sync_sched_expedited_done);

2038

if (UINT_CMP_GE((unsigned)s, (unsigned)firstsnap)) {

2070

if (UINT_CMP_GE((unsigned)s, (unsigned)firstsnap)) {

2039

smp_mb(); /* ensure test happens before caller kfree */

2071

smp_mb(); /* ensure test happens before caller kfree */

2040

return;

2072

return;

2041

}

2073

}

2042

2074

2043

/*

2075

/*

2044

* Refetching sync_sched_expedited_started allows later

2076

* Refetching sync_sched_expedited_started allows later

2045

* callers to piggyback on our grace period. We subtract

2077

* callers to piggyback on our grace period. We subtract

2046

* 1 to get the same token that the last incrementer got.

2078

* 1 to get the same token that the last incrementer got.

2047

* We retry after they started, so our grace period works

2079

* We retry after they started, so our grace period works

2048

* for them, and they started after our first try, so their

2080

* for them, and they started after our first try, so their

2049

* grace period works for us.

2081

* grace period works for us.

2050

*/

2082

*/

2051

get_online_cpus();

2083

get_online_cpus();

2052

snap = atomic_read(&sync_sched_expedited_started);

2084

snap = atomic_read(&sync_sched_expedited_started);

2053

smp_mb(); /* ensure read is before try_stop_cpus(). */

2085

smp_mb(); /* ensure read is before try_stop_cpus(). */

2054

}

2086

}

2055

2087

2056

/*

2088

/*

2057

* Everyone up to our most recent fetch is covered by our grace

2089

* Everyone up to our most recent fetch is covered by our grace

2058

* period. Update the counter, but only if our work is still

2090

* period. Update the counter, but only if our work is still

2059

* relevant -- which it won't be if someone who started later

2091

* relevant -- which it won't be if someone who started later

2060

* than we did beat us to the punch.

2092

* than we did beat us to the punch.

2061

*/

2093

*/

2062

do {

2094

do {

2063

s = atomic_read(&sync_sched_expedited_done);

2095

s = atomic_read(&sync_sched_expedited_done);

2064

if (UINT_CMP_GE((unsigned)s, (unsigned)snap)) {

2096

if (UINT_CMP_GE((unsigned)s, (unsigned)snap)) {

2065

smp_mb(); /* ensure test happens before caller kfree */

2097

smp_mb(); /* ensure test happens before caller kfree */

2066

break;

2098

break;

2067

}

2099

}

2068

} while (atomic_cmpxchg(&sync_sched_expedited_done, s, snap) != s);

2100

} while (atomic_cmpxchg(&sync_sched_expedited_done, s, snap) != s);

2069

2101

2070

put_online_cpus();

2102

put_online_cpus();

2071

}

2103

}

2072

EXPORT_SYMBOL_GPL(synchronize_sched_expedited);

2104

EXPORT_SYMBOL_GPL(synchronize_sched_expedited);

2073

2105

2074

/*

2106

/*

2075

* Check to see if there is any immediate RCU-related work to be done

2107

* Check to see if there is any immediate RCU-related work to be done

2076

* by the current CPU, for the specified type of RCU, returning 1 if so.

2108

* by the current CPU, for the specified type of RCU, returning 1 if so.

2077

* The checks are in order of increasing expense: checks that can be

2109

* The checks are in order of increasing expense: checks that can be

2078

* carried out against CPU-local state are performed first. However,

2110

* carried out against CPU-local state are performed first. However,

2079

* we must check for CPU stalls first, else we might not get a chance.

2111

* we must check for CPU stalls first, else we might not get a chance.

2080

*/

2112

*/

2081

static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)

2113

static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)

2082

{

2114

{

2083

struct rcu_node *rnp = rdp->mynode;

2115

struct rcu_node *rnp = rdp->mynode;

2084

2116

2085

rdp->n_rcu_pending++;

2117

rdp->n_rcu_pending++;

2086

2118

2087

/* Check for CPU stalls, if enabled. */

2119

/* Check for CPU stalls, if enabled. */

2088

check_cpu_stall(rsp, rdp);

2120

check_cpu_stall(rsp, rdp);

2089

2121

2090

/* Is the RCU core waiting for a quiescent state from this CPU? */

2122

/* Is the RCU core waiting for a quiescent state from this CPU? */

2091

if (rcu_scheduler_fully_active &&

2123

if (rcu_scheduler_fully_active &&

2092

rdp->qs_pending && !rdp->passed_quiesce) {

2124

rdp->qs_pending && !rdp->passed_quiesce) {

2093

2125

2094

/*

2126

/*

2095

* If force_quiescent_state() coming soon and this CPU

2127

* If force_quiescent_state() coming soon and this CPU

2096

* needs a quiescent state, and this is either RCU-sched

2128

* needs a quiescent state, and this is either RCU-sched

2097

* or RCU-bh, force a local reschedule.

2129

* or RCU-bh, force a local reschedule.

2098

*/

2130

*/

2099

rdp->n_rp_qs_pending++;

2131

rdp->n_rp_qs_pending++;

2100

if (!rdp->preemptible &&

2132

if (!rdp->preemptible &&

2101

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,

2133

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,

2102

jiffies))

2134

jiffies))

2103

set_need_resched();

2135

set_need_resched();

2104

} else if (rdp->qs_pending && rdp->passed_quiesce) {

2136

} else if (rdp->qs_pending && rdp->passed_quiesce) {

2105

rdp->n_rp_report_qs++;

2137

rdp->n_rp_report_qs++;

2106

return 1;

2138

return 1;

2107

}

2139

}

2108

2140

2109

/* Does this CPU have callbacks ready to invoke? */

2141

/* Does this CPU have callbacks ready to invoke? */

2110

if (cpu_has_callbacks_ready_to_invoke(rdp)) {

2142

if (cpu_has_callbacks_ready_to_invoke(rdp)) {

2111

rdp->n_rp_cb_ready++;

2143

rdp->n_rp_cb_ready++;

2112

return 1;

2144

return 1;

2113

}

2145

}

2114

2146

2115

/* Has RCU gone idle with this CPU needing another grace period? */

2147

/* Has RCU gone idle with this CPU needing another grace period? */

2116

if (cpu_needs_another_gp(rsp, rdp)) {

2148

if (cpu_needs_another_gp(rsp, rdp)) {

2117

rdp->n_rp_cpu_needs_gp++;

2149

rdp->n_rp_cpu_needs_gp++;

2118

return 1;

2150

return 1;

2119

}

2151

}

2120

2152

2121

/* Has another RCU grace period completed? */

2153

/* Has another RCU grace period completed? */

2122

if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */

2154

if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */

2123

rdp->n_rp_gp_completed++;

2155

rdp->n_rp_gp_completed++;

2124

return 1;

2156

return 1;

2125

}

2157

}

2126

2158

2127

/* Has a new RCU grace period started? */

2159

/* Has a new RCU grace period started? */

2128

if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */

2160

if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */

2129

rdp->n_rp_gp_started++;

2161

rdp->n_rp_gp_started++;

2130

return 1;

2162

return 1;

2131

}

2163

}

2132

2164

2133

/* Has an RCU GP gone long enough to send resched IPIs &c? */

2165

/* Has an RCU GP gone long enough to send resched IPIs &c? */

2134

if (rcu_gp_in_progress(rsp) &&

2166

if (rcu_gp_in_progress(rsp) &&

2135

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {

2167

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {

2136

rdp->n_rp_need_fqs++;

2168

rdp->n_rp_need_fqs++;

2137

return 1;

2169

return 1;

2138

}

2170

}

2139

2171

2140

/* nothing to do */

2172

/* nothing to do */

2141

rdp->n_rp_need_nothing++;

2173

rdp->n_rp_need_nothing++;

2142

return 0;

2174

return 0;

2143

}

2175

}

2144

2176

2145

/*

2177

/*

2146

* Check to see if there is any immediate RCU-related work to be done

2178

* Check to see if there is any immediate RCU-related work to be done

2147

* by the current CPU, returning 1 if so. This function is part of the

2179

* by the current CPU, returning 1 if so. This function is part of the

2148

* RCU implementation; it is -not- an exported member of the RCU API.

2180

* RCU implementation; it is -not- an exported member of the RCU API.

2149

*/

2181

*/

2150

static int rcu_pending(int cpu)

2182

static int rcu_pending(int cpu)

2151

{

2183

{

2152

return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||

2184

return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||

2153

__rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||

2185

__rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||

2154

rcu_preempt_pending(cpu);

2186

rcu_preempt_pending(cpu);

2155

}

2187

}

2156

2188

2157

/*

2189

/*

2158

* Check to see if any future RCU-related work will need to be done

2190

* Check to see if any future RCU-related work will need to be done

2159

* by the current CPU, even if none need be done immediately, returning

2191

* by the current CPU, even if none need be done immediately, returning

2160

* 1 if so.

2192

* 1 if so.

2161

*/

2193

*/

2162

static int rcu_cpu_has_callbacks(int cpu)

2194

static int rcu_cpu_has_callbacks(int cpu)

2163

{

2195

{

2164

/* RCU callbacks either ready or pending? */

2196

/* RCU callbacks either ready or pending? */

2165

return per_cpu(rcu_sched_data, cpu).nxtlist ||

2197

return per_cpu(rcu_sched_data, cpu).nxtlist ||

2166

per_cpu(rcu_bh_data, cpu).nxtlist ||

2198

per_cpu(rcu_bh_data, cpu).nxtlist ||

2167

rcu_preempt_cpu_has_callbacks(cpu);

2199

rcu_preempt_cpu_has_callbacks(cpu);

2168

}

2200

}

2169

2201

2170

static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};

2202

static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};

2171

static atomic_t rcu_barrier_cpu_count;

2203

static atomic_t rcu_barrier_cpu_count;

2172

static DEFINE_MUTEX(rcu_barrier_mutex);

2204

static DEFINE_MUTEX(rcu_barrier_mutex);

2173

static struct completion rcu_barrier_completion;

2205

static struct completion rcu_barrier_completion;

2174

2206

2175

static void rcu_barrier_callback(struct rcu_head *notused)

2207

static void rcu_barrier_callback(struct rcu_head *notused)

2176

{

2208

{

2177

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

2209

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

2178

complete(&rcu_barrier_completion);

2210

complete(&rcu_barrier_completion);

2179

}

2211

}

2180

2212

2181

/*

2213

/*

2182

* Called with preemption disabled, and from cross-cpu IRQ context.

2214

* Called with preemption disabled, and from cross-cpu IRQ context.

2183

*/

2215

*/

2184

static void rcu_barrier_func(void *type)

2216

static void rcu_barrier_func(void *type)

2185

{

2217

{

2186

int cpu = smp_processor_id();

2218

int cpu = smp_processor_id();

2187

struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);

2219

struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);

2188

void (*call_rcu_func)(struct rcu_head *head,

2220

void (*call_rcu_func)(struct rcu_head *head,

2189

void (*func)(struct rcu_head *head));

2221

void (*func)(struct rcu_head *head));

2190

2222

2191

atomic_inc(&rcu_barrier_cpu_count);

2223

atomic_inc(&rcu_barrier_cpu_count);

2192

call_rcu_func = type;

2224

call_rcu_func = type;

2193

call_rcu_func(head, rcu_barrier_callback);

2225

call_rcu_func(head, rcu_barrier_callback);

2194

}

2226

}

2195

2227

2196

/*

2228

/*

2197

* Orchestrate the specified type of RCU barrier, waiting for all

2229

* Orchestrate the specified type of RCU barrier, waiting for all

2198

* RCU callbacks of the specified type to complete.

2230

* RCU callbacks of the specified type to complete.

2199

*/

2231

*/

2200

static void _rcu_barrier(struct rcu_state *rsp,

2232

static void _rcu_barrier(struct rcu_state *rsp,

2201

void (*call_rcu_func)(struct rcu_head *head,

2233

void (*call_rcu_func)(struct rcu_head *head,

2202

void (*func)(struct rcu_head *head)))

2234

void (*func)(struct rcu_head *head)))

2203

{

2235

{

2204

BUG_ON(in_interrupt());

2236

BUG_ON(in_interrupt());

2205

/* Take mutex to serialize concurrent rcu_barrier() requests. */

2237

/* Take mutex to serialize concurrent rcu_barrier() requests. */

2206

mutex_lock(&rcu_barrier_mutex);

2238

mutex_lock(&rcu_barrier_mutex);

2207

init_completion(&rcu_barrier_completion);

2239

init_completion(&rcu_barrier_completion);

2208

/*

2240

/*

2209

* Initialize rcu_barrier_cpu_count to 1, then invoke

2241

* Initialize rcu_barrier_cpu_count to 1, then invoke

2210

* rcu_barrier_func() on each CPU, so that each CPU also has

2242

* rcu_barrier_func() on each CPU, so that each CPU also has

2211

* incremented rcu_barrier_cpu_count. Only then is it safe to

2243

* incremented rcu_barrier_cpu_count. Only then is it safe to

2212

* decrement rcu_barrier_cpu_count -- otherwise the first CPU

2244

* decrement rcu_barrier_cpu_count -- otherwise the first CPU

2213

* might complete its grace period before all of the other CPUs

2245

* might complete its grace period before all of the other CPUs

2214

* did their increment, causing this function to return too

2246

* did their increment, causing this function to return too

2215

* early. Note that on_each_cpu() disables irqs, which prevents

2247

* early. Note that on_each_cpu() disables irqs, which prevents

2216

* any CPUs from coming online or going offline until each online

2248

* any CPUs from coming online or going offline until each online

2217

* CPU has queued its RCU-barrier callback.

2249

* CPU has queued its RCU-barrier callback.

2218

*/

2250

*/

2219

atomic_set(&rcu_barrier_cpu_count, 1);

2251

atomic_set(&rcu_barrier_cpu_count, 1);

2220

on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);

2252

on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);

2221

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

2253

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

2222

complete(&rcu_barrier_completion);

2254

complete(&rcu_barrier_completion);

2223

wait_for_completion(&rcu_barrier_completion);

2255

wait_for_completion(&rcu_barrier_completion);

2224

mutex_unlock(&rcu_barrier_mutex);

2256

mutex_unlock(&rcu_barrier_mutex);

2225

}

2257

}

2226

2258

2227

/**

2259

/**

2228

* rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.

2260

* rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.

2229

*/

2261

*/

2230

void rcu_barrier_bh(void)

2262

void rcu_barrier_bh(void)

2231

{

2263

{

2232

_rcu_barrier(&rcu_bh_state, call_rcu_bh);

2264

_rcu_barrier(&rcu_bh_state, call_rcu_bh);

2233

}

2265

}

2234

EXPORT_SYMBOL_GPL(rcu_barrier_bh);

2266

EXPORT_SYMBOL_GPL(rcu_barrier_bh);

2235

2267

2236

/**

2268

/**

2237

* rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.

2269

* rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.

2238

*/

2270

*/

2239

void rcu_barrier_sched(void)

2271

void rcu_barrier_sched(void)

2240

{

2272

{

2241

_rcu_barrier(&rcu_sched_state, call_rcu_sched);

2273

_rcu_barrier(&rcu_sched_state, call_rcu_sched);

2242

}

2274

}

2243

EXPORT_SYMBOL_GPL(rcu_barrier_sched);

2275

EXPORT_SYMBOL_GPL(rcu_barrier_sched);

2244

2276

2245

/*

2277

/*

2246

* Do boot-time initialization of a CPU's per-CPU RCU data.

2278

* Do boot-time initialization of a CPU's per-CPU RCU data.

2247

*/

2279

*/

2248

static void __init

2280

static void __init

2249

rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)

2281

rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)

2250

{

2282

{

2251

unsigned long flags;

2283

unsigned long flags;

2252

int i;

2284

int i;

2253

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

2285

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

2254

struct rcu_node *rnp = rcu_get_root(rsp);

2286

struct rcu_node *rnp = rcu_get_root(rsp);

2255

2287

2256

/* Set up local state, ensuring consistent view of global state. */

2288

/* Set up local state, ensuring consistent view of global state. */

2257

raw_spin_lock_irqsave(&rnp->lock, flags);

2289

raw_spin_lock_irqsave(&rnp->lock, flags);

2258

rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);

2290

rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);

2259

rdp->nxtlist = NULL;

2291

rdp->nxtlist = NULL;

2260

for (i = 0; i < RCU_NEXT_SIZE; i++)

2292

for (i = 0; i < RCU_NEXT_SIZE; i++)

2261

rdp->nxttail[i] = &rdp->nxtlist;

2293

rdp->nxttail[i] = &rdp->nxtlist;

2262

rdp->qlen_lazy = 0;

2294

rdp->qlen_lazy = 0;

2263

rdp->qlen = 0;

2295

rdp->qlen = 0;

2264

rdp->dynticks = &per_cpu(rcu_dynticks, cpu);

2296

rdp->dynticks = &per_cpu(rcu_dynticks, cpu);

2265

WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);

2297

WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);

2266

WARN_ON_ONCE(atomic_read(&rdp->dynticks->dynticks) != 1);

2298

WARN_ON_ONCE(atomic_read(&rdp->dynticks->dynticks) != 1);

2267

rdp->cpu = cpu;

2299

rdp->cpu = cpu;

2268

rdp->rsp = rsp;

2300

rdp->rsp = rsp;

2269

raw_spin_unlock_irqrestore(&rnp->lock, flags);

2301

raw_spin_unlock_irqrestore(&rnp->lock, flags);

2270

}

2302

}

2271

2303

2272

/*

2304

/*

2273

* Initialize a CPU's per-CPU RCU data. Note that only one online or

2305

* Initialize a CPU's per-CPU RCU data. Note that only one online or

2274

* offline event can be happening at a given time. Note also that we

2306

* offline event can be happening at a given time. Note also that we

2275

* can accept some slop in the rsp->completed access due to the fact

2307

* can accept some slop in the rsp->completed access due to the fact

2276

* that this CPU cannot possibly have any RCU callbacks in flight yet.

2308

* that this CPU cannot possibly have any RCU callbacks in flight yet.

2277

*/

2309

*/

2278

static void __cpuinit

2310

static void __cpuinit

2279

rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptible)

2311

rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptible)

2280

{

2312

{

2281

unsigned long flags;

2313

unsigned long flags;

2282

unsigned long mask;

2314

unsigned long mask;

2283

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

2315

struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);

2284

struct rcu_node *rnp = rcu_get_root(rsp);

2316

struct rcu_node *rnp = rcu_get_root(rsp);

2285

2317

2286

/* Set up local state, ensuring consistent view of global state. */

2318

/* Set up local state, ensuring consistent view of global state. */

2287

raw_spin_lock_irqsave(&rnp->lock, flags);

2319

raw_spin_lock_irqsave(&rnp->lock, flags);

2288

rdp->beenonline = 1; /* We have now been online. */

2320

rdp->beenonline = 1; /* We have now been online. */

2289

rdp->preemptible = preemptible;

2321

rdp->preemptible = preemptible;

2290

rdp->qlen_last_fqs_check = 0;

2322

rdp->qlen_last_fqs_check = 0;

2291

rdp->n_force_qs_snap = rsp->n_force_qs;

2323

rdp->n_force_qs_snap = rsp->n_force_qs;

2292

rdp->blimit = blimit;

2324

rdp->blimit = blimit;

2293

rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;

2325

rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;

2294

atomic_set(&rdp->dynticks->dynticks,

2326

atomic_set(&rdp->dynticks->dynticks,

2295

(atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);

2327

(atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);

2296

rcu_prepare_for_idle_init(cpu);

2328

rcu_prepare_for_idle_init(cpu);

2297

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

2329

raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */

2298

2330

2299

/*

2331

/*

2300

* A new grace period might start here. If so, we won't be part

2332

* A new grace period might start here. If so, we won't be part

2301

* of it, but that is OK, as we are currently in a quiescent state.

2333

* of it, but that is OK, as we are currently in a quiescent state.

2302

*/

2334

*/

2303

2335

2304

/* Exclude any attempts to start a new GP on large systems. */

2336

/* Exclude any attempts to start a new GP on large systems. */

2305

raw_spin_lock(&rsp->onofflock); /* irqs already disabled. */

2337

raw_spin_lock(&rsp->onofflock); /* irqs already disabled. */

2306

2338

2307

/* Add CPU to rcu_node bitmasks. */

2339

/* Add CPU to rcu_node bitmasks. */

2308

rnp = rdp->mynode;

2340

rnp = rdp->mynode;

2309

mask = rdp->grpmask;

2341

mask = rdp->grpmask;

2310

do {

2342

do {

2311

/* Exclude any attempts to start a new GP on small systems. */

2343

/* Exclude any attempts to start a new GP on small systems. */

2312

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

2344

raw_spin_lock(&rnp->lock); /* irqs already disabled. */

2313

rnp->qsmaskinit |= mask;

2345

rnp->qsmaskinit |= mask;

2314

mask = rnp->grpmask;

2346

mask = rnp->grpmask;

2315

if (rnp == rdp->mynode) {

2347

if (rnp == rdp->mynode) {

2316

/*

2348

/*

2317

* If there is a grace period in progress, we will

2349

* If there is a grace period in progress, we will

2318

* set up to wait for it next time we run the

2350

* set up to wait for it next time we run the

2319

* RCU core code.

2351

* RCU core code.

2320

*/

2352

*/

2321

rdp->gpnum = rnp->completed;

2353

rdp->gpnum = rnp->completed;

2322

rdp->completed = rnp->completed;

2354

rdp->completed = rnp->completed;

2323

rdp->passed_quiesce = 0;

2355

rdp->passed_quiesce = 0;

2324

rdp->qs_pending = 0;

2356

rdp->qs_pending = 0;

2325

rdp->passed_quiesce_gpnum = rnp->gpnum - 1;

2357

rdp->passed_quiesce_gpnum = rnp->gpnum - 1;

2326

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuonl");

2358

trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuonl");

2327

}

2359

}

2328

raw_spin_unlock(&rnp->lock); /* irqs already disabled. */

2360

raw_spin_unlock(&rnp->lock); /* irqs already disabled. */

2329

rnp = rnp->parent;

2361

rnp = rnp->parent;

2330

} while (rnp != NULL && !(rnp->qsmaskinit & mask));

2362

} while (rnp != NULL && !(rnp->qsmaskinit & mask));

2331

2363

2332

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

2364

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

2333

}

2365

}

2334

2366

2335

static void __cpuinit rcu_prepare_cpu(int cpu)

2367

static void __cpuinit rcu_prepare_cpu(int cpu)

2336

{

2368

{

2337

rcu_init_percpu_data(cpu, &rcu_sched_state, 0);

2369

rcu_init_percpu_data(cpu, &rcu_sched_state, 0);

2338

rcu_init_percpu_data(cpu, &rcu_bh_state, 0);

2370

rcu_init_percpu_data(cpu, &rcu_bh_state, 0);

2339

rcu_preempt_init_percpu_data(cpu);

2371

rcu_preempt_init_percpu_data(cpu);

2340

}

2372

}

2341

2373

2342

/*

2374

/*

2343

* Handle CPU online/offline notification events.

2375

* Handle CPU online/offline notification events.

2344

*/

2376

*/

2345

static int __cpuinit rcu_cpu_notify(struct notifier_block *self,

2377

static int __cpuinit rcu_cpu_notify(struct notifier_block *self,

2346

unsigned long action, void *hcpu)

2378

unsigned long action, void *hcpu)

2347

{

2379

{

2348

long cpu = (long)hcpu;

2380

long cpu = (long)hcpu;

2349

struct rcu_data *rdp = per_cpu_ptr(rcu_state->rda, cpu);

2381

struct rcu_data *rdp = per_cpu_ptr(rcu_state->rda, cpu);

2350

struct rcu_node *rnp = rdp->mynode;

2382

struct rcu_node *rnp = rdp->mynode;

2351

2383

2352

trace_rcu_utilization("Start CPU hotplug");

2384

trace_rcu_utilization("Start CPU hotplug");

2353

switch (action) {

2385

switch (action) {

2354

case CPU_UP_PREPARE:

2386

case CPU_UP_PREPARE:

2355

case CPU_UP_PREPARE_FROZEN:

2387

case CPU_UP_PREPARE_FROZEN:

2356

rcu_prepare_cpu(cpu);

2388

rcu_prepare_cpu(cpu);

2357

rcu_prepare_kthreads(cpu);

2389

rcu_prepare_kthreads(cpu);

2358

break;

2390

break;

2359

case CPU_ONLINE:

2391

case CPU_ONLINE:

2360

case CPU_DOWN_FAILED:

2392

case CPU_DOWN_FAILED:

2361

rcu_node_kthread_setaffinity(rnp, -1);

2393

rcu_node_kthread_setaffinity(rnp, -1);

2362

rcu_cpu_kthread_setrt(cpu, 1);

2394

rcu_cpu_kthread_setrt(cpu, 1);

2363

break;

2395

break;

2364

case CPU_DOWN_PREPARE:

2396

case CPU_DOWN_PREPARE:

2365

rcu_node_kthread_setaffinity(rnp, cpu);

2397

rcu_node_kthread_setaffinity(rnp, cpu);

2366

rcu_cpu_kthread_setrt(cpu, 0);

2398

rcu_cpu_kthread_setrt(cpu, 0);

2367

break;

2399

break;

2368

case CPU_DYING:

2400

case CPU_DYING:

2369

case CPU_DYING_FROZEN:

2401

case CPU_DYING_FROZEN:

2370

/*

2402

/*

2371

* The whole machine is "stopped" except this CPU, so we can

2403

* The whole machine is "stopped" except this CPU, so we can

2372

* touch any data without introducing corruption. We send the

2404

* touch any data without introducing corruption. We send the

2373

* dying CPU's callbacks to an arbitrarily chosen online CPU.

2405

* dying CPU's callbacks to an arbitrarily chosen online CPU.

2374

*/

2406

*/

2375

rcu_cleanup_dying_cpu(&rcu_bh_state);

2407

rcu_cleanup_dying_cpu(&rcu_bh_state);

2376

rcu_cleanup_dying_cpu(&rcu_sched_state);

2408

rcu_cleanup_dying_cpu(&rcu_sched_state);

2377

rcu_preempt_cleanup_dying_cpu();

2409

rcu_preempt_cleanup_dying_cpu();

2378

rcu_cleanup_after_idle(cpu);

2410

rcu_cleanup_after_idle(cpu);

2379

break;

2411

break;

2380

case CPU_DEAD:

2412

case CPU_DEAD:

2381

case CPU_DEAD_FROZEN:

2413

case CPU_DEAD_FROZEN:

2382

case CPU_UP_CANCELED:

2414

case CPU_UP_CANCELED:

2383

case CPU_UP_CANCELED_FROZEN:

2415

case CPU_UP_CANCELED_FROZEN:

2384

rcu_cleanup_dead_cpu(cpu, &rcu_bh_state);

2416

rcu_cleanup_dead_cpu(cpu, &rcu_bh_state);

2385

rcu_cleanup_dead_cpu(cpu, &rcu_sched_state);

2417

rcu_cleanup_dead_cpu(cpu, &rcu_sched_state);

2386

rcu_preempt_cleanup_dead_cpu(cpu);

2418

rcu_preempt_cleanup_dead_cpu(cpu);

2387

break;

2419

break;

2388

default:

2420

default:

2389

break;

2421

break;

2390

}

2422

}

2391

trace_rcu_utilization("End CPU hotplug");

2423

trace_rcu_utilization("End CPU hotplug");

2392

return NOTIFY_OK;

2424

return NOTIFY_OK;

2393

}

2425

}

2394

2426

2395

/*

2427

/*

2396

* This function is invoked towards the end of the scheduler's initialization

2428

* This function is invoked towards the end of the scheduler's initialization

2397

* process. Before this is called, the idle task might contain

2429

* process. Before this is called, the idle task might contain

2398

* RCU read-side critical sections (during which time, this idle

2430

* RCU read-side critical sections (during which time, this idle

2399

* task is booting the system). After this function is called, the

2431

* task is booting the system). After this function is called, the

2400

* idle tasks are prohibited from containing RCU read-side critical

2432

* idle tasks are prohibited from containing RCU read-side critical

2401

* sections. This function also enables RCU lockdep checking.

2433

* sections. This function also enables RCU lockdep checking.

2402

*/

2434

*/

2403

void rcu_scheduler_starting(void)

2435

void rcu_scheduler_starting(void)

2404

{

2436

{

2405

WARN_ON(num_online_cpus() != 1);

2437

WARN_ON(num_online_cpus() != 1);

2406

WARN_ON(nr_context_switches() > 0);

2438

WARN_ON(nr_context_switches() > 0);

2407

rcu_scheduler_active = 1;

2439

rcu_scheduler_active = 1;

2408

}

2440

}

2409

2441

2410

/*

2442

/*

2411

* Compute the per-level fanout, either using the exact fanout specified

2443

* Compute the per-level fanout, either using the exact fanout specified

2412

* or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.

2444

* or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.

2413

*/

2445

*/

2414

#ifdef CONFIG_RCU_FANOUT_EXACT

2446

#ifdef CONFIG_RCU_FANOUT_EXACT

2415

static void __init rcu_init_levelspread(struct rcu_state *rsp)

2447

static void __init rcu_init_levelspread(struct rcu_state *rsp)

2416

{

2448

{

2417

int i;

2449

int i;

2418

2450

2419

for (i = NUM_RCU_LVLS - 1; i > 0; i--)

2451

for (i = NUM_RCU_LVLS - 1; i > 0; i--)

2420

rsp->levelspread[i] = CONFIG_RCU_FANOUT;

2452

rsp->levelspread[i] = CONFIG_RCU_FANOUT;

2421

rsp->levelspread[0] = CONFIG_RCU_FANOUT_LEAF;

2453

rsp->levelspread[0] = CONFIG_RCU_FANOUT_LEAF;

2422

}

2454

}

2423

#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */

2455

#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */

2424

static void __init rcu_init_levelspread(struct rcu_state *rsp)

2456

static void __init rcu_init_levelspread(struct rcu_state *rsp)

2425

{

2457

{

2426

int ccur;

2458

int ccur;

2427

int cprv;

2459

int cprv;

2428

int i;

2460

int i;

2429

2461

2430

cprv = NR_CPUS;

2462

cprv = NR_CPUS;

2431

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

2463

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

2432

ccur = rsp->levelcnt[i];

2464

ccur = rsp->levelcnt[i];

2433

rsp->levelspread[i] = (cprv + ccur - 1) / ccur;

2465

rsp->levelspread[i] = (cprv + ccur - 1) / ccur;

2434

cprv = ccur;

2466

cprv = ccur;

2435

}

2467

}

2436

}

2468

}

2437

#endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */

2469

#endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */

2438

2470

2439

/*

2471

/*

2440

* Helper function for rcu_init() that initializes one rcu_state structure.

2472

* Helper function for rcu_init() that initializes one rcu_state structure.

2441

*/

2473

*/

2442

static void __init rcu_init_one(struct rcu_state *rsp,

2474

static void __init rcu_init_one(struct rcu_state *rsp,

2443

struct rcu_data __percpu *rda)

2475

struct rcu_data __percpu *rda)

2444

{

2476

{

2445

static char *buf[] = { "rcu_node_level_0",

2477

static char *buf[] = { "rcu_node_level_0",

2446

"rcu_node_level_1",

2478

"rcu_node_level_1",

2447

"rcu_node_level_2",

2479

"rcu_node_level_2",

2448

"rcu_node_level_3" }; /* Match MAX_RCU_LVLS */

2480

"rcu_node_level_3" }; /* Match MAX_RCU_LVLS */

2449

int cpustride = 1;

2481

int cpustride = 1;

2450

int i;

2482

int i;

2451

int j;

2483

int j;

2452

struct rcu_node *rnp;

2484

struct rcu_node *rnp;

2453

2485

2454

BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf)); /* Fix buf[] init! */

2486

BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf)); /* Fix buf[] init! */

2455

2487

2456

/* Initialize the level-tracking arrays. */

2488

/* Initialize the level-tracking arrays. */

2457

2489

2458

for (i = 1; i < NUM_RCU_LVLS; i++)

2490

for (i = 1; i < NUM_RCU_LVLS; i++)

2459

rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];

2491

rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];

2460

rcu_init_levelspread(rsp);

2492

rcu_init_levelspread(rsp);

2461

2493

2462

/* Initialize the elements themselves, starting from the leaves. */

2494

/* Initialize the elements themselves, starting from the leaves. */

2463

2495

2464

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

2496

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

2465

cpustride *= rsp->levelspread[i];

2497

cpustride *= rsp->levelspread[i];

2466

rnp = rsp->level[i];

2498

rnp = rsp->level[i];

2467

for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {

2499

for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {

2468

raw_spin_lock_init(&rnp->lock);

2500

raw_spin_lock_init(&rnp->lock);

2469

lockdep_set_class_and_name(&rnp->lock,

2501

lockdep_set_class_and_name(&rnp->lock,

2470

&rcu_node_class[i], buf[i]);

2502

&rcu_node_class[i], buf[i]);

2471

rnp->gpnum = 0;

2503

rnp->gpnum = 0;

2472

rnp->qsmask = 0;

2504

rnp->qsmask = 0;

2473

rnp->qsmaskinit = 0;

2505

rnp->qsmaskinit = 0;

2474

rnp->grplo = j * cpustride;

2506

rnp->grplo = j * cpustride;

2475

rnp->grphi = (j + 1) * cpustride - 1;

2507

rnp->grphi = (j + 1) * cpustride - 1;

2476

if (rnp->grphi >= NR_CPUS)

2508

if (rnp->grphi >= NR_CPUS)

2477

rnp->grphi = NR_CPUS - 1;

2509

rnp->grphi = NR_CPUS - 1;

2478

if (i == 0) {

2510

if (i == 0) {

2479

rnp->grpnum = 0;

2511

rnp->grpnum = 0;

2480

rnp->grpmask = 0;

2512

rnp->grpmask = 0;

2481

rnp->parent = NULL;

2513

rnp->parent = NULL;

2482

} else {

2514

} else {

2483

rnp->grpnum = j % rsp->levelspread[i - 1];

2515

rnp->grpnum = j % rsp->levelspread[i - 1];

2484

rnp->grpmask = 1UL << rnp->grpnum;

2516

rnp->grpmask = 1UL << rnp->grpnum;

2485

rnp->parent = rsp->level[i - 1] +

2517

rnp->parent = rsp->level[i - 1] +

2486

j / rsp->levelspread[i - 1];

2518

j / rsp->levelspread[i - 1];

2487

}

2519

}

2488

rnp->level = i;

2520

rnp->level = i;

2489

INIT_LIST_HEAD(&rnp->blkd_tasks);

2521

INIT_LIST_HEAD(&rnp->blkd_tasks);

2490

}

2522

}

2491

}

2523

}

2492

2524

2493

rsp->rda = rda;

2525

rsp->rda = rda;

2494

rnp = rsp->level[NUM_RCU_LVLS - 1];

2526

rnp = rsp->level[NUM_RCU_LVLS - 1];

2495

for_each_possible_cpu(i) {

2527

for_each_possible_cpu(i) {

2496

while (i > rnp->grphi)

2528

while (i > rnp->grphi)

2497

rnp++;

2529

rnp++;

2498

per_cpu_ptr(rsp->rda, i)->mynode = rnp;

2530

per_cpu_ptr(rsp->rda, i)->mynode = rnp;

2499

rcu_boot_init_percpu_data(i, rsp);

2531

rcu_boot_init_percpu_data(i, rsp);

2500

}

2532

}

2501

}

2533

}

2502

2534

2503

void __init rcu_init(void)

2535

void __init rcu_init(void)

2504

{

2536

{

2505

int cpu;

2537

int cpu;

2506

2538

2507

rcu_bootup_announce();

2539

rcu_bootup_announce();

2508

rcu_init_one(&rcu_sched_state, &rcu_sched_data);

2540

rcu_init_one(&rcu_sched_state, &rcu_sched_data);

2509

rcu_init_one(&rcu_bh_state, &rcu_bh_data);

2541

rcu_init_one(&rcu_bh_state, &rcu_bh_data);

2510

__rcu_init_preempt();

2542

__rcu_init_preempt();

2511

open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);

2543

open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);

2512

2544

2513

/*

2545

/*

2514

* We don't need protection against CPU-hotplug here because

2546

* We don't need protection against CPU-hotplug here because

2515

* this is called early in boot, before either interrupts

2547

* this is called early in boot, before either interrupts

2516

* or the scheduler are operational.

2548

* or the scheduler are operational.

2517

*/

2549

*/

2518

cpu_notifier(rcu_cpu_notify, 0);

2550

cpu_notifier(rcu_cpu_notify, 0);

2519

for_each_online_cpu(cpu)

2551

for_each_online_cpu(cpu)

2520

rcu_cpu_notify(NULL, CPU_UP_PREPARE, (void *)(long)cpu);

2552

rcu_cpu_notify(NULL, CPU_UP_PREPARE, (void *)(long)cpu);

2521

check_cpu_stall_init();

2553

check_cpu_stall_init();

2522

}

2554

}

2523

2555

2524

#include "rcutree_plugin.h"

2556

#include "rcutree_plugin.h"

2525

2557

GITLAB

rcu: Document why rcu_blocking_is_gp() is safe

 /*
  * Read-Copy Update mechanism for mutual exclusion (tree-based version)
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation; either version 2 of the License, or
  * (at your option) any later version.
  *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
  * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  *
  * Copyright IBM Corporation, 2008
  *
  * Author: Dipankar Sarma <dipankar@in.ibm.com>
  *	   Paul E. McKenney <paulmck@linux.vnet.ibm.com> Hierarchical algorithm
  *
  * Based on the original work by Paul McKenney <paulmck@us.ibm.com>
  * and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
  *
  * For detailed explanation of Read-Copy Update mechanism see -
  *	Documentation/RCU
  */
 #ifndef __LINUX_RCUTREE_H
 #define __LINUX_RCUTREE_H
 extern void rcu_init(void);
 extern void rcu_note_context_switch(int cpu);
 extern int rcu_needs_cpu(int cpu);
 extern void rcu_cpu_stall_reset(void);
 /*
  * Note a virtualization-based context switch.  This is simply a
  * wrapper around rcu_note_context_switch(), which allows TINY_RCU
  * to save a few bytes.
  */
 static inline void rcu_virt_note_context_switch(int cpu)
 {
 	rcu_note_context_switch(cpu);
 }
 #ifdef CONFIG_TREE_PREEMPT_RCU
 extern void exit_rcu(void);
 #else /* #ifdef CONFIG_TREE_PREEMPT_RCU */
 static inline void exit_rcu(void)
 {
 }
 #endif /* #else #ifdef CONFIG_TREE_PREEMPT_RCU */
 extern void synchronize_rcu_bh(void);
 extern void synchronize_sched_expedited(void);
 extern void synchronize_rcu_expedited(void);
 void kfree_call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu));
 /**
  * synchronize_rcu_bh_expedited - Brute-force RCU-bh grace period
  *
  * Wait for an RCU-bh grace period to elapse, but use a "big hammer"
  * approach to force the grace period to end quickly.  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_bh_expedited() in a loop, please
  * restructure your code to batch your updates, and then use a single
  * synchronize_rcu_bh() 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.
  */
 static inline void synchronize_rcu_bh_expedited(void)
 {
 	synchronize_sched_expedited();
 }
 extern void rcu_barrier(void);
 extern void rcu_barrier_bh(void);
 extern void rcu_barrier_sched(void);
 extern unsigned long rcutorture_testseq;
 extern unsigned long rcutorture_vernum;
 extern long rcu_batches_completed(void);
 extern long rcu_batches_completed_bh(void);
 extern long rcu_batches_completed_sched(void);
 extern void rcu_force_quiescent_state(void);
 extern void rcu_bh_force_quiescent_state(void);
 extern void rcu_sched_force_quiescent_state(void);
-/* A context switch is a grace period for RCU-sched and RCU-bh. */
-static inline int rcu_blocking_is_gp(void)
-{
-	might_sleep();  /* Check for RCU read-side critical section. */
-	return num_online_cpus() == 1;
-}
 extern void rcu_scheduler_starting(void);
 extern int rcu_scheduler_active __read_mostly;
 #endif /* __LINUX_RCUTREE_H */

 /*
  * Read-Copy Update mechanism for mutual exclusion
  *
  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License as published by
  * the Free Software Foundation; either version 2 of the License, or
  * (at your option) any later version.
  *
  * This program is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  *
  * You should have received a copy of the GNU General Public License
  * along with this program; if not, write to the Free Software
  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
  *
  * Copyright IBM Corporation, 2008
  *
  * Authors: Dipankar Sarma <dipankar@in.ibm.com>
  *	    Manfred Spraul <manfred@colorfullife.com>
  *	    Paul E. McKenney <paulmck@linux.vnet.ibm.com> Hierarchical version
  *
  * Based on the original work by Paul McKenney <paulmck@us.ibm.com>
  * and inputs from Rusty Russell, Andrea Arcangeli and Andi Kleen.
  *
  * For detailed explanation of Read-Copy Update mechanism see -
  *	Documentation/RCU
  */
 #include <linux/types.h>
 #include <linux/kernel.h>
 #include <linux/init.h>
 #include <linux/spinlock.h>
 #include <linux/smp.h>
 #include <linux/rcupdate.h>
 #include <linux/interrupt.h>
 #include <linux/sched.h>
 #include <linux/nmi.h>
 #include <linux/atomic.h>
 #include <linux/bitops.h>
 #include <linux/export.h>
 #include <linux/completion.h>
 #include <linux/moduleparam.h>
 #include <linux/percpu.h>
 #include <linux/notifier.h>
 #include <linux/cpu.h>
 #include <linux/mutex.h>
 #include <linux/time.h>
 #include <linux/kernel_stat.h>
 #include <linux/wait.h>
 #include <linux/kthread.h>
 #include <linux/prefetch.h>
 #include <linux/delay.h>
 #include <linux/stop_machine.h>
 #include "rcutree.h"
 #include <trace/events/rcu.h>
 #include "rcu.h"
 /* Data structures. */
 static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];
 #define RCU_STATE_INITIALIZER(structname) { \
 	.level = { &structname##_state.node[0] }, \
 	.levelcnt = { \
 		NUM_RCU_LVL_0,  /* root of hierarchy. */ \
 		NUM_RCU_LVL_1, \
 		NUM_RCU_LVL_2, \
 		NUM_RCU_LVL_3, \
 		NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \
 	}, \
 	.fqs_state = RCU_GP_IDLE, \
 	.gpnum = -300, \
 	.completed = -300, \
 	.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.onofflock), \
 	.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&structname##_state.fqslock), \
 	.n_force_qs = 0, \
 	.n_force_qs_ngp = 0, \
 	.name = #structname, \
 }
 struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched);
 DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
 struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh);
 DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
 static struct rcu_state *rcu_state;
 /*
  * The rcu_scheduler_active variable transitions from zero to one just
  * before the first task is spawned.  So when this variable is zero, RCU
  * can assume that there is but one task, allowing RCU to (for example)
  * optimized synchronize_sched() to a simple barrier().  When this variable
  * is one, RCU must actually do all the hard work required to detect real
  * grace periods.  This variable is also used to suppress boot-time false
  * positives from lockdep-RCU error checking.
  */
 int rcu_scheduler_active __read_mostly;
 EXPORT_SYMBOL_GPL(rcu_scheduler_active);
 /*
  * The rcu_scheduler_fully_active variable transitions from zero to one
  * during the early_initcall() processing, which is after the scheduler
  * is capable of creating new tasks.  So RCU processing (for example,
  * creating tasks for RCU priority boosting) must be delayed until after
  * rcu_scheduler_fully_active transitions from zero to one.  We also
  * currently delay invocation of any RCU callbacks until after this point.
  *
  * It might later prove better for people registering RCU callbacks during
  * early boot to take responsibility for these callbacks, but one step at
  * a time.
  */
 static int rcu_scheduler_fully_active __read_mostly;
 #ifdef CONFIG_RCU_BOOST
 /*
  * Control variables for per-CPU and per-rcu_node kthreads.  These
  * handle all flavors of RCU.
  */
 static DEFINE_PER_CPU(struct task_struct *, rcu_cpu_kthread_task);
 DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_status);
 DEFINE_PER_CPU(int, rcu_cpu_kthread_cpu);
 DEFINE_PER_CPU(unsigned int, rcu_cpu_kthread_loops);
 DEFINE_PER_CPU(char, rcu_cpu_has_work);
 #endif /* #ifdef CONFIG_RCU_BOOST */
 static void rcu_node_kthread_setaffinity(struct rcu_node *rnp, int outgoingcpu);
 static void invoke_rcu_core(void);
 static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp);
 /*
  * Track the rcutorture test sequence number and the update version
  * number within a given test.  The rcutorture_testseq is incremented
  * on every rcutorture module load and unload, so has an odd value
  * when a test is running.  The rcutorture_vernum is set to zero
  * when rcutorture starts and is incremented on each rcutorture update.
  * These variables enable correlating rcutorture output with the
  * RCU tracing information.
  */
 unsigned long rcutorture_testseq;
 unsigned long rcutorture_vernum;
 /*
  * Return true if an RCU grace period is in progress.  The ACCESS_ONCE()s
  * permit this function to be invoked without holding the root rcu_node
  * structure's ->lock, but of course results can be subject to change.
  */
 static int rcu_gp_in_progress(struct rcu_state *rsp)
 {
 	return ACCESS_ONCE(rsp->completed) != ACCESS_ONCE(rsp->gpnum);
 }
 /*
  * Note a quiescent state.  Because we do not need to know
  * how many quiescent states passed, just if there was at least
  * one since the start of the grace period, this just sets a flag.
  * The caller must have disabled preemption.
  */
 void rcu_sched_qs(int cpu)
 {
 	struct rcu_data *rdp = &per_cpu(rcu_sched_data, cpu);
 	rdp->passed_quiesce_gpnum = rdp->gpnum;
 	barrier();
 	if (rdp->passed_quiesce == 0)
 		trace_rcu_grace_period("rcu_sched", rdp->gpnum, "cpuqs");
 	rdp->passed_quiesce = 1;
 }
 void rcu_bh_qs(int cpu)
 {
 	struct rcu_data *rdp = &per_cpu(rcu_bh_data, cpu);
 	rdp->passed_quiesce_gpnum = rdp->gpnum;
 	barrier();
 	if (rdp->passed_quiesce == 0)
 		trace_rcu_grace_period("rcu_bh", rdp->gpnum, "cpuqs");
 	rdp->passed_quiesce = 1;
 }
 /*
  * Note a context switch.  This is a quiescent state for RCU-sched,
  * and requires special handling for preemptible RCU.
  * The caller must have disabled preemption.
  */
 void rcu_note_context_switch(int cpu)
 {
 	trace_rcu_utilization("Start context switch");
 	rcu_sched_qs(cpu);
 	rcu_preempt_note_context_switch(cpu);
 	trace_rcu_utilization("End context switch");
 }
 EXPORT_SYMBOL_GPL(rcu_note_context_switch);
 DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 	.dynticks_nesting = DYNTICK_TASK_EXIT_IDLE,
 	.dynticks = ATOMIC_INIT(1),
 };
 static int blimit = 10;		/* Maximum callbacks per rcu_do_batch. */
 static int qhimark = 10000;	/* If this many pending, ignore blimit. */
 static int qlowmark = 100;	/* Once only this many pending, use blimit. */
 module_param(blimit, int, 0);
 module_param(qhimark, int, 0);
 module_param(qlowmark, int, 0);
 int rcu_cpu_stall_suppress __read_mostly; /* 1 = suppress stall warnings. */
 int rcu_cpu_stall_timeout __read_mostly = CONFIG_RCU_CPU_STALL_TIMEOUT;
 module_param(rcu_cpu_stall_suppress, int, 0644);
 module_param(rcu_cpu_stall_timeout, int, 0644);
 static void force_quiescent_state(struct rcu_state *rsp, int relaxed);
 static int rcu_pending(int cpu);
 /*
  * Return the number of RCU-sched batches processed thus far for debug & stats.
  */
 long rcu_batches_completed_sched(void)
 {
 	return rcu_sched_state.completed;
 }
 EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);
 /*
  * Return the number of RCU BH batches processed thus far for debug & stats.
  */
 long rcu_batches_completed_bh(void)
 {
 	return rcu_bh_state.completed;
 }
 EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);
 /*
  * Force a quiescent state for RCU BH.
  */
 void rcu_bh_force_quiescent_state(void)
 {
 	force_quiescent_state(&rcu_bh_state, 0);
 }
 EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);
 /*
  * Record the number of times rcutorture tests have been initiated and
  * terminated.  This information allows the debugfs tracing stats to be
  * correlated to the rcutorture messages, even when the rcutorture module
  * is being repeatedly loaded and unloaded.  In other words, we cannot
  * store this state in rcutorture itself.
  */
 void rcutorture_record_test_transition(void)
 {
 	rcutorture_testseq++;
 	rcutorture_vernum = 0;
 }
 EXPORT_SYMBOL_GPL(rcutorture_record_test_transition);
 /*
  * Record the number of writer passes through the current rcutorture test.
  * This is also used to correlate debugfs tracing stats with the rcutorture
  * messages.
  */
 void rcutorture_record_progress(unsigned long vernum)
 {
 	rcutorture_vernum++;
 }
 EXPORT_SYMBOL_GPL(rcutorture_record_progress);
 /*
  * Force a quiescent state for RCU-sched.
  */
 void rcu_sched_force_quiescent_state(void)
 {
 	force_quiescent_state(&rcu_sched_state, 0);
 }
 EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);
 /*
  * Does the CPU have callbacks ready to be invoked?
  */
 static int
 cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)
 {
 	return &rdp->nxtlist != rdp->nxttail[RCU_DONE_TAIL];
 }
 /*
  * Does the current CPU require a yet-as-unscheduled grace period?
  */
 static int
 cpu_needs_another_gp(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	return *rdp->nxttail[RCU_DONE_TAIL] && !rcu_gp_in_progress(rsp);
 }
 /*
  * Return the root node of the specified rcu_state structure.
  */
 static struct rcu_node *rcu_get_root(struct rcu_state *rsp)
 {
 	return &rsp->node[0];
 }
 /*
  * If the specified CPU is offline, tell the caller that it is in
  * a quiescent state.  Otherwise, whack it with a reschedule IPI.
  * Grace periods can end up waiting on an offline CPU when that
  * CPU is in the process of coming online -- it will be added to the
  * rcu_node bitmasks before it actually makes it online.  The same thing
  * can happen while a CPU is in the process of coming online.  Because this
  * race is quite rare, we check for it after detecting that the grace
  * period has been delayed rather than checking each and every CPU
  * each and every time we start a new grace period.
  */
 static int rcu_implicit_offline_qs(struct rcu_data *rdp)
 {
 	/*
 	 * If the CPU is offline for more than a jiffy, it is in a quiescent
 	 * state.  We can trust its state not to change because interrupts
 	 * are disabled.  The reason for the jiffy's worth of slack is to
 	 * handle CPUs initializing on the way up and finding their way
 	 * to the idle loop on the way down.
 	 */
 	if (cpu_is_offline(rdp->cpu) &&
 	    ULONG_CMP_LT(rdp->rsp->gp_start + 2, jiffies)) {
 		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "ofl");
 		rdp->offline_fqs++;
 		return 1;
 	}
 	return 0;
 }
 /*
  * rcu_idle_enter_common - inform RCU that current CPU is moving towards idle
  *
  * If the new value of the ->dynticks_nesting counter now is zero,
  * we really have entered idle, and must do the appropriate accounting.
  * The caller must have disabled interrupts.
  */
 static void rcu_idle_enter_common(struct rcu_dynticks *rdtp, long long oldval)
 {
 	trace_rcu_dyntick("Start", oldval, 0);
 	if (!is_idle_task(current)) {
 		struct task_struct *idle = idle_task(smp_processor_id());
 		trace_rcu_dyntick("Error on entry: not idle task", oldval, 0);
 		ftrace_dump(DUMP_ALL);
 		WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",
 			  current->pid, current->comm,
 			  idle->pid, idle->comm); /* must be idle task! */
 	}
 	rcu_prepare_for_idle(smp_processor_id());
 	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
 	smp_mb__before_atomic_inc();  /* See above. */
 	atomic_inc(&rdtp->dynticks);
 	smp_mb__after_atomic_inc();  /* Force ordering with next sojourn. */
 	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
 	/*
 	 * The idle task is not permitted to enter the idle loop while
 	 * in an RCU read-side critical section.
 	 */
 	rcu_lockdep_assert(!lock_is_held(&rcu_lock_map),
 			   "Illegal idle entry in RCU read-side critical section.");
 	rcu_lockdep_assert(!lock_is_held(&rcu_bh_lock_map),
 			   "Illegal idle entry in RCU-bh read-side critical section.");
 	rcu_lockdep_assert(!lock_is_held(&rcu_sched_lock_map),
 			   "Illegal idle entry in RCU-sched read-side critical section.");
 }
 /**
  * rcu_idle_enter - inform RCU that current CPU is entering idle
  *
  * Enter idle mode, in other words, -leave- the mode in which RCU
  * read-side critical sections can occur.  (Though RCU read-side
  * critical sections can occur in irq handlers in idle, a possibility
  * handled by irq_enter() and irq_exit().)
  *
  * We crowbar the ->dynticks_nesting field to zero to allow for
  * the possibility of usermode upcalls having messed up our count
  * of interrupt nesting level during the prior busy period.
  */
 void rcu_idle_enter(void)
 {
 	unsigned long flags;
 	long long oldval;
 	struct rcu_dynticks *rdtp;
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	oldval = rdtp->dynticks_nesting;
 	WARN_ON_ONCE((oldval & DYNTICK_TASK_NEST_MASK) == 0);
 	if ((oldval & DYNTICK_TASK_NEST_MASK) == DYNTICK_TASK_NEST_VALUE)
 		rdtp->dynticks_nesting = 0;
 	else
 		rdtp->dynticks_nesting -= DYNTICK_TASK_NEST_VALUE;
 	rcu_idle_enter_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
 EXPORT_SYMBOL_GPL(rcu_idle_enter);
 /**
  * rcu_irq_exit - inform RCU that current CPU is exiting irq towards idle
  *
  * Exit from an interrupt handler, which might possibly result in entering
  * idle mode, in other words, leaving the mode in which read-side critical
  * sections can occur.
  *
  * This code assumes that the idle loop never does anything that might
  * result in unbalanced calls to irq_enter() and irq_exit().  If your
  * architecture violates this assumption, RCU will give you what you
  * deserve, good and hard.  But very infrequently and irreproducibly.
  *
  * Use things like work queues to work around this limitation.
  *
  * You have been warned.
  */
 void rcu_irq_exit(void)
 {
 	unsigned long flags;
 	long long oldval;
 	struct rcu_dynticks *rdtp;
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	oldval = rdtp->dynticks_nesting;
 	rdtp->dynticks_nesting--;
 	WARN_ON_ONCE(rdtp->dynticks_nesting < 0);
 	if (rdtp->dynticks_nesting)
 		trace_rcu_dyntick("--=", oldval, rdtp->dynticks_nesting);
 	else
 		rcu_idle_enter_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
 /*
  * rcu_idle_exit_common - inform RCU that current CPU is moving away from idle
  *
  * If the new value of the ->dynticks_nesting counter was previously zero,
  * we really have exited idle, and must do the appropriate accounting.
  * The caller must have disabled interrupts.
  */
 static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)
 {
 	smp_mb__before_atomic_inc();  /* Force ordering w/previous sojourn. */
 	atomic_inc(&rdtp->dynticks);
 	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
 	smp_mb__after_atomic_inc();  /* See above. */
 	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
 	rcu_cleanup_after_idle(smp_processor_id());
 	trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);
 	if (!is_idle_task(current)) {
 		struct task_struct *idle = idle_task(smp_processor_id());
 		trace_rcu_dyntick("Error on exit: not idle task",
 				  oldval, rdtp->dynticks_nesting);
 		ftrace_dump(DUMP_ALL);
 		WARN_ONCE(1, "Current pid: %d comm: %s / Idle pid: %d comm: %s",
 			  current->pid, current->comm,
 			  idle->pid, idle->comm); /* must be idle task! */
 	}
 }
 /**
  * rcu_idle_exit - inform RCU that current CPU is leaving idle
  *
  * Exit idle mode, in other words, -enter- the mode in which RCU
  * read-side critical sections can occur.
  *
  * We crowbar the ->dynticks_nesting field to DYNTICK_TASK_NEST to
  * allow for the possibility of usermode upcalls messing up our count
  * of interrupt nesting level during the busy period that is just
  * now starting.
  */
 void rcu_idle_exit(void)
 {
 	unsigned long flags;
 	struct rcu_dynticks *rdtp;
 	long long oldval;
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	oldval = rdtp->dynticks_nesting;
 	WARN_ON_ONCE(oldval < 0);
 	if (oldval & DYNTICK_TASK_NEST_MASK)
 		rdtp->dynticks_nesting += DYNTICK_TASK_NEST_VALUE;
 	else
 		rdtp->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
 	rcu_idle_exit_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
 EXPORT_SYMBOL_GPL(rcu_idle_exit);
 /**
  * rcu_irq_enter - inform RCU that current CPU is entering irq away from idle
  *
  * Enter an interrupt handler, which might possibly result in exiting
  * idle mode, in other words, entering the mode in which read-side critical
  * sections can occur.
  *
  * Note that the Linux kernel is fully capable of entering an interrupt
  * handler that it never exits, for example when doing upcalls to
  * user mode!  This code assumes that the idle loop never does upcalls to
  * user mode.  If your architecture does do upcalls from the idle loop (or
  * does anything else that results in unbalanced calls to the irq_enter()
  * and irq_exit() functions), RCU will give you what you deserve, good
  * and hard.  But very infrequently and irreproducibly.
  *
  * Use things like work queues to work around this limitation.
  *
  * You have been warned.
  */
 void rcu_irq_enter(void)
 {
 	unsigned long flags;
 	struct rcu_dynticks *rdtp;
 	long long oldval;
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	oldval = rdtp->dynticks_nesting;
 	rdtp->dynticks_nesting++;
 	WARN_ON_ONCE(rdtp->dynticks_nesting == 0);
 	if (oldval)
 		trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);
 	else
 		rcu_idle_exit_common(rdtp, oldval);
 	local_irq_restore(flags);
 }
 /**
  * rcu_nmi_enter - inform RCU of entry to NMI context
  *
  * If the CPU was idle with dynamic ticks active, and there is no
  * irq handler running, this updates rdtp->dynticks_nmi to let the
  * RCU grace-period handling know that the CPU is active.
  */
 void rcu_nmi_enter(void)
 {
 	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);
 	if (rdtp->dynticks_nmi_nesting == 0 &&
 	    (atomic_read(&rdtp->dynticks) & 0x1))
 		return;
 	rdtp->dynticks_nmi_nesting++;
 	smp_mb__before_atomic_inc();  /* Force delay from prior write. */
 	atomic_inc(&rdtp->dynticks);
 	/* CPUs seeing atomic_inc() must see later RCU read-side crit sects */
 	smp_mb__after_atomic_inc();  /* See above. */
 	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
 }
 /**
  * rcu_nmi_exit - inform RCU of exit from NMI context
  *
  * If the CPU was idle with dynamic ticks active, and there is no
  * irq handler running, this updates rdtp->dynticks_nmi to let the
  * RCU grace-period handling know that the CPU is no longer active.
  */
 void rcu_nmi_exit(void)
 {
 	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);
 	if (rdtp->dynticks_nmi_nesting == 0 ||
 	    --rdtp->dynticks_nmi_nesting != 0)
 		return;
 	/* CPUs seeing atomic_inc() must see prior RCU read-side crit sects */
 	smp_mb__before_atomic_inc();  /* See above. */
 	atomic_inc(&rdtp->dynticks);
 	smp_mb__after_atomic_inc();  /* Force delay to next write. */
 	WARN_ON_ONCE(atomic_read(&rdtp->dynticks) & 0x1);
 }
 #ifdef CONFIG_PROVE_RCU
 /**
  * rcu_is_cpu_idle - see if RCU thinks that the current CPU is idle
  *
  * If the current CPU is in its idle loop and is neither in an interrupt
  * or NMI handler, return true.
  */
 int rcu_is_cpu_idle(void)
 {
 	int ret;
 	preempt_disable();
 	ret = (atomic_read(&__get_cpu_var(rcu_dynticks).dynticks) & 0x1) == 0;
 	preempt_enable();
 	return ret;
 }
 EXPORT_SYMBOL(rcu_is_cpu_idle);
 #ifdef CONFIG_HOTPLUG_CPU
 /*
  * Is the current CPU online?  Disable preemption to avoid false positives
  * that could otherwise happen due to the current CPU number being sampled,
  * this task being preempted, its old CPU being taken offline, resuming
  * on some other CPU, then determining that its old CPU is now offline.
  * It is OK to use RCU on an offline processor during initial boot, hence
  * the check for rcu_scheduler_fully_active.  Note also that it is OK
  * for a CPU coming online to use RCU for one jiffy prior to marking itself
  * online in the cpu_online_mask.  Similarly, it is OK for a CPU going
  * offline to continue to use RCU for one jiffy after marking itself
  * offline in the cpu_online_mask.  This leniency is necessary given the
  * non-atomic nature of the online and offline processing, for example,
  * the fact that a CPU enters the scheduler after completing the CPU_DYING
  * notifiers.
  *
  * This is also why RCU internally marks CPUs online during the
  * CPU_UP_PREPARE phase and offline during the CPU_DEAD phase.
  *
  * Disable checking if in an NMI handler because we cannot safely report
  * errors from NMI handlers anyway.
  */
 bool rcu_lockdep_current_cpu_online(void)
 {
 	struct rcu_data *rdp;
 	struct rcu_node *rnp;
 	bool ret;
 	if (in_nmi())
 		return 1;
 	preempt_disable();
 	rdp = &__get_cpu_var(rcu_sched_data);
 	rnp = rdp->mynode;
 	ret = (rdp->grpmask & rnp->qsmaskinit) ||
 	      !rcu_scheduler_fully_active;
 	preempt_enable();
 	return ret;
 }
 EXPORT_SYMBOL_GPL(rcu_lockdep_current_cpu_online);
 #endif /* #ifdef CONFIG_HOTPLUG_CPU */
 #endif /* #ifdef CONFIG_PROVE_RCU */
 /**
  * rcu_is_cpu_rrupt_from_idle - see if idle or immediately interrupted from idle
  *
  * If the current CPU is idle or running at a first-level (not nested)
  * interrupt from idle, return true.  The caller must have at least
  * disabled preemption.
  */
 int rcu_is_cpu_rrupt_from_idle(void)
 {
 	return __get_cpu_var(rcu_dynticks).dynticks_nesting <= 1;
 }
 /*
  * Snapshot the specified CPU's dynticks counter so that we can later
  * credit them with an implicit quiescent state.  Return 1 if this CPU
  * is in dynticks idle mode, which is an extended quiescent state.
  */
 static int dyntick_save_progress_counter(struct rcu_data *rdp)
 {
 	rdp->dynticks_snap = atomic_add_return(0, &rdp->dynticks->dynticks);
 	return (rdp->dynticks_snap & 0x1) == 0;
 }
 /*
  * Return true if the specified CPU has passed through a quiescent
  * state by virtue of being in or having passed through an dynticks
  * idle state since the last call to dyntick_save_progress_counter()
  * for this same CPU.
  */
 static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
 {
 	unsigned int curr;
 	unsigned int snap;
 	curr = (unsigned int)atomic_add_return(0, &rdp->dynticks->dynticks);
 	snap = (unsigned int)rdp->dynticks_snap;
 	/*
 	 * If the CPU passed through or entered a dynticks idle phase with
 	 * no active irq/NMI handlers, then we can safely pretend that the CPU
 	 * already acknowledged the request to pass through a quiescent
 	 * state.  Either way, that CPU cannot possibly be in an RCU
 	 * read-side critical section that started before the beginning
 	 * of the current RCU grace period.
 	 */
 	if ((curr & 0x1) == 0 || UINT_CMP_GE(curr, snap + 2)) {
 		trace_rcu_fqs(rdp->rsp->name, rdp->gpnum, rdp->cpu, "dti");
 		rdp->dynticks_fqs++;
 		return 1;
 	}
 	/* Go check for the CPU being offline. */
 	return rcu_implicit_offline_qs(rdp);
 }
 static int jiffies_till_stall_check(void)
 {
 	int till_stall_check = ACCESS_ONCE(rcu_cpu_stall_timeout);
 	/*
 	 * Limit check must be consistent with the Kconfig limits
 	 * for CONFIG_RCU_CPU_STALL_TIMEOUT.
 	 */
 	if (till_stall_check < 3) {
 		ACCESS_ONCE(rcu_cpu_stall_timeout) = 3;
 		till_stall_check = 3;
 	} else if (till_stall_check > 300) {
 		ACCESS_ONCE(rcu_cpu_stall_timeout) = 300;
 		till_stall_check = 300;
 	}
 	return till_stall_check * HZ + RCU_STALL_DELAY_DELTA;
 }
 static void record_gp_stall_check_time(struct rcu_state *rsp)
 {
 	rsp->gp_start = jiffies;
 	rsp->jiffies_stall = jiffies + jiffies_till_stall_check();
 }
 static void print_other_cpu_stall(struct rcu_state *rsp)
 {
 	int cpu;
 	long delta;
 	unsigned long flags;
 	int ndetected;
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	/* Only let one CPU complain about others per time interval. */
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	delta = jiffies - rsp->jiffies_stall;
 	if (delta < RCU_STALL_RAT_DELAY || !rcu_gp_in_progress(rsp)) {
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		return;
 	}
 	rsp->jiffies_stall = jiffies + 3 * jiffies_till_stall_check() + 3;
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	/*
 	 * OK, time to rat on our buddy...
 	 * See Documentation/RCU/stallwarn.txt for info on how to debug
 	 * RCU CPU stall warnings.
 	 */
 	printk(KERN_ERR "INFO: %s detected stalls on CPUs/tasks:",
 	       rsp->name);
 	print_cpu_stall_info_begin();
 	rcu_for_each_leaf_node(rsp, rnp) {
 		raw_spin_lock_irqsave(&rnp->lock, flags);
 		ndetected += rcu_print_task_stall(rnp);
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		if (rnp->qsmask == 0)
 			continue;
 		for (cpu = 0; cpu <= rnp->grphi - rnp->grplo; cpu++)
 			if (rnp->qsmask & (1UL << cpu)) {
 				print_cpu_stall_info(rsp, rnp->grplo + cpu);
 				ndetected++;
 			}
 	}
 	/*
 	 * Now rat on any tasks that got kicked up to the root rcu_node
 	 * due to CPU offlining.
 	 */
 	rnp = rcu_get_root(rsp);
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	ndetected = rcu_print_task_stall(rnp);
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	print_cpu_stall_info_end();
 	printk(KERN_CONT "(detected by %d, t=%ld jiffies)\n",
 	       smp_processor_id(), (long)(jiffies - rsp->gp_start));
 	if (ndetected == 0)
 		printk(KERN_ERR "INFO: Stall ended before state dump start\n");
 	else if (!trigger_all_cpu_backtrace())
 		dump_stack();
 	/* If so configured, complain about tasks blocking the grace period. */
 	rcu_print_detail_task_stall(rsp);
 	force_quiescent_state(rsp, 0);  /* Kick them all. */
 }
 static void print_cpu_stall(struct rcu_state *rsp)
 {
 	unsigned long flags;
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	/*
 	 * OK, time to rat on ourselves...
 	 * See Documentation/RCU/stallwarn.txt for info on how to debug
 	 * RCU CPU stall warnings.
 	 */
 	printk(KERN_ERR "INFO: %s self-detected stall on CPU", rsp->name);
 	print_cpu_stall_info_begin();
 	print_cpu_stall_info(rsp, smp_processor_id());
 	print_cpu_stall_info_end();
 	printk(KERN_CONT " (t=%lu jiffies)\n", jiffies - rsp->gp_start);
 	if (!trigger_all_cpu_backtrace())
 		dump_stack();
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))
 		rsp->jiffies_stall = jiffies +
 				     3 * jiffies_till_stall_check() + 3;
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	set_need_resched();  /* kick ourselves to get things going. */
 }
 static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long j;
 	unsigned long js;
 	struct rcu_node *rnp;
 	if (rcu_cpu_stall_suppress)
 		return;
 	j = ACCESS_ONCE(jiffies);
 	js = ACCESS_ONCE(rsp->jiffies_stall);
 	rnp = rdp->mynode;
 	if ((ACCESS_ONCE(rnp->qsmask) & rdp->grpmask) && ULONG_CMP_GE(j, js)) {
 		/* We haven't checked in, so go dump stack. */
 		print_cpu_stall(rsp);
 	} else if (rcu_gp_in_progress(rsp) &&
 		   ULONG_CMP_GE(j, js + RCU_STALL_RAT_DELAY)) {
 		/* They had a few time units to dump stack, so complain. */
 		print_other_cpu_stall(rsp);
 	}
 }
 static int rcu_panic(struct notifier_block *this, unsigned long ev, void *ptr)
 {
 	rcu_cpu_stall_suppress = 1;
 	return NOTIFY_DONE;
 }
 /**
  * rcu_cpu_stall_reset - prevent further stall warnings in current grace period
  *
  * Set the stall-warning timeout way off into the future, thus preventing
  * any RCU CPU stall-warning messages from appearing in the current set of
  * RCU grace periods.
  *
  * The caller must disable hard irqs.
  */
 void rcu_cpu_stall_reset(void)
 {
 	rcu_sched_state.jiffies_stall = jiffies + ULONG_MAX / 2;
 	rcu_bh_state.jiffies_stall = jiffies + ULONG_MAX / 2;
 	rcu_preempt_stall_reset();
 }
 static struct notifier_block rcu_panic_block = {
 	.notifier_call = rcu_panic,
 };
 static void __init check_cpu_stall_init(void)
 {
 	atomic_notifier_chain_register(&panic_notifier_list, &rcu_panic_block);
 }
 /*
  * Update CPU-local rcu_data state to record the newly noticed grace period.
  * This is used both when we started the grace period and when we notice
  * that someone else started the grace period.  The caller must hold the
  * ->lock of the leaf rcu_node structure corresponding to the current CPU,
  *  and must have irqs disabled.
  */
 static void __note_new_gpnum(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
 {
 	if (rdp->gpnum != rnp->gpnum) {
 		/*
 		 * If the current grace period is waiting for this CPU,
 		 * set up to detect a quiescent state, otherwise don't
 		 * go looking for one.
 		 */
 		rdp->gpnum = rnp->gpnum;
 		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpustart");
 		if (rnp->qsmask & rdp->grpmask) {
 			rdp->qs_pending = 1;
 			rdp->passed_quiesce = 0;
 		} else
 			rdp->qs_pending = 0;
 		zero_cpu_stall_ticks(rdp);
 	}
 }
 static void note_new_gpnum(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long flags;
 	struct rcu_node *rnp;
 	local_irq_save(flags);
 	rnp = rdp->mynode;
 	if (rdp->gpnum == ACCESS_ONCE(rnp->gpnum) || /* outside lock. */
 	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
 		local_irq_restore(flags);
 		return;
 	}
 	__note_new_gpnum(rsp, rnp, rdp);
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 }
 /*
  * Did someone else start a new RCU grace period start since we last
  * checked?  Update local state appropriately if so.  Must be called
  * on the CPU corresponding to rdp.
  */
 static int
 check_for_new_grace_period(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long flags;
 	int ret = 0;
 	local_irq_save(flags);
 	if (rdp->gpnum != rsp->gpnum) {
 		note_new_gpnum(rsp, rdp);
 		ret = 1;
 	}
 	local_irq_restore(flags);
 	return ret;
 }
 /*
  * Advance this CPU's callbacks, but only if the current grace period
  * has ended.  This may be called only from the CPU to whom the rdp
  * belongs.  In addition, the corresponding leaf rcu_node structure's
  * ->lock must be held by the caller, with irqs disabled.
  */
 static void
 __rcu_process_gp_end(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
 {
 	/* Did another grace period end? */
 	if (rdp->completed != rnp->completed) {
 		/* Advance callbacks.  No harm if list empty. */
 		rdp->nxttail[RCU_DONE_TAIL] = rdp->nxttail[RCU_WAIT_TAIL];
 		rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_READY_TAIL];
 		rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
 		/* Remember that we saw this grace-period completion. */
 		rdp->completed = rnp->completed;
 		trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuend");
 		/*
 		 * If we were in an extended quiescent state, we may have
 		 * missed some grace periods that others CPUs handled on
 		 * our behalf. Catch up with this state to avoid noting
 		 * spurious new grace periods.  If another grace period
 		 * has started, then rnp->gpnum will have advanced, so
 		 * we will detect this later on.
 		 */
 		if (ULONG_CMP_LT(rdp->gpnum, rdp->completed))
 			rdp->gpnum = rdp->completed;
 		/*
 		 * If RCU does not need a quiescent state from this CPU,
 		 * then make sure that this CPU doesn't go looking for one.
 		 */
 		if ((rnp->qsmask & rdp->grpmask) == 0)
 			rdp->qs_pending = 0;
 	}
 }
 /*
  * Advance this CPU's callbacks, but only if the current grace period
  * has ended.  This may be called only from the CPU to whom the rdp
  * belongs.
  */
 static void
 rcu_process_gp_end(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long flags;
 	struct rcu_node *rnp;
 	local_irq_save(flags);
 	rnp = rdp->mynode;
 	if (rdp->completed == ACCESS_ONCE(rnp->completed) || /* outside lock. */
 	    !raw_spin_trylock(&rnp->lock)) { /* irqs already off, so later. */
 		local_irq_restore(flags);
 		return;
 	}
 	__rcu_process_gp_end(rsp, rnp, rdp);
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 }
 /*
  * Do per-CPU grace-period initialization for running CPU.  The caller
  * must hold the lock of the leaf rcu_node structure corresponding to
  * this CPU.
  */
 static void
 rcu_start_gp_per_cpu(struct rcu_state *rsp, struct rcu_node *rnp, struct rcu_data *rdp)
 {
 	/* Prior grace period ended, so advance callbacks for current CPU. */
 	__rcu_process_gp_end(rsp, rnp, rdp);
 	/*
 	 * Because this CPU just now started the new grace period, we know
 	 * that all of its callbacks will be covered by this upcoming grace
 	 * period, even the ones that were registered arbitrarily recently.
 	 * Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.
 	 *
 	 * Other CPUs cannot be sure exactly when the grace period started.
 	 * Therefore, their recently registered callbacks must pass through
 	 * an additional RCU_NEXT_READY stage, so that they will be handled
 	 * by the next RCU grace period.
 	 */
 	rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
 	rdp->nxttail[RCU_WAIT_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
 	/* Set state so that this CPU will detect the next quiescent state. */
 	__note_new_gpnum(rsp, rnp, rdp);
 }
 /*
  * Start a new RCU grace period if warranted, re-initializing the hierarchy
  * in preparation for detecting the next grace period.  The caller must hold
  * the root node's ->lock, which is released before return.  Hard irqs must
  * be disabled.
  *
  * Note that it is legal for a dying CPU (which is marked as offline) to
  * invoke this function.  This can happen when the dying CPU reports its
  * quiescent state.
  */
 static void
 rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
 	__releases(rcu_get_root(rsp)->lock)
 {
 	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	if (!rcu_scheduler_fully_active ||
 	    !cpu_needs_another_gp(rsp, rdp)) {
 		/*
 		 * Either the scheduler hasn't yet spawned the first
 		 * non-idle task or this CPU does not need another
 		 * grace period.  Either way, don't start a new grace
 		 * period.
 		 */
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		return;
 	}
 	if (rsp->fqs_active) {
 		/*
 		 * This CPU needs a grace period, but force_quiescent_state()
 		 * is running.  Tell it to start one on this CPU's behalf.
 		 */
 		rsp->fqs_need_gp = 1;
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		return;
 	}
 	/* Advance to a new grace period and initialize state. */
 	rsp->gpnum++;
 	trace_rcu_grace_period(rsp->name, rsp->gpnum, "start");
 	WARN_ON_ONCE(rsp->fqs_state == RCU_GP_INIT);
 	rsp->fqs_state = RCU_GP_INIT; /* Hold off force_quiescent_state. */
 	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
 	record_gp_stall_check_time(rsp);
 	raw_spin_unlock(&rnp->lock);  /* leave irqs disabled. */
 	/* Exclude any concurrent CPU-hotplug operations. */
 	raw_spin_lock(&rsp->onofflock);  /* irqs already disabled. */
 	/*
 	 * Set the quiescent-state-needed bits in all the rcu_node
 	 * structures for all currently online CPUs in breadth-first
 	 * order, starting from the root rcu_node structure.  This
 	 * operation relies on the layout of the hierarchy within the
 	 * rsp->node[] array.  Note that other CPUs will access only
 	 * the leaves of the hierarchy, which still indicate that no
 	 * grace period is in progress, at least until the corresponding
 	 * leaf node has been initialized.  In addition, we have excluded
 	 * CPU-hotplug operations.
 	 *
 	 * Note that the grace period cannot complete until we finish
 	 * the initialization process, as there will be at least one
 	 * qsmask bit set in the root node until that time, namely the
 	 * one corresponding to this CPU, due to the fact that we have
 	 * irqs disabled.
 	 */
 	rcu_for_each_node_breadth_first(rsp, rnp) {
 		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
 		rcu_preempt_check_blocked_tasks(rnp);
 		rnp->qsmask = rnp->qsmaskinit;
 		rnp->gpnum = rsp->gpnum;
 		rnp->completed = rsp->completed;
 		if (rnp == rdp->mynode)
 			rcu_start_gp_per_cpu(rsp, rnp, rdp);
 		rcu_preempt_boost_start_gp(rnp);
 		trace_rcu_grace_period_init(rsp->name, rnp->gpnum,
 					    rnp->level, rnp->grplo,
 					    rnp->grphi, rnp->qsmask);
 		raw_spin_unlock(&rnp->lock);	/* irqs remain disabled. */
 	}
 	rnp = rcu_get_root(rsp);
 	raw_spin_lock(&rnp->lock);		/* irqs already disabled. */
 	rsp->fqs_state = RCU_SIGNAL_INIT; /* force_quiescent_state now OK. */
 	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
 	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
 }
 /*
  * Report a full set of quiescent states to the specified rcu_state
  * data structure.  This involves cleaning up after the prior grace
  * period and letting rcu_start_gp() start up the next grace period
  * if one is needed.  Note that the caller must hold rnp->lock, as
  * required by rcu_start_gp(), which will release it.
  */
 static void rcu_report_qs_rsp(struct rcu_state *rsp, unsigned long flags)
 	__releases(rcu_get_root(rsp)->lock)
 {
 	unsigned long gp_duration;
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
 	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
 	/*
 	 * Ensure that all grace-period and pre-grace-period activity
 	 * is seen before the assignment to rsp->completed.
 	 */
 	smp_mb(); /* See above block comment. */
 	gp_duration = jiffies - rsp->gp_start;
 	if (gp_duration > rsp->gp_max)
 		rsp->gp_max = gp_duration;
 	/*
 	 * We know the grace period is complete, but to everyone else
 	 * it appears to still be ongoing.  But it is also the case
 	 * that to everyone else it looks like there is nothing that
 	 * they can do to advance the grace period.  It is therefore
 	 * safe for us to drop the lock in order to mark the grace
 	 * period as completed in all of the rcu_node structures.
 	 *
 	 * But if this CPU needs another grace period, it will take
 	 * care of this while initializing the next grace period.
 	 * We use RCU_WAIT_TAIL instead of the usual RCU_DONE_TAIL
 	 * because the callbacks have not yet been advanced: Those
 	 * callbacks are waiting on the grace period that just now
 	 * completed.
 	 */
 	if (*rdp->nxttail[RCU_WAIT_TAIL] == NULL) {
 		raw_spin_unlock(&rnp->lock);	 /* irqs remain disabled. */
 		/*
 		 * Propagate new ->completed value to rcu_node structures
 		 * so that other CPUs don't have to wait until the start
 		 * of the next grace period to process their callbacks.
 		 */
 		rcu_for_each_node_breadth_first(rsp, rnp) {
 			raw_spin_lock(&rnp->lock); /* irqs already disabled. */
 			rnp->completed = rsp->gpnum;
 			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
 		}
 		rnp = rcu_get_root(rsp);
 		raw_spin_lock(&rnp->lock); /* irqs already disabled. */
 	}
 	rsp->completed = rsp->gpnum;  /* Declare the grace period complete. */
 	trace_rcu_grace_period(rsp->name, rsp->completed, "end");
 	rsp->fqs_state = RCU_GP_IDLE;
 	rcu_start_gp(rsp, flags);  /* releases root node's rnp->lock. */
 }
 /*
  * Similar to rcu_report_qs_rdp(), for which it is a helper function.
  * Allows quiescent states for a group of CPUs to be reported at one go
  * to the specified rcu_node structure, though all the CPUs in the group
  * must be represented by the same rcu_node structure (which need not be
  * a leaf rcu_node structure, though it often will be).  That structure's
  * lock must be held upon entry, and it is released before return.
  */
 static void
 rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,
 		  struct rcu_node *rnp, unsigned long flags)
 	__releases(rnp->lock)
 {
 	struct rcu_node *rnp_c;
 	/* Walk up the rcu_node hierarchy. */
 	for (;;) {
 		if (!(rnp->qsmask & mask)) {
 			/* Our bit has already been cleared, so done. */
 			raw_spin_unlock_irqrestore(&rnp->lock, flags);
 			return;
 		}
 		rnp->qsmask &= ~mask;
 		trace_rcu_quiescent_state_report(rsp->name, rnp->gpnum,
 						 mask, rnp->qsmask, rnp->level,
 						 rnp->grplo, rnp->grphi,
 						 !!rnp->gp_tasks);
 		if (rnp->qsmask != 0 || rcu_preempt_blocked_readers_cgp(rnp)) {
 			/* Other bits still set at this level, so done. */
 			raw_spin_unlock_irqrestore(&rnp->lock, flags);
 			return;
 		}
 		mask = rnp->grpmask;
 		if (rnp->parent == NULL) {
 			/* No more levels.  Exit loop holding root lock. */
 			break;
 		}
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		rnp_c = rnp;
 		rnp = rnp->parent;
 		raw_spin_lock_irqsave(&rnp->lock, flags);
 		WARN_ON_ONCE(rnp_c->qsmask);
 	}
 	/*
 	 * Get here if we are the last CPU to pass through a quiescent
 	 * state for this grace period.  Invoke rcu_report_qs_rsp()
 	 * to clean up and start the next grace period if one is needed.
 	 */
 	rcu_report_qs_rsp(rsp, flags); /* releases rnp->lock. */
 }
 /*
  * Record a quiescent state for the specified CPU to that CPU's rcu_data
  * structure.  This must be either called from the specified CPU, or
  * called when the specified CPU is known to be offline (and when it is
  * also known that no other CPU is concurrently trying to help the offline
  * CPU).  The lastcomp argument is used to make sure we are still in the
  * grace period of interest.  We don't want to end the current grace period
  * based on quiescent states detected in an earlier grace period!
  */
 static void
 rcu_report_qs_rdp(int cpu, struct rcu_state *rsp, struct rcu_data *rdp, long lastgp)
 {
 	unsigned long flags;
 	unsigned long mask;
 	struct rcu_node *rnp;
 	rnp = rdp->mynode;
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	if (lastgp != rnp->gpnum || rnp->completed == rnp->gpnum) {
 		/*
 		 * The grace period in which this quiescent state was
 		 * recorded has ended, so don't report it upwards.
 		 * We will instead need a new quiescent state that lies
 		 * within the current grace period.
 		 */
 		rdp->passed_quiesce = 0;	/* need qs for new gp. */
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		return;
 	}
 	mask = rdp->grpmask;
 	if ((rnp->qsmask & mask) == 0) {
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	} else {
 		rdp->qs_pending = 0;
 		/*
 		 * This GP can't end until cpu checks in, so all of our
 		 * callbacks can be processed during the next GP.
 		 */
 		rdp->nxttail[RCU_NEXT_READY_TAIL] = rdp->nxttail[RCU_NEXT_TAIL];
 		rcu_report_qs_rnp(mask, rsp, rnp, flags); /* rlses rnp->lock */
 	}
 }
 /*
  * Check to see if there is a new grace period of which this CPU
  * is not yet aware, and if so, set up local rcu_data state for it.
  * Otherwise, see if this CPU has just passed through its first
  * quiescent state for this grace period, and record that fact if so.
  */
 static void
 rcu_check_quiescent_state(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	/* If there is now a new grace period, record and return. */
 	if (check_for_new_grace_period(rsp, rdp))
 		return;
 	/*
 	 * Does this CPU still need to do its part for current grace period?
 	 * If no, return and let the other CPUs do their part as well.
 	 */
 	if (!rdp->qs_pending)
 		return;
 	/*
 	 * Was there a quiescent state since the beginning of the grace
 	 * period? If no, then exit and wait for the next call.
 	 */
 	if (!rdp->passed_quiesce)
 		return;
 	/*
 	 * Tell RCU we are done (but rcu_report_qs_rdp() will be the
 	 * judge of that).
 	 */
 	rcu_report_qs_rdp(rdp->cpu, rsp, rdp, rdp->passed_quiesce_gpnum);
 }
 #ifdef CONFIG_HOTPLUG_CPU
 /*
  * Move a dying CPU's RCU callbacks to online CPU's callback list.
  * Also record a quiescent state for this CPU for the current grace period.
  * Synchronization and interrupt disabling are not required because
  * this function executes in stop_machine() context.  Therefore, cleanup
  * operations that might block must be done later from the CPU_DEAD
  * notifier.
  *
  * Note that the outgoing CPU's bit has already been cleared in the
  * cpu_online_mask.  This allows us to randomly pick a callback
  * destination from the bits set in that mask.
  */
 static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
 {
 	int i;
 	unsigned long mask;
 	int receive_cpu = cpumask_any(cpu_online_mask);
 	struct rcu_data *rdp = this_cpu_ptr(rsp->rda);
 	struct rcu_data *receive_rdp = per_cpu_ptr(rsp->rda, receive_cpu);
 	RCU_TRACE(struct rcu_node *rnp = rdp->mynode); /* For dying CPU. */
 	/* First, adjust the counts. */
 	if (rdp->nxtlist != NULL) {
 		receive_rdp->qlen_lazy += rdp->qlen_lazy;
 		receive_rdp->qlen += rdp->qlen;
 		rdp->qlen_lazy = 0;
 		rdp->qlen = 0;
 	}
 	/*
 	 * Next, move ready-to-invoke callbacks to be invoked on some
 	 * other CPU.  These will not be required to pass through another
 	 * grace period:  They are done, regardless of CPU.
 	 */
 	if (rdp->nxtlist != NULL &&
 	    rdp->nxttail[RCU_DONE_TAIL] != &rdp->nxtlist) {
 		struct rcu_head *oldhead;
 		struct rcu_head **oldtail;
 		struct rcu_head **newtail;
 		oldhead = rdp->nxtlist;
 		oldtail = receive_rdp->nxttail[RCU_DONE_TAIL];
 		rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];
 		*rdp->nxttail[RCU_DONE_TAIL] = *oldtail;
 		*receive_rdp->nxttail[RCU_DONE_TAIL] = oldhead;
 		newtail = rdp->nxttail[RCU_DONE_TAIL];
 		for (i = RCU_DONE_TAIL; i < RCU_NEXT_SIZE; i++) {
 			if (receive_rdp->nxttail[i] == oldtail)
 				receive_rdp->nxttail[i] = newtail;
 			if (rdp->nxttail[i] == newtail)
 				rdp->nxttail[i] = &rdp->nxtlist;
 		}
 	}
 	/*
 	 * Finally, put the rest of the callbacks at the end of the list.
 	 * The ones that made it partway through get to start over:  We
 	 * cannot assume that grace periods are synchronized across CPUs.
 	 * (We could splice RCU_WAIT_TAIL into RCU_NEXT_READY_TAIL, but
 	 * this does not seem compelling.  Not yet, anyway.)
 	 */
 	if (rdp->nxtlist != NULL) {
 		*receive_rdp->nxttail[RCU_NEXT_TAIL] = rdp->nxtlist;
 		receive_rdp->nxttail[RCU_NEXT_TAIL] =
 				rdp->nxttail[RCU_NEXT_TAIL];
 		receive_rdp->n_cbs_adopted += rdp->qlen;
 		rdp->n_cbs_orphaned += rdp->qlen;
 		rdp->nxtlist = NULL;
 		for (i = 0; i < RCU_NEXT_SIZE; i++)
 			rdp->nxttail[i] = &rdp->nxtlist;
 	}
 	/*
 	 * Record a quiescent state for the dying CPU.  This is safe
 	 * only because we have already cleared out the callbacks.
 	 * (Otherwise, the RCU core might try to schedule the invocation
 	 * of callbacks on this now-offline CPU, which would be bad.)
 	 */
 	mask = rdp->grpmask;	/* rnp->grplo is constant. */
 	trace_rcu_grace_period(rsp->name,
 			       rnp->gpnum + 1 - !!(rnp->qsmask & mask),
 			       "cpuofl");
 	rcu_report_qs_rdp(smp_processor_id(), rsp, rdp, rsp->gpnum);
 	/* Note that rcu_report_qs_rdp() might call trace_rcu_grace_period(). */
 }
 /*
  * The CPU has been completely removed, and some other CPU is reporting
  * this fact from process context.  Do the remainder of the cleanup.
  * There can only be one CPU hotplug operation at a time, so no other
  * CPU can be attempting to update rcu_cpu_kthread_task.
  */
 static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
 {
 	unsigned long flags;
 	unsigned long mask;
 	int need_report = 0;
 	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
 	struct rcu_node *rnp = rdp->mynode;  /* Outgoing CPU's rnp. */
 	/* Adjust any no-longer-needed kthreads. */
 	rcu_stop_cpu_kthread(cpu);
 	rcu_node_kthread_setaffinity(rnp, -1);
 	/* Remove the dying CPU from the bitmasks in the rcu_node hierarchy. */
 	/* Exclude any attempts to start a new grace period. */
 	raw_spin_lock_irqsave(&rsp->onofflock, flags);
 	/* Remove the outgoing CPU from the masks in the rcu_node hierarchy. */
 	mask = rdp->grpmask;	/* rnp->grplo is constant. */
 	do {
 		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
 		rnp->qsmaskinit &= ~mask;
 		if (rnp->qsmaskinit != 0) {
 			if (rnp != rdp->mynode)
 				raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
 			break;
 		}
 		if (rnp == rdp->mynode)
 			need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);
 		else
 			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
 		mask = rnp->grpmask;
 		rnp = rnp->parent;
 	} while (rnp != NULL);
 	/*
 	 * We still hold the leaf rcu_node structure lock here, and
 	 * irqs are still disabled.  The reason for this subterfuge is
 	 * because invoking rcu_report_unblock_qs_rnp() with ->onofflock
 	 * held leads to deadlock.
 	 */
 	raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */
 	rnp = rdp->mynode;
 	if (need_report & RCU_OFL_TASKS_NORM_GP)
 		rcu_report_unblock_qs_rnp(rnp, flags);
 	else
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	if (need_report & RCU_OFL_TASKS_EXP_GP)
 		rcu_report_exp_rnp(rsp, rnp, true);
 }
 #else /* #ifdef CONFIG_HOTPLUG_CPU */
 static void rcu_cleanup_dying_cpu(struct rcu_state *rsp)
 {
 }
 static void rcu_cleanup_dead_cpu(int cpu, struct rcu_state *rsp)
 {
 }
 #endif /* #else #ifdef CONFIG_HOTPLUG_CPU */
 /*
  * Invoke any RCU callbacks that have made it to the end of their grace
  * period.  Thottle as specified by rdp->blimit.
  */
 static void rcu_do_batch(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long flags;
 	struct rcu_head *next, *list, **tail;
 	int bl, count, count_lazy;
 	/* If no callbacks are ready, just return.*/
 	if (!cpu_has_callbacks_ready_to_invoke(rdp)) {
 		trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, 0);
 		trace_rcu_batch_end(rsp->name, 0, !!ACCESS_ONCE(rdp->nxtlist),
 				    need_resched(), is_idle_task(current),
 				    rcu_is_callbacks_kthread());
 		return;
 	}
 	/*
 	 * Extract the list of ready callbacks, disabling to prevent
 	 * races with call_rcu() from interrupt handlers.
 	 */
 	local_irq_save(flags);
 	WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));
 	bl = rdp->blimit;
 	trace_rcu_batch_start(rsp->name, rdp->qlen_lazy, rdp->qlen, bl);
 	list = rdp->nxtlist;
 	rdp->nxtlist = *rdp->nxttail[RCU_DONE_TAIL];
 	*rdp->nxttail[RCU_DONE_TAIL] = NULL;
 	tail = rdp->nxttail[RCU_DONE_TAIL];
 	for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)
 		if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])
 			rdp->nxttail[count] = &rdp->nxtlist;
 	local_irq_restore(flags);
 	/* Invoke callbacks. */
 	count = count_lazy = 0;
 	while (list) {
 		next = list->next;
 		prefetch(next);
 		debug_rcu_head_unqueue(list);
 		if (__rcu_reclaim(rsp->name, list))
 			count_lazy++;
 		list = next;
 		/* Stop only if limit reached and CPU has something to do. */
 		if (++count >= bl &&
 		    (need_resched() ||
 		     (!is_idle_task(current) && !rcu_is_callbacks_kthread())))
 			break;
 	}
 	local_irq_save(flags);
 	trace_rcu_batch_end(rsp->name, count, !!list, need_resched(),
 			    is_idle_task(current),
 			    rcu_is_callbacks_kthread());
 	/* Update count, and requeue any remaining callbacks. */
 	rdp->qlen_lazy -= count_lazy;
 	rdp->qlen -= count;
 	rdp->n_cbs_invoked += count;
 	if (list != NULL) {
 		*tail = rdp->nxtlist;
 		rdp->nxtlist = list;
 		for (count = 0; count < RCU_NEXT_SIZE; count++)
 			if (&rdp->nxtlist == rdp->nxttail[count])
 				rdp->nxttail[count] = tail;
 			else
 				break;
 	}
 	/* Reinstate batch limit if we have worked down the excess. */
 	if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)
 		rdp->blimit = blimit;
 	/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */
 	if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {
 		rdp->qlen_last_fqs_check = 0;
 		rdp->n_force_qs_snap = rsp->n_force_qs;
 	} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)
 		rdp->qlen_last_fqs_check = rdp->qlen;
 	local_irq_restore(flags);
 	/* Re-invoke RCU core processing if there are callbacks remaining. */
 	if (cpu_has_callbacks_ready_to_invoke(rdp))
 		invoke_rcu_core();
 }
 /*
  * Check to see if this CPU is in a non-context-switch quiescent state
  * (user mode or idle loop for rcu, non-softirq execution for rcu_bh).
  * Also schedule RCU core processing.
  *
  * This function must be called from hardirq context.  It is normally
  * invoked from the scheduling-clock interrupt.  If rcu_pending returns
  * false, there is no point in invoking rcu_check_callbacks().
  */
 void rcu_check_callbacks(int cpu, int user)
 {
 	trace_rcu_utilization("Start scheduler-tick");
 	increment_cpu_stall_ticks();
 	if (user || rcu_is_cpu_rrupt_from_idle()) {
 		/*
 		 * Get here if this CPU took its interrupt from user
 		 * mode or from the idle loop, and if this is not a
 		 * nested interrupt.  In this case, the CPU is in
 		 * a quiescent state, so note it.
 		 *
 		 * No memory barrier is required here because both
 		 * rcu_sched_qs() and rcu_bh_qs() reference only CPU-local
 		 * variables that other CPUs neither access nor modify,
 		 * at least not while the corresponding CPU is online.
 		 */
 		rcu_sched_qs(cpu);
 		rcu_bh_qs(cpu);
 	} else if (!in_softirq()) {
 		/*
 		 * Get here if this CPU did not take its interrupt from
 		 * softirq, in other words, if it is not interrupting
 		 * a rcu_bh read-side critical section.  This is an _bh
 		 * critical section, so note it.
 		 */
 		rcu_bh_qs(cpu);
 	}
 	rcu_preempt_check_callbacks(cpu);
 	if (rcu_pending(cpu))
 		invoke_rcu_core();
 	trace_rcu_utilization("End scheduler-tick");
 }
 /*
  * Scan the leaf rcu_node structures, processing dyntick state for any that
  * have not yet encountered a quiescent state, using the function specified.
  * Also initiate boosting for any threads blocked on the root rcu_node.
  *
  * The caller must have suppressed start of new grace periods.
  */
 static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))
 {
 	unsigned long bit;
 	int cpu;
 	unsigned long flags;
 	unsigned long mask;
 	struct rcu_node *rnp;
 	rcu_for_each_leaf_node(rsp, rnp) {
 		mask = 0;
 		raw_spin_lock_irqsave(&rnp->lock, flags);
 		if (!rcu_gp_in_progress(rsp)) {
 			raw_spin_unlock_irqrestore(&rnp->lock, flags);
 			return;
 		}
 		if (rnp->qsmask == 0) {
 			rcu_initiate_boost(rnp, flags); /* releases rnp->lock */
 			continue;
 		}
 		cpu = rnp->grplo;
 		bit = 1;
 		for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
 			if ((rnp->qsmask & bit) != 0 &&
 			    f(per_cpu_ptr(rsp->rda, cpu)))
 				mask |= bit;
 		}
 		if (mask != 0) {
 			/* rcu_report_qs_rnp() releases rnp->lock. */
 			rcu_report_qs_rnp(mask, rsp, rnp, flags);
 			continue;
 		}
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	}
 	rnp = rcu_get_root(rsp);
 	if (rnp->qsmask == 0) {
 		raw_spin_lock_irqsave(&rnp->lock, flags);
 		rcu_initiate_boost(rnp, flags); /* releases rnp->lock. */
 	}
 }
 /*
  * Force quiescent states on reluctant CPUs, and also detect which
  * CPUs are in dyntick-idle mode.
  */
 static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
 {
 	unsigned long flags;
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	trace_rcu_utilization("Start fqs");
 	if (!rcu_gp_in_progress(rsp)) {
 		trace_rcu_utilization("End fqs");
 		return;  /* No grace period in progress, nothing to force. */
 	}
 	if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {
 		rsp->n_force_qs_lh++; /* Inexact, can lose counts.  Tough! */
 		trace_rcu_utilization("End fqs");
 		return;	/* Someone else is already on the job. */
 	}
 	if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))
 		goto unlock_fqs_ret; /* no emergency and done recently. */
 	rsp->n_force_qs++;
 	raw_spin_lock(&rnp->lock);  /* irqs already disabled */
 	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
 	if(!rcu_gp_in_progress(rsp)) {
 		rsp->n_force_qs_ngp++;
 		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
 		goto unlock_fqs_ret;  /* no GP in progress, time updated. */
 	}
 	rsp->fqs_active = 1;
 	switch (rsp->fqs_state) {
 	case RCU_GP_IDLE:
 	case RCU_GP_INIT:
 		break; /* grace period idle or initializing, ignore. */
 	case RCU_SAVE_DYNTICK:
 		if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)
 			break; /* So gcc recognizes the dead code. */
 		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
 		/* Record dyntick-idle state. */
 		force_qs_rnp(rsp, dyntick_save_progress_counter);
 		raw_spin_lock(&rnp->lock);  /* irqs already disabled */
 		if (rcu_gp_in_progress(rsp))
 			rsp->fqs_state = RCU_FORCE_QS;
 		break;
 	case RCU_FORCE_QS:
 		/* Check dyntick-idle state, send IPI to laggarts. */
 		raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
 		force_qs_rnp(rsp, rcu_implicit_dynticks_qs);
 		/* Leave state in case more forcing is required. */
 		raw_spin_lock(&rnp->lock);  /* irqs already disabled */
 		break;
 	}
 	rsp->fqs_active = 0;
 	if (rsp->fqs_need_gp) {
 		raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */
 		rsp->fqs_need_gp = 0;
 		rcu_start_gp(rsp, flags); /* releases rnp->lock */
 		trace_rcu_utilization("End fqs");
 		return;
 	}
 	raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
 unlock_fqs_ret:
 	raw_spin_unlock_irqrestore(&rsp->fqslock, flags);
 	trace_rcu_utilization("End fqs");
 }
 /*
  * This does the RCU core processing work for the specified rcu_state
  * and rcu_data structures.  This may be called only from the CPU to
  * whom the rdp belongs.
  */
 static void
 __rcu_process_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	unsigned long flags;
 	WARN_ON_ONCE(rdp->beenonline == 0);
 	/*
 	 * If an RCU GP has gone long enough, go check for dyntick
 	 * idle CPUs and, if needed, send resched IPIs.
 	 */
 	if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))
 		force_quiescent_state(rsp, 1);
 	/*
 	 * Advance callbacks in response to end of earlier grace
 	 * period that some other CPU ended.
 	 */
 	rcu_process_gp_end(rsp, rdp);
 	/* Update RCU state based on any recent quiescent states. */
 	rcu_check_quiescent_state(rsp, rdp);
 	/* Does this CPU require a not-yet-started grace period? */
 	if (cpu_needs_another_gp(rsp, rdp)) {
 		raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);
 		rcu_start_gp(rsp, flags);  /* releases above lock */
 	}
 	/* If there are callbacks ready, invoke them. */
 	if (cpu_has_callbacks_ready_to_invoke(rdp))
 		invoke_rcu_callbacks(rsp, rdp);
 }
 /*
  * Do RCU core processing for the current CPU.
  */
 static void rcu_process_callbacks(struct softirq_action *unused)
 {
 	trace_rcu_utilization("Start RCU core");
 	__rcu_process_callbacks(&rcu_sched_state,
 				&__get_cpu_var(rcu_sched_data));
 	__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));
 	rcu_preempt_process_callbacks();
 	trace_rcu_utilization("End RCU core");
 }
 /*
  * Schedule RCU callback invocation.  If the specified type of RCU
  * does not support RCU priority boosting, just do a direct call,
  * otherwise wake up the per-CPU kernel kthread.  Note that because we
  * are running on the current CPU with interrupts disabled, the
  * rcu_cpu_kthread_task cannot disappear out from under us.
  */
 static void invoke_rcu_callbacks(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	if (unlikely(!ACCESS_ONCE(rcu_scheduler_fully_active)))
 		return;
 	if (likely(!rsp->boost)) {
 		rcu_do_batch(rsp, rdp);
 		return;
 	}
 	invoke_rcu_callbacks_kthread();
 }
 static void invoke_rcu_core(void)
 {
 	raise_softirq(RCU_SOFTIRQ);
 }
 static void
 __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
 	   struct rcu_state *rsp, bool lazy)
 {
 	unsigned long flags;
 	struct rcu_data *rdp;
 	WARN_ON_ONCE((unsigned long)head & 0x3); /* Misaligned rcu_head! */
 	debug_rcu_head_queue(head);
 	head->func = func;
 	head->next = NULL;
 	smp_mb(); /* Ensure RCU update seen before callback registry. */
 	/*
 	 * Opportunistically note grace-period endings and beginnings.
 	 * Note that we might see a beginning right after we see an
 	 * end, but never vice versa, since this CPU has to pass through
 	 * a quiescent state betweentimes.
 	 */
 	local_irq_save(flags);
 	WARN_ON_ONCE(cpu_is_offline(smp_processor_id()));
 	rdp = this_cpu_ptr(rsp->rda);
 	/* Add the callback to our list. */
 	*rdp->nxttail[RCU_NEXT_TAIL] = head;
 	rdp->nxttail[RCU_NEXT_TAIL] = &head->next;
 	rdp->qlen++;
 	if (lazy)
 		rdp->qlen_lazy++;
 	if (__is_kfree_rcu_offset((unsigned long)func))
 		trace_rcu_kfree_callback(rsp->name, head, (unsigned long)func,
 					 rdp->qlen_lazy, rdp->qlen);
 	else
 		trace_rcu_callback(rsp->name, head, rdp->qlen_lazy, rdp->qlen);
 	/* If interrupts were disabled, don't dive into RCU core. */
 	if (irqs_disabled_flags(flags)) {
 		local_irq_restore(flags);
 		return;
 	}
 	/*
 	 * Force the grace period if too many callbacks or too long waiting.
 	 * Enforce hysteresis, and don't invoke force_quiescent_state()
 	 * if some other CPU has recently done so.  Also, don't bother
 	 * invoking force_quiescent_state() if the newly enqueued callback
 	 * is the only one waiting for a grace period to complete.
 	 */
 	if (unlikely(rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {
 		/* Are we ignoring a completed grace period? */
 		rcu_process_gp_end(rsp, rdp);
 		check_for_new_grace_period(rsp, rdp);
 		/* Start a new grace period if one not already started. */
 		if (!rcu_gp_in_progress(rsp)) {
 			unsigned long nestflag;
 			struct rcu_node *rnp_root = rcu_get_root(rsp);
 			raw_spin_lock_irqsave(&rnp_root->lock, nestflag);
 			rcu_start_gp(rsp, nestflag);  /* rlses rnp_root->lock */
 		} else {
 			/* Give the grace period a kick. */
 			rdp->blimit = LONG_MAX;
 			if (rsp->n_force_qs == rdp->n_force_qs_snap &&
 			    *rdp->nxttail[RCU_DONE_TAIL] != head)
 				force_quiescent_state(rsp, 0);
 			rdp->n_force_qs_snap = rsp->n_force_qs;
 			rdp->qlen_last_fqs_check = rdp->qlen;
 		}
 	} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))
 		force_quiescent_state(rsp, 1);
 	local_irq_restore(flags);
 }
 /*
  * Queue an RCU-sched callback for invocation after a grace period.
  */
 void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
 {
 	__call_rcu(head, func, &rcu_sched_state, 0);
 }
 EXPORT_SYMBOL_GPL(call_rcu_sched);
 /*
  * Queue an RCU callback for invocation after a quicker grace period.
  */
 void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))
 {
 	__call_rcu(head, func, &rcu_bh_state, 0);
 }
 EXPORT_SYMBOL_GPL(call_rcu_bh);
+/*
+ * Because a context switch is a grace period for RCU-sched and RCU-bh,
+ * any blocking grace-period wait automatically implies a grace period
+ * if there is only one CPU online at any point time during execution
+ * of either synchronize_sched() or synchronize_rcu_bh().  It is OK to
+ * occasionally incorrectly indicate that there are multiple CPUs online
+ * when there was in fact only one the whole time, as this just adds
+ * some overhead: RCU still operates correctly.
+ *
+ * Of course, sampling num_online_cpus() with preemption enabled can
+ * give erroneous results if there are concurrent CPU-hotplug operations.
+ * For example, given a demonic sequence of preemptions in num_online_cpus()
+ * and CPU-hotplug operations, there could be two or more CPUs online at
+ * all times, but num_online_cpus() might well return one (or even zero).
+ *
+ * However, all such demonic sequences require at least one CPU-offline
+ * operation.  Furthermore, rcu_blocking_is_gp() giving the wrong answer
+ * is only a problem if there is an RCU read-side critical section executing
+ * throughout.  But RCU-sched and RCU-bh read-side critical sections
+ * disable either preemption or bh, which prevents a CPU from going offline.
+ * Therefore, the only way that rcu_blocking_is_gp() can incorrectly return
+ * that there is only one CPU when in fact there was more than one throughout
+ * is when there were no RCU readers in the system.  If there are no
+ * RCU readers, the grace period by definition can be of zero length,
+ * regardless of the number of online CPUs.
+ */
+static inline int rcu_blocking_is_gp(void)
+{
+	might_sleep();  /* Check for RCU read-side critical section. */
+	return num_online_cpus() <= 1;
+}
 /**
  * synchronize_sched - wait until an rcu-sched grace period has elapsed.
  *
  * Control will return to the caller some time after a full rcu-sched
  * grace period has elapsed, in other words after all currently executing
  * rcu-sched read-side critical sections have completed.   These read-side
  * critical sections are delimited by rcu_read_lock_sched() and
  * rcu_read_unlock_sched(), and may be nested.  Note that preempt_disable(),
  * local_irq_disable(), and so on may be used in place of
  * rcu_read_lock_sched().
  *
  * This means that all preempt_disable code sequences, including NMI and
  * hardware-interrupt handlers, in progress on entry will have completed
  * before this primitive returns.  However, this does not guarantee that
  * softirq handlers will have completed, since in some kernels, these
  * handlers can run in process context, and can block.
  *
  * This primitive provides the guarantees made by the (now removed)
  * synchronize_kernel() API.  In contrast, synchronize_rcu() only
  * guarantees that rcu_read_lock() sections will have completed.
  * In "classic RCU", these two guarantees happen to be one and
  * the same, but can differ in realtime RCU implementations.
  */
 void synchronize_sched(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_sched() in RCU-sched read-side critical section");
 	if (rcu_blocking_is_gp())
 		return;
 	wait_rcu_gp(call_rcu_sched);
 }
 EXPORT_SYMBOL_GPL(synchronize_sched);
 /**
  * synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.
  *
  * Control will return to the caller some time after a full rcu_bh grace
  * period has elapsed, in other words after all currently executing rcu_bh
  * read-side critical sections have completed.  RCU read-side critical
  * sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),
  * and may be nested.
  */
 void synchronize_rcu_bh(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_bh() in RCU-bh read-side critical section");
 	if (rcu_blocking_is_gp())
 		return;
 	wait_rcu_gp(call_rcu_bh);
 }
 EXPORT_SYMBOL_GPL(synchronize_rcu_bh);
 static atomic_t sync_sched_expedited_started = ATOMIC_INIT(0);
 static atomic_t sync_sched_expedited_done = ATOMIC_INIT(0);
 static int synchronize_sched_expedited_cpu_stop(void *data)
 {
 	/*
 	 * 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;
 }
 /**
  * synchronize_sched_expedited - Brute-force RCU-sched grace period
  *
  * Wait for an RCU-sched grace period to elapse, but use a "big hammer"
  * approach to force the grace period to end quickly.  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_sched_expedited() in a loop, please
  * restructure your code to batch your updates, and then use a single
  * synchronize_sched() 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.
  *
  * 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.
  *
  * 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().
  */
 void synchronize_sched_expedited(void)
 {
 	int firstsnap, s, snap, trycount = 0;
 	/* Note that atomic_inc_return() implies full memory barrier. */
 	firstsnap = snap = atomic_inc_return(&sync_sched_expedited_started);
 	get_online_cpus();
 	WARN_ON_ONCE(cpu_is_offline(raw_smp_processor_id()));
 	/*
 	 * 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();
 		/* No joy, try again later.  Or just synchronize_sched(). */
 		if (trycount++ < 10)
 			udelay(trycount * num_online_cpus());
 		else {
 			synchronize_sched();
 			return;
 		}
 		/* 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;
 		}
 		/*
 		 * 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);
 		smp_mb(); /* ensure read is before try_stop_cpus(). */
 	}
 	/*
 	 * 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);
 	put_online_cpus();
 }
 EXPORT_SYMBOL_GPL(synchronize_sched_expedited);
 /*
  * Check to see if there is any immediate RCU-related work to be done
  * by the current CPU, for the specified type of RCU, returning 1 if so.
  * The checks are in order of increasing expense: checks that can be
  * carried out against CPU-local state are performed first.  However,
  * we must check for CPU stalls first, else we might not get a chance.
  */
 static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 	struct rcu_node *rnp = rdp->mynode;
 	rdp->n_rcu_pending++;
 	/* Check for CPU stalls, if enabled. */
 	check_cpu_stall(rsp, rdp);
 	/* Is the RCU core waiting for a quiescent state from this CPU? */
 	if (rcu_scheduler_fully_active &&
 	    rdp->qs_pending && !rdp->passed_quiesce) {
 		/*
 		 * If force_quiescent_state() coming soon and this CPU
 		 * needs a quiescent state, and this is either RCU-sched
 		 * or RCU-bh, force a local reschedule.
 		 */
 		rdp->n_rp_qs_pending++;
 		if (!rdp->preemptible &&
 		    ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,
 				 jiffies))
 			set_need_resched();
 	} else if (rdp->qs_pending && rdp->passed_quiesce) {
 		rdp->n_rp_report_qs++;
 		return 1;
 	}
 	/* Does this CPU have callbacks ready to invoke? */
 	if (cpu_has_callbacks_ready_to_invoke(rdp)) {
 		rdp->n_rp_cb_ready++;
 		return 1;
 	}
 	/* Has RCU gone idle with this CPU needing another grace period? */
 	if (cpu_needs_another_gp(rsp, rdp)) {
 		rdp->n_rp_cpu_needs_gp++;
 		return 1;
 	}
 	/* Has another RCU grace period completed?  */
 	if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */
 		rdp->n_rp_gp_completed++;
 		return 1;
 	}
 	/* Has a new RCU grace period started? */
 	if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */
 		rdp->n_rp_gp_started++;
 		return 1;
 	}
 	/* Has an RCU GP gone long enough to send resched IPIs &c? */
 	if (rcu_gp_in_progress(rsp) &&
 	    ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {
 		rdp->n_rp_need_fqs++;
 		return 1;
 	}
 	/* nothing to do */
 	rdp->n_rp_need_nothing++;
 	return 0;
 }
 /*
  * Check to see if there is any immediate RCU-related work to be done
  * by the current CPU, returning 1 if so.  This function is part of the
  * RCU implementation; it is -not- an exported member of the RCU API.
  */
 static int rcu_pending(int cpu)
 {
 	return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||
 	       __rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||
 	       rcu_preempt_pending(cpu);
 }
 /*
  * 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.
  */
 static int rcu_cpu_has_callbacks(int cpu)
 {
 	/* RCU callbacks either ready or pending? */
 	return per_cpu(rcu_sched_data, cpu).nxtlist ||
 	       per_cpu(rcu_bh_data, cpu).nxtlist ||
 	       rcu_preempt_cpu_has_callbacks(cpu);
 }
 static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};
 static atomic_t rcu_barrier_cpu_count;
 static DEFINE_MUTEX(rcu_barrier_mutex);
 static struct completion rcu_barrier_completion;
 static void rcu_barrier_callback(struct rcu_head *notused)
 {
 	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
 		complete(&rcu_barrier_completion);
 }
 /*
  * Called with preemption disabled, and from cross-cpu IRQ context.
  */
 static void rcu_barrier_func(void *type)
 {
 	int cpu = smp_processor_id();
 	struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);
 	void (*call_rcu_func)(struct rcu_head *head,
 			      void (*func)(struct rcu_head *head));
 	atomic_inc(&rcu_barrier_cpu_count);
 	call_rcu_func = type;
 	call_rcu_func(head, rcu_barrier_callback);
 }
 /*
  * Orchestrate the specified type of RCU barrier, waiting for all
  * RCU callbacks of the specified type to complete.
  */
 static void _rcu_barrier(struct rcu_state *rsp,
 			 void (*call_rcu_func)(struct rcu_head *head,
 					       void (*func)(struct rcu_head *head)))
 {
 	BUG_ON(in_interrupt());
 	/* Take mutex to serialize concurrent rcu_barrier() requests. */
 	mutex_lock(&rcu_barrier_mutex);
 	init_completion(&rcu_barrier_completion);
 	/*
 	 * Initialize rcu_barrier_cpu_count to 1, then invoke
 	 * rcu_barrier_func() on each CPU, so that each CPU also has
 	 * incremented rcu_barrier_cpu_count.  Only then is it safe to
 	 * decrement rcu_barrier_cpu_count -- otherwise the first CPU
 	 * might complete its grace period before all of the other CPUs
 	 * did their increment, causing this function to return too
 	 * early.  Note that on_each_cpu() disables irqs, which prevents
 	 * any CPUs from coming online or going offline until each online
 	 * CPU has queued its RCU-barrier callback.
 	 */
 	atomic_set(&rcu_barrier_cpu_count, 1);
 	on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);
 	if (atomic_dec_and_test(&rcu_barrier_cpu_count))
 		complete(&rcu_barrier_completion);
 	wait_for_completion(&rcu_barrier_completion);
 	mutex_unlock(&rcu_barrier_mutex);
 }
 /**
  * rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.
  */
 void rcu_barrier_bh(void)
 {
 	_rcu_barrier(&rcu_bh_state, call_rcu_bh);
 }
 EXPORT_SYMBOL_GPL(rcu_barrier_bh);
 /**
  * rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.
  */
 void rcu_barrier_sched(void)
 {
 	_rcu_barrier(&rcu_sched_state, call_rcu_sched);
 }
 EXPORT_SYMBOL_GPL(rcu_barrier_sched);
 /*
  * Do boot-time initialization of a CPU's per-CPU RCU data.
  */
 static void __init
 rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)
 {
 	unsigned long flags;
 	int i;
 	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	/* Set up local state, ensuring consistent view of global state. */
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);
 	rdp->nxtlist = NULL;
 	for (i = 0; i < RCU_NEXT_SIZE; i++)
 		rdp->nxttail[i] = &rdp->nxtlist;
 	rdp->qlen_lazy = 0;
 	rdp->qlen = 0;
 	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
 	WARN_ON_ONCE(rdp->dynticks->dynticks_nesting != DYNTICK_TASK_EXIT_IDLE);
 	WARN_ON_ONCE(atomic_read(&rdp->dynticks->dynticks) != 1);
 	rdp->cpu = cpu;
 	rdp->rsp = rsp;
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 }
 /*
  * Initialize a CPU's per-CPU RCU data.  Note that only one online or
  * offline event can be happening at a given time.  Note also that we
  * can accept some slop in the rsp->completed access due to the fact
  * that this CPU cannot possibly have any RCU callbacks in flight yet.
  */
 static void __cpuinit
 rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptible)
 {
 	unsigned long flags;
 	unsigned long mask;
 	struct rcu_data *rdp = per_cpu_ptr(rsp->rda, cpu);
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	/* Set up local state, ensuring consistent view of global state. */
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	rdp->beenonline = 1;	 /* We have now been online. */
 	rdp->preemptible = preemptible;
 	rdp->qlen_last_fqs_check = 0;
 	rdp->n_force_qs_snap = rsp->n_force_qs;
 	rdp->blimit = blimit;
 	rdp->dynticks->dynticks_nesting = DYNTICK_TASK_EXIT_IDLE;
 	atomic_set(&rdp->dynticks->dynticks,
 		   (atomic_read(&rdp->dynticks->dynticks) & ~0x1) + 1);
 	rcu_prepare_for_idle_init(cpu);
 	raw_spin_unlock(&rnp->lock);		/* irqs remain disabled. */
 	/*
 	 * A new grace period might start here.  If so, we won't be part
 	 * of it, but that is OK, as we are currently in a quiescent state.
 	 */
 	/* Exclude any attempts to start a new GP on large systems. */
 	raw_spin_lock(&rsp->onofflock);		/* irqs already disabled. */
 	/* Add CPU to rcu_node bitmasks. */
 	rnp = rdp->mynode;
 	mask = rdp->grpmask;
 	do {
 		/* Exclude any attempts to start a new GP on small systems. */
 		raw_spin_lock(&rnp->lock);	/* irqs already disabled. */
 		rnp->qsmaskinit |= mask;
 		mask = rnp->grpmask;
 		if (rnp == rdp->mynode) {
 			/*
 			 * If there is a grace period in progress, we will
 			 * set up to wait for it next time we run the
 			 * RCU core code.
 			 */
 			rdp->gpnum = rnp->completed;
 			rdp->completed = rnp->completed;
 			rdp->passed_quiesce = 0;
 			rdp->qs_pending = 0;
 			rdp->passed_quiesce_gpnum = rnp->gpnum - 1;
 			trace_rcu_grace_period(rsp->name, rdp->gpnum, "cpuonl");
 		}
 		raw_spin_unlock(&rnp->lock); /* irqs already disabled. */
 		rnp = rnp->parent;
 	} while (rnp != NULL && !(rnp->qsmaskinit & mask));
 	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
 }
 static void __cpuinit rcu_prepare_cpu(int cpu)
 {
 	rcu_init_percpu_data(cpu, &rcu_sched_state, 0);
 	rcu_init_percpu_data(cpu, &rcu_bh_state, 0);
 	rcu_preempt_init_percpu_data(cpu);
 }
 /*
  * Handle CPU online/offline notification events.
  */
 static int __cpuinit rcu_cpu_notify(struct notifier_block *self,
 				    unsigned long action, void *hcpu)
 {
 	long cpu = (long)hcpu;
 	struct rcu_data *rdp = per_cpu_ptr(rcu_state->rda, cpu);
 	struct rcu_node *rnp = rdp->mynode;
 	trace_rcu_utilization("Start CPU hotplug");
 	switch (action) {
 	case CPU_UP_PREPARE:
 	case CPU_UP_PREPARE_FROZEN:
 		rcu_prepare_cpu(cpu);
 		rcu_prepare_kthreads(cpu);
 		break;
 	case CPU_ONLINE:
 	case CPU_DOWN_FAILED:
 		rcu_node_kthread_setaffinity(rnp, -1);
 		rcu_cpu_kthread_setrt(cpu, 1);
 		break;
 	case CPU_DOWN_PREPARE:
 		rcu_node_kthread_setaffinity(rnp, cpu);
 		rcu_cpu_kthread_setrt(cpu, 0);
 		break;
 	case CPU_DYING:
 	case CPU_DYING_FROZEN:
 		/*
 		 * The whole machine is "stopped" except this CPU, so we can
 		 * touch any data without introducing corruption. We send the
 		 * dying CPU's callbacks to an arbitrarily chosen online CPU.
 		 */
 		rcu_cleanup_dying_cpu(&rcu_bh_state);
 		rcu_cleanup_dying_cpu(&rcu_sched_state);
 		rcu_preempt_cleanup_dying_cpu();
 		rcu_cleanup_after_idle(cpu);
 		break;
 	case CPU_DEAD:
 	case CPU_DEAD_FROZEN:
 	case CPU_UP_CANCELED:
 	case CPU_UP_CANCELED_FROZEN:
 		rcu_cleanup_dead_cpu(cpu, &rcu_bh_state);
 		rcu_cleanup_dead_cpu(cpu, &rcu_sched_state);
 		rcu_preempt_cleanup_dead_cpu(cpu);
 		break;
 	default:
 		break;
 	}
 	trace_rcu_utilization("End CPU hotplug");
 	return NOTIFY_OK;
 }
 /*
  * This function is invoked towards the end of the scheduler's initialization
  * process.  Before this is called, the idle task might contain
  * RCU read-side critical sections (during which time, this idle
  * task is booting the system).  After this function is called, the
  * idle tasks are prohibited from containing RCU read-side critical
  * sections.  This function also enables RCU lockdep checking.
  */
 void rcu_scheduler_starting(void)
 {
 	WARN_ON(num_online_cpus() != 1);
 	WARN_ON(nr_context_switches() > 0);
 	rcu_scheduler_active = 1;
 }
 /*
  * Compute the per-level fanout, either using the exact fanout specified
  * or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.
  */
 #ifdef CONFIG_RCU_FANOUT_EXACT
 static void __init rcu_init_levelspread(struct rcu_state *rsp)
 {
 	int i;
 	for (i = NUM_RCU_LVLS - 1; i > 0; i--)
 		rsp->levelspread[i] = CONFIG_RCU_FANOUT;
 	rsp->levelspread[0] = CONFIG_RCU_FANOUT_LEAF;
 }
 #else /* #ifdef CONFIG_RCU_FANOUT_EXACT */
 static void __init rcu_init_levelspread(struct rcu_state *rsp)
 {
 	int ccur;
 	int cprv;
 	int i;
 	cprv = NR_CPUS;
 	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
 		ccur = rsp->levelcnt[i];
 		rsp->levelspread[i] = (cprv + ccur - 1) / ccur;
 		cprv = ccur;
 	}
 }
 #endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */
 /*
  * Helper function for rcu_init() that initializes one rcu_state structure.
  */
 static void __init rcu_init_one(struct rcu_state *rsp,
 		struct rcu_data __percpu *rda)
 {
 	static char *buf[] = { "rcu_node_level_0",
 			       "rcu_node_level_1",
 			       "rcu_node_level_2",
 			       "rcu_node_level_3" };  /* Match MAX_RCU_LVLS */
 	int cpustride = 1;
 	int i;
 	int j;
 	struct rcu_node *rnp;
 	BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf));  /* Fix buf[] init! */
 	/* Initialize the level-tracking arrays. */
 	for (i = 1; i < NUM_RCU_LVLS; i++)
 		rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];
 	rcu_init_levelspread(rsp);
 	/* Initialize the elements themselves, starting from the leaves. */
 	for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {
 		cpustride *= rsp->levelspread[i];
 		rnp = rsp->level[i];
 		for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {
 			raw_spin_lock_init(&rnp->lock);
 			lockdep_set_class_and_name(&rnp->lock,
 						   &rcu_node_class[i], buf[i]);
 			rnp->gpnum = 0;
 			rnp->qsmask = 0;
 			rnp->qsmaskinit = 0;
 			rnp->grplo = j * cpustride;
 			rnp->grphi = (j + 1) * cpustride - 1;
 			if (rnp->grphi >= NR_CPUS)
 				rnp->grphi = NR_CPUS - 1;
 			if (i == 0) {
 				rnp->grpnum = 0;
 				rnp->grpmask = 0;
 				rnp->parent = NULL;
 			} else {
 				rnp->grpnum = j % rsp->levelspread[i - 1];
 				rnp->grpmask = 1UL << rnp->grpnum;
 				rnp->parent = rsp->level[i - 1] +
 					      j / rsp->levelspread[i - 1];
 			}
 			rnp->level = i;
 			INIT_LIST_HEAD(&rnp->blkd_tasks);
 		}
 	}
 	rsp->rda = rda;
 	rnp = rsp->level[NUM_RCU_LVLS - 1];
 	for_each_possible_cpu(i) {
 		while (i > rnp->grphi)
 			rnp++;
 		per_cpu_ptr(rsp->rda, i)->mynode = rnp;
 		rcu_boot_init_percpu_data(i, rsp);
 	}
 }
 void __init rcu_init(void)
 {
 	int cpu;
 	rcu_bootup_announce();
 	rcu_init_one(&rcu_sched_state, &rcu_sched_data);
 	rcu_init_one(&rcu_bh_state, &rcu_bh_data);
 	__rcu_init_preempt();
 	 open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);
 	/*
 	 * We don't need protection against CPU-hotplug here because
 	 * this is called early in boot, before either interrupts
 	 * or the scheduler are operational.
 	 */
 	cpu_notifier(rcu_cpu_notify, 0);
 	for_each_online_cpu(cpu)
 		rcu_cpu_notify(NULL, CPU_UP_PREPARE, (void *)(long)cpu);
 	check_cpu_stall_init();
 }
 #include "rcutree_plugin.h"