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 <asm/atomic.h>

39

#include <asm/atomic.h>

40

#include <linux/bitops.h>

40

#include <linux/bitops.h>

41

#include <linux/module.h>

41

#include <linux/module.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

50

#include "rcutree.h"

50

#include "rcutree.h"

51

52

/* Data structures. */

52

/* Data structures. */

53

54

static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];

54

static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];

55

56

#define RCU_STATE_INITIALIZER(name) { \

56

#define RCU_STATE_INITIALIZER(name) { \

57

.level = { &name.node[0] }, \

57

.level = { &name.node[0] }, \

58

.levelcnt = { \

58

.levelcnt = { \

59

NUM_RCU_LVL_0, /* root of hierarchy. */ \

59

NUM_RCU_LVL_0, /* root of hierarchy. */ \

60

NUM_RCU_LVL_1, \

60

NUM_RCU_LVL_1, \

61

NUM_RCU_LVL_2, \

61

NUM_RCU_LVL_2, \

62

NUM_RCU_LVL_3, \

62

NUM_RCU_LVL_3, \

63

NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \

63

NUM_RCU_LVL_4, /* == MAX_RCU_LVLS */ \

64

}, \

64

}, \

65

.signaled = RCU_GP_IDLE, \

65

.signaled = RCU_GP_IDLE, \

66

.gpnum = -300, \

66

.gpnum = -300, \

67

.completed = -300, \

67

.completed = -300, \

68

.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&name.onofflock), \

68

.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&name.onofflock), \

69

.orphan_cbs_list = NULL, \

69

.orphan_cbs_list = NULL, \

70

.orphan_cbs_tail = &name.orphan_cbs_list, \

70

.orphan_cbs_tail = &name.orphan_cbs_list, \

71

.orphan_qlen = 0, \

71

.orphan_qlen = 0, \

72

.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&name.fqslock), \

72

.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&name.fqslock), \

73

.n_force_qs = 0, \

73

.n_force_qs = 0, \

74

.n_force_qs_ngp = 0, \

74

.n_force_qs_ngp = 0, \

75

}

75

}

76

77

struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched_state);

77

struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched_state);

78

DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);

78

DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);

79

80

struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state);

80

struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state);

81

DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);

81

DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);

82

83

/*

83

/*

84

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

84

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

85

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

85

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

86

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

86

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

87

*/

87

*/

88

static int rcu_gp_in_progress(struct rcu_state *rsp)

88

static int rcu_gp_in_progress(struct rcu_state *rsp)

89

{

89

{

90

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

90

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

91

}

91

}

92

93

/*

93

/*

94

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

94

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

95

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

95

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

96

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

96

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

97

*/

97

*/

98

void rcu_sched_qs(int cpu)

98

void rcu_sched_qs(int cpu)

99

{

99

{

100

struct rcu_data *rdp;

100

struct rcu_data *rdp;

101

102

rdp = &per_cpu(rcu_sched_data, cpu);

102

rdp = &per_cpu(rcu_sched_data, cpu);

103

rdp->passed_quiesc_completed = rdp->gpnum - 1;

103

rdp->passed_quiesc_completed = rdp->gpnum - 1;

104

barrier();

104

barrier();

105

rdp->passed_quiesc = 1;

105

rdp->passed_quiesc = 1;

106

rcu_preempt_note_context_switch(cpu);

106

rcu_preempt_note_context_switch(cpu);

107

}

107

}

108

109

void rcu_bh_qs(int cpu)

109

void rcu_bh_qs(int cpu)

110

{

110

{

111

struct rcu_data *rdp;

111

struct rcu_data *rdp;

112

113

rdp = &per_cpu(rcu_bh_data, cpu);

113

rdp = &per_cpu(rcu_bh_data, cpu);

114

rdp->passed_quiesc_completed = rdp->gpnum - 1;

114

rdp->passed_quiesc_completed = rdp->gpnum - 1;

115

barrier();

115

barrier();

116

rdp->passed_quiesc = 1;

116

rdp->passed_quiesc = 1;

117

}

117

}

118

119

#ifdef CONFIG_NO_HZ

119

#ifdef CONFIG_NO_HZ

120

DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {

120

DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {

121

.dynticks_nesting = 1,

121

.dynticks_nesting = 1,

122

.dynticks = 1,

122

.dynticks = 1,

123

};

123

};

124

#endif /* #ifdef CONFIG_NO_HZ */

124

#endif /* #ifdef CONFIG_NO_HZ */

125

126

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

126

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

127

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

127

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

128

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

128

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

129

130

module_param(blimit, int, 0);

130

module_param(blimit, int, 0);

131

module_param(qhimark, int, 0);

131

module_param(qhimark, int, 0);

132

module_param(qlowmark, int, 0);

132

module_param(qlowmark, int, 0);

133

134

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

134

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

135

static int rcu_pending(int cpu);

135

static int rcu_pending(int cpu);

136

137

/*

137

/*

138

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

138

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

139

*/

139

*/

140

long rcu_batches_completed_sched(void)

140

long rcu_batches_completed_sched(void)

141

{

141

{

142

return rcu_sched_state.completed;

142

return rcu_sched_state.completed;

143

}

143

}

144

EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);

144

EXPORT_SYMBOL_GPL(rcu_batches_completed_sched);

145

146

/*

146

/*

147

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

147

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

148

*/

148

*/

149

long rcu_batches_completed_bh(void)

149

long rcu_batches_completed_bh(void)

150

{

150

{

151

return rcu_bh_state.completed;

151

return rcu_bh_state.completed;

152

}

152

}

153

EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

153

EXPORT_SYMBOL_GPL(rcu_batches_completed_bh);

154

155

/*

155

/*

156

* Force a quiescent state for RCU BH.

156

* Force a quiescent state for RCU BH.

157

*/

157

*/

158

void rcu_bh_force_quiescent_state(void)

158

void rcu_bh_force_quiescent_state(void)

159

{

159

{

160

force_quiescent_state(&rcu_bh_state, 0);

160

force_quiescent_state(&rcu_bh_state, 0);

161

}

161

}

162

EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

162

EXPORT_SYMBOL_GPL(rcu_bh_force_quiescent_state);

163

164

/*

164

/*

165

* Force a quiescent state for RCU-sched.

165

* Force a quiescent state for RCU-sched.

166

*/

166

*/

167

void rcu_sched_force_quiescent_state(void)

167

void rcu_sched_force_quiescent_state(void)

168

{

168

{

169

force_quiescent_state(&rcu_sched_state, 0);

169

force_quiescent_state(&rcu_sched_state, 0);

170

}

170

}

171

EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

171

EXPORT_SYMBOL_GPL(rcu_sched_force_quiescent_state);

172

173

/*

173

/*

174

* Does the CPU have callbacks ready to be invoked?

174

* Does the CPU have callbacks ready to be invoked?

175

*/

175

*/

176

static int

176

static int

177

cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)

177

cpu_has_callbacks_ready_to_invoke(struct rcu_data *rdp)

178

{

178

{

179

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

179

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

180

}

180

}

181

182

/*

182

/*

183

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

183

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

184

*/

184

*/

185

static int

185

static int

186

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

186

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

187

{

187

{

188

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

188

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

189

}

189

}

190

191

/*

191

/*

192

* Return the root node of the specified rcu_state structure.

192

* Return the root node of the specified rcu_state structure.

193

*/

193

*/

194

static struct rcu_node *rcu_get_root(struct rcu_state *rsp)

194

static struct rcu_node *rcu_get_root(struct rcu_state *rsp)

195

{

195

{

196

return &rsp->node[0];

196

return &rsp->node[0];

197

}

197

}

198

199

#ifdef CONFIG_SMP

199

#ifdef CONFIG_SMP

200

201

/*

201

/*

202

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

202

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

203

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

203

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

204

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

204

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

205

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

205

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

206

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

206

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

207

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

207

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

208

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

208

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

209

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

209

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

210

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

210

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

211

*/

211

*/

212

static int rcu_implicit_offline_qs(struct rcu_data *rdp)

212

static int rcu_implicit_offline_qs(struct rcu_data *rdp)

213

{

213

{

214

/*

214

/*

215

* If the CPU is offline, it is in a quiescent state. We can

215

* If the CPU is offline, it is in a quiescent state. We can

216

* trust its state not to change because interrupts are disabled.

216

* trust its state not to change because interrupts are disabled.

217

*/

217

*/

218

if (cpu_is_offline(rdp->cpu)) {

218

if (cpu_is_offline(rdp->cpu)) {

219

rdp->offline_fqs++;

219

rdp->offline_fqs++;

220

return 1;

220

return 1;

221

}

221

}

222

223

/* If preemptable RCU, no point in sending reschedule IPI. */

223

/* If preemptable RCU, no point in sending reschedule IPI. */

224

if (rdp->preemptable)

224

if (rdp->preemptable)

225

return 0;

225

return 0;

226

227

/* The CPU is online, so send it a reschedule IPI. */

227

/* The CPU is online, so send it a reschedule IPI. */

228

if (rdp->cpu != smp_processor_id())

228

if (rdp->cpu != smp_processor_id())

229

smp_send_reschedule(rdp->cpu);

229

smp_send_reschedule(rdp->cpu);

230

else

230

else

231

set_need_resched();

231

set_need_resched();

232

rdp->resched_ipi++;

232

rdp->resched_ipi++;

233

return 0;

233

return 0;

234

}

234

}

235

236

#endif /* #ifdef CONFIG_SMP */

236

#endif /* #ifdef CONFIG_SMP */

237

238

#ifdef CONFIG_NO_HZ

238

#ifdef CONFIG_NO_HZ

239

240

/**

240

/**

241

* rcu_enter_nohz - inform RCU that current CPU is entering nohz

241

* rcu_enter_nohz - inform RCU that current CPU is entering nohz

242

*

242

*

243

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

243

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

244

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

244

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

245

* critical sections can occur in irq handlers in nohz mode, a possibility

245

* critical sections can occur in irq handlers in nohz mode, a possibility

246

* handled by rcu_irq_enter() and rcu_irq_exit()).

246

* handled by rcu_irq_enter() and rcu_irq_exit()).

247

*/

247

*/

248

void rcu_enter_nohz(void)

248

void rcu_enter_nohz(void)

249

{

249

{

250

unsigned long flags;

250

unsigned long flags;

251

struct rcu_dynticks *rdtp;

251

struct rcu_dynticks *rdtp;

252

253

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

253

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

254

local_irq_save(flags);

254

local_irq_save(flags);

255

rdtp = &__get_cpu_var(rcu_dynticks);

255

rdtp = &__get_cpu_var(rcu_dynticks);

256

rdtp->dynticks++;

256

rdtp->dynticks++;

257

rdtp->dynticks_nesting--;

257

rdtp->dynticks_nesting--;

258

WARN_ON_ONCE(rdtp->dynticks & 0x1);

258

WARN_ON_ONCE(rdtp->dynticks & 0x1);

259

local_irq_restore(flags);

259

local_irq_restore(flags);

260

}

260

}

261

262

/*

262

/*

263

* rcu_exit_nohz - inform RCU that current CPU is leaving nohz

263

* rcu_exit_nohz - inform RCU that current CPU is leaving nohz

264

*

264

*

265

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

265

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

266

* read-side critical sections normally occur.

266

* read-side critical sections normally occur.

267

*/

267

*/

268

void rcu_exit_nohz(void)

268

void rcu_exit_nohz(void)

269

{

269

{

270

unsigned long flags;

270

unsigned long flags;

271

struct rcu_dynticks *rdtp;

271

struct rcu_dynticks *rdtp;

272

273

local_irq_save(flags);

273

local_irq_save(flags);

274

rdtp = &__get_cpu_var(rcu_dynticks);

274

rdtp = &__get_cpu_var(rcu_dynticks);

275

rdtp->dynticks++;

275

rdtp->dynticks++;

276

rdtp->dynticks_nesting++;

276

rdtp->dynticks_nesting++;

277

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

277

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

278

local_irq_restore(flags);

278

local_irq_restore(flags);

279

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

279

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

280

}

280

}

281

282

/**

282

/**

283

* rcu_nmi_enter - inform RCU of entry to NMI context

283

* rcu_nmi_enter - inform RCU of entry to NMI context

284

*

284

*

285

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

285

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

286

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

286

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

287

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

287

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

288

*/

288

*/

289

void rcu_nmi_enter(void)

289

void rcu_nmi_enter(void)

290

{

290

{

291

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

291

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

292

293

if (rdtp->dynticks & 0x1)

293

if (rdtp->dynticks & 0x1)

294

return;

294

return;

295

rdtp->dynticks_nmi++;

295

rdtp->dynticks_nmi++;

296

WARN_ON_ONCE(!(rdtp->dynticks_nmi & 0x1));

296

WARN_ON_ONCE(!(rdtp->dynticks_nmi & 0x1));

297

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

297

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

298

}

298

}

299

300

/**

300

/**

301

* rcu_nmi_exit - inform RCU of exit from NMI context

301

* rcu_nmi_exit - inform RCU of exit from NMI context

302

*

302

*

303

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

303

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

304

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

304

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

305

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

305

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

306

*/

306

*/

307

void rcu_nmi_exit(void)

307

void rcu_nmi_exit(void)

308

{

308

{

309

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

309

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

310

311

if (rdtp->dynticks & 0x1)

311

if (rdtp->dynticks & 0x1)

312

return;

312

return;

313

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

313

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

314

rdtp->dynticks_nmi++;

314

rdtp->dynticks_nmi++;

315

WARN_ON_ONCE(rdtp->dynticks_nmi & 0x1);

315

WARN_ON_ONCE(rdtp->dynticks_nmi & 0x1);

316

}

316

}

317

318

/**

318

/**

319

* rcu_irq_enter - inform RCU of entry to hard irq context

319

* rcu_irq_enter - inform RCU of entry to hard irq context

320

*

320

*

321

* If the CPU was idle with dynamic ticks active, this updates the

321

* If the CPU was idle with dynamic ticks active, this updates the

322

* rdtp->dynticks to let the RCU handling know that the CPU is active.

322

* rdtp->dynticks to let the RCU handling know that the CPU is active.

323

*/

323

*/

324

void rcu_irq_enter(void)

324

void rcu_irq_enter(void)

325

{

325

{

326

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

326

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

327

328

if (rdtp->dynticks_nesting++)

328

if (rdtp->dynticks_nesting++)

329

return;

329

return;

330

rdtp->dynticks++;

330

rdtp->dynticks++;

331

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

331

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

332

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

332

smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */

333

}

333

}

334

335

/**

335

/**

336

* rcu_irq_exit - inform RCU of exit from hard irq context

336

* rcu_irq_exit - inform RCU of exit from hard irq context

337

*

337

*

338

* If the CPU was idle with dynamic ticks active, update the rdp->dynticks

338

* If the CPU was idle with dynamic ticks active, update the rdp->dynticks

339

* to put let the RCU handling be aware that the CPU is going back to idle

339

* to put let the RCU handling be aware that the CPU is going back to idle

340

* with no ticks.

340

* with no ticks.

341

*/

341

*/

342

void rcu_irq_exit(void)

342

void rcu_irq_exit(void)

343

{

343

{

344

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

344

struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);

345

346

if (--rdtp->dynticks_nesting)

346

if (--rdtp->dynticks_nesting)

347

return;

347

return;

348

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

348

smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */

349

rdtp->dynticks++;

349

rdtp->dynticks++;

350

WARN_ON_ONCE(rdtp->dynticks & 0x1);

350

WARN_ON_ONCE(rdtp->dynticks & 0x1);

351

352

/* If the interrupt queued a callback, get out of dyntick mode. */

352

/* If the interrupt queued a callback, get out of dyntick mode. */

353

if (__get_cpu_var(rcu_sched_data).nxtlist ||

353

if (__get_cpu_var(rcu_sched_data).nxtlist ||

354

__get_cpu_var(rcu_bh_data).nxtlist)

354

__get_cpu_var(rcu_bh_data).nxtlist)

355

set_need_resched();

355

set_need_resched();

356

}

356

}

357

358

#ifdef CONFIG_SMP

358

#ifdef CONFIG_SMP

359

360

/*

360

/*

361

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

361

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

362

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

362

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

363

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

363

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

364

*/

364

*/

365

static int dyntick_save_progress_counter(struct rcu_data *rdp)

365

static int dyntick_save_progress_counter(struct rcu_data *rdp)

366

{

366

{

367

int ret;

367

int ret;

368

int snap;

368

int snap;

369

int snap_nmi;

369

int snap_nmi;

370

371

snap = rdp->dynticks->dynticks;

371

snap = rdp->dynticks->dynticks;

372

snap_nmi = rdp->dynticks->dynticks_nmi;

372

snap_nmi = rdp->dynticks->dynticks_nmi;

373

smp_mb(); /* Order sampling of snap with end of grace period. */

373

smp_mb(); /* Order sampling of snap with end of grace period. */

374

rdp->dynticks_snap = snap;

374

rdp->dynticks_snap = snap;

375

rdp->dynticks_nmi_snap = snap_nmi;

375

rdp->dynticks_nmi_snap = snap_nmi;

376

ret = ((snap & 0x1) == 0) && ((snap_nmi & 0x1) == 0);

376

ret = ((snap & 0x1) == 0) && ((snap_nmi & 0x1) == 0);

377

if (ret)

377

if (ret)

378

rdp->dynticks_fqs++;

378

rdp->dynticks_fqs++;

379

return ret;

379

return ret;

380

}

380

}

381

382

/*

382

/*

383

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

383

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

384

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

384

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

385

* idle state since the last call to dyntick_save_progress_counter()

385

* idle state since the last call to dyntick_save_progress_counter()

386

* for this same CPU.

386

* for this same CPU.

387

*/

387

*/

388

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

388

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

389

{

389

{

390

long curr;

390

long curr;

391

long curr_nmi;

391

long curr_nmi;

392

long snap;

392

long snap;

393

long snap_nmi;

393

long snap_nmi;

394

395

curr = rdp->dynticks->dynticks;

395

curr = rdp->dynticks->dynticks;

396

snap = rdp->dynticks_snap;

396

snap = rdp->dynticks_snap;

397

curr_nmi = rdp->dynticks->dynticks_nmi;

397

curr_nmi = rdp->dynticks->dynticks_nmi;

398

snap_nmi = rdp->dynticks_nmi_snap;

398

snap_nmi = rdp->dynticks_nmi_snap;

399

smp_mb(); /* force ordering with cpu entering/leaving dynticks. */

399

smp_mb(); /* force ordering with cpu entering/leaving dynticks. */

400

401

/*

401

/*

402

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

402

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

403

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

403

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

404

* already acknowledged the request to pass through a quiescent

404

* already acknowledged the request to pass through a quiescent

405

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

405

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

406

* read-side critical section that started before the beginning

406

* read-side critical section that started before the beginning

407

* of the current RCU grace period.

407

* of the current RCU grace period.

408

*/

408

*/

409

if ((curr != snap || (curr & 0x1) == 0) &&

409

if ((curr != snap || (curr & 0x1) == 0) &&

410

(curr_nmi != snap_nmi || (curr_nmi & 0x1) == 0)) {

410

(curr_nmi != snap_nmi || (curr_nmi & 0x1) == 0)) {

411

rdp->dynticks_fqs++;

411

rdp->dynticks_fqs++;

412

return 1;

412

return 1;

413

}

413

}

414

415

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

415

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

416

return rcu_implicit_offline_qs(rdp);

416

return rcu_implicit_offline_qs(rdp);

417

}

417

}

418

419

#endif /* #ifdef CONFIG_SMP */

419

#endif /* #ifdef CONFIG_SMP */

420

421

#else /* #ifdef CONFIG_NO_HZ */

421

#else /* #ifdef CONFIG_NO_HZ */

422

423

#ifdef CONFIG_SMP

423

#ifdef CONFIG_SMP

424

425

static int dyntick_save_progress_counter(struct rcu_data *rdp)

425

static int dyntick_save_progress_counter(struct rcu_data *rdp)

426

{

426

{

427

return 0;

427

return 0;

428

}

428

}

429

430

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

430

static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)

431

{

431

{

432

return rcu_implicit_offline_qs(rdp);

432

return rcu_implicit_offline_qs(rdp);

433

}

433

}

434

435

#endif /* #ifdef CONFIG_SMP */

435

#endif /* #ifdef CONFIG_SMP */

436

437

#endif /* #else #ifdef CONFIG_NO_HZ */

437

#endif /* #else #ifdef CONFIG_NO_HZ */

438

439

#ifdef CONFIG_RCU_CPU_STALL_DETECTOR

439

#ifdef CONFIG_RCU_CPU_STALL_DETECTOR

440

441

static void record_gp_stall_check_time(struct rcu_state *rsp)

441

static void record_gp_stall_check_time(struct rcu_state *rsp)

442

{

442

{

443

rsp->gp_start = jiffies;

443

rsp->gp_start = jiffies;

444

rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_CHECK;

444

rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_CHECK;

445

}

445

}

446

447

static void print_other_cpu_stall(struct rcu_state *rsp)

447

static void print_other_cpu_stall(struct rcu_state *rsp)

448

{

448

{

449

int cpu;

449

int cpu;

450

long delta;

450

long delta;

451

unsigned long flags;

451

unsigned long flags;

452

struct rcu_node *rnp = rcu_get_root(rsp);

452

struct rcu_node *rnp = rcu_get_root(rsp);

453

454

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

454

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

455

456

raw_spin_lock_irqsave(&rnp->lock, flags);

456

raw_spin_lock_irqsave(&rnp->lock, flags);

457

delta = jiffies - rsp->jiffies_stall;

457

delta = jiffies - rsp->jiffies_stall;

458

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

458

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

459

raw_spin_unlock_irqrestore(&rnp->lock, flags);

459

raw_spin_unlock_irqrestore(&rnp->lock, flags);

460

return;

460

return;

461

}

461

}

462

rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_RECHECK;

462

rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_RECHECK;

463

464

/*

464

/*

465

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

465

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

466

* due to CPU offlining.

466

* due to CPU offlining.

467

*/

467

*/

468

rcu_print_task_stall(rnp);

468

rcu_print_task_stall(rnp);

469

raw_spin_unlock_irqrestore(&rnp->lock, flags);

469

raw_spin_unlock_irqrestore(&rnp->lock, flags);

470

471

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

471

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

472

473

printk(KERN_ERR "INFO: RCU detected CPU stalls:");

473

printk(KERN_ERR "INFO: RCU detected CPU stalls:");

474

rcu_for_each_leaf_node(rsp, rnp) {

474

rcu_for_each_leaf_node(rsp, rnp) {

475

raw_spin_lock_irqsave(&rnp->lock, flags);

475

raw_spin_lock_irqsave(&rnp->lock, flags);

476

rcu_print_task_stall(rnp);

476

rcu_print_task_stall(rnp);

477

raw_spin_unlock_irqrestore(&rnp->lock, flags);

477

raw_spin_unlock_irqrestore(&rnp->lock, flags);

478

if (rnp->qsmask == 0)

478

if (rnp->qsmask == 0)

479

continue;

479

continue;

480

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

480

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

481

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

481

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

482

printk(" %d", rnp->grplo + cpu);

482

printk(" %d", rnp->grplo + cpu);

483

}

483

}

484

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

484

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

485

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

485

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

486

trigger_all_cpu_backtrace();

486

trigger_all_cpu_backtrace();

487

488

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

488

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

489

490

rcu_print_detail_task_stall(rsp);

490

rcu_print_detail_task_stall(rsp);

491

492

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

492

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

493

}

493

}

494

495

static void print_cpu_stall(struct rcu_state *rsp)

495

static void print_cpu_stall(struct rcu_state *rsp)

496

{

496

{

497

unsigned long flags;

497

unsigned long flags;

498

struct rcu_node *rnp = rcu_get_root(rsp);

498

struct rcu_node *rnp = rcu_get_root(rsp);

499

500

printk(KERN_ERR "INFO: RCU detected CPU %d stall (t=%lu jiffies)\n",

500

printk(KERN_ERR "INFO: RCU detected CPU %d stall (t=%lu jiffies)\n",

501

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

501

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

502

trigger_all_cpu_backtrace();

502

trigger_all_cpu_backtrace();

503

504

raw_spin_lock_irqsave(&rnp->lock, flags);

504

raw_spin_lock_irqsave(&rnp->lock, flags);

505

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

505

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

506

rsp->jiffies_stall =

506

rsp->jiffies_stall =

507

jiffies + RCU_SECONDS_TILL_STALL_RECHECK;

507

jiffies + RCU_SECONDS_TILL_STALL_RECHECK;

508

raw_spin_unlock_irqrestore(&rnp->lock, flags);

508

raw_spin_unlock_irqrestore(&rnp->lock, flags);

509

510

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

510

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

511

}

511

}

512

513

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

513

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

514

{

514

{

515

long delta;

515

long delta;

516

struct rcu_node *rnp;

516

struct rcu_node *rnp;

517

518

delta = jiffies - rsp->jiffies_stall;

518

delta = jiffies - rsp->jiffies_stall;

519

rnp = rdp->mynode;

519

rnp = rdp->mynode;

520

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

520

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

521

522

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

522

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

523

print_cpu_stall(rsp);

523

print_cpu_stall(rsp);

524

525

} else if (rcu_gp_in_progress(rsp) && delta >= RCU_STALL_RAT_DELAY) {

525

} else if (rcu_gp_in_progress(rsp) && delta >= RCU_STALL_RAT_DELAY) {

526

527

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

527

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

528

print_other_cpu_stall(rsp);

528

print_other_cpu_stall(rsp);

529

}

529

}

530

}

530

}

531

532

#else /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

532

#else /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

533

534

static void record_gp_stall_check_time(struct rcu_state *rsp)

534

static void record_gp_stall_check_time(struct rcu_state *rsp)

535

{

535

{

536

}

536

}

537

538

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

538

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

539

{

539

{

540

}

540

}

541

542

#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

542

#endif /* #else #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

543

544

/*

544

/*

545

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

545

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

546

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

546

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

547

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

547

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

548

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

548

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

549

* and must have irqs disabled.

549

* and must have irqs disabled.

550

*/

550

*/

551

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

551

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

552

{

552

{

553

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

553

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

554

rdp->qs_pending = 1;

554

rdp->qs_pending = 1;

555

rdp->passed_quiesc = 0;

555

rdp->passed_quiesc = 0;

556

rdp->gpnum = rnp->gpnum;

556

rdp->gpnum = rnp->gpnum;

557

}

557

}

558

}

558

}

559

560

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

560

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

561

{

561

{

562

unsigned long flags;

562

unsigned long flags;

563

struct rcu_node *rnp;

563

struct rcu_node *rnp;

564

565

local_irq_save(flags);

565

local_irq_save(flags);

566

rnp = rdp->mynode;

566

rnp = rdp->mynode;

567

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

567

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

568

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

568

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

569

local_irq_restore(flags);

569

local_irq_restore(flags);

570

return;

570

return;

571

}

571

}

572

__note_new_gpnum(rsp, rnp, rdp);

572

__note_new_gpnum(rsp, rnp, rdp);

573

raw_spin_unlock_irqrestore(&rnp->lock, flags);

573

raw_spin_unlock_irqrestore(&rnp->lock, flags);

574

}

574

}

575

576

/*

576

/*

577

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

577

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

578

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

578

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

579

* on the CPU corresponding to rdp.

579

* on the CPU corresponding to rdp.

580

*/

580

*/

581

static int

581

static int

582

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

582

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

583

{

583

{

584

unsigned long flags;

584

unsigned long flags;

585

int ret = 0;

585

int ret = 0;

586

587

local_irq_save(flags);

587

local_irq_save(flags);

588

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

588

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

589

note_new_gpnum(rsp, rdp);

589

note_new_gpnum(rsp, rdp);

590

ret = 1;

590

ret = 1;

591

}

591

}

592

local_irq_restore(flags);

592

local_irq_restore(flags);

593

return ret;

593

return ret;

594

}

594

}

595

596

/*

596

/*

597

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

597

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

598

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

598

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

599

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

599

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

600

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

600

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

601

*/

601

*/

602

static void

602

static void

603

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

603

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

604

{

604

{

605

/* Did another grace period end? */

605

/* Did another grace period end? */

606

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

606

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

607

608

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

608

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

609

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

609

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

610

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

610

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

611

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

611

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

612

613

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

613

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

614

rdp->completed = rnp->completed;

614

rdp->completed = rnp->completed;

615

}

615

}

616

}

616

}

617

618

/*

618

/*

619

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

619

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

620

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

620

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

621

* belongs.

621

* belongs.

622

*/

622

*/

623

static void

623

static void

624

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

624

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

625

{

625

{

626

unsigned long flags;

626

unsigned long flags;

627

struct rcu_node *rnp;

627

struct rcu_node *rnp;

628

629

local_irq_save(flags);

629

local_irq_save(flags);

630

rnp = rdp->mynode;

630

rnp = rdp->mynode;

631

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

631

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

632

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

632

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

633

local_irq_restore(flags);

633

local_irq_restore(flags);

634

return;

634

return;

635

}

635

}

636

__rcu_process_gp_end(rsp, rnp, rdp);

636

__rcu_process_gp_end(rsp, rnp, rdp);

637

raw_spin_unlock_irqrestore(&rnp->lock, flags);

637

raw_spin_unlock_irqrestore(&rnp->lock, flags);

638

}

638

}

639

640

/*

640

/*

641

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

641

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

642

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

642

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

643

* this CPU.

643

* this CPU.

644

*/

644

*/

645

static void

645

static void

646

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

646

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

647

{

647

{

648

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

648

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

649

__rcu_process_gp_end(rsp, rnp, rdp);

649

__rcu_process_gp_end(rsp, rnp, rdp);

650

651

/*

651

/*

652

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

652

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

653

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

653

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

654

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

654

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

655

* Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.

655

* Therefore, advance all outstanding callbacks to RCU_WAIT_TAIL.

656

*

656

*

657

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

657

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

658

* Therefore, their recently registered callbacks must pass through

658

* Therefore, their recently registered callbacks must pass through

659

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

659

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

660

* by the next RCU grace period.

660

* by the next RCU grace period.

661

*/

661

*/

662

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

662

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

663

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

663

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

664

665

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

665

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

666

__note_new_gpnum(rsp, rnp, rdp);

666

__note_new_gpnum(rsp, rnp, rdp);

667

}

667

}

668

669

/*

669

/*

670

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

670

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

671

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

671

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

672

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

672

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

673

* be disabled.

673

* be disabled.

674

*/

674

*/

675

static void

675

static void

676

rcu_start_gp(struct rcu_state *rsp, unsigned long flags)

676

rcu_start_gp(struct rcu_state *rsp, unsigned long flags)

677

__releases(rcu_get_root(rsp)->lock)

677

__releases(rcu_get_root(rsp)->lock)

678

{

678

{

679

struct rcu_data *rdp = rsp->rda[smp_processor_id()];

679

struct rcu_data *rdp = rsp->rda[smp_processor_id()];

680

struct rcu_node *rnp = rcu_get_root(rsp);

680

struct rcu_node *rnp = rcu_get_root(rsp);

681

682

if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {

682

if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {

683

if (cpu_needs_another_gp(rsp, rdp))

683

if (cpu_needs_another_gp(rsp, rdp))

684

rsp->fqs_need_gp = 1;

684

rsp->fqs_need_gp = 1;

685

if (rnp->completed == rsp->completed) {

685

if (rnp->completed == rsp->completed) {

686

raw_spin_unlock_irqrestore(&rnp->lock, flags);

686

raw_spin_unlock_irqrestore(&rnp->lock, flags);

687

return;

687

return;

688

}

688

}

689

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

689

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

690

691

/*

691

/*

692

* Propagate new ->completed value to rcu_node structures

692

* Propagate new ->completed value to rcu_node structures

693

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

693

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

694

* of the next grace period to process their callbacks.

694

* of the next grace period to process their callbacks.

695

*/

695

*/

696

rcu_for_each_node_breadth_first(rsp, rnp) {

696

rcu_for_each_node_breadth_first(rsp, rnp) {

697

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

697

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

698

rnp->completed = rsp->completed;

698

rnp->completed = rsp->completed;

699

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

699

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

700

}

700

}

701

local_irq_restore(flags);

701

local_irq_restore(flags);

702

return;

702

return;

703

}

703

}

704

705

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

705

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

706

rsp->gpnum++;

706

rsp->gpnum++;

707

WARN_ON_ONCE(rsp->signaled == RCU_GP_INIT);

707

WARN_ON_ONCE(rsp->signaled == RCU_GP_INIT);

708

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

708

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

709

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

709

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

710

record_gp_stall_check_time(rsp);

710

record_gp_stall_check_time(rsp);

711

712

/* Special-case the common single-level case. */

712

/* Special-case the common single-level case. */

713

if (NUM_RCU_NODES == 1) {

713

if (NUM_RCU_NODES == 1) {

714

rcu_preempt_check_blocked_tasks(rnp);

714

rcu_preempt_check_blocked_tasks(rnp);

715

rnp->qsmask = rnp->qsmaskinit;

715

rnp->qsmask = rnp->qsmaskinit;

716

rnp->gpnum = rsp->gpnum;

716

rnp->gpnum = rsp->gpnum;

717

rnp->completed = rsp->completed;

717

rnp->completed = rsp->completed;

718

rsp->signaled = RCU_SIGNAL_INIT; /* force_quiescent_state OK. */

718

rsp->signaled = RCU_SIGNAL_INIT; /* force_quiescent_state OK. */

719

rcu_start_gp_per_cpu(rsp, rnp, rdp);

719

rcu_start_gp_per_cpu(rsp, rnp, rdp);

720

raw_spin_unlock_irqrestore(&rnp->lock, flags);

720

raw_spin_unlock_irqrestore(&rnp->lock, flags);

721

return;

721

return;

722

}

722

}

723

724

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

724

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

725

726

727

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

727

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

728

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

728

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

729

730

/*

730

/*

731

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

731

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

732

* structures for all currently online CPUs in breadth-first

732

* structures for all currently online CPUs in breadth-first

733

* order, starting from the root rcu_node structure. This

733

* order, starting from the root rcu_node structure. This

734

* operation relies on the layout of the hierarchy within the

734

* operation relies on the layout of the hierarchy within the

735

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

735

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

736

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

736

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

737

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

737

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

738

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

738

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

739

* CPU-hotplug operations.

739

* CPU-hotplug operations.

740

*

740

*

741

* Note that the grace period cannot complete until we finish

741

* Note that the grace period cannot complete until we finish

742

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

742

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

743

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

743

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

744

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

744

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

745

* irqs disabled.

745

* irqs disabled.

746

*/

746

*/

747

rcu_for_each_node_breadth_first(rsp, rnp) {

747

rcu_for_each_node_breadth_first(rsp, rnp) {

748

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

748

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

749

rcu_preempt_check_blocked_tasks(rnp);

749

rcu_preempt_check_blocked_tasks(rnp);

750

rnp->qsmask = rnp->qsmaskinit;

750

rnp->qsmask = rnp->qsmaskinit;

751

rnp->gpnum = rsp->gpnum;

751

rnp->gpnum = rsp->gpnum;

752

rnp->completed = rsp->completed;

752

rnp->completed = rsp->completed;

753

if (rnp == rdp->mynode)

753

if (rnp == rdp->mynode)

754

rcu_start_gp_per_cpu(rsp, rnp, rdp);

754

rcu_start_gp_per_cpu(rsp, rnp, rdp);

755

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

755

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

756

}

756

}

757

758

rnp = rcu_get_root(rsp);

758

rnp = rcu_get_root(rsp);

759

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

759

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

760

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

760

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

761

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

761

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

762

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

762

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

763

}

763

}

764

765

/*

765

/*

766

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

766

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

767

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

767

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

768

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

768

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

769

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

769

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

770

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

770

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

771

*/

771

*/

772

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

772

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

773

__releases(rcu_get_root(rsp)->lock)

773

__releases(rcu_get_root(rsp)->lock)

774

{

774

{

775

WARN_ON_ONCE(!rcu_gp_in_progress(rsp));

775

WARN_ON_ONCE(!rcu_gp_in_progress(rsp));

776

rsp->completed = rsp->gpnum;

776

rsp->completed = rsp->gpnum;

777

rsp->signaled = RCU_GP_IDLE;

777

rsp->signaled = RCU_GP_IDLE;

778

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

778

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

779

}

779

}

780

781

/*

781

/*

782

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

782

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

783

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

783

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

784

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

784

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

785

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

785

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

786

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

786

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

787

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

787

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

788

*/

788

*/

789

static void

789

static void

790

rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,

790

rcu_report_qs_rnp(unsigned long mask, struct rcu_state *rsp,

791

struct rcu_node *rnp, unsigned long flags)

791

struct rcu_node *rnp, unsigned long flags)

792

__releases(rnp->lock)

792

__releases(rnp->lock)

793

{

793

{

794

struct rcu_node *rnp_c;

794

struct rcu_node *rnp_c;

795

796

/* Walk up the rcu_node hierarchy. */

796

/* Walk up the rcu_node hierarchy. */

797

for (;;) {

797

for (;;) {

798

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

798

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

799

800

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

800

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

801

raw_spin_unlock_irqrestore(&rnp->lock, flags);

801

raw_spin_unlock_irqrestore(&rnp->lock, flags);

802

return;

802

return;

803

}

803

}

804

rnp->qsmask &= ~mask;

804

rnp->qsmask &= ~mask;

805

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

805

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

806

807

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

807

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

808

raw_spin_unlock_irqrestore(&rnp->lock, flags);

808

raw_spin_unlock_irqrestore(&rnp->lock, flags);

809

return;

809

return;

810

}

810

}

811

mask = rnp->grpmask;

811

mask = rnp->grpmask;

812

if (rnp->parent == NULL) {

812

if (rnp->parent == NULL) {

813

814

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

814

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

815

816

break;

816

break;

817

}

817

}

818

raw_spin_unlock_irqrestore(&rnp->lock, flags);

818

raw_spin_unlock_irqrestore(&rnp->lock, flags);

819

rnp_c = rnp;

819

rnp_c = rnp;

820

rnp = rnp->parent;

820

rnp = rnp->parent;

821

raw_spin_lock_irqsave(&rnp->lock, flags);

821

raw_spin_lock_irqsave(&rnp->lock, flags);

822

WARN_ON_ONCE(rnp_c->qsmask);

822

WARN_ON_ONCE(rnp_c->qsmask);

823

}

823

}

824

825

/*

825

/*

826

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

826

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

827

* state for this grace period. Invoke rcu_report_qs_rsp()

827

* state for this grace period. Invoke rcu_report_qs_rsp()

828

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

828

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

829

*/

829

*/

830

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

830

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

831

}

831

}

832

833

/*

833

/*

834

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

834

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

835

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

835

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

836

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

836

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

837

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

837

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

838

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

838

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

839

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

839

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

840

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

840

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

841

*/

841

*/

842

static void

842

static void

843

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

843

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

844

{

844

{

845

unsigned long flags;

845

unsigned long flags;

846

unsigned long mask;

846

unsigned long mask;

847

struct rcu_node *rnp;

847

struct rcu_node *rnp;

848

849

rnp = rdp->mynode;

849

rnp = rdp->mynode;

850

raw_spin_lock_irqsave(&rnp->lock, flags);

850

raw_spin_lock_irqsave(&rnp->lock, flags);

851

if (lastcomp != rnp->completed) {

851

if (lastcomp != rnp->completed) {

852

853

/*

853

/*

854

* Someone beat us to it for this grace period, so leave.

854

* Someone beat us to it for this grace period, so leave.

855

* The race with GP start is resolved by the fact that we

855

* The race with GP start is resolved by the fact that we

856

* hold the leaf rcu_node lock, so that the per-CPU bits

856

* hold the leaf rcu_node lock, so that the per-CPU bits

857

* cannot yet be initialized -- so we would simply find our

857

* cannot yet be initialized -- so we would simply find our

858

* CPU's bit already cleared in rcu_report_qs_rnp() if this

858

* CPU's bit already cleared in rcu_report_qs_rnp() if this

859

* race occurred.

859

* race occurred.

860

*/

860

*/

861

rdp->passed_quiesc = 0; /* try again later! */

861

rdp->passed_quiesc = 0; /* try again later! */

862

raw_spin_unlock_irqrestore(&rnp->lock, flags);

862

raw_spin_unlock_irqrestore(&rnp->lock, flags);

863

return;

863

return;

864

}

864

}

865

mask = rdp->grpmask;

865

mask = rdp->grpmask;

866

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

866

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

867

raw_spin_unlock_irqrestore(&rnp->lock, flags);

867

raw_spin_unlock_irqrestore(&rnp->lock, flags);

868

} else {

868

} else {

869

rdp->qs_pending = 0;

869

rdp->qs_pending = 0;

870

871

/*

871

/*

872

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

872

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

873

* callbacks can be processed during the next GP.

873

* callbacks can be processed during the next GP.

874

*/

874

*/

875

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

875

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

876

877

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

877

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

878

}

878

}

879

}

879

}

880

881

/*

881

/*

882

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

882

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

883

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

883

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

884

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

884

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

885

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

885

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

886

*/

886

*/

887

static void

887

static void

888

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

888

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

889

{

889

{

890

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

890

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

891

if (check_for_new_grace_period(rsp, rdp))

891

if (check_for_new_grace_period(rsp, rdp))

892

return;

892

return;

893

894

/*

894

/*

895

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

895

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

896

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

896

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

897

*/

897

*/

898

if (!rdp->qs_pending)

898

if (!rdp->qs_pending)

899

return;

899

return;

900

901

/*

901

/*

902

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

902

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

903

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

903

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

904

*/

904

*/

905

if (!rdp->passed_quiesc)

905

if (!rdp->passed_quiesc)

906

return;

906

return;

907

908

/*

908

/*

909

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

909

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

910

* judge of that).

910

* judge of that).

911

*/

911

*/

912

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

912

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

913

}

913

}

914

915

#ifdef CONFIG_HOTPLUG_CPU

915

#ifdef CONFIG_HOTPLUG_CPU

916

917

/*

917

/*

918

* Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the

918

* Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the

919

* specified flavor of RCU. The callbacks will be adopted by the next

919

* specified flavor of RCU. The callbacks will be adopted by the next

920

* _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever

920

* _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever

921

* comes first. Because this is invoked from the CPU_DYING notifier,

921

* comes first. Because this is invoked from the CPU_DYING notifier,

922

* irqs are already disabled.

922

* irqs are already disabled.

923

*/

923

*/

924

static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)

924

static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)

925

{

925

{

926

int i;

926

int i;

927

struct rcu_data *rdp = rsp->rda[smp_processor_id()];

927

struct rcu_data *rdp = rsp->rda[smp_processor_id()];

928

929

if (rdp->nxtlist == NULL)

929

if (rdp->nxtlist == NULL)

930

return; /* irqs disabled, so comparison is stable. */

930

return; /* irqs disabled, so comparison is stable. */

931

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

931

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

932

*rsp->orphan_cbs_tail = rdp->nxtlist;

932

*rsp->orphan_cbs_tail = rdp->nxtlist;

933

rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL];

933

rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL];

934

rdp->nxtlist = NULL;

934

rdp->nxtlist = NULL;

935

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

935

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

936

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

936

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

937

rsp->orphan_qlen += rdp->qlen;

937

rsp->orphan_qlen += rdp->qlen;

938

rdp->qlen = 0;

938

rdp->qlen = 0;

939

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

939

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

940

}

940

}

941

942

/*

942

/*

943

* Adopt previously orphaned RCU callbacks.

943

* Adopt previously orphaned RCU callbacks.

944

*/

944

*/

945

static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)

945

static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)

946

{

946

{

947

unsigned long flags;

947

unsigned long flags;

948

struct rcu_data *rdp;

948

struct rcu_data *rdp;

949

950

raw_spin_lock_irqsave(&rsp->onofflock, flags);

950

raw_spin_lock_irqsave(&rsp->onofflock, flags);

951

rdp = rsp->rda[smp_processor_id()];

951

rdp = rsp->rda[smp_processor_id()];

952

if (rsp->orphan_cbs_list == NULL) {

952

if (rsp->orphan_cbs_list == NULL) {

953

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

953

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

954

return;

954

return;

955

}

955

}

956

*rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list;

956

*rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list;

957

rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail;

957

rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail;

958

rdp->qlen += rsp->orphan_qlen;

958

rdp->qlen += rsp->orphan_qlen;

959

rsp->orphan_cbs_list = NULL;

959

rsp->orphan_cbs_list = NULL;

960

rsp->orphan_cbs_tail = &rsp->orphan_cbs_list;

960

rsp->orphan_cbs_tail = &rsp->orphan_cbs_list;

961

rsp->orphan_qlen = 0;

961

rsp->orphan_qlen = 0;

962

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

962

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

963

}

963

}

964

965

/*

965

/*

966

* Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy

966

* Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy

967

* and move all callbacks from the outgoing CPU to the current one.

967

* and move all callbacks from the outgoing CPU to the current one.

968

*/

968

*/

969

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

969

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

970

{

970

{

971

unsigned long flags;

971

unsigned long flags;

972

unsigned long mask;

972

unsigned long mask;

973

int need_report = 0;

973

int need_report = 0;

974

struct rcu_data *rdp = rsp->rda[cpu];

974

struct rcu_data *rdp = rsp->rda[cpu];

975

struct rcu_node *rnp;

975

struct rcu_node *rnp;

976

977

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

977

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

978

raw_spin_lock_irqsave(&rsp->onofflock, flags);

978

raw_spin_lock_irqsave(&rsp->onofflock, flags);

979

980

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

980

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

981

rnp = rdp->mynode; /* this is the outgoing CPU's rnp. */

981

rnp = rdp->mynode; /* this is the outgoing CPU's rnp. */

982

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

982

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

983

do {

983

do {

984

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

984

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

985

rnp->qsmaskinit &= ~mask;

985

rnp->qsmaskinit &= ~mask;

986

if (rnp->qsmaskinit != 0) {

986

if (rnp->qsmaskinit != 0) {

987

if (rnp != rdp->mynode)

987

if (rnp != rdp->mynode)

988

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

988

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

989

break;

989

break;

990

}

990

}

991

if (rnp == rdp->mynode)

991

if (rnp == rdp->mynode)

992

need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);

992

need_report = rcu_preempt_offline_tasks(rsp, rnp, rdp);

993

else

993

else

994

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

994

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

995

mask = rnp->grpmask;

995

mask = rnp->grpmask;

996

rnp = rnp->parent;

996

rnp = rnp->parent;

997

} while (rnp != NULL);

997

} while (rnp != NULL);

998

999

/*

999

/*

1000

* We still hold the leaf rcu_node structure lock here, and

1000

* We still hold the leaf rcu_node structure lock here, and

1001

* irqs are still disabled. The reason for this subterfuge is

1001

* irqs are still disabled. The reason for this subterfuge is

1002

* because invoking rcu_report_unblock_qs_rnp() with ->onofflock

1002

* because invoking rcu_report_unblock_qs_rnp() with ->onofflock

1003

* held leads to deadlock.

1003

* held leads to deadlock.

1004

*/

1004

*/

1005

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

1005

raw_spin_unlock(&rsp->onofflock); /* irqs remain disabled. */

1006

rnp = rdp->mynode;

1006

rnp = rdp->mynode;

1007

if (need_report & RCU_OFL_TASKS_NORM_GP)

1007

if (need_report & RCU_OFL_TASKS_NORM_GP)

1008

rcu_report_unblock_qs_rnp(rnp, flags);

1008

rcu_report_unblock_qs_rnp(rnp, flags);

1009

else

1009

else

1010

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1010

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1011

if (need_report & RCU_OFL_TASKS_EXP_GP)

1011

if (need_report & RCU_OFL_TASKS_EXP_GP)

1012

rcu_report_exp_rnp(rsp, rnp);

1012

rcu_report_exp_rnp(rsp, rnp);

1013

1014

rcu_adopt_orphan_cbs(rsp);

1014

rcu_adopt_orphan_cbs(rsp);

1015

}

1015

}

1016

1017

/*

1017

/*

1018

* Remove the specified CPU from the RCU hierarchy and move any pending

1018

* Remove the specified CPU from the RCU hierarchy and move any pending

1019

* callbacks that it might have to the current CPU. This code assumes

1019

* callbacks that it might have to the current CPU. This code assumes

1020

* that at least one CPU in the system will remain running at all times.

1020

* that at least one CPU in the system will remain running at all times.

1021

* Any attempt to offline -all- CPUs is likely to strand RCU callbacks.

1021

* Any attempt to offline -all- CPUs is likely to strand RCU callbacks.

1022

*/

1022

*/

1023

static void rcu_offline_cpu(int cpu)

1023

static void rcu_offline_cpu(int cpu)

1024

{

1024

{

1025

__rcu_offline_cpu(cpu, &rcu_sched_state);

1025

__rcu_offline_cpu(cpu, &rcu_sched_state);

1026

__rcu_offline_cpu(cpu, &rcu_bh_state);

1026

__rcu_offline_cpu(cpu, &rcu_bh_state);

1027

rcu_preempt_offline_cpu(cpu);

1027

rcu_preempt_offline_cpu(cpu);

1028

}

1028

}

1029

1030

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1030

#else /* #ifdef CONFIG_HOTPLUG_CPU */

1031

1032

static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)

1032

static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)

1033

{

1033

{

1034

}

1034

}

1035

1036

static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)

1036

static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)

1037

{

1037

{

1038

}

1038

}

1039

1040

static void rcu_offline_cpu(int cpu)

1040

static void rcu_offline_cpu(int cpu)

1041

{

1041

{

1042

}

1042

}

1043

1044

#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */

1044

#endif /* #else #ifdef CONFIG_HOTPLUG_CPU */

1045

1046

/*

1046

/*

1047

* Invoke any RCU callbacks that have made it to the end of their grace

1047

* Invoke any RCU callbacks that have made it to the end of their grace

1048

* period. Thottle as specified by rdp->blimit.

1048

* period. Thottle as specified by rdp->blimit.

1049

*/

1049

*/

1050

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

1050

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

1051

{

1051

{

1052

unsigned long flags;

1052

unsigned long flags;

1053

struct rcu_head *next, *list, **tail;

1053

struct rcu_head *next, *list, **tail;

1054

int count;

1054

int count;

1055

1056

/* If no callbacks are ready, just return.*/

1056

/* If no callbacks are ready, just return.*/

1057

if (!cpu_has_callbacks_ready_to_invoke(rdp))

1057

if (!cpu_has_callbacks_ready_to_invoke(rdp))

1058

return;

1058

return;

1059

1060

/*

1060

/*

1061

* Extract the list of ready callbacks, disabling to prevent

1061

* Extract the list of ready callbacks, disabling to prevent

1062

* races with call_rcu() from interrupt handlers.

1062

* races with call_rcu() from interrupt handlers.

1063

*/

1063

*/

1064

local_irq_save(flags);

1064

local_irq_save(flags);

1065

list = rdp->nxtlist;

1065

list = rdp->nxtlist;

1066

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

1066

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

1067

*rdp->nxttail[RCU_DONE_TAIL] = NULL;

1067

*rdp->nxttail[RCU_DONE_TAIL] = NULL;

1068

tail = rdp->nxttail[RCU_DONE_TAIL];

1068

tail = rdp->nxttail[RCU_DONE_TAIL];

1069

for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)

1069

for (count = RCU_NEXT_SIZE - 1; count >= 0; count--)

1070

if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])

1070

if (rdp->nxttail[count] == rdp->nxttail[RCU_DONE_TAIL])

1071

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

1071

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

1072

local_irq_restore(flags);

1072

local_irq_restore(flags);

1073

1074

/* Invoke callbacks. */

1074

/* Invoke callbacks. */

1075

count = 0;

1075

count = 0;

1076

while (list) {

1076

while (list) {

1077

next = list->next;

1077

next = list->next;

1078

prefetch(next);

1078

prefetch(next);

1079

list->func(list);

1079

list->func(list);

1080

list = next;

1080

list = next;

1081

if (++count >= rdp->blimit)

1081

if (++count >= rdp->blimit)

1082

break;

1082

break;

1083

}

1083

}

1084

1085

local_irq_save(flags);

1085

local_irq_save(flags);

1086

1087

/* Update count, and requeue any remaining callbacks. */

1087

/* Update count, and requeue any remaining callbacks. */

1088

rdp->qlen -= count;

1088

rdp->qlen -= count;

1089

if (list != NULL) {

1089

if (list != NULL) {

1090

*tail = rdp->nxtlist;

1090

*tail = rdp->nxtlist;

1091

rdp->nxtlist = list;

1091

rdp->nxtlist = list;

1092

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

1092

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

1093

if (&rdp->nxtlist == rdp->nxttail[count])

1093

if (&rdp->nxtlist == rdp->nxttail[count])

1094

rdp->nxttail[count] = tail;

1094

rdp->nxttail[count] = tail;

1095

else

1095

else

1096

break;

1096

break;

1097

}

1097

}

1098

1099

/* Reinstate batch limit if we have worked down the excess. */

1099

/* Reinstate batch limit if we have worked down the excess. */

1100

if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)

1100

if (rdp->blimit == LONG_MAX && rdp->qlen <= qlowmark)

1101

rdp->blimit = blimit;

1101

rdp->blimit = blimit;

1102

1103

/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */

1103

/* Reset ->qlen_last_fqs_check trigger if enough CBs have drained. */

1104

if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {

1104

if (rdp->qlen == 0 && rdp->qlen_last_fqs_check != 0) {

1105

rdp->qlen_last_fqs_check = 0;

1105

rdp->qlen_last_fqs_check = 0;

1106

rdp->n_force_qs_snap = rsp->n_force_qs;

1106

rdp->n_force_qs_snap = rsp->n_force_qs;

1107

} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)

1107

} else if (rdp->qlen < rdp->qlen_last_fqs_check - qhimark)

1108

rdp->qlen_last_fqs_check = rdp->qlen;

1108

rdp->qlen_last_fqs_check = rdp->qlen;

1109

1110

local_irq_restore(flags);

1110

local_irq_restore(flags);

1111

1112

/* Re-raise the RCU softirq if there are callbacks remaining. */

1112

/* Re-raise the RCU softirq if there are callbacks remaining. */

1113

if (cpu_has_callbacks_ready_to_invoke(rdp))

1113

if (cpu_has_callbacks_ready_to_invoke(rdp))

1114

raise_softirq(RCU_SOFTIRQ);

1114

raise_softirq(RCU_SOFTIRQ);

1115

}

1115

}

1116

1117

/*

1117

/*

1118

* Check to see if this CPU is in a non-context-switch quiescent state

1118

* Check to see if this CPU is in a non-context-switch quiescent state

1119

* (user mode or idle loop for rcu, non-softirq execution for rcu_bh).

1119

* (user mode or idle loop for rcu, non-softirq execution for rcu_bh).

1120

* Also schedule the RCU softirq handler.

1120

* Also schedule the RCU softirq handler.

1121

*

1121

*

1122

* This function must be called with hardirqs disabled. It is normally

1122

* This function must be called with hardirqs disabled. It is normally

1123

* invoked from the scheduling-clock interrupt. If rcu_pending returns

1123

* invoked from the scheduling-clock interrupt. If rcu_pending returns

1124

* false, there is no point in invoking rcu_check_callbacks().

1124

* false, there is no point in invoking rcu_check_callbacks().

1125

*/

1125

*/

1126

void rcu_check_callbacks(int cpu, int user)

1126

void rcu_check_callbacks(int cpu, int user)

1127

{

1127

{

1128

if (!rcu_pending(cpu))

1128

if (!rcu_pending(cpu))

1129

return; /* if nothing for RCU to do. */

1129

return; /* if nothing for RCU to do. */

1130

if (user ||

1130

if (user ||

1131

(idle_cpu(cpu) && rcu_scheduler_active &&

1131

(idle_cpu(cpu) && rcu_scheduler_active &&

1132

!in_softirq() && hardirq_count() <= (1 << HARDIRQ_SHIFT))) {

1132

!in_softirq() && hardirq_count() <= (1 << HARDIRQ_SHIFT))) {

1133

1134

/*

1134

/*

1135

* Get here if this CPU took its interrupt from user

1135

* Get here if this CPU took its interrupt from user

1136

* mode or from the idle loop, and if this is not a

1136

* mode or from the idle loop, and if this is not a

1137

* nested interrupt. In this case, the CPU is in

1137

* nested interrupt. In this case, the CPU is in

1138

* a quiescent state, so note it.

1138

* a quiescent state, so note it.

1139

*

1139

*

1140

* No memory barrier is required here because both

1140

* No memory barrier is required here because both

1141

* rcu_sched_qs() and rcu_bh_qs() reference only CPU-local

1141

* rcu_sched_qs() and rcu_bh_qs() reference only CPU-local

1142

* variables that other CPUs neither access nor modify,

1142

* variables that other CPUs neither access nor modify,

1143

* at least not while the corresponding CPU is online.

1143

* at least not while the corresponding CPU is online.

1144

*/

1144

*/

1145

1146

rcu_sched_qs(cpu);

1146

rcu_sched_qs(cpu);

1147

rcu_bh_qs(cpu);

1147

rcu_bh_qs(cpu);

1148

1149

} else if (!in_softirq()) {

1149

} else if (!in_softirq()) {

1150

1151

/*

1151

/*

1152

* Get here if this CPU did not take its interrupt from

1152

* Get here if this CPU did not take its interrupt from

1153

* softirq, in other words, if it is not interrupting

1153

* softirq, in other words, if it is not interrupting

1154

* a rcu_bh read-side critical section. This is an _bh

1154

* a rcu_bh read-side critical section. This is an _bh

1155

* critical section, so note it.

1155

* critical section, so note it.

1156

*/

1156

*/

1157

1158

rcu_bh_qs(cpu);

1158

rcu_bh_qs(cpu);

1159

}

1159

}

1160

rcu_preempt_check_callbacks(cpu);

1160

rcu_preempt_check_callbacks(cpu);

1161

raise_softirq(RCU_SOFTIRQ);

1161

raise_softirq(RCU_SOFTIRQ);

1162

}

1162

}

1163

1164

#ifdef CONFIG_SMP

1164

#ifdef CONFIG_SMP

1165

1166

/*

1166

/*

1167

* Scan the leaf rcu_node structures, processing dyntick state for any that

1167

* Scan the leaf rcu_node structures, processing dyntick state for any that

1168

* have not yet encountered a quiescent state, using the function specified.

1168

* have not yet encountered a quiescent state, using the function specified.

1169

* The caller must have suppressed start of new grace periods.

1169

* The caller must have suppressed start of new grace periods.

1170

*/

1170

*/

1171

static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))

1171

static void force_qs_rnp(struct rcu_state *rsp, int (*f)(struct rcu_data *))

1172

{

1172

{

1173

unsigned long bit;

1173

unsigned long bit;

1174

int cpu;

1174

int cpu;

1175

unsigned long flags;

1175

unsigned long flags;

1176

unsigned long mask;

1176

unsigned long mask;

1177

struct rcu_node *rnp;

1177

struct rcu_node *rnp;

1178

1179

rcu_for_each_leaf_node(rsp, rnp) {

1179

rcu_for_each_leaf_node(rsp, rnp) {

1180

mask = 0;

1180

mask = 0;

1181

raw_spin_lock_irqsave(&rnp->lock, flags);

1181

raw_spin_lock_irqsave(&rnp->lock, flags);

1182

if (!rcu_gp_in_progress(rsp)) {

1182

if (!rcu_gp_in_progress(rsp)) {

1183

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1183

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1184

return;

1184

return;

1185

}

1185

}

1186

if (rnp->qsmask == 0) {

1186

if (rnp->qsmask == 0) {

1187

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1187

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1188

continue;

1188

continue;

1189

}

1189

}

1190

cpu = rnp->grplo;

1190

cpu = rnp->grplo;

1191

bit = 1;

1191

bit = 1;

1192

for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {

1192

for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {

1193

if ((rnp->qsmask & bit) != 0 && f(rsp->rda[cpu]))

1193

if ((rnp->qsmask & bit) != 0 && f(rsp->rda[cpu]))

1194

mask |= bit;

1194

mask |= bit;

1195

}

1195

}

1196

if (mask != 0) {

1196

if (mask != 0) {

1197

1198

/* rcu_report_qs_rnp() releases rnp->lock. */

1198

/* rcu_report_qs_rnp() releases rnp->lock. */

1199

rcu_report_qs_rnp(mask, rsp, rnp, flags);

1199

rcu_report_qs_rnp(mask, rsp, rnp, flags);

1200

continue;

1200

continue;

1201

}

1201

}

1202

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1202

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1203

}

1203

}

1204

}

1204

}

1205

1206

/*

1206

/*

1207

* Force quiescent states on reluctant CPUs, and also detect which

1207

* Force quiescent states on reluctant CPUs, and also detect which

1208

* CPUs are in dyntick-idle mode.

1208

* CPUs are in dyntick-idle mode.

1209

*/

1209

*/

1210

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

1210

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

1211

{

1211

{

1212

unsigned long flags;

1212

unsigned long flags;

1213

struct rcu_node *rnp = rcu_get_root(rsp);

1213

struct rcu_node *rnp = rcu_get_root(rsp);

1214

1215

if (!rcu_gp_in_progress(rsp))

1215

if (!rcu_gp_in_progress(rsp))

1216

return; /* No grace period in progress, nothing to force. */

1216

return; /* No grace period in progress, nothing to force. */

1217

if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {

1217

if (!raw_spin_trylock_irqsave(&rsp->fqslock, flags)) {

1218

rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */

1218

rsp->n_force_qs_lh++; /* Inexact, can lose counts. Tough! */

1219

return; /* Someone else is already on the job. */

1219

return; /* Someone else is already on the job. */

1220

}

1220

}

1221

if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))

1221

if (relaxed && ULONG_CMP_GE(rsp->jiffies_force_qs, jiffies))

1222

goto unlock_fqs_ret; /* no emergency and done recently. */

1222

goto unlock_fqs_ret; /* no emergency and done recently. */

1223

rsp->n_force_qs++;

1223

rsp->n_force_qs++;

1224

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

1224

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

1225

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1225

rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;

1226

if(!rcu_gp_in_progress(rsp)) {

1226

if(!rcu_gp_in_progress(rsp)) {

1227

rsp->n_force_qs_ngp++;

1227

rsp->n_force_qs_ngp++;

1228

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

1228

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

1229

goto unlock_fqs_ret; /* no GP in progress, time updated. */

1229

goto unlock_fqs_ret; /* no GP in progress, time updated. */

1230

}

1230

}

1231

rsp->fqs_active = 1;

1231

rsp->fqs_active = 1;

1232

switch (rsp->signaled) {

1232

switch (rsp->signaled) {

1233

case RCU_GP_IDLE:

1233

case RCU_GP_IDLE:

1234

case RCU_GP_INIT:

1234

case RCU_GP_INIT:

1235

1236

break; /* grace period idle or initializing, ignore. */

1236

break; /* grace period idle or initializing, ignore. */

1237

1238

case RCU_SAVE_DYNTICK:

1238

case RCU_SAVE_DYNTICK:

1239

if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)

1239

if (RCU_SIGNAL_INIT != RCU_SAVE_DYNTICK)

1240

break; /* So gcc recognizes the dead code. */

1240

break; /* So gcc recognizes the dead code. */

1241

1242

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

1242

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

1243

1244

/* Record dyntick-idle state. */

1244

/* Record dyntick-idle state. */

1245

force_qs_rnp(rsp, dyntick_save_progress_counter);

1245

force_qs_rnp(rsp, dyntick_save_progress_counter);

1246

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

1246

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

1247

if (rcu_gp_in_progress(rsp))

1247

if (rcu_gp_in_progress(rsp))

1248

rsp->signaled = RCU_FORCE_QS;

1248

rsp->signaled = RCU_FORCE_QS;

1249

break;

1249

break;

1250

1251

case RCU_FORCE_QS:

1251

case RCU_FORCE_QS:

1252

1253

/* Check dyntick-idle state, send IPI to laggarts. */

1253

/* Check dyntick-idle state, send IPI to laggarts. */

1254

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

1254

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

1255

force_qs_rnp(rsp, rcu_implicit_dynticks_qs);

1255

force_qs_rnp(rsp, rcu_implicit_dynticks_qs);

1256

1257

/* Leave state in case more forcing is required. */

1257

/* Leave state in case more forcing is required. */

1258

1259

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

1259

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

1260

break;

1260

break;

1261

}

1261

}

1262

rsp->fqs_active = 0;

1262

rsp->fqs_active = 0;

1263

if (rsp->fqs_need_gp) {

1263

if (rsp->fqs_need_gp) {

1264

raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */

1264

raw_spin_unlock(&rsp->fqslock); /* irqs remain disabled */

1265

rsp->fqs_need_gp = 0;

1265

rsp->fqs_need_gp = 0;

1266

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

1266

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

1267

return;

1267

return;

1268

}

1268

}

1269

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

1269

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

1270

unlock_fqs_ret:

1270

unlock_fqs_ret:

1271

raw_spin_unlock_irqrestore(&rsp->fqslock, flags);

1271

raw_spin_unlock_irqrestore(&rsp->fqslock, flags);

1272

}

1272

}

1273

1274

#else /* #ifdef CONFIG_SMP */

1274

#else /* #ifdef CONFIG_SMP */

1275

1276

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

1276

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

1277

{

1277

{

1278

set_need_resched();

1278

set_need_resched();

1279

}

1279

}

1280

1281

#endif /* #else #ifdef CONFIG_SMP */

1281

#endif /* #else #ifdef CONFIG_SMP */

1282

1283

/*

1283

/*

1284

* This does the RCU processing work from softirq context for the

1284

* This does the RCU processing work from softirq context for the

1285

* specified rcu_state and rcu_data structures. This may be called

1285

* specified rcu_state and rcu_data structures. This may be called

1286

* only from the CPU to whom the rdp belongs.

1286

* only from the CPU to whom the rdp belongs.

1287

*/

1287

*/

1288

static void

1288

static void

1289

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

1289

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

1290

{

1290

{

1291

unsigned long flags;

1291

unsigned long flags;

1292

1293

WARN_ON_ONCE(rdp->beenonline == 0);

1293

WARN_ON_ONCE(rdp->beenonline == 0);

1294

1295

/*

1295

/*

1296

* If an RCU GP has gone long enough, go check for dyntick

1296

* If an RCU GP has gone long enough, go check for dyntick

1297

* idle CPUs and, if needed, send resched IPIs.

1297

* idle CPUs and, if needed, send resched IPIs.

1298

*/

1298

*/

1299

if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1299

if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1300

force_quiescent_state(rsp, 1);

1300

force_quiescent_state(rsp, 1);

1301

1302

/*

1302

/*

1303

* Advance callbacks in response to end of earlier grace

1303

* Advance callbacks in response to end of earlier grace

1304

* period that some other CPU ended.

1304

* period that some other CPU ended.

1305

*/

1305

*/

1306

rcu_process_gp_end(rsp, rdp);

1306

rcu_process_gp_end(rsp, rdp);

1307

1308

/* Update RCU state based on any recent quiescent states. */

1308

/* Update RCU state based on any recent quiescent states. */

1309

rcu_check_quiescent_state(rsp, rdp);

1309

rcu_check_quiescent_state(rsp, rdp);

1310

1311

/* Does this CPU require a not-yet-started grace period? */

1311

/* Does this CPU require a not-yet-started grace period? */

1312

if (cpu_needs_another_gp(rsp, rdp)) {

1312

if (cpu_needs_another_gp(rsp, rdp)) {

1313

raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);

1313

raw_spin_lock_irqsave(&rcu_get_root(rsp)->lock, flags);

1314

rcu_start_gp(rsp, flags); /* releases above lock */

1314

rcu_start_gp(rsp, flags); /* releases above lock */

1315

}

1315

}

1316

1317

/* If there are callbacks ready, invoke them. */

1317

/* If there are callbacks ready, invoke them. */

1318

rcu_do_batch(rsp, rdp);

1318

rcu_do_batch(rsp, rdp);

1319

}

1319

}

1320

1321

/*

1321

/*

1322

* Do softirq processing for the current CPU.

1322

* Do softirq processing for the current CPU.

1323

*/

1323

*/

1324

static void rcu_process_callbacks(struct softirq_action *unused)

1324

static void rcu_process_callbacks(struct softirq_action *unused)

1325

{

1325

{

1326

/*

1326

/*

1327

* Memory references from any prior RCU read-side critical sections

1327

* Memory references from any prior RCU read-side critical sections

1328

* executed by the interrupted code must be seen before any RCU

1328

* executed by the interrupted code must be seen before any RCU

1329

* grace-period manipulations below.

1329

* grace-period manipulations below.

1330

*/

1330

*/

1331

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

1331

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

1332

1333

__rcu_process_callbacks(&rcu_sched_state,

1333

__rcu_process_callbacks(&rcu_sched_state,

1334

&__get_cpu_var(rcu_sched_data));

1334

&__get_cpu_var(rcu_sched_data));

1335

__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));

1335

__rcu_process_callbacks(&rcu_bh_state, &__get_cpu_var(rcu_bh_data));

1336

rcu_preempt_process_callbacks();

1336

rcu_preempt_process_callbacks();

1337

1338

/*

1338

/*

1339

* Memory references from any later RCU read-side critical sections

1339

* Memory references from any later RCU read-side critical sections

1340

* executed by the interrupted code must be seen after any RCU

1340

* executed by the interrupted code must be seen after any RCU

1341

* grace-period manipulations above.

1341

* grace-period manipulations above.

1342

*/

1342

*/

1343

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

1343

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

1344

1345

/* If we are last CPU on way to dyntick-idle mode, accelerate it. */

1345

/* If we are last CPU on way to dyntick-idle mode, accelerate it. */

1346

rcu_needs_cpu_flush();

1346

rcu_needs_cpu_flush();

1347

}

1347

}

1348

1349

static void

1349

static void

1350

__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),

1350

__call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),

1351

struct rcu_state *rsp)

1351

struct rcu_state *rsp)

1352

{

1352

{

1353

unsigned long flags;

1353

unsigned long flags;

1354

struct rcu_data *rdp;

1354

struct rcu_data *rdp;

1355

1356

head->func = func;

1356

head->func = func;

1357

head->next = NULL;

1357

head->next = NULL;

1358

1359

smp_mb(); /* Ensure RCU update seen before callback registry. */

1359

smp_mb(); /* Ensure RCU update seen before callback registry. */

1360

1361

/*

1361

/*

1362

* Opportunistically note grace-period endings and beginnings.

1362

* Opportunistically note grace-period endings and beginnings.

1363

* Note that we might see a beginning right after we see an

1363

* Note that we might see a beginning right after we see an

1364

* end, but never vice versa, since this CPU has to pass through

1364

* end, but never vice versa, since this CPU has to pass through

1365

* a quiescent state betweentimes.

1365

* a quiescent state betweentimes.

1366

*/

1366

*/

1367

local_irq_save(flags);

1367

local_irq_save(flags);

1368

rdp = rsp->rda[smp_processor_id()];

1368

rdp = rsp->rda[smp_processor_id()];

1369

rcu_process_gp_end(rsp, rdp);

1369

rcu_process_gp_end(rsp, rdp);

1370

check_for_new_grace_period(rsp, rdp);

1370

check_for_new_grace_period(rsp, rdp);

1371

1372

/* Add the callback to our list. */

1372

/* Add the callback to our list. */

1373

*rdp->nxttail[RCU_NEXT_TAIL] = head;

1373

*rdp->nxttail[RCU_NEXT_TAIL] = head;

1374

rdp->nxttail[RCU_NEXT_TAIL] = &head->next;

1374

rdp->nxttail[RCU_NEXT_TAIL] = &head->next;

1375

1376

/* Start a new grace period if one not already started. */

1376

/* Start a new grace period if one not already started. */

1377

if (!rcu_gp_in_progress(rsp)) {

1377

if (!rcu_gp_in_progress(rsp)) {

1378

unsigned long nestflag;

1378

unsigned long nestflag;

1379

struct rcu_node *rnp_root = rcu_get_root(rsp);

1379

struct rcu_node *rnp_root = rcu_get_root(rsp);

1380

1381

raw_spin_lock_irqsave(&rnp_root->lock, nestflag);

1381

raw_spin_lock_irqsave(&rnp_root->lock, nestflag);

1382

rcu_start_gp(rsp, nestflag); /* releases rnp_root->lock. */

1382

rcu_start_gp(rsp, nestflag); /* releases rnp_root->lock. */

1383

}

1383

}

1384

1385

/*

1385

/*

1386

* Force the grace period if too many callbacks or too long waiting.

1386

* Force the grace period if too many callbacks or too long waiting.

1387

* Enforce hysteresis, and don't invoke force_quiescent_state()

1387

* Enforce hysteresis, and don't invoke force_quiescent_state()

1388

* if some other CPU has recently done so. Also, don't bother

1388

* if some other CPU has recently done so. Also, don't bother

1389

* invoking force_quiescent_state() if the newly enqueued callback

1389

* invoking force_quiescent_state() if the newly enqueued callback

1390

* is the only one waiting for a grace period to complete.

1390

* is the only one waiting for a grace period to complete.

1391

*/

1391

*/

1392

if (unlikely(++rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {

1392

if (unlikely(++rdp->qlen > rdp->qlen_last_fqs_check + qhimark)) {

1393

rdp->blimit = LONG_MAX;

1393

rdp->blimit = LONG_MAX;

1394

if (rsp->n_force_qs == rdp->n_force_qs_snap &&

1394

if (rsp->n_force_qs == rdp->n_force_qs_snap &&

1395

*rdp->nxttail[RCU_DONE_TAIL] != head)

1395

*rdp->nxttail[RCU_DONE_TAIL] != head)

1396

force_quiescent_state(rsp, 0);

1396

force_quiescent_state(rsp, 0);

1397

rdp->n_force_qs_snap = rsp->n_force_qs;

1397

rdp->n_force_qs_snap = rsp->n_force_qs;

1398

rdp->qlen_last_fqs_check = rdp->qlen;

1398

rdp->qlen_last_fqs_check = rdp->qlen;

1399

} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1399

} else if (ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies))

1400

force_quiescent_state(rsp, 1);

1400

force_quiescent_state(rsp, 1);

1401

local_irq_restore(flags);

1401

local_irq_restore(flags);

1402

}

1402

}

1403

1404

/*

1404

/*

1405

* Queue an RCU-sched callback for invocation after a grace period.

1405

* Queue an RCU-sched callback for invocation after a grace period.

1406

*/

1406

*/

1407

void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1407

void call_rcu_sched(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1408

{

1408

{

1409

__call_rcu(head, func, &rcu_sched_state);

1409

__call_rcu(head, func, &rcu_sched_state);

1410

}

1410

}

1411

EXPORT_SYMBOL_GPL(call_rcu_sched);

1411

EXPORT_SYMBOL_GPL(call_rcu_sched);

1412

1413

/*

1413

/*

1414

* Queue an RCU for invocation after a quicker grace period.

1414

* Queue an RCU for invocation after a quicker grace period.

1415

*/

1415

*/

1416

void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1416

void call_rcu_bh(struct rcu_head *head, void (*func)(struct rcu_head *rcu))

1417

{

1417

{

1418

__call_rcu(head, func, &rcu_bh_state);

1418

__call_rcu(head, func, &rcu_bh_state);

1419

}

1419

}

1420

EXPORT_SYMBOL_GPL(call_rcu_bh);

1420

EXPORT_SYMBOL_GPL(call_rcu_bh);

1421

1422

/**

1422

/**

1423

* synchronize_sched - wait until an rcu-sched grace period has elapsed.

1423

* synchronize_sched - wait until an rcu-sched grace period has elapsed.

1424

*

1424

*

1425

* Control will return to the caller some time after a full rcu-sched

1425

* Control will return to the caller some time after a full rcu-sched

1426

* grace period has elapsed, in other words after all currently executing

1426

* grace period has elapsed, in other words after all currently executing

1427

* rcu-sched read-side critical sections have completed. These read-side

1427

* rcu-sched read-side critical sections have completed. These read-side

1428

* critical sections are delimited by rcu_read_lock_sched() and

1428

* critical sections are delimited by rcu_read_lock_sched() and

1429

* rcu_read_unlock_sched(), and may be nested. Note that preempt_disable(),

1429

* rcu_read_unlock_sched(), and may be nested. Note that preempt_disable(),

1430

* local_irq_disable(), and so on may be used in place of

1430

* local_irq_disable(), and so on may be used in place of

1431

* rcu_read_lock_sched().

1431

* rcu_read_lock_sched().

1432

*

1432

*

1433

* This means that all preempt_disable code sequences, including NMI and

1433

* This means that all preempt_disable code sequences, including NMI and

1434

* hardware-interrupt handlers, in progress on entry will have completed

1434

* hardware-interrupt handlers, in progress on entry will have completed

1435

* before this primitive returns. However, this does not guarantee that

1435

* before this primitive returns. However, this does not guarantee that

1436

* softirq handlers will have completed, since in some kernels, these

1436

* softirq handlers will have completed, since in some kernels, these

1437

* handlers can run in process context, and can block.

1437

* handlers can run in process context, and can block.

1438

*

1438

*

1439

* This primitive provides the guarantees made by the (now removed)

1439

* This primitive provides the guarantees made by the (now removed)

1440

* synchronize_kernel() API. In contrast, synchronize_rcu() only

1440

* synchronize_kernel() API. In contrast, synchronize_rcu() only

1441

* guarantees that rcu_read_lock() sections will have completed.

1441

* guarantees that rcu_read_lock() sections will have completed.

1442

* In "classic RCU", these two guarantees happen to be one and

1442

* In "classic RCU", these two guarantees happen to be one and

1443

* the same, but can differ in realtime RCU implementations.

1443

* the same, but can differ in realtime RCU implementations.

1444

*/

1444

*/

1445

void synchronize_sched(void)

1445

void synchronize_sched(void)

1446

{

1446

{

1447

struct rcu_synchronize rcu;

1447

struct rcu_synchronize rcu;

1448

1449

if (rcu_blocking_is_gp())

1449

if (rcu_blocking_is_gp())

1450

return;

1450

return;

1451

1452

init_completion(&rcu.completion);

1452

init_completion(&rcu.completion);

1453

/* Will wake me after RCU finished. */

1453

/* Will wake me after RCU finished. */

1454

call_rcu_sched(&rcu.head, wakeme_after_rcu);

1454

call_rcu_sched(&rcu.head, wakeme_after_rcu);

1455

/* Wait for it. */

1455

/* Wait for it. */

1456

wait_for_completion(&rcu.completion);

1456

wait_for_completion(&rcu.completion);

1457

}

1457

}

1458

EXPORT_SYMBOL_GPL(synchronize_sched);

1458

EXPORT_SYMBOL_GPL(synchronize_sched);

1459

1460

/**

1460

/**

1461

* synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.

1461

* synchronize_rcu_bh - wait until an rcu_bh grace period has elapsed.

1462

*

1462

*

1463

* Control will return to the caller some time after a full rcu_bh grace

1463

* Control will return to the caller some time after a full rcu_bh grace

1464

* period has elapsed, in other words after all currently executing rcu_bh

1464

* period has elapsed, in other words after all currently executing rcu_bh

1465

* read-side critical sections have completed. RCU read-side critical

1465

* read-side critical sections have completed. RCU read-side critical

1466

* sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),

1466

* sections are delimited by rcu_read_lock_bh() and rcu_read_unlock_bh(),

1467

* and may be nested.

1467

* and may be nested.

1468

*/

1468

*/

1469

void synchronize_rcu_bh(void)

1469

void synchronize_rcu_bh(void)

1470

{

1470

{

1471

struct rcu_synchronize rcu;

1471

struct rcu_synchronize rcu;

1472

1473

if (rcu_blocking_is_gp())

1473

if (rcu_blocking_is_gp())

1474

return;

1474

return;

1475

1476

init_completion(&rcu.completion);

1476

init_completion(&rcu.completion);

1477

/* Will wake me after RCU finished. */

1477

/* Will wake me after RCU finished. */

1478

call_rcu_bh(&rcu.head, wakeme_after_rcu);

1478

call_rcu_bh(&rcu.head, wakeme_after_rcu);

1479

/* Wait for it. */

1479

/* Wait for it. */

1480

wait_for_completion(&rcu.completion);

1480

wait_for_completion(&rcu.completion);

1481

}

1481

}

1482

EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

1482

EXPORT_SYMBOL_GPL(synchronize_rcu_bh);

1483

1484

/*

1484

/*

1485

* Check to see if there is any immediate RCU-related work to be done

1485

* Check to see if there is any immediate RCU-related work to be done

1486

* by the current CPU, for the specified type of RCU, returning 1 if so.

1486

* by the current CPU, for the specified type of RCU, returning 1 if so.

1487

* The checks are in order of increasing expense: checks that can be

1487

* The checks are in order of increasing expense: checks that can be

1488

* carried out against CPU-local state are performed first. However,

1488

* carried out against CPU-local state are performed first. However,

1489

* we must check for CPU stalls first, else we might not get a chance.

1489

* we must check for CPU stalls first, else we might not get a chance.

1490

*/

1490

*/

1491

static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)

1491

static int __rcu_pending(struct rcu_state *rsp, struct rcu_data *rdp)

1492

{

1492

{

1493

struct rcu_node *rnp = rdp->mynode;

1493

struct rcu_node *rnp = rdp->mynode;

1494

1495

rdp->n_rcu_pending++;

1495

rdp->n_rcu_pending++;

1496

1497

/* Check for CPU stalls, if enabled. */

1497

/* Check for CPU stalls, if enabled. */

1498

check_cpu_stall(rsp, rdp);

1498

check_cpu_stall(rsp, rdp);

1499

1500

/* Is the RCU core waiting for a quiescent state from this CPU? */

1500

/* Is the RCU core waiting for a quiescent state from this CPU? */

1501

if (rdp->qs_pending) {

1501

if (rdp->qs_pending) {

1502

1503

/*

1503

/*

1504

* If force_quiescent_state() coming soon and this CPU

1504

* If force_quiescent_state() coming soon and this CPU

1505

* needs a quiescent state, and this is either RCU-sched

1505

* needs a quiescent state, and this is either RCU-sched

1506

* or RCU-bh, force a local reschedule.

1506

* or RCU-bh, force a local reschedule.

1507

*/

1507

*/

1508

if (!rdp->preemptable &&

1508

if (!rdp->preemptable &&

1509

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,

1509

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,

1510

jiffies))

1510

jiffies))

1511

set_need_resched();

1511

set_need_resched();

1512

rdp->n_rp_qs_pending++;

1512

rdp->n_rp_qs_pending++;

1513

return 1;

1513

return 1;

1514

}

1514

}

1515

1516

/* Does this CPU have callbacks ready to invoke? */

1516

/* Does this CPU have callbacks ready to invoke? */

1517

if (cpu_has_callbacks_ready_to_invoke(rdp)) {

1517

if (cpu_has_callbacks_ready_to_invoke(rdp)) {

1518

rdp->n_rp_cb_ready++;

1518

rdp->n_rp_cb_ready++;

1519

return 1;

1519

return 1;

1520

}

1520

}

1521

1522

/* Has RCU gone idle with this CPU needing another grace period? */

1522

/* Has RCU gone idle with this CPU needing another grace period? */

1523

if (cpu_needs_another_gp(rsp, rdp)) {

1523

if (cpu_needs_another_gp(rsp, rdp)) {

1524

rdp->n_rp_cpu_needs_gp++;

1524

rdp->n_rp_cpu_needs_gp++;

1525

return 1;

1525

return 1;

1526

}

1526

}

1527

1528

/* Has another RCU grace period completed? */

1528

/* Has another RCU grace period completed? */

1529

if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */

1529

if (ACCESS_ONCE(rnp->completed) != rdp->completed) { /* outside lock */

1530

rdp->n_rp_gp_completed++;

1530

rdp->n_rp_gp_completed++;

1531

return 1;

1531

return 1;

1532

}

1532

}

1533

1534

/* Has a new RCU grace period started? */

1534

/* Has a new RCU grace period started? */

1535

if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */

1535

if (ACCESS_ONCE(rnp->gpnum) != rdp->gpnum) { /* outside lock */

1536

rdp->n_rp_gp_started++;

1536

rdp->n_rp_gp_started++;

1537

return 1;

1537

return 1;

1538

}

1538

}

1539

1540

/* Has an RCU GP gone long enough to send resched IPIs &c? */

1540

/* Has an RCU GP gone long enough to send resched IPIs &c? */

1541

if (rcu_gp_in_progress(rsp) &&

1541

if (rcu_gp_in_progress(rsp) &&

1542

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {

1542

ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs), jiffies)) {

1543

rdp->n_rp_need_fqs++;

1543

rdp->n_rp_need_fqs++;

1544

return 1;

1544

return 1;

1545

}

1545

}

1546

1547

/* nothing to do */

1547

/* nothing to do */

1548

rdp->n_rp_need_nothing++;

1548

rdp->n_rp_need_nothing++;

1549

return 0;

1549

return 0;

1550

}

1550

}

1551

1552

/*

1552

/*

1553

* Check to see if there is any immediate RCU-related work to be done

1553

* Check to see if there is any immediate RCU-related work to be done

1554

* by the current CPU, returning 1 if so. This function is part of the

1554

* by the current CPU, returning 1 if so. This function is part of the

1555

* RCU implementation; it is -not- an exported member of the RCU API.

1555

* RCU implementation; it is -not- an exported member of the RCU API.

1556

*/

1556

*/

1557

static int rcu_pending(int cpu)

1557

static int rcu_pending(int cpu)

1558

{

1558

{

1559

return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||

1559

return __rcu_pending(&rcu_sched_state, &per_cpu(rcu_sched_data, cpu)) ||

1560

__rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||

1560

__rcu_pending(&rcu_bh_state, &per_cpu(rcu_bh_data, cpu)) ||

1561

rcu_preempt_pending(cpu);

1561

rcu_preempt_pending(cpu);

1562

}

1562

}

1563

1564

/*

1564

/*

1565

* Check to see if any future RCU-related work will need to be done

1565

* Check to see if any future RCU-related work will need to be done

1566

* by the current CPU, even if none need be done immediately, returning

1566

* by the current CPU, even if none need be done immediately, returning

1567

* 1 if so.

1567

* 1 if so.

1568

*/

1568

*/

1569

static int rcu_needs_cpu_quick_check(int cpu)

1569

static int rcu_needs_cpu_quick_check(int cpu)

1570

{

1570

{

1571

/* RCU callbacks either ready or pending? */

1571

/* RCU callbacks either ready or pending? */

1572

return per_cpu(rcu_sched_data, cpu).nxtlist ||

1572

return per_cpu(rcu_sched_data, cpu).nxtlist ||

1573

per_cpu(rcu_bh_data, cpu).nxtlist ||

1573

per_cpu(rcu_bh_data, cpu).nxtlist ||

1574

rcu_preempt_needs_cpu(cpu);

1574

rcu_preempt_needs_cpu(cpu);

1575

}

1575

}

1576

1577

static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};

1577

static DEFINE_PER_CPU(struct rcu_head, rcu_barrier_head) = {NULL};

1578

static atomic_t rcu_barrier_cpu_count;

1578

static atomic_t rcu_barrier_cpu_count;

1579

static DEFINE_MUTEX(rcu_barrier_mutex);

1579

static DEFINE_MUTEX(rcu_barrier_mutex);

1580

static struct completion rcu_barrier_completion;

1580

static struct completion rcu_barrier_completion;

1581

1582

static void rcu_barrier_callback(struct rcu_head *notused)

1582

static void rcu_barrier_callback(struct rcu_head *notused)

1583

{

1583

{

1584

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

1584

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

1585

complete(&rcu_barrier_completion);

1585

complete(&rcu_barrier_completion);

1586

}

1586

}

1587

1588

/*

1588

/*

1589

* Called with preemption disabled, and from cross-cpu IRQ context.

1589

* Called with preemption disabled, and from cross-cpu IRQ context.

1590

*/

1590

*/

1591

static void rcu_barrier_func(void *type)

1591

static void rcu_barrier_func(void *type)

1592

{

1592

{

1593

int cpu = smp_processor_id();

1593

int cpu = smp_processor_id();

1594

struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);

1594

struct rcu_head *head = &per_cpu(rcu_barrier_head, cpu);

1595

void (*call_rcu_func)(struct rcu_head *head,

1595

void (*call_rcu_func)(struct rcu_head *head,

1596

void (*func)(struct rcu_head *head));

1596

void (*func)(struct rcu_head *head));

1597

1598

atomic_inc(&rcu_barrier_cpu_count);

1598

atomic_inc(&rcu_barrier_cpu_count);

1599

call_rcu_func = type;

1599

call_rcu_func = type;

1600

call_rcu_func(head, rcu_barrier_callback);

1600

call_rcu_func(head, rcu_barrier_callback);

1601

}

1601

}

1602

1603

/*

1603

/*

1604

* Orchestrate the specified type of RCU barrier, waiting for all

1604

* Orchestrate the specified type of RCU barrier, waiting for all

1605

* RCU callbacks of the specified type to complete.

1605

* RCU callbacks of the specified type to complete.

1606

*/

1606

*/

1607

static void _rcu_barrier(struct rcu_state *rsp,

1607

static void _rcu_barrier(struct rcu_state *rsp,

1608

void (*call_rcu_func)(struct rcu_head *head,

1608

void (*call_rcu_func)(struct rcu_head *head,

1609

void (*func)(struct rcu_head *head)))

1609

void (*func)(struct rcu_head *head)))

1610

{

1610

{

1611

BUG_ON(in_interrupt());

1611

BUG_ON(in_interrupt());

1612

/* Take mutex to serialize concurrent rcu_barrier() requests. */

1612

/* Take mutex to serialize concurrent rcu_barrier() requests. */

1613

mutex_lock(&rcu_barrier_mutex);

1613

mutex_lock(&rcu_barrier_mutex);

1614

init_completion(&rcu_barrier_completion);

1614

init_completion(&rcu_barrier_completion);

1615

/*

1615

/*

1616

* Initialize rcu_barrier_cpu_count to 1, then invoke

1616

* Initialize rcu_barrier_cpu_count to 1, then invoke

1617

* rcu_barrier_func() on each CPU, so that each CPU also has

1617

* rcu_barrier_func() on each CPU, so that each CPU also has

1618

* incremented rcu_barrier_cpu_count. Only then is it safe to

1618

* incremented rcu_barrier_cpu_count. Only then is it safe to

1619

* decrement rcu_barrier_cpu_count -- otherwise the first CPU

1619

* decrement rcu_barrier_cpu_count -- otherwise the first CPU

1620

* might complete its grace period before all of the other CPUs

1620

* might complete its grace period before all of the other CPUs

1621

* did their increment, causing this function to return too

1621

* did their increment, causing this function to return too

1622

* early.

1622

* early.

1623

*/

1623

*/

1624

atomic_set(&rcu_barrier_cpu_count, 1);

1624

atomic_set(&rcu_barrier_cpu_count, 1);

1625

preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */

1625

preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */

1626

rcu_adopt_orphan_cbs(rsp);

1626

rcu_adopt_orphan_cbs(rsp);

1627

on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);

1627

on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);

1628

preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */

1628

preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */

1629

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

1629

if (atomic_dec_and_test(&rcu_barrier_cpu_count))

1630

complete(&rcu_barrier_completion);

1630

complete(&rcu_barrier_completion);

1631

wait_for_completion(&rcu_barrier_completion);

1631

wait_for_completion(&rcu_barrier_completion);

1632

mutex_unlock(&rcu_barrier_mutex);

1632

mutex_unlock(&rcu_barrier_mutex);

1633

}

1633

}

1634

1635

/**

1635

/**

1636

* rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.

1636

* rcu_barrier_bh - Wait until all in-flight call_rcu_bh() callbacks complete.

1637

*/

1637

*/

1638

void rcu_barrier_bh(void)

1638

void rcu_barrier_bh(void)

1639

{

1639

{

1640

_rcu_barrier(&rcu_bh_state, call_rcu_bh);

1640

_rcu_barrier(&rcu_bh_state, call_rcu_bh);

1641

}

1641

}

1642

EXPORT_SYMBOL_GPL(rcu_barrier_bh);

1642

EXPORT_SYMBOL_GPL(rcu_barrier_bh);

1643

1644

/**

1644

/**

1645

* rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.

1645

* rcu_barrier_sched - Wait for in-flight call_rcu_sched() callbacks.

1646

*/

1646

*/

1647

void rcu_barrier_sched(void)

1647

void rcu_barrier_sched(void)

1648

{

1648

{

1649

_rcu_barrier(&rcu_sched_state, call_rcu_sched);

1649

_rcu_barrier(&rcu_sched_state, call_rcu_sched);

1650

}

1650

}

1651

EXPORT_SYMBOL_GPL(rcu_barrier_sched);

1651

EXPORT_SYMBOL_GPL(rcu_barrier_sched);

1652

1653

/*

1653

/*

1654

* Do boot-time initialization of a CPU's per-CPU RCU data.

1654

* Do boot-time initialization of a CPU's per-CPU RCU data.

1655

*/

1655

*/

1656

static void __init

1656

static void __init

1657

rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)

1657

rcu_boot_init_percpu_data(int cpu, struct rcu_state *rsp)

1658

{

1658

{

1659

unsigned long flags;

1659

unsigned long flags;

1660

int i;

1660

int i;

1661

struct rcu_data *rdp = rsp->rda[cpu];

1661

struct rcu_data *rdp = rsp->rda[cpu];

1662

struct rcu_node *rnp = rcu_get_root(rsp);

1662

struct rcu_node *rnp = rcu_get_root(rsp);

1663

1664

/* Set up local state, ensuring consistent view of global state. */

1664

/* Set up local state, ensuring consistent view of global state. */

1665

raw_spin_lock_irqsave(&rnp->lock, flags);

1665

raw_spin_lock_irqsave(&rnp->lock, flags);

1666

rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);

1666

rdp->grpmask = 1UL << (cpu - rdp->mynode->grplo);

1667

rdp->nxtlist = NULL;

1667

rdp->nxtlist = NULL;

1668

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

1668

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

1669

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

1669

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

1670

rdp->qlen = 0;

1670

rdp->qlen = 0;

1671

#ifdef CONFIG_NO_HZ

1671

#ifdef CONFIG_NO_HZ

1672

rdp->dynticks = &per_cpu(rcu_dynticks, cpu);

1672

rdp->dynticks = &per_cpu(rcu_dynticks, cpu);

1673

#endif /* #ifdef CONFIG_NO_HZ */

1673

#endif /* #ifdef CONFIG_NO_HZ */

1674

rdp->cpu = cpu;

1674

rdp->cpu = cpu;

1675

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1675

raw_spin_unlock_irqrestore(&rnp->lock, flags);

1676

}

1676

}

1677

1678

/*

1678

/*

1679

* Initialize a CPU's per-CPU RCU data. Note that only one online or

1679

* Initialize a CPU's per-CPU RCU data. Note that only one online or

1680

* offline event can be happening at a given time. Note also that we

1680

* offline event can be happening at a given time. Note also that we

1681

* can accept some slop in the rsp->completed access due to the fact

1681

* can accept some slop in the rsp->completed access due to the fact

1682

* that this CPU cannot possibly have any RCU callbacks in flight yet.

1682

* that this CPU cannot possibly have any RCU callbacks in flight yet.

1683

*/

1683

*/

1684

static void __cpuinit

1684

static void __cpuinit

1685

rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptable)

1685

rcu_init_percpu_data(int cpu, struct rcu_state *rsp, int preemptable)

1686

{

1686

{

1687

unsigned long flags;

1687

unsigned long flags;

1688

unsigned long mask;

1688

unsigned long mask;

1689

struct rcu_data *rdp = rsp->rda[cpu];

1689

struct rcu_data *rdp = rsp->rda[cpu];

1690

struct rcu_node *rnp = rcu_get_root(rsp);

1690

struct rcu_node *rnp = rcu_get_root(rsp);

1691

1692

/* Set up local state, ensuring consistent view of global state. */

1692

/* Set up local state, ensuring consistent view of global state. */

1693

raw_spin_lock_irqsave(&rnp->lock, flags);

1693

raw_spin_lock_irqsave(&rnp->lock, flags);

1694

rdp->passed_quiesc = 0; /* We could be racing with new GP, */

1694

rdp->passed_quiesc = 0; /* We could be racing with new GP, */

1695

rdp->qs_pending = 1; /* so set up to respond to current GP. */

1695

rdp->qs_pending = 1; /* so set up to respond to current GP. */

1696

rdp->beenonline = 1; /* We have now been online. */

1696

rdp->beenonline = 1; /* We have now been online. */

1697

rdp->preemptable = preemptable;

1697

rdp->preemptable = preemptable;

1698

rdp->qlen_last_fqs_check = 0;

1698

rdp->qlen_last_fqs_check = 0;

1699

rdp->n_force_qs_snap = rsp->n_force_qs;

1699

rdp->n_force_qs_snap = rsp->n_force_qs;

1700

rdp->blimit = blimit;

1700

rdp->blimit = blimit;

1701

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

1701

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

1702

1703

/*

1703

/*

1704

* A new grace period might start here. If so, we won't be part

1704

* A new grace period might start here. If so, we won't be part

1705

* of it, but that is OK, as we are currently in a quiescent state.

1705

* of it, but that is OK, as we are currently in a quiescent state.

1706

*/

1706

*/

1707

1708

/* Exclude any attempts to start a new GP on large systems. */

1708

/* Exclude any attempts to start a new GP on large systems. */

1709

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

1709

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

1710

1711

/* Add CPU to rcu_node bitmasks. */

1711

/* Add CPU to rcu_node bitmasks. */

1712

rnp = rdp->mynode;

1712

rnp = rdp->mynode;

1713

mask = rdp->grpmask;

1713

mask = rdp->grpmask;

1714

do {

1714

do {

1715

/* Exclude any attempts to start a new GP on small systems. */

1715

/* Exclude any attempts to start a new GP on small systems. */

1716

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

1716

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

1717

rnp->qsmaskinit |= mask;

1717

rnp->qsmaskinit |= mask;

1718

mask = rnp->grpmask;

1718

mask = rnp->grpmask;

1719

if (rnp == rdp->mynode) {

1719

if (rnp == rdp->mynode) {

1720

rdp->gpnum = rnp->completed; /* if GP in progress... */

1720

rdp->gpnum = rnp->completed; /* if GP in progress... */

1721

rdp->completed = rnp->completed;

1721

rdp->completed = rnp->completed;

1722

rdp->passed_quiesc_completed = rnp->completed - 1;

1722

rdp->passed_quiesc_completed = rnp->completed - 1;

1723

}

1723

}

1724

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

1724

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

1725

rnp = rnp->parent;

1725

rnp = rnp->parent;

1726

} while (rnp != NULL && !(rnp->qsmaskinit & mask));

1726

} while (rnp != NULL && !(rnp->qsmaskinit & mask));

1727

1728

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

1728

raw_spin_unlock_irqrestore(&rsp->onofflock, flags);

1729

}

1729

}

1730

1731

static void __cpuinit rcu_online_cpu(int cpu)

1731

static void __cpuinit rcu_online_cpu(int cpu)

1732

{

1732

{

1733

rcu_init_percpu_data(cpu, &rcu_sched_state, 0);

1733

rcu_init_percpu_data(cpu, &rcu_sched_state, 0);

1734

rcu_init_percpu_data(cpu, &rcu_bh_state, 0);

1734

rcu_init_percpu_data(cpu, &rcu_bh_state, 0);

1735

rcu_preempt_init_percpu_data(cpu);

1735

rcu_preempt_init_percpu_data(cpu);

1736

}

1736

}

1737

1738

/*

1738

/*

1739

* Handle CPU online/offline notification events.

1739

* Handle CPU online/offline notification events.

1740

*/

1740

*/

1741

static int __cpuinit rcu_cpu_notify(struct notifier_block *self,

1741

static int __cpuinit rcu_cpu_notify(struct notifier_block *self,

1742

unsigned long action, void *hcpu)

1742

unsigned long action, void *hcpu)

1743

{

1743

{

1744

long cpu = (long)hcpu;

1744

long cpu = (long)hcpu;

1745

1746

switch (action) {

1746

switch (action) {

1747

case CPU_UP_PREPARE:

1747

case CPU_UP_PREPARE:

1748

case CPU_UP_PREPARE_FROZEN:

1748

case CPU_UP_PREPARE_FROZEN:

1749

rcu_online_cpu(cpu);

1749

rcu_online_cpu(cpu);

1750

break;

1750

break;

1751

case CPU_DYING:

1751

case CPU_DYING:

1752

case CPU_DYING_FROZEN:

1752

case CPU_DYING_FROZEN:

1753

/*

1753

/*

1754

* preempt_disable() in _rcu_barrier() prevents stop_machine(),

1754

* preempt_disable() in _rcu_barrier() prevents stop_machine(),

1755

* so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"

1755

* so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"

1756

* returns, all online cpus have queued rcu_barrier_func().

1756

* returns, all online cpus have queued rcu_barrier_func().

1757

* The dying CPU clears its cpu_online_mask bit and

1757

* The dying CPU clears its cpu_online_mask bit and

1758

* moves all of its RCU callbacks to ->orphan_cbs_list

1758

* moves all of its RCU callbacks to ->orphan_cbs_list

1759

* in the context of stop_machine(), so subsequent calls

1759

* in the context of stop_machine(), so subsequent calls

1760

* to _rcu_barrier() will adopt these callbacks and only

1760

* to _rcu_barrier() will adopt these callbacks and only

1761

* then queue rcu_barrier_func() on all remaining CPUs.

1761

* then queue rcu_barrier_func() on all remaining CPUs.

1762

*/

1762

*/

1763

rcu_send_cbs_to_orphanage(&rcu_bh_state);

1763

rcu_send_cbs_to_orphanage(&rcu_bh_state);

1764

rcu_send_cbs_to_orphanage(&rcu_sched_state);

1764

rcu_send_cbs_to_orphanage(&rcu_sched_state);

1765

rcu_preempt_send_cbs_to_orphanage();

1765

rcu_preempt_send_cbs_to_orphanage();

1766

break;

1766

break;

1767

case CPU_DEAD:

1767

case CPU_DEAD:

1768

case CPU_DEAD_FROZEN:

1768

case CPU_DEAD_FROZEN:

1769

case CPU_UP_CANCELED:

1769

case CPU_UP_CANCELED:

1770

case CPU_UP_CANCELED_FROZEN:

1770

case CPU_UP_CANCELED_FROZEN:

1771

rcu_offline_cpu(cpu);

1771

rcu_offline_cpu(cpu);

1772

break;

1772

break;

1773

default:

1773

default:

1774

break;

1774

break;

1775

}

1775

}

1776

return NOTIFY_OK;

1776

return NOTIFY_OK;

1777

}

1777

}

1778

1779

/*

1779

/*

1780

* Compute the per-level fanout, either using the exact fanout specified

1780

* Compute the per-level fanout, either using the exact fanout specified

1781

* or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.

1781

* or balancing the tree, depending on CONFIG_RCU_FANOUT_EXACT.

1782

*/

1782

*/

1783

#ifdef CONFIG_RCU_FANOUT_EXACT

1783

#ifdef CONFIG_RCU_FANOUT_EXACT

1784

static void __init rcu_init_levelspread(struct rcu_state *rsp)

1784

static void __init rcu_init_levelspread(struct rcu_state *rsp)

1785

{

1785

{

1786

int i;

1786

int i;

1787

1788

for (i = NUM_RCU_LVLS - 1; i >= 0; i--)

1788

for (i = NUM_RCU_LVLS - 1; i >= 0; i--)

1789

rsp->levelspread[i] = CONFIG_RCU_FANOUT;

1789

rsp->levelspread[i] = CONFIG_RCU_FANOUT;

1790

}

1790

}

1791

#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */

1791

#else /* #ifdef CONFIG_RCU_FANOUT_EXACT */

1792

static void __init rcu_init_levelspread(struct rcu_state *rsp)

1792

static void __init rcu_init_levelspread(struct rcu_state *rsp)

1793

{

1793

{

1794

int ccur;

1794

int ccur;

1795

int cprv;

1795

int cprv;

1796

int i;

1796

int i;

1797

1798

cprv = NR_CPUS;

1798

cprv = NR_CPUS;

1799

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

1799

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

1800

ccur = rsp->levelcnt[i];

1800

ccur = rsp->levelcnt[i];

1801

rsp->levelspread[i] = (cprv + ccur - 1) / ccur;

1801

rsp->levelspread[i] = (cprv + ccur - 1) / ccur;

1802

cprv = ccur;

1802

cprv = ccur;

1803

}

1803

}

1804

}

1804

}

1805

#endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */

1805

#endif /* #else #ifdef CONFIG_RCU_FANOUT_EXACT */

1806

1807

/*

1807

/*

1808

* Helper function for rcu_init() that initializes one rcu_state structure.

1808

* Helper function for rcu_init() that initializes one rcu_state structure.

1809

*/

1809

*/

1810

static void __init rcu_init_one(struct rcu_state *rsp)

1810

static void __init rcu_init_one(struct rcu_state *rsp)

1811

{

1811

{

1812

static char *buf[] = { "rcu_node_level_0",

1812

static char *buf[] = { "rcu_node_level_0",

1813

"rcu_node_level_1",

1813

"rcu_node_level_1",

1814

"rcu_node_level_2",

1814

"rcu_node_level_2",

1815

"rcu_node_level_3" }; /* Match MAX_RCU_LVLS */

1815

"rcu_node_level_3" }; /* Match MAX_RCU_LVLS */

1816

int cpustride = 1;

1816

int cpustride = 1;

1817

int i;

1817

int i;

1818

int j;

1818

int j;

1819

struct rcu_node *rnp;

1819

struct rcu_node *rnp;

1820

1821

BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf)); /* Fix buf[] init! */

1821

BUILD_BUG_ON(MAX_RCU_LVLS > ARRAY_SIZE(buf)); /* Fix buf[] init! */

1822

1823

/* Initialize the level-tracking arrays. */

1823

/* Initialize the level-tracking arrays. */

1824

1825

for (i = 1; i < NUM_RCU_LVLS; i++)

1825

for (i = 1; i < NUM_RCU_LVLS; i++)

1826

rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];

1826

rsp->level[i] = rsp->level[i - 1] + rsp->levelcnt[i - 1];

1827

rcu_init_levelspread(rsp);

1827

rcu_init_levelspread(rsp);

1828

1829

/* Initialize the elements themselves, starting from the leaves. */

1829

/* Initialize the elements themselves, starting from the leaves. */

1830

1831

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

1831

for (i = NUM_RCU_LVLS - 1; i >= 0; i--) {

1832

cpustride *= rsp->levelspread[i];

1832

cpustride *= rsp->levelspread[i];

1833

rnp = rsp->level[i];

1833

rnp = rsp->level[i];

1834

for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {

1834

for (j = 0; j < rsp->levelcnt[i]; j++, rnp++) {

1835

raw_spin_lock_init(&rnp->lock);

1835

raw_spin_lock_init(&rnp->lock);

1836

lockdep_set_class_and_name(&rnp->lock,

1836

lockdep_set_class_and_name(&rnp->lock,

1837

&rcu_node_class[i], buf[i]);

1837

&rcu_node_class[i], buf[i]);

1838

rnp->gpnum = 0;

1838

rnp->gpnum = 0;

1839

rnp->qsmask = 0;

1839

rnp->qsmask = 0;

1840

rnp->qsmaskinit = 0;

1840

rnp->qsmaskinit = 0;

1841

rnp->grplo = j * cpustride;

1841

rnp->grplo = j * cpustride;

1842

rnp->grphi = (j + 1) * cpustride - 1;

1842

rnp->grphi = (j + 1) * cpustride - 1;

1843

if (rnp->grphi >= NR_CPUS)

1843

if (rnp->grphi >= NR_CPUS)

1844

rnp->grphi = NR_CPUS - 1;

1844

rnp->grphi = NR_CPUS - 1;

1845

if (i == 0) {

1845

if (i == 0) {

1846

rnp->grpnum = 0;

1846

rnp->grpnum = 0;

1847

rnp->grpmask = 0;

1847

rnp->grpmask = 0;

1848

rnp->parent = NULL;

1848

rnp->parent = NULL;

1849

} else {

1849

} else {

1850

rnp->grpnum = j % rsp->levelspread[i - 1];

1850

rnp->grpnum = j % rsp->levelspread[i - 1];

1851

rnp->grpmask = 1UL << rnp->grpnum;

1851

rnp->grpmask = 1UL << rnp->grpnum;

1852

rnp->parent = rsp->level[i - 1] +

1852

rnp->parent = rsp->level[i - 1] +

1853

j / rsp->levelspread[i - 1];

1853

j / rsp->levelspread[i - 1];

1854

}

1854

}

1855

rnp->level = i;

1855

rnp->level = i;

1856

INIT_LIST_HEAD(&rnp->blocked_tasks[0]);

1856

INIT_LIST_HEAD(&rnp->blocked_tasks[0]);

1857

INIT_LIST_HEAD(&rnp->blocked_tasks[1]);

1857

INIT_LIST_HEAD(&rnp->blocked_tasks[1]);

1858

INIT_LIST_HEAD(&rnp->blocked_tasks[2]);

1858

INIT_LIST_HEAD(&rnp->blocked_tasks[2]);

1859

INIT_LIST_HEAD(&rnp->blocked_tasks[3]);

1859

INIT_LIST_HEAD(&rnp->blocked_tasks[3]);

1860

}

1860

}

1861

}

1861

}

1862

1863

rnp = rsp->level[NUM_RCU_LVLS - 1];

1864

for_each_possible_cpu(i) {

1865

if (i > rnp->grphi)

1866

rnp++;

1867

rsp->rda[i]->mynode = rnp;

1868

rcu_boot_init_percpu_data(i, rsp);

1869

}

1862

}

1870

}

1863

1871

1864

/*

1872

/*

1865

* Helper macro for __rcu_init() and __rcu_init_preempt(). To be used

1873

* Helper macro for __rcu_init() and __rcu_init_preempt(). To be used

1866

* nowhere else! Assigns leaf node pointers into each CPU's rcu_data

1874

* nowhere else! Assigns leaf node pointers into each CPU's rcu_data

1867

* structure.

1875

* structure.

1868

*/

1876

*/

1869

#define RCU_INIT_FLAVOR(rsp, rcu_data) \

1877

#define RCU_INIT_FLAVOR(rsp, rcu_data) \

1870

do { \

1878

do { \

1871

int i; \

1879

int i; \

1872

int j; \

1873

struct rcu_node *rnp; \

1874

\

1880

\

1875

rcu_init_one(rsp); \

1876

rnp = (rsp)->level[NUM_RCU_LVLS - 1]; \

1877

j = 0; \

1878

for_each_possible_cpu(i) { \

1881

for_each_possible_cpu(i) { \

1879

if (i > rnp[j].grphi) \

1880

j++; \

1881

per_cpu(rcu_data, i).mynode = &rnp[j]; \

1882

(rsp)->rda[i] = &per_cpu(rcu_data, i); \

1882

(rsp)->rda[i] = &per_cpu(rcu_data, i); \

1883

rcu_boot_init_percpu_data(i, rsp); \

1884

} \

1883

} \

1884

rcu_init_one(rsp); \

1885

} while (0)

1885

} while (0)

1886

1887

void __init rcu_init(void)

1887

void __init rcu_init(void)

1888

{

1888

{

1889

int cpu;

1889

int cpu;

1890

1891

rcu_bootup_announce();

1891

rcu_bootup_announce();

1892

#ifdef CONFIG_RCU_CPU_STALL_DETECTOR

1892

#ifdef CONFIG_RCU_CPU_STALL_DETECTOR

1893

printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n");

1893

printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n");

1894

#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

1894

#endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */

1895

#if NUM_RCU_LVL_4 != 0

1895

#if NUM_RCU_LVL_4 != 0

1896

printk(KERN_INFO "Experimental four-level hierarchy is enabled.\n");

1896

printk(KERN_INFO "Experimental four-level hierarchy is enabled.\n");

1897

#endif /* #if NUM_RCU_LVL_4 != 0 */

1897

#endif /* #if NUM_RCU_LVL_4 != 0 */

1898

RCU_INIT_FLAVOR(&rcu_sched_state, rcu_sched_data);

1898

RCU_INIT_FLAVOR(&rcu_sched_state, rcu_sched_data);

1899

RCU_INIT_FLAVOR(&rcu_bh_state, rcu_bh_data);

1899

RCU_INIT_FLAVOR(&rcu_bh_state, rcu_bh_data);

1900

__rcu_init_preempt();

1900

__rcu_init_preempt();

1901

open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);

1901

open_softirq(RCU_SOFTIRQ, rcu_process_callbacks);

1902

1903

/*

1903

/*

1904

* We don't need protection against CPU-hotplug here because

1904

* We don't need protection against CPU-hotplug here because

1905

* this is called early in boot, before either interrupts

1905

* this is called early in boot, before either interrupts

GITLAB

rcu: move some code from macro to function

 /*
  * 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 <asm/atomic.h>
 #include <linux/bitops.h>
 #include <linux/module.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 "rcutree.h"
 /* Data structures. */
 static struct lock_class_key rcu_node_class[NUM_RCU_LVLS];
 #define RCU_STATE_INITIALIZER(name) { \
 	.level = { &name.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 */ \
 	}, \
 	.signaled = RCU_GP_IDLE, \
 	.gpnum = -300, \
 	.completed = -300, \
 	.onofflock = __RAW_SPIN_LOCK_UNLOCKED(&name.onofflock), \
 	.orphan_cbs_list = NULL, \
 	.orphan_cbs_tail = &name.orphan_cbs_list, \
 	.orphan_qlen = 0, \
 	.fqslock = __RAW_SPIN_LOCK_UNLOCKED(&name.fqslock), \
 	.n_force_qs = 0, \
 	.n_force_qs_ngp = 0, \
 }
 struct rcu_state rcu_sched_state = RCU_STATE_INITIALIZER(rcu_sched_state);
 DEFINE_PER_CPU(struct rcu_data, rcu_sched_data);
 struct rcu_state rcu_bh_state = RCU_STATE_INITIALIZER(rcu_bh_state);
 DEFINE_PER_CPU(struct rcu_data, rcu_bh_data);
 /*
  * 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.
  */
 void rcu_sched_qs(int cpu)
 {
 	struct rcu_data *rdp;
 	rdp = &per_cpu(rcu_sched_data, cpu);
 	rdp->passed_quiesc_completed = rdp->gpnum - 1;
 	barrier();
 	rdp->passed_quiesc = 1;
 	rcu_preempt_note_context_switch(cpu);
 }
 void rcu_bh_qs(int cpu)
 {
 	struct rcu_data *rdp;
 	rdp = &per_cpu(rcu_bh_data, cpu);
 	rdp->passed_quiesc_completed = rdp->gpnum - 1;
 	barrier();
 	rdp->passed_quiesc = 1;
 }
 #ifdef CONFIG_NO_HZ
 DEFINE_PER_CPU(struct rcu_dynticks, rcu_dynticks) = {
 	.dynticks_nesting = 1,
 	.dynticks = 1,
 };
 #endif /* #ifdef CONFIG_NO_HZ */
 static int blimit = 10;		/* Maximum callbacks per softirq. */
 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);
 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);
 /*
  * 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];
 }
 #ifdef CONFIG_SMP
 /*
  * 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, it is in a quiescent state.  We can
 	 * trust its state not to change because interrupts are disabled.
 	 */
 	if (cpu_is_offline(rdp->cpu)) {
 		rdp->offline_fqs++;
 		return 1;
 	}
 	/* If preemptable RCU, no point in sending reschedule IPI. */
 	if (rdp->preemptable)
 		return 0;
 	/* The CPU is online, so send it a reschedule IPI. */
 	if (rdp->cpu != smp_processor_id())
 		smp_send_reschedule(rdp->cpu);
 	else
 		set_need_resched();
 	rdp->resched_ipi++;
 	return 0;
 }
 #endif /* #ifdef CONFIG_SMP */
 #ifdef CONFIG_NO_HZ
 /**
  * rcu_enter_nohz - inform RCU that current CPU is entering nohz
  *
  * Enter nohz 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 nohz mode, a possibility
  * handled by rcu_irq_enter() and rcu_irq_exit()).
  */
 void rcu_enter_nohz(void)
 {
 	unsigned long flags;
 	struct rcu_dynticks *rdtp;
 	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	rdtp->dynticks++;
 	rdtp->dynticks_nesting--;
 	WARN_ON_ONCE(rdtp->dynticks & 0x1);
 	local_irq_restore(flags);
 }
 /*
  * rcu_exit_nohz - inform RCU that current CPU is leaving nohz
  *
  * Exit nohz mode, in other words, -enter- the mode in which RCU
  * read-side critical sections normally occur.
  */
 void rcu_exit_nohz(void)
 {
 	unsigned long flags;
 	struct rcu_dynticks *rdtp;
 	local_irq_save(flags);
 	rdtp = &__get_cpu_var(rcu_dynticks);
 	rdtp->dynticks++;
 	rdtp->dynticks_nesting++;
 	WARN_ON_ONCE(!(rdtp->dynticks & 0x1));
 	local_irq_restore(flags);
 	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 }
 /**
  * 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 & 0x1)
 		return;
 	rdtp->dynticks_nmi++;
 	WARN_ON_ONCE(!(rdtp->dynticks_nmi & 0x1));
 	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 }
 /**
  * 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 & 0x1)
 		return;
 	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
 	rdtp->dynticks_nmi++;
 	WARN_ON_ONCE(rdtp->dynticks_nmi & 0x1);
 }
 /**
  * rcu_irq_enter - inform RCU of entry to hard irq context
  *
  * If the CPU was idle with dynamic ticks active, this updates the
  * rdtp->dynticks to let the RCU handling know that the CPU is active.
  */
 void rcu_irq_enter(void)
 {
 	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);
 	if (rdtp->dynticks_nesting++)
 		return;
 	rdtp->dynticks++;
 	WARN_ON_ONCE(!(rdtp->dynticks & 0x1));
 	smp_mb(); /* CPUs seeing ++ must see later RCU read-side crit sects */
 }
 /**
  * rcu_irq_exit - inform RCU of exit from hard irq context
  *
  * If the CPU was idle with dynamic ticks active, update the rdp->dynticks
  * to put let the RCU handling be aware that the CPU is going back to idle
  * with no ticks.
  */
 void rcu_irq_exit(void)
 {
 	struct rcu_dynticks *rdtp = &__get_cpu_var(rcu_dynticks);
 	if (--rdtp->dynticks_nesting)
 		return;
 	smp_mb(); /* CPUs seeing ++ must see prior RCU read-side crit sects */
 	rdtp->dynticks++;
 	WARN_ON_ONCE(rdtp->dynticks & 0x1);
 	/* If the interrupt queued a callback, get out of dyntick mode. */
 	if (__get_cpu_var(rcu_sched_data).nxtlist ||
 	    __get_cpu_var(rcu_bh_data).nxtlist)
 		set_need_resched();
 }
 #ifdef CONFIG_SMP
 /*
  * 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)
 {
 	int ret;
 	int snap;
 	int snap_nmi;
 	snap = rdp->dynticks->dynticks;
 	snap_nmi = rdp->dynticks->dynticks_nmi;
 	smp_mb();	/* Order sampling of snap with end of grace period. */
 	rdp->dynticks_snap = snap;
 	rdp->dynticks_nmi_snap = snap_nmi;
 	ret = ((snap & 0x1) == 0) && ((snap_nmi & 0x1) == 0);
 	if (ret)
 		rdp->dynticks_fqs++;
 	return ret;
 }
 /*
  * 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)
 {
 	long curr;
 	long curr_nmi;
 	long snap;
 	long snap_nmi;
 	curr = rdp->dynticks->dynticks;
 	snap = rdp->dynticks_snap;
 	curr_nmi = rdp->dynticks->dynticks_nmi;
 	snap_nmi = rdp->dynticks_nmi_snap;
 	smp_mb(); /* force ordering with cpu entering/leaving dynticks. */
 	/*
 	 * 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 != snap || (curr & 0x1) == 0) &&
 	    (curr_nmi != snap_nmi || (curr_nmi & 0x1) == 0)) {
 		rdp->dynticks_fqs++;
 		return 1;
 	}
 	/* Go check for the CPU being offline. */
 	return rcu_implicit_offline_qs(rdp);
 }
 #endif /* #ifdef CONFIG_SMP */
 #else /* #ifdef CONFIG_NO_HZ */
 #ifdef CONFIG_SMP
 static int dyntick_save_progress_counter(struct rcu_data *rdp)
 {
 	return 0;
 }
 static int rcu_implicit_dynticks_qs(struct rcu_data *rdp)
 {
 	return rcu_implicit_offline_qs(rdp);
 }
 #endif /* #ifdef CONFIG_SMP */
 #endif /* #else #ifdef CONFIG_NO_HZ */
 #ifdef CONFIG_RCU_CPU_STALL_DETECTOR
 static void record_gp_stall_check_time(struct rcu_state *rsp)
 {
 	rsp->gp_start = jiffies;
 	rsp->jiffies_stall = jiffies + RCU_SECONDS_TILL_STALL_CHECK;
 }
 static void print_other_cpu_stall(struct rcu_state *rsp)
 {
 	int cpu;
 	long delta;
 	unsigned long flags;
 	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 + RCU_SECONDS_TILL_STALL_RECHECK;
 	/*
 	 * Now rat on any tasks that got kicked up to the root rcu_node
 	 * due to CPU offlining.
 	 */
 	rcu_print_task_stall(rnp);
 	raw_spin_unlock_irqrestore(&rnp->lock, flags);
 	/* OK, time to rat on our buddy... */
 	printk(KERN_ERR "INFO: RCU detected CPU stalls:");
 	rcu_for_each_leaf_node(rsp, rnp) {
 		raw_spin_lock_irqsave(&rnp->lock, flags);
 		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))
 				printk(" %d", rnp->grplo + cpu);
 	}
 	printk(" (detected by %d, t=%ld jiffies)\n",
 	       smp_processor_id(), (long)(jiffies - rsp->gp_start));
 	trigger_all_cpu_backtrace();
 	/* 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);
 	printk(KERN_ERR "INFO: RCU detected CPU %d stall (t=%lu jiffies)\n",
 			smp_processor_id(), jiffies - rsp->gp_start);
 	trigger_all_cpu_backtrace();
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	if (ULONG_CMP_GE(jiffies, rsp->jiffies_stall))
 		rsp->jiffies_stall =
 			jiffies + RCU_SECONDS_TILL_STALL_RECHECK;
 	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)
 {
 	long delta;
 	struct rcu_node *rnp;
 	delta = jiffies - rsp->jiffies_stall;
 	rnp = rdp->mynode;
 	if ((rnp->qsmask & rdp->grpmask) && delta >= 0) {
 		/* We haven't checked in, so go dump stack. */
 		print_cpu_stall(rsp);
 	} else if (rcu_gp_in_progress(rsp) && delta >= RCU_STALL_RAT_DELAY) {
 		/* They had two time units to dump stack, so complain. */
 		print_other_cpu_stall(rsp);
 	}
 }
 #else /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
 static void record_gp_stall_check_time(struct rcu_state *rsp)
 {
 }
 static void check_cpu_stall(struct rcu_state *rsp, struct rcu_data *rdp)
 {
 }
 #endif /* #else #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
 /*
  * 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) {
 		rdp->qs_pending = 1;
 		rdp->passed_quiesc = 0;
 		rdp->gpnum = rnp->gpnum;
 	}
 }
 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;
 	}
 }
 /*
  * 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.
  */
 static void
 rcu_start_gp(struct rcu_state *rsp, unsigned long flags)
 	__releases(rcu_get_root(rsp)->lock)
 {
 	struct rcu_data *rdp = rsp->rda[smp_processor_id()];
 	struct rcu_node *rnp = rcu_get_root(rsp);
 	if (!cpu_needs_another_gp(rsp, rdp) || rsp->fqs_active) {
 		if (cpu_needs_another_gp(rsp, rdp))
 			rsp->fqs_need_gp = 1;
 		if (rnp->completed == rsp->completed) {
 			raw_spin_unlock_irqrestore(&rnp->lock, flags);
 			return;
 		}
 		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->completed;
 			raw_spin_unlock(&rnp->lock); /* irqs remain disabled. */
 		}
 		local_irq_restore(flags);
 		return;
 	}
 	/* Advance to a new grace period and initialize state. */
 	rsp->gpnum++;
 	WARN_ON_ONCE(rsp->signaled == RCU_GP_INIT);
 	rsp->signaled = RCU_GP_INIT; /* Hold off force_quiescent_state. */
 	rsp->jiffies_force_qs = jiffies + RCU_JIFFIES_TILL_FORCE_QS;
 	record_gp_stall_check_time(rsp);
 	/* Special-case the common single-level case. */
 	if (NUM_RCU_NODES == 1) {
 		rcu_preempt_check_blocked_tasks(rnp);
 		rnp->qsmask = rnp->qsmaskinit;
 		rnp->gpnum = rsp->gpnum;
 		rnp->completed = rsp->completed;
 		rsp->signaled = RCU_SIGNAL_INIT; /* force_quiescent_state OK. */
 		rcu_start_gp_per_cpu(rsp, rnp, rdp);
 		raw_spin_unlock_irqrestore(&rnp->lock, flags);
 		return;
 	}
 	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);
 		raw_spin_unlock(&rnp->lock);	/* irqs remain disabled. */
 	}
 	rnp = rcu_get_root(rsp);
 	raw_spin_lock(&rnp->lock);		/* irqs already disabled. */
 	rsp->signaled = 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)
 {
 	WARN_ON_ONCE(!rcu_gp_in_progress(rsp));
 	rsp->completed = rsp->gpnum;
 	rsp->signaled = 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;
 		if (rnp->qsmask != 0 || rcu_preempted_readers(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 lastcomp)
 {
 	unsigned long flags;
 	unsigned long mask;
 	struct rcu_node *rnp;
 	rnp = rdp->mynode;
 	raw_spin_lock_irqsave(&rnp->lock, flags);
 	if (lastcomp != rnp->completed) {
 		/*
 		 * Someone beat us to it for this grace period, so leave.
 		 * The race with GP start is resolved by the fact that we
 		 * hold the leaf rcu_node lock, so that the per-CPU bits
 		 * cannot yet be initialized -- so we would simply find our
 		 * CPU's bit already cleared in rcu_report_qs_rnp() if this
 		 * race occurred.
 		 */
 		rdp->passed_quiesc = 0;	/* try again later! */
 		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_quiesc)
 		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_quiesc_completed);
 }
 #ifdef CONFIG_HOTPLUG_CPU
 /*
  * Move a dying CPU's RCU callbacks to the ->orphan_cbs_list for the
  * specified flavor of RCU.  The callbacks will be adopted by the next
  * _rcu_barrier() invocation or by the CPU_DEAD notifier, whichever
  * comes first.  Because this is invoked from the CPU_DYING notifier,
  * irqs are already disabled.
  */
 static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
 {
 	int i;
 	struct rcu_data *rdp = rsp->rda[smp_processor_id()];
 	if (rdp->nxtlist == NULL)
 		return;  /* irqs disabled, so comparison is stable. */
 	raw_spin_lock(&rsp->onofflock);  /* irqs already disabled. */
 	*rsp->orphan_cbs_tail = rdp->nxtlist;
 	rsp->orphan_cbs_tail = rdp->nxttail[RCU_NEXT_TAIL];
 	rdp->nxtlist = NULL;
 	for (i = 0; i < RCU_NEXT_SIZE; i++)
 		rdp->nxttail[i] = &rdp->nxtlist;
 	rsp->orphan_qlen += rdp->qlen;
 	rdp->qlen = 0;
 	raw_spin_unlock(&rsp->onofflock);  /* irqs remain disabled. */
 }
 /*
  * Adopt previously orphaned RCU callbacks.
  */
 static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
 {
 	unsigned long flags;
 	struct rcu_data *rdp;
 	raw_spin_lock_irqsave(&rsp->onofflock, flags);
 	rdp = rsp->rda[smp_processor_id()];
 	if (rsp->orphan_cbs_list == NULL) {
 		raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
 		return;
 	}
 	*rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_list;
 	rdp->nxttail[RCU_NEXT_TAIL] = rsp->orphan_cbs_tail;
 	rdp->qlen += rsp->orphan_qlen;
 	rsp->orphan_cbs_list = NULL;
 	rsp->orphan_cbs_tail = &rsp->orphan_cbs_list;
 	rsp->orphan_qlen = 0;
 	raw_spin_unlock_irqrestore(&rsp->onofflock, flags);
 }
 /*
  * Remove the outgoing CPU from the bitmasks in the rcu_node hierarchy
  * and move all callbacks from the outgoing CPU to the current one.
  */
 static void __rcu_offline_cpu(int cpu, struct rcu_state *rsp)
 {
 	unsigned long flags;
 	unsigned long mask;
 	int need_report = 0;
 	struct rcu_data *rdp = rsp->rda[cpu];
 	struct rcu_node *rnp;
 	/* 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. */
 	rnp = rdp->mynode;	/* this is the outgoing CPU's rnp. */
 	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);
 	rcu_adopt_orphan_cbs(rsp);
 }
 /*
  * Remove the specified CPU from the RCU hierarchy and move any pending
  * callbacks that it might have to the current CPU.  This code assumes
  * that at least one CPU in the system will remain running at all times.
  * Any attempt to offline -all- CPUs is likely to strand RCU callbacks.
  */
 static void rcu_offline_cpu(int cpu)
 {
 	__rcu_offline_cpu(cpu, &rcu_sched_state);
 	__rcu_offline_cpu(cpu, &rcu_bh_state);
 	rcu_preempt_offline_cpu(cpu);
 }
 #else /* #ifdef CONFIG_HOTPLUG_CPU */
 static void rcu_send_cbs_to_orphanage(struct rcu_state *rsp)
 {
 }
 static void rcu_adopt_orphan_cbs(struct rcu_state *rsp)
 {
 }
 static void rcu_offline_cpu(int cpu)
 {
 }
 #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 count;
 	/* If no callbacks are ready, just return.*/
 	if (!cpu_has_callbacks_ready_to_invoke(rdp))
 		return;
 	/*
 	 * Extract the list of ready callbacks, disabling to prevent
 	 * races with call_rcu() from interrupt handlers.
 	 */
 	local_irq_save(flags);
 	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 = 0;
 	while (list) {
 		next = list->next;
 		prefetch(next);
 		list->func(list);
 		list = next;
 		if (++count >= rdp->blimit)
 			break;
 	}
 	local_irq_save(flags);
 	/* Update count, and requeue any remaining callbacks. */
 	rdp->qlen -= 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-raise the RCU softirq if there are callbacks remaining. */
 	if (cpu_has_callbacks_ready_to_invoke(rdp))
 		raise_softirq(RCU_SOFTIRQ);
 }
 /*
  * 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 the RCU softirq handler.
  *
  * This function must be called with hardirqs disabled.  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)
 {
 	if (!rcu_pending(cpu))
 		return; /* if nothing for RCU to do. */
 	if (user ||
 	    (idle_cpu(cpu) && rcu_scheduler_active &&
 	     !in_softirq() && hardirq_count() <= (1 << HARDIRQ_SHIFT))) {
 		/*
 		 * 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);
 	raise_softirq(RCU_SOFTIRQ);
 }
 #ifdef CONFIG_SMP
 /*
  * Scan the leaf rcu_node structures, processing dyntick state for any that
  * have not yet encountered a quiescent state, using the function specified.
  * 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) {
 			raw_spin_unlock_irqrestore(&rnp->lock, flags);
 			continue;
 		}
 		cpu = rnp->grplo;
 		bit = 1;
 		for (; cpu <= rnp->grphi; cpu++, bit <<= 1) {
 			if ((rnp->qsmask & bit) != 0 && f(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);
 	}
 }
 /*
  * 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);
 	if (!rcu_gp_in_progress(rsp))
 		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! */
 		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->signaled) {
 	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->signaled = 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 */
 		return;
 	}
 	raw_spin_unlock(&rnp->lock);  /* irqs remain disabled */
 unlock_fqs_ret:
 	raw_spin_unlock_irqrestore(&rsp->fqslock, flags);
 }
 #else /* #ifdef CONFIG_SMP */
 static void force_quiescent_state(struct rcu_state *rsp, int relaxed)
 {
 	set_need_resched();
 }
 #endif /* #else #ifdef CONFIG_SMP */
 /*
  * This does the RCU processing work from softirq context 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. */
 	rcu_do_batch(rsp, rdp);
 }
 /*
  * Do softirq processing for the current CPU.
  */
 static void rcu_process_callbacks(struct softirq_action *unused)
 {
 	/*
 	 * Memory references from any prior RCU read-side critical sections
 	 * executed by the interrupted code must be seen before any RCU
 	 * grace-period manipulations below.
 	 */
 	smp_mb(); /* See above block comment. */
 	__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();
 	/*
 	 * Memory references from any later RCU read-side critical sections
 	 * executed by the interrupted code must be seen after any RCU
 	 * grace-period manipulations above.
 	 */
 	smp_mb(); /* See above block comment. */
 	/* If we are last CPU on way to dyntick-idle mode, accelerate it. */
 	rcu_needs_cpu_flush();
 }
 static void
 __call_rcu(struct rcu_head *head, void (*func)(struct rcu_head *rcu),
 	   struct rcu_state *rsp)
 {
 	unsigned long flags;
 	struct rcu_data *rdp;
 	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);
 	rdp = rsp->rda[smp_processor_id()];
 	rcu_process_gp_end(rsp, rdp);
 	check_for_new_grace_period(rsp, rdp);
 	/* Add the callback to our list. */
 	*rdp->nxttail[RCU_NEXT_TAIL] = head;
 	rdp->nxttail[RCU_NEXT_TAIL] = &head->next;
 	/* 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);  /* releases rnp_root->lock. */
 	}
 	/*
 	 * 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)) {
 		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);
 }
 EXPORT_SYMBOL_GPL(call_rcu_sched);
 /*
  * Queue an RCU 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);
 }
 EXPORT_SYMBOL_GPL(call_rcu_bh);
 /**
  * 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)
 {
 	struct rcu_synchronize rcu;
 	if (rcu_blocking_is_gp())
 		return;
 	init_completion(&rcu.completion);
 	/* Will wake me after RCU finished. */
 	call_rcu_sched(&rcu.head, wakeme_after_rcu);
 	/* Wait for it. */
 	wait_for_completion(&rcu.completion);
 }
 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)
 {
 	struct rcu_synchronize rcu;
 	if (rcu_blocking_is_gp())
 		return;
 	init_completion(&rcu.completion);
 	/* Will wake me after RCU finished. */
 	call_rcu_bh(&rcu.head, wakeme_after_rcu);
 	/* Wait for it. */
 	wait_for_completion(&rcu.completion);
 }
 EXPORT_SYMBOL_GPL(synchronize_rcu_bh);
 /*
  * 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 (rdp->qs_pending) {
 		/*
 		 * 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.
 		 */
 		if (!rdp->preemptable &&
 		    ULONG_CMP_LT(ACCESS_ONCE(rsp->jiffies_force_qs) - 1,
 				 jiffies))
 			set_need_resched();
 		rdp->n_rp_qs_pending++;
 		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_needs_cpu_quick_check(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_needs_cpu(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.
 	 */
 	atomic_set(&rcu_barrier_cpu_count, 1);
 	preempt_disable(); /* stop CPU_DYING from filling orphan_cbs_list */
 	rcu_adopt_orphan_cbs(rsp);
 	on_each_cpu(rcu_barrier_func, (void *)call_rcu_func, 1);
 	preempt_enable(); /* CPU_DYING can again fill orphan_cbs_list */
 	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 = 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 = 0;
 #ifdef CONFIG_NO_HZ
 	rdp->dynticks = &per_cpu(rcu_dynticks, cpu);
 #endif /* #ifdef CONFIG_NO_HZ */
 	rdp->cpu = cpu;
 	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 preemptable)
 {
 	unsigned long flags;
 	unsigned long mask;
 	struct rcu_data *rdp = 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->passed_quiesc = 0;  /* We could be racing with new GP, */
 	rdp->qs_pending = 1;	 /*  so set up to respond to current GP. */
 	rdp->beenonline = 1;	 /* We have now been online. */
 	rdp->preemptable = preemptable;
 	rdp->qlen_last_fqs_check = 0;
 	rdp->n_force_qs_snap = rsp->n_force_qs;
 	rdp->blimit = blimit;
 	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) {
 			rdp->gpnum = rnp->completed; /* if GP in progress... */
 			rdp->completed = rnp->completed;
 			rdp->passed_quiesc_completed = rnp->completed - 1;
 		}
 		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_online_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;
 	switch (action) {
 	case CPU_UP_PREPARE:
 	case CPU_UP_PREPARE_FROZEN:
 		rcu_online_cpu(cpu);
 		break;
 	case CPU_DYING:
 	case CPU_DYING_FROZEN:
 		/*
 		 * preempt_disable() in _rcu_barrier() prevents stop_machine(),
 		 * so when "on_each_cpu(rcu_barrier_func, (void *)type, 1);"
 		 * returns, all online cpus have queued rcu_barrier_func().
 		 * The dying CPU clears its cpu_online_mask bit and
 		 * moves all of its RCU callbacks to ->orphan_cbs_list
 		 * in the context of stop_machine(), so subsequent calls
 		 * to _rcu_barrier() will adopt these callbacks and only
 		 * then queue rcu_barrier_func() on all remaining CPUs.
 		 */
 		rcu_send_cbs_to_orphanage(&rcu_bh_state);
 		rcu_send_cbs_to_orphanage(&rcu_sched_state);
 		rcu_preempt_send_cbs_to_orphanage();
 		break;
 	case CPU_DEAD:
 	case CPU_DEAD_FROZEN:
 	case CPU_UP_CANCELED:
 	case CPU_UP_CANCELED_FROZEN:
 		rcu_offline_cpu(cpu);
 		break;
 	default:
 		break;
 	}
 	return NOTIFY_OK;
 }
 /*
  * 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;
 }
 #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)
 {
 	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->blocked_tasks[0]);
 			INIT_LIST_HEAD(&rnp->blocked_tasks[1]);
 			INIT_LIST_HEAD(&rnp->blocked_tasks[2]);
 			INIT_LIST_HEAD(&rnp->blocked_tasks[3]);
 		}
 	}
+	rnp = rsp->level[NUM_RCU_LVLS - 1];
+	for_each_possible_cpu(i) {
+		if (i > rnp->grphi)
+			rnp++;
+		rsp->rda[i]->mynode = rnp;
+		rcu_boot_init_percpu_data(i, rsp);
+	}
 }
 /*
  * Helper macro for __rcu_init() and __rcu_init_preempt().  To be used
  * nowhere else!  Assigns leaf node pointers into each CPU's rcu_data
  * structure.
  */
 #define RCU_INIT_FLAVOR(rsp, rcu_data) \
 do { \
 	int i; \
-	int j; \
-	struct rcu_node *rnp; \
 	\
-	rcu_init_one(rsp); \
-	rnp = (rsp)->level[NUM_RCU_LVLS - 1]; \
-	j = 0; \
 	for_each_possible_cpu(i) { \
-		if (i > rnp[j].grphi) \
-			j++; \
-		per_cpu(rcu_data, i).mynode = &rnp[j]; \
 		(rsp)->rda[i] = &per_cpu(rcu_data, i); \
-		rcu_boot_init_percpu_data(i, rsp); \
 	} \
+	rcu_init_one(rsp); \
 } while (0)
 void __init rcu_init(void)
 {
 	int cpu;
 	rcu_bootup_announce();
 #ifdef CONFIG_RCU_CPU_STALL_DETECTOR
 	printk(KERN_INFO "RCU-based detection of stalled CPUs is enabled.\n");
 #endif /* #ifdef CONFIG_RCU_CPU_STALL_DETECTOR */
 #if NUM_RCU_LVL_4 != 0
 	printk(KERN_INFO "Experimental four-level hierarchy is enabled.\n");
 #endif /* #if NUM_RCU_LVL_4 != 0 */
 	RCU_INIT_FLAVOR(&rcu_sched_state, rcu_sched_data);
 	RCU_INIT_FLAVOR(&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