aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--arch/s390/kernel/kprobes.c6
-rw-r--r--drivers/char/hw_random/intel-rng.c6
-rw-r--r--include/linux/stop_machine.h50
-rw-r--r--kernel/cpu.c16
-rw-r--r--kernel/module.c33
-rw-r--r--kernel/rcuclassic.c4
-rw-r--r--kernel/stop_machine.c288
-rw-r--r--kernel/trace/ftrace.c6
-rw-r--r--mm/page_alloc.c4
9 files changed, 199 insertions, 214 deletions
diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c
index 4f82e5b5f879..569079ec4ff0 100644
--- a/arch/s390/kernel/kprobes.c
+++ b/arch/s390/kernel/kprobes.c
@@ -197,7 +197,7 @@ void __kprobes arch_arm_kprobe(struct kprobe *p)
197 args.new = BREAKPOINT_INSTRUCTION; 197 args.new = BREAKPOINT_INSTRUCTION;
198 198
199 kcb->kprobe_status = KPROBE_SWAP_INST; 199 kcb->kprobe_status = KPROBE_SWAP_INST;
200 stop_machine_run(swap_instruction, &args, NR_CPUS); 200 stop_machine(swap_instruction, &args, NULL);
201 kcb->kprobe_status = status; 201 kcb->kprobe_status = status;
202} 202}
203 203
@@ -212,7 +212,7 @@ void __kprobes arch_disarm_kprobe(struct kprobe *p)
212 args.new = p->opcode; 212 args.new = p->opcode;
213 213
214 kcb->kprobe_status = KPROBE_SWAP_INST; 214 kcb->kprobe_status = KPROBE_SWAP_INST;
215 stop_machine_run(swap_instruction, &args, NR_CPUS); 215 stop_machine(swap_instruction, &args, NULL);
216 kcb->kprobe_status = status; 216 kcb->kprobe_status = status;
217} 217}
218 218
@@ -331,7 +331,7 @@ static int __kprobes kprobe_handler(struct pt_regs *regs)
331 * No kprobe at this address. The fault has not been 331 * No kprobe at this address. The fault has not been
332 * caused by a kprobe breakpoint. The race of breakpoint 332 * caused by a kprobe breakpoint. The race of breakpoint
333 * vs. kprobe remove does not exist because on s390 we 333 * vs. kprobe remove does not exist because on s390 we
334 * use stop_machine_run to arm/disarm the breakpoints. 334 * use stop_machine to arm/disarm the breakpoints.
335 */ 335 */
336 goto no_kprobe; 336 goto no_kprobe;
337 337
diff --git a/drivers/char/hw_random/intel-rng.c b/drivers/char/hw_random/intel-rng.c
index 27fdc0866496..8a2fce0756ec 100644
--- a/drivers/char/hw_random/intel-rng.c
+++ b/drivers/char/hw_random/intel-rng.c
@@ -241,7 +241,7 @@ static int __init intel_rng_hw_init(void *_intel_rng_hw)
241 struct intel_rng_hw *intel_rng_hw = _intel_rng_hw; 241 struct intel_rng_hw *intel_rng_hw = _intel_rng_hw;
242 u8 mfc, dvc; 242 u8 mfc, dvc;
243 243
244 /* interrupts disabled in stop_machine_run call */ 244 /* interrupts disabled in stop_machine call */
245 245
246 if (!(intel_rng_hw->fwh_dec_en1_val & FWH_F8_EN_MASK)) 246 if (!(intel_rng_hw->fwh_dec_en1_val & FWH_F8_EN_MASK))
247 pci_write_config_byte(intel_rng_hw->dev, 247 pci_write_config_byte(intel_rng_hw->dev,
@@ -365,10 +365,10 @@ static int __init mod_init(void)
365 * location with the Read ID command, all activity on the system 365 * location with the Read ID command, all activity on the system
366 * must be stopped until the state is back to normal. 366 * must be stopped until the state is back to normal.
367 * 367 *
368 * Use stop_machine_run because IPIs can be blocked by disabling 368 * Use stop_machine because IPIs can be blocked by disabling
369 * interrupts. 369 * interrupts.
370 */ 370 */
371 err = stop_machine_run(intel_rng_hw_init, intel_rng_hw, NR_CPUS); 371 err = stop_machine(intel_rng_hw_init, intel_rng_hw, NULL);
372 pci_dev_put(dev); 372 pci_dev_put(dev);
373 iounmap(intel_rng_hw->mem); 373 iounmap(intel_rng_hw->mem);
374 kfree(intel_rng_hw); 374 kfree(intel_rng_hw);
diff --git a/include/linux/stop_machine.h b/include/linux/stop_machine.h
index 5bfc553bdb21..f1cb0ba6d715 100644
--- a/include/linux/stop_machine.h
+++ b/include/linux/stop_machine.h
@@ -5,41 +5,43 @@
5 (and more). So the "read" side to such a lock is anything which 5 (and more). So the "read" side to such a lock is anything which
6 diables preeempt. */ 6 diables preeempt. */
7#include <linux/cpu.h> 7#include <linux/cpu.h>
8#include <linux/cpumask.h>
8#include <asm/system.h> 9#include <asm/system.h>
9 10
10#if defined(CONFIG_STOP_MACHINE) && defined(CONFIG_SMP) 11#if defined(CONFIG_STOP_MACHINE) && defined(CONFIG_SMP)
12
13/* Deprecated, but useful for transition. */
14#define ALL_CPUS ~0U
15
11/** 16/**
12 * stop_machine_run: freeze the machine on all CPUs and run this function 17 * stop_machine: freeze the machine on all CPUs and run this function
13 * @fn: the function to run 18 * @fn: the function to run
14 * @data: the data ptr for the @fn() 19 * @data: the data ptr for the @fn()
15 * @cpu: the cpu to run @fn() on (or any, if @cpu == NR_CPUS. 20 * @cpus: the cpus to run the @fn() on (NULL = any online cpu)
16 * 21 *
17 * Description: This causes a thread to be scheduled on every other cpu, 22 * Description: This causes a thread to be scheduled on every cpu,
18 * each of which disables interrupts, and finally interrupts are disabled 23 * each of which disables interrupts. The result is that noone is
19 * on the current CPU. The result is that noone is holding a spinlock 24 * holding a spinlock or inside any other preempt-disabled region when
20 * or inside any other preempt-disabled region when @fn() runs. 25 * @fn() runs.
21 * 26 *
22 * This can be thought of as a very heavy write lock, equivalent to 27 * This can be thought of as a very heavy write lock, equivalent to
23 * grabbing every spinlock in the kernel. */ 28 * grabbing every spinlock in the kernel. */
24int stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu); 29int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus);
25 30
26/** 31/**
27 * __stop_machine_run: freeze the machine on all CPUs and run this function 32 * __stop_machine: freeze the machine on all CPUs and run this function
28 * @fn: the function to run 33 * @fn: the function to run
29 * @data: the data ptr for the @fn 34 * @data: the data ptr for the @fn
30 * @cpu: the cpu to run @fn on (or any, if @cpu == NR_CPUS. 35 * @cpus: the cpus to run the @fn() on (NULL = any online cpu)
31 * 36 *
32 * Description: This is a special version of the above, which returns the 37 * Description: This is a special version of the above, which assumes cpus
33 * thread which has run @fn(): kthread_stop will return the return value 38 * won't come or go while it's being called. Used by hotplug cpu.
34 * of @fn(). Used by hotplug cpu.
35 */ 39 */
36struct task_struct *__stop_machine_run(int (*fn)(void *), void *data, 40int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus);
37 unsigned int cpu);
38
39#else 41#else
40 42
41static inline int stop_machine_run(int (*fn)(void *), void *data, 43static inline int stop_machine(int (*fn)(void *), void *data,
42 unsigned int cpu) 44 const cpumask_t *cpus)
43{ 45{
44 int ret; 46 int ret;
45 local_irq_disable(); 47 local_irq_disable();
@@ -48,4 +50,18 @@ static inline int stop_machine_run(int (*fn)(void *), void *data,
48 return ret; 50 return ret;
49} 51}
50#endif /* CONFIG_SMP */ 52#endif /* CONFIG_SMP */
53
54static inline int __deprecated stop_machine_run(int (*fn)(void *), void *data,
55 unsigned int cpu)
56{
57 /* If they don't care which cpu fn runs on, just pick one. */
58 if (cpu == NR_CPUS)
59 return stop_machine(fn, data, NULL);
60 else if (cpu == ~0U)
61 return stop_machine(fn, data, &cpu_possible_map);
62 else {
63 cpumask_t cpus = cpumask_of_cpu(cpu);
64 return stop_machine(fn, data, &cpus);
65 }
66}
51#endif /* _LINUX_STOP_MACHINE */ 67#endif /* _LINUX_STOP_MACHINE */
diff --git a/kernel/cpu.c b/kernel/cpu.c
index 10ba5f1004a5..29510d68338a 100644
--- a/kernel/cpu.c
+++ b/kernel/cpu.c
@@ -216,7 +216,6 @@ static int __ref take_cpu_down(void *_param)
216static int __ref _cpu_down(unsigned int cpu, int tasks_frozen) 216static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
217{ 217{
218 int err, nr_calls = 0; 218 int err, nr_calls = 0;
219 struct task_struct *p;
220 cpumask_t old_allowed, tmp; 219 cpumask_t old_allowed, tmp;
221 void *hcpu = (void *)(long)cpu; 220 void *hcpu = (void *)(long)cpu;
222 unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0; 221 unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0;
@@ -249,21 +248,18 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
249 cpus_setall(tmp); 248 cpus_setall(tmp);
250 cpu_clear(cpu, tmp); 249 cpu_clear(cpu, tmp);
251 set_cpus_allowed_ptr(current, &tmp); 250 set_cpus_allowed_ptr(current, &tmp);
251 tmp = cpumask_of_cpu(cpu);
252 252
253 p = __stop_machine_run(take_cpu_down, &tcd_param, cpu); 253 err = __stop_machine(take_cpu_down, &tcd_param, &tmp);
254 254 if (err) {
255 if (IS_ERR(p) || cpu_online(cpu)) {
256 /* CPU didn't die: tell everyone. Can't complain. */ 255 /* CPU didn't die: tell everyone. Can't complain. */
257 if (raw_notifier_call_chain(&cpu_chain, CPU_DOWN_FAILED | mod, 256 if (raw_notifier_call_chain(&cpu_chain, CPU_DOWN_FAILED | mod,
258 hcpu) == NOTIFY_BAD) 257 hcpu) == NOTIFY_BAD)
259 BUG(); 258 BUG();
260 259
261 if (IS_ERR(p)) { 260 goto out_allowed;
262 err = PTR_ERR(p);
263 goto out_allowed;
264 }
265 goto out_thread;
266 } 261 }
262 BUG_ON(cpu_online(cpu));
267 263
268 /* Wait for it to sleep (leaving idle task). */ 264 /* Wait for it to sleep (leaving idle task). */
269 while (!idle_cpu(cpu)) 265 while (!idle_cpu(cpu))
@@ -279,8 +275,6 @@ static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
279 275
280 check_for_tasks(cpu); 276 check_for_tasks(cpu);
281 277
282out_thread:
283 err = kthread_stop(p);
284out_allowed: 278out_allowed:
285 set_cpus_allowed_ptr(current, &old_allowed); 279 set_cpus_allowed_ptr(current, &old_allowed);
286out_release: 280out_release:
diff --git a/kernel/module.c b/kernel/module.c
index d8b5605132a0..61d212120df4 100644
--- a/kernel/module.c
+++ b/kernel/module.c
@@ -325,18 +325,6 @@ static unsigned long find_symbol(const char *name,
325 return -ENOENT; 325 return -ENOENT;
326} 326}
327 327
328/* lookup symbol in given range of kernel_symbols */
329static const struct kernel_symbol *lookup_symbol(const char *name,
330 const struct kernel_symbol *start,
331 const struct kernel_symbol *stop)
332{
333 const struct kernel_symbol *ks = start;
334 for (; ks < stop; ks++)
335 if (strcmp(ks->name, name) == 0)
336 return ks;
337 return NULL;
338}
339
340/* Search for module by name: must hold module_mutex. */ 328/* Search for module by name: must hold module_mutex. */
341static struct module *find_module(const char *name) 329static struct module *find_module(const char *name)
342{ 330{
@@ -690,7 +678,7 @@ static int try_stop_module(struct module *mod, int flags, int *forced)
690 if (flags & O_NONBLOCK) { 678 if (flags & O_NONBLOCK) {
691 struct stopref sref = { mod, flags, forced }; 679 struct stopref sref = { mod, flags, forced };
692 680
693 return stop_machine_run(__try_stop_module, &sref, NR_CPUS); 681 return stop_machine(__try_stop_module, &sref, NULL);
694 } else { 682 } else {
695 /* We don't need to stop the machine for this. */ 683 /* We don't need to stop the machine for this. */
696 mod->state = MODULE_STATE_GOING; 684 mod->state = MODULE_STATE_GOING;
@@ -1428,7 +1416,7 @@ static int __unlink_module(void *_mod)
1428static void free_module(struct module *mod) 1416static void free_module(struct module *mod)
1429{ 1417{
1430 /* Delete from various lists */ 1418 /* Delete from various lists */
1431 stop_machine_run(__unlink_module, mod, NR_CPUS); 1419 stop_machine(__unlink_module, mod, NULL);
1432 remove_notes_attrs(mod); 1420 remove_notes_attrs(mod);
1433 remove_sect_attrs(mod); 1421 remove_sect_attrs(mod);
1434 mod_kobject_remove(mod); 1422 mod_kobject_remove(mod);
@@ -1703,6 +1691,19 @@ static void setup_modinfo(struct module *mod, Elf_Shdr *sechdrs,
1703} 1691}
1704 1692
1705#ifdef CONFIG_KALLSYMS 1693#ifdef CONFIG_KALLSYMS
1694
1695/* lookup symbol in given range of kernel_symbols */
1696static const struct kernel_symbol *lookup_symbol(const char *name,
1697 const struct kernel_symbol *start,
1698 const struct kernel_symbol *stop)
1699{
1700 const struct kernel_symbol *ks = start;
1701 for (; ks < stop; ks++)
1702 if (strcmp(ks->name, name) == 0)
1703 return ks;
1704 return NULL;
1705}
1706
1706static int is_exported(const char *name, const struct module *mod) 1707static int is_exported(const char *name, const struct module *mod)
1707{ 1708{
1708 if (!mod && lookup_symbol(name, __start___ksymtab, __stop___ksymtab)) 1709 if (!mod && lookup_symbol(name, __start___ksymtab, __stop___ksymtab))
@@ -2196,7 +2197,7 @@ static struct module *load_module(void __user *umod,
2196 /* Now sew it into the lists so we can get lockdep and oops 2197 /* Now sew it into the lists so we can get lockdep and oops
2197 * info during argument parsing. Noone should access us, since 2198 * info during argument parsing. Noone should access us, since
2198 * strong_try_module_get() will fail. */ 2199 * strong_try_module_get() will fail. */
2199 stop_machine_run(__link_module, mod, NR_CPUS); 2200 stop_machine(__link_module, mod, NULL);
2200 2201
2201 /* Size of section 0 is 0, so this works well if no params */ 2202 /* Size of section 0 is 0, so this works well if no params */
2202 err = parse_args(mod->name, mod->args, 2203 err = parse_args(mod->name, mod->args,
@@ -2230,7 +2231,7 @@ static struct module *load_module(void __user *umod,
2230 return mod; 2231 return mod;
2231 2232
2232 unlink: 2233 unlink:
2233 stop_machine_run(__unlink_module, mod, NR_CPUS); 2234 stop_machine(__unlink_module, mod, NULL);
2234 module_arch_cleanup(mod); 2235 module_arch_cleanup(mod);
2235 cleanup: 2236 cleanup:
2236 kobject_del(&mod->mkobj.kobj); 2237 kobject_del(&mod->mkobj.kobj);
diff --git a/kernel/rcuclassic.c b/kernel/rcuclassic.c
index 6f8696c502f4..aad93cdc9f68 100644
--- a/kernel/rcuclassic.c
+++ b/kernel/rcuclassic.c
@@ -91,8 +91,8 @@ static void force_quiescent_state(struct rcu_data *rdp,
91 * rdp->cpu is the current cpu. 91 * rdp->cpu is the current cpu.
92 * 92 *
93 * cpu_online_map is updated by the _cpu_down() 93 * cpu_online_map is updated by the _cpu_down()
94 * using stop_machine_run(). Since we're in irqs disabled 94 * using __stop_machine(). Since we're in irqs disabled
95 * section, stop_machine_run() is not exectuting, hence 95 * section, __stop_machine() is not exectuting, hence
96 * the cpu_online_map is stable. 96 * the cpu_online_map is stable.
97 * 97 *
98 * However, a cpu might have been offlined _just_ before 98 * However, a cpu might have been offlined _just_ before
diff --git a/kernel/stop_machine.c b/kernel/stop_machine.c
index 738b411ff2d3..e446c7c7d6a9 100644
--- a/kernel/stop_machine.c
+++ b/kernel/stop_machine.c
@@ -1,4 +1,4 @@
1/* Copyright 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation. 1/* Copyright 2008, 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation.
2 * GPL v2 and any later version. 2 * GPL v2 and any later version.
3 */ 3 */
4#include <linux/cpu.h> 4#include <linux/cpu.h>
@@ -13,204 +13,178 @@
13#include <asm/atomic.h> 13#include <asm/atomic.h>
14#include <asm/uaccess.h> 14#include <asm/uaccess.h>
15 15
16/* Since we effect priority and affinity (both of which are visible 16/* This controls the threads on each CPU. */
17 * to, and settable by outside processes) we do indirection via a
18 * kthread. */
19
20/* Thread to stop each CPU in user context. */
21enum stopmachine_state { 17enum stopmachine_state {
22 STOPMACHINE_WAIT, 18 /* Dummy starting state for thread. */
19 STOPMACHINE_NONE,
20 /* Awaiting everyone to be scheduled. */
23 STOPMACHINE_PREPARE, 21 STOPMACHINE_PREPARE,
22 /* Disable interrupts. */
24 STOPMACHINE_DISABLE_IRQ, 23 STOPMACHINE_DISABLE_IRQ,
24 /* Run the function */
25 STOPMACHINE_RUN,
26 /* Exit */
25 STOPMACHINE_EXIT, 27 STOPMACHINE_EXIT,
26}; 28};
29static enum stopmachine_state state;
27 30
28static enum stopmachine_state stopmachine_state; 31struct stop_machine_data {
29static unsigned int stopmachine_num_threads; 32 int (*fn)(void *);
30static atomic_t stopmachine_thread_ack; 33 void *data;
31 34 int fnret;
32static int stopmachine(void *cpu) 35};
33{
34 int irqs_disabled = 0;
35 int prepared = 0;
36 cpumask_of_cpu_ptr(cpumask, (int)(long)cpu);
37
38 set_cpus_allowed_ptr(current, cpumask);
39
40 /* Ack: we are alive */
41 smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */
42 atomic_inc(&stopmachine_thread_ack);
43
44 /* Simple state machine */
45 while (stopmachine_state != STOPMACHINE_EXIT) {
46 if (stopmachine_state == STOPMACHINE_DISABLE_IRQ
47 && !irqs_disabled) {
48 local_irq_disable();
49 hard_irq_disable();
50 irqs_disabled = 1;
51 /* Ack: irqs disabled. */
52 smp_mb(); /* Must read state first. */
53 atomic_inc(&stopmachine_thread_ack);
54 } else if (stopmachine_state == STOPMACHINE_PREPARE
55 && !prepared) {
56 /* Everyone is in place, hold CPU. */
57 preempt_disable();
58 prepared = 1;
59 smp_mb(); /* Must read state first. */
60 atomic_inc(&stopmachine_thread_ack);
61 }
62 /* Yield in first stage: migration threads need to
63 * help our sisters onto their CPUs. */
64 if (!prepared && !irqs_disabled)
65 yield();
66 cpu_relax();
67 }
68
69 /* Ack: we are exiting. */
70 smp_mb(); /* Must read state first. */
71 atomic_inc(&stopmachine_thread_ack);
72
73 if (irqs_disabled)
74 local_irq_enable();
75 if (prepared)
76 preempt_enable();
77 36
78 return 0; 37/* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */
79} 38static unsigned int num_threads;
39static atomic_t thread_ack;
40static struct completion finished;
41static DEFINE_MUTEX(lock);
80 42
81/* Change the thread state */ 43static void set_state(enum stopmachine_state newstate)
82static void stopmachine_set_state(enum stopmachine_state state)
83{ 44{
84 atomic_set(&stopmachine_thread_ack, 0); 45 /* Reset ack counter. */
46 atomic_set(&thread_ack, num_threads);
85 smp_wmb(); 47 smp_wmb();
86 stopmachine_state = state; 48 state = newstate;
87 while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads)
88 cpu_relax();
89} 49}
90 50
91static int stop_machine(void) 51/* Last one to ack a state moves to the next state. */
52static void ack_state(void)
92{ 53{
93 int i, ret = 0; 54 if (atomic_dec_and_test(&thread_ack)) {
94 55 /* If we're the last one to ack the EXIT, we're finished. */
95 atomic_set(&stopmachine_thread_ack, 0); 56 if (state == STOPMACHINE_EXIT)
96 stopmachine_num_threads = 0; 57 complete(&finished);
97 stopmachine_state = STOPMACHINE_WAIT; 58 else
98 59 set_state(state + 1);
99 for_each_online_cpu(i) {
100 if (i == raw_smp_processor_id())
101 continue;
102 ret = kernel_thread(stopmachine, (void *)(long)i,CLONE_KERNEL);
103 if (ret < 0)
104 break;
105 stopmachine_num_threads++;
106 }
107
108 /* Wait for them all to come to life. */
109 while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads) {
110 yield();
111 cpu_relax();
112 } 60 }
61}
113 62
114 /* If some failed, kill them all. */ 63/* This is the actual thread which stops the CPU. It exits by itself rather
115 if (ret < 0) { 64 * than waiting for kthread_stop(), because it's easier for hotplug CPU. */
116 stopmachine_set_state(STOPMACHINE_EXIT); 65static int stop_cpu(struct stop_machine_data *smdata)
117 return ret; 66{
118 } 67 enum stopmachine_state curstate = STOPMACHINE_NONE;
68 int uninitialized_var(ret);
119 69
120 /* Now they are all started, make them hold the CPUs, ready. */ 70 /* Simple state machine */
121 preempt_disable(); 71 do {
122 stopmachine_set_state(STOPMACHINE_PREPARE); 72 /* Chill out and ensure we re-read stopmachine_state. */
73 cpu_relax();
74 if (state != curstate) {
75 curstate = state;
76 switch (curstate) {
77 case STOPMACHINE_DISABLE_IRQ:
78 local_irq_disable();
79 hard_irq_disable();
80 break;
81 case STOPMACHINE_RUN:
82 /* |= allows error detection if functions on
83 * multiple CPUs. */
84 smdata->fnret |= smdata->fn(smdata->data);
85 break;
86 default:
87 break;
88 }
89 ack_state();
90 }
91 } while (curstate != STOPMACHINE_EXIT);
123 92
124 /* Make them disable irqs. */ 93 local_irq_enable();
125 local_irq_disable(); 94 do_exit(0);
126 hard_irq_disable(); 95}
127 stopmachine_set_state(STOPMACHINE_DISABLE_IRQ);
128 96
97/* Callback for CPUs which aren't supposed to do anything. */
98static int chill(void *unused)
99{
129 return 0; 100 return 0;
130} 101}
131 102
132static void restart_machine(void) 103int __stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus)
133{ 104{
134 stopmachine_set_state(STOPMACHINE_EXIT); 105 int i, err;
135 local_irq_enable(); 106 struct stop_machine_data active, idle;
136 preempt_enable_no_resched(); 107 struct task_struct **threads;
137} 108
109 active.fn = fn;
110 active.data = data;
111 active.fnret = 0;
112 idle.fn = chill;
113 idle.data = NULL;
114
115 /* This could be too big for stack on large machines. */
116 threads = kcalloc(NR_CPUS, sizeof(threads[0]), GFP_KERNEL);
117 if (!threads)
118 return -ENOMEM;
119
120 /* Set up initial state. */
121 mutex_lock(&lock);
122 init_completion(&finished);
123 num_threads = num_online_cpus();
124 set_state(STOPMACHINE_PREPARE);
138 125
139struct stop_machine_data { 126 for_each_online_cpu(i) {
140 int (*fn)(void *); 127 struct stop_machine_data *smdata = &idle;
141 void *data; 128 struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
142 struct completion done;
143};
144 129
145static int do_stop(void *_smdata) 130 if (!cpus) {
146{ 131 if (i == first_cpu(cpu_online_map))
147 struct stop_machine_data *smdata = _smdata; 132 smdata = &active;
148 int ret; 133 } else {
134 if (cpu_isset(i, *cpus))
135 smdata = &active;
136 }
149 137
150 ret = stop_machine(); 138 threads[i] = kthread_create((void *)stop_cpu, smdata, "kstop%u",
151 if (ret == 0) { 139 i);
152 ret = smdata->fn(smdata->data); 140 if (IS_ERR(threads[i])) {
153 restart_machine(); 141 err = PTR_ERR(threads[i]);
154 } 142 threads[i] = NULL;
143 goto kill_threads;
144 }
155 145
156 /* We're done: you can kthread_stop us now */ 146 /* Place it onto correct cpu. */
157 complete(&smdata->done); 147 kthread_bind(threads[i], i);
158 148
159 /* Wait for kthread_stop */ 149 /* Make it highest prio. */
160 set_current_state(TASK_INTERRUPTIBLE); 150 if (sched_setscheduler_nocheck(threads[i], SCHED_FIFO, &param))
161 while (!kthread_should_stop()) { 151 BUG();
162 schedule();
163 set_current_state(TASK_INTERRUPTIBLE);
164 } 152 }
165 __set_current_state(TASK_RUNNING);
166 return ret;
167}
168 153
169struct task_struct *__stop_machine_run(int (*fn)(void *), void *data, 154 /* We've created all the threads. Wake them all: hold this CPU so one
170 unsigned int cpu) 155 * doesn't hit this CPU until we're ready. */
171{ 156 get_cpu();
172 static DEFINE_MUTEX(stopmachine_mutex); 157 for_each_online_cpu(i)
173 struct stop_machine_data smdata; 158 wake_up_process(threads[i]);
174 struct task_struct *p;
175 159
176 smdata.fn = fn; 160 /* This will release the thread on our CPU. */
177 smdata.data = data; 161 put_cpu();
178 init_completion(&smdata.done); 162 wait_for_completion(&finished);
163 mutex_unlock(&lock);
179 164
180 mutex_lock(&stopmachine_mutex); 165 kfree(threads);
181 166
182 /* If they don't care which CPU fn runs on, bind to any online one. */ 167 return active.fnret;
183 if (cpu == NR_CPUS)
184 cpu = raw_smp_processor_id();
185 168
186 p = kthread_create(do_stop, &smdata, "kstopmachine"); 169kill_threads:
187 if (!IS_ERR(p)) { 170 for_each_online_cpu(i)
188 struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; 171 if (threads[i])
172 kthread_stop(threads[i]);
173 mutex_unlock(&lock);
189 174
190 /* One high-prio thread per cpu. We'll do this one. */ 175 kfree(threads);
191 sched_setscheduler_nocheck(p, SCHED_FIFO, &param); 176 return err;
192 kthread_bind(p, cpu);
193 wake_up_process(p);
194 wait_for_completion(&smdata.done);
195 }
196 mutex_unlock(&stopmachine_mutex);
197 return p;
198} 177}
199 178
200int stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu) 179int stop_machine(int (*fn)(void *), void *data, const cpumask_t *cpus)
201{ 180{
202 struct task_struct *p;
203 int ret; 181 int ret;
204 182
205 /* No CPUs can come up or down during this. */ 183 /* No CPUs can come up or down during this. */
206 get_online_cpus(); 184 get_online_cpus();
207 p = __stop_machine_run(fn, data, cpu); 185 ret = __stop_machine(fn, data, cpus);
208 if (!IS_ERR(p))
209 ret = kthread_stop(p);
210 else
211 ret = PTR_ERR(p);
212 put_online_cpus(); 186 put_online_cpus();
213 187
214 return ret; 188 return ret;
215} 189}
216EXPORT_SYMBOL_GPL(stop_machine_run); 190EXPORT_SYMBOL_GPL(stop_machine);
diff --git a/kernel/trace/ftrace.c b/kernel/trace/ftrace.c
index 4231a3dc224a..f6e3af31b403 100644
--- a/kernel/trace/ftrace.c
+++ b/kernel/trace/ftrace.c
@@ -587,7 +587,7 @@ static int __ftrace_modify_code(void *data)
587 587
588static void ftrace_run_update_code(int command) 588static void ftrace_run_update_code(int command)
589{ 589{
590 stop_machine_run(__ftrace_modify_code, &command, NR_CPUS); 590 stop_machine(__ftrace_modify_code, &command, NULL);
591} 591}
592 592
593void ftrace_disable_daemon(void) 593void ftrace_disable_daemon(void)
@@ -787,7 +787,7 @@ static int ftrace_update_code(void)
787 !ftrace_enabled || !ftraced_trigger) 787 !ftrace_enabled || !ftraced_trigger)
788 return 0; 788 return 0;
789 789
790 stop_machine_run(__ftrace_update_code, NULL, NR_CPUS); 790 stop_machine(__ftrace_update_code, NULL, NULL);
791 791
792 return 1; 792 return 1;
793} 793}
@@ -1564,7 +1564,7 @@ static int __init ftrace_dynamic_init(void)
1564 1564
1565 addr = (unsigned long)ftrace_record_ip; 1565 addr = (unsigned long)ftrace_record_ip;
1566 1566
1567 stop_machine_run(ftrace_dyn_arch_init, &addr, NR_CPUS); 1567 stop_machine(ftrace_dyn_arch_init, &addr, NULL);
1568 1568
1569 /* ftrace_dyn_arch_init places the return code in addr */ 1569 /* ftrace_dyn_arch_init places the return code in addr */
1570 if (addr) { 1570 if (addr) {
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 6da667274df5..3cf3d05b6bd4 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -2372,7 +2372,7 @@ static void build_zonelist_cache(pg_data_t *pgdat)
2372 2372
2373#endif /* CONFIG_NUMA */ 2373#endif /* CONFIG_NUMA */
2374 2374
2375/* return values int ....just for stop_machine_run() */ 2375/* return values int ....just for stop_machine() */
2376static int __build_all_zonelists(void *dummy) 2376static int __build_all_zonelists(void *dummy)
2377{ 2377{
2378 int nid; 2378 int nid;
@@ -2397,7 +2397,7 @@ void build_all_zonelists(void)
2397 } else { 2397 } else {
2398 /* we have to stop all cpus to guarantee there is no user 2398 /* we have to stop all cpus to guarantee there is no user
2399 of zonelist */ 2399 of zonelist */
2400 stop_machine_run(__build_all_zonelists, NULL, NR_CPUS); 2400 stop_machine(__build_all_zonelists, NULL, NULL);
2401 /* cpuset refresh routine should be here */ 2401 /* cpuset refresh routine should be here */
2402 } 2402 }
2403 vm_total_pages = nr_free_pagecache_pages(); 2403 vm_total_pages = nr_free_pagecache_pages();