-/* Copyright 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation.
+/* Copyright 2008, 2005 Rusty Russell rusty@rustcorp.com.au IBM Corporation.
* GPL v2 and any later version.
*/
#include <linux/cpu.h>
#include <asm/atomic.h>
#include <asm/uaccess.h>
-/* Since we effect priority and affinity (both of which are visible
- * to, and settable by outside processes) we do indirection via a
- * kthread. */
-
-/* Thread to stop each CPU in user context. */
+/* This controls the threads on each CPU. */
enum stopmachine_state {
- STOPMACHINE_WAIT,
+ /* Dummy starting state for thread. */
+ STOPMACHINE_NONE,
+ /* Awaiting everyone to be scheduled. */
STOPMACHINE_PREPARE,
+ /* Disable interrupts. */
STOPMACHINE_DISABLE_IRQ,
+ /* Run the function */
STOPMACHINE_RUN,
+ /* Exit */
STOPMACHINE_EXIT,
};
+static enum stopmachine_state state;
struct stop_machine_data {
int (*fn)(void *);
void *data;
- struct completion done;
- int run_all;
-} smdata;
-
-static enum stopmachine_state stopmachine_state;
-static unsigned int stopmachine_num_threads;
-static atomic_t stopmachine_thread_ack;
+ int fnret;
+};
-static int stopmachine(void *cpu)
+/* Like num_online_cpus(), but hotplug cpu uses us, so we need this. */
+static unsigned int num_threads;
+static atomic_t thread_ack;
+static DEFINE_MUTEX(lock);
+/* setup_lock protects refcount, stop_machine_wq and stop_machine_work. */
+static DEFINE_MUTEX(setup_lock);
+/* Users of stop_machine. */
+static int refcount;
+static struct workqueue_struct *stop_machine_wq;
+static struct stop_machine_data active, idle;
+static const struct cpumask *active_cpus;
+static void *stop_machine_work;
+
+static void set_state(enum stopmachine_state newstate)
{
- int irqs_disabled = 0;
- int prepared = 0;
- int ran = 0;
- cpumask_of_cpu_ptr(cpumask, (int)(long)cpu);
-
- set_cpus_allowed_ptr(current, cpumask);
-
- /* Ack: we are alive */
- smp_mb(); /* Theoretically the ack = 0 might not be on this CPU yet. */
- atomic_inc(&stopmachine_thread_ack);
-
- /* Simple state machine */
- while (stopmachine_state != STOPMACHINE_EXIT) {
- if (stopmachine_state == STOPMACHINE_DISABLE_IRQ
- && !irqs_disabled) {
- local_irq_disable();
- hard_irq_disable();
- irqs_disabled = 1;
- /* Ack: irqs disabled. */
- smp_mb(); /* Must read state first. */
- atomic_inc(&stopmachine_thread_ack);
- } else if (stopmachine_state == STOPMACHINE_PREPARE
- && !prepared) {
- /* Everyone is in place, hold CPU. */
- preempt_disable();
- prepared = 1;
- smp_mb(); /* Must read state first. */
- atomic_inc(&stopmachine_thread_ack);
- } else if (stopmachine_state == STOPMACHINE_RUN && !ran) {
- smdata.fn(smdata.data);
- ran = 1;
- smp_mb(); /* Must read state first. */
- atomic_inc(&stopmachine_thread_ack);
- }
- /* Yield in first stage: migration threads need to
- * help our sisters onto their CPUs. */
- if (!prepared && !irqs_disabled)
- yield();
- cpu_relax();
- }
-
- /* Ack: we are exiting. */
- smp_mb(); /* Must read state first. */
- atomic_inc(&stopmachine_thread_ack);
-
- if (irqs_disabled)
- local_irq_enable();
- if (prepared)
- preempt_enable();
-
- return 0;
+ /* Reset ack counter. */
+ atomic_set(&thread_ack, num_threads);
+ smp_wmb();
+ state = newstate;
}
-/* Change the thread state */
-static void stopmachine_set_state(enum stopmachine_state state)
+/* Last one to ack a state moves to the next state. */
+static void ack_state(void)
{
- atomic_set(&stopmachine_thread_ack, 0);
- smp_wmb();
- stopmachine_state = state;
- while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads)
- cpu_relax();
+ if (atomic_dec_and_test(&thread_ack))
+ set_state(state + 1);
}
-static int stop_machine(void)
+/* This is the actual function which stops the CPU. It runs
+ * in the context of a dedicated stopmachine workqueue. */
+static void stop_cpu(struct work_struct *unused)
{
- int i, ret = 0;
-
- atomic_set(&stopmachine_thread_ack, 0);
- stopmachine_num_threads = 0;
- stopmachine_state = STOPMACHINE_WAIT;
-
- for_each_online_cpu(i) {
- if (i == raw_smp_processor_id())
- continue;
- ret = kernel_thread(stopmachine, (void *)(long)i,CLONE_KERNEL);
- if (ret < 0)
- break;
- stopmachine_num_threads++;
+ enum stopmachine_state curstate = STOPMACHINE_NONE;
+ struct stop_machine_data *smdata = &idle;
+ int cpu = smp_processor_id();
+ int err;
+
+ if (!active_cpus) {
+ if (cpu == cpumask_first(cpu_online_mask))
+ smdata = &active;
+ } else {
+ if (cpumask_test_cpu(cpu, active_cpus))
+ smdata = &active;
}
-
- /* Wait for them all to come to life. */
- while (atomic_read(&stopmachine_thread_ack) != stopmachine_num_threads) {
- yield();
+ /* Simple state machine */
+ do {
+ /* Chill out and ensure we re-read stopmachine_state. */
cpu_relax();
- }
-
- /* If some failed, kill them all. */
- if (ret < 0) {
- stopmachine_set_state(STOPMACHINE_EXIT);
- return ret;
- }
-
- /* Now they are all started, make them hold the CPUs, ready. */
- preempt_disable();
- stopmachine_set_state(STOPMACHINE_PREPARE);
-
- /* Make them disable irqs. */
- local_irq_disable();
- hard_irq_disable();
- stopmachine_set_state(STOPMACHINE_DISABLE_IRQ);
-
- return 0;
-}
+ if (state != curstate) {
+ curstate = state;
+ switch (curstate) {
+ case STOPMACHINE_DISABLE_IRQ:
+ local_irq_disable();
+ hard_irq_disable();
+ break;
+ case STOPMACHINE_RUN:
+ /* On multiple CPUs only a single error code
+ * is needed to tell that something failed. */
+ err = smdata->fn(smdata->data);
+ if (err)
+ smdata->fnret = err;
+ break;
+ default:
+ break;
+ }
+ ack_state();
+ }
+ } while (curstate != STOPMACHINE_EXIT);
-static void restart_machine(void)
-{
- stopmachine_set_state(STOPMACHINE_EXIT);
local_irq_enable();
- preempt_enable_no_resched();
}
-static void run_other_cpus(void)
+/* Callback for CPUs which aren't supposed to do anything. */
+static int chill(void *unused)
{
- stopmachine_set_state(STOPMACHINE_RUN);
+ return 0;
}
-static int do_stop(void *_smdata)
+int stop_machine_create(void)
{
- struct stop_machine_data *smdata = _smdata;
- int ret;
-
- ret = stop_machine();
- if (ret == 0) {
- ret = smdata->fn(smdata->data);
- if (smdata->run_all)
- run_other_cpus();
- restart_machine();
- }
-
- /* We're done: you can kthread_stop us now */
- complete(&smdata->done);
+ mutex_lock(&setup_lock);
+ if (refcount)
+ goto done;
+ stop_machine_wq = create_rt_workqueue("kstop");
+ if (!stop_machine_wq)
+ goto err_out;
+ stop_machine_work = alloc_percpu(struct work_struct);
+ if (!stop_machine_work)
+ goto err_out;
+done:
+ refcount++;
+ mutex_unlock(&setup_lock);
+ return 0;
- /* Wait for kthread_stop */
- set_current_state(TASK_INTERRUPTIBLE);
- while (!kthread_should_stop()) {
- schedule();
- set_current_state(TASK_INTERRUPTIBLE);
- }
- __set_current_state(TASK_RUNNING);
- return ret;
+err_out:
+ if (stop_machine_wq)
+ destroy_workqueue(stop_machine_wq);
+ mutex_unlock(&setup_lock);
+ return -ENOMEM;
}
+EXPORT_SYMBOL_GPL(stop_machine_create);
-struct task_struct *__stop_machine_run(int (*fn)(void *), void *data,
- unsigned int cpu)
+void stop_machine_destroy(void)
{
- static DEFINE_MUTEX(stopmachine_mutex);
- struct stop_machine_data smdata;
- struct task_struct *p;
-
- mutex_lock(&stopmachine_mutex);
-
- smdata.fn = fn;
- smdata.data = data;
- smdata.run_all = (cpu == ALL_CPUS) ? 1 : 0;
- init_completion(&smdata.done);
-
- smp_wmb(); /* make sure other cpus see smdata updates */
-
- /* If they don't care which CPU fn runs on, bind to any online one. */
- if (cpu == NR_CPUS || cpu == ALL_CPUS)
- cpu = raw_smp_processor_id();
-
- p = kthread_create(do_stop, &smdata, "kstopmachine");
- if (!IS_ERR(p)) {
- struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 };
+ mutex_lock(&setup_lock);
+ refcount--;
+ if (refcount)
+ goto done;
+ destroy_workqueue(stop_machine_wq);
+ free_percpu(stop_machine_work);
+done:
+ mutex_unlock(&setup_lock);
+}
+EXPORT_SYMBOL_GPL(stop_machine_destroy);
- /* One high-prio thread per cpu. We'll do this one. */
- sched_setscheduler_nocheck(p, SCHED_FIFO, ¶m);
- kthread_bind(p, cpu);
- wake_up_process(p);
- wait_for_completion(&smdata.done);
+int __stop_machine(int (*fn)(void *), void *data, const struct cpumask *cpus)
+{
+ struct work_struct *sm_work;
+ int i, ret;
+
+ /* Set up initial state. */
+ mutex_lock(&lock);
+ num_threads = num_online_cpus();
+ active_cpus = cpus;
+ active.fn = fn;
+ active.data = data;
+ active.fnret = 0;
+ idle.fn = chill;
+ idle.data = NULL;
+
+ set_state(STOPMACHINE_PREPARE);
+
+ /* Schedule the stop_cpu work on all cpus: hold this CPU so one
+ * doesn't hit this CPU until we're ready. */
+ get_cpu();
+ for_each_online_cpu(i) {
+ sm_work = per_cpu_ptr(stop_machine_work, i);
+ INIT_WORK(sm_work, stop_cpu);
+ queue_work_on(i, stop_machine_wq, sm_work);
}
- mutex_unlock(&stopmachine_mutex);
- return p;
+ /* This will release the thread on our CPU. */
+ put_cpu();
+ flush_workqueue(stop_machine_wq);
+ ret = active.fnret;
+ mutex_unlock(&lock);
+ return ret;
}
-int stop_machine_run(int (*fn)(void *), void *data, unsigned int cpu)
+int stop_machine(int (*fn)(void *), void *data, const struct cpumask *cpus)
{
- struct task_struct *p;
int ret;
+ ret = stop_machine_create();
+ if (ret)
+ return ret;
/* No CPUs can come up or down during this. */
get_online_cpus();
- p = __stop_machine_run(fn, data, cpu);
- if (!IS_ERR(p))
- ret = kthread_stop(p);
- else
- ret = PTR_ERR(p);
+ ret = __stop_machine(fn, data, cpus);
put_online_cpus();
-
+ stop_machine_destroy();
return ret;
}
-EXPORT_SYMBOL_GPL(stop_machine_run);
+EXPORT_SYMBOL_GPL(stop_machine);
git://ftp.safe.ca / safe / jmp / linux-2.6 / blobdiff