2005-04-16 18:20:36 -04:00
|
|
|
/* CPU control.
|
|
|
|
* (C) 2001, 2002, 2003, 2004 Rusty Russell
|
|
|
|
*
|
|
|
|
* This code is licenced under the GPL.
|
|
|
|
*/
|
|
|
|
#include <linux/proc_fs.h>
|
|
|
|
#include <linux/smp.h>
|
|
|
|
#include <linux/init.h>
|
|
|
|
#include <linux/notifier.h>
|
|
|
|
#include <linux/sched.h>
|
|
|
|
#include <linux/unistd.h>
|
|
|
|
#include <linux/cpu.h>
|
|
|
|
#include <linux/module.h>
|
|
|
|
#include <linux/kthread.h>
|
|
|
|
#include <linux/stop_machine.h>
|
2006-06-26 03:24:32 -04:00
|
|
|
#include <linux/mutex.h>
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2008-05-29 14:17:02 -04:00
|
|
|
/*
|
|
|
|
* Represents all cpu's present in the system
|
|
|
|
* In systems capable of hotplug, this map could dynamically grow
|
|
|
|
* as new cpu's are detected in the system via any platform specific
|
|
|
|
* method, such as ACPI for e.g.
|
|
|
|
*/
|
|
|
|
cpumask_t cpu_present_map __read_mostly;
|
|
|
|
EXPORT_SYMBOL(cpu_present_map);
|
|
|
|
|
|
|
|
#ifndef CONFIG_SMP
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Represents all cpu's that are currently online.
|
|
|
|
*/
|
|
|
|
cpumask_t cpu_online_map __read_mostly = CPU_MASK_ALL;
|
|
|
|
EXPORT_SYMBOL(cpu_online_map);
|
|
|
|
|
|
|
|
cpumask_t cpu_possible_map __read_mostly = CPU_MASK_ALL;
|
|
|
|
EXPORT_SYMBOL(cpu_possible_map);
|
|
|
|
|
|
|
|
#else /* CONFIG_SMP */
|
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
/* Serializes the updates to cpu_online_map, cpu_present_map */
|
2006-07-23 15:12:16 -04:00
|
|
|
static DEFINE_MUTEX(cpu_add_remove_lock);
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2006-10-17 03:10:35 -04:00
|
|
|
static __cpuinitdata RAW_NOTIFIER_HEAD(cpu_chain);
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
/* If set, cpu_up and cpu_down will return -EBUSY and do nothing.
|
|
|
|
* Should always be manipulated under cpu_add_remove_lock
|
|
|
|
*/
|
|
|
|
static int cpu_hotplug_disabled;
|
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
static struct {
|
|
|
|
struct task_struct *active_writer;
|
|
|
|
struct mutex lock; /* Synchronizes accesses to refcount, */
|
|
|
|
/*
|
|
|
|
* Also blocks the new readers during
|
|
|
|
* an ongoing cpu hotplug operation.
|
|
|
|
*/
|
|
|
|
int refcount;
|
|
|
|
} cpu_hotplug;
|
2005-11-09 00:34:24 -05:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
void __init cpu_hotplug_init(void)
|
|
|
|
{
|
|
|
|
cpu_hotplug.active_writer = NULL;
|
|
|
|
mutex_init(&cpu_hotplug.lock);
|
|
|
|
cpu_hotplug.refcount = 0;
|
|
|
|
}
|
|
|
|
|
2008-07-15 07:43:49 -04:00
|
|
|
cpumask_t cpu_active_map;
|
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
2005-11-09 00:34:24 -05:00
|
|
|
|
2008-01-25 15:08:02 -05:00
|
|
|
void get_online_cpus(void)
|
2005-11-28 16:43:46 -05:00
|
|
|
{
|
2008-01-25 15:08:01 -05:00
|
|
|
might_sleep();
|
|
|
|
if (cpu_hotplug.active_writer == current)
|
2006-07-23 15:12:16 -04:00
|
|
|
return;
|
2008-01-25 15:08:01 -05:00
|
|
|
mutex_lock(&cpu_hotplug.lock);
|
|
|
|
cpu_hotplug.refcount++;
|
|
|
|
mutex_unlock(&cpu_hotplug.lock);
|
|
|
|
|
2005-11-28 16:43:46 -05:00
|
|
|
}
|
2008-01-25 15:08:02 -05:00
|
|
|
EXPORT_SYMBOL_GPL(get_online_cpus);
|
2005-11-09 00:34:24 -05:00
|
|
|
|
2008-01-25 15:08:02 -05:00
|
|
|
void put_online_cpus(void)
|
2005-11-28 16:43:46 -05:00
|
|
|
{
|
2008-01-25 15:08:01 -05:00
|
|
|
if (cpu_hotplug.active_writer == current)
|
2006-07-23 15:12:16 -04:00
|
|
|
return;
|
2008-01-25 15:08:01 -05:00
|
|
|
mutex_lock(&cpu_hotplug.lock);
|
2008-04-29 04:00:29 -04:00
|
|
|
if (!--cpu_hotplug.refcount && unlikely(cpu_hotplug.active_writer))
|
|
|
|
wake_up_process(cpu_hotplug.active_writer);
|
2008-01-25 15:08:01 -05:00
|
|
|
mutex_unlock(&cpu_hotplug.lock);
|
|
|
|
|
2005-11-28 16:43:46 -05:00
|
|
|
}
|
2008-01-25 15:08:02 -05:00
|
|
|
EXPORT_SYMBOL_GPL(put_online_cpus);
|
2005-11-28 16:43:46 -05:00
|
|
|
|
|
|
|
#endif /* CONFIG_HOTPLUG_CPU */
|
2005-11-09 00:34:24 -05:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
/*
|
|
|
|
* The following two API's must be used when attempting
|
|
|
|
* to serialize the updates to cpu_online_map, cpu_present_map.
|
|
|
|
*/
|
|
|
|
void cpu_maps_update_begin(void)
|
|
|
|
{
|
|
|
|
mutex_lock(&cpu_add_remove_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
void cpu_maps_update_done(void)
|
|
|
|
{
|
|
|
|
mutex_unlock(&cpu_add_remove_lock);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* This ensures that the hotplug operation can begin only when the
|
|
|
|
* refcount goes to zero.
|
|
|
|
*
|
|
|
|
* Note that during a cpu-hotplug operation, the new readers, if any,
|
|
|
|
* will be blocked by the cpu_hotplug.lock
|
|
|
|
*
|
2008-04-29 04:00:29 -04:00
|
|
|
* Since cpu_hotplug_begin() is always called after invoking
|
|
|
|
* cpu_maps_update_begin(), we can be sure that only one writer is active.
|
2008-01-25 15:08:01 -05:00
|
|
|
*
|
|
|
|
* Note that theoretically, there is a possibility of a livelock:
|
|
|
|
* - Refcount goes to zero, last reader wakes up the sleeping
|
|
|
|
* writer.
|
|
|
|
* - Last reader unlocks the cpu_hotplug.lock.
|
|
|
|
* - A new reader arrives at this moment, bumps up the refcount.
|
|
|
|
* - The writer acquires the cpu_hotplug.lock finds the refcount
|
|
|
|
* non zero and goes to sleep again.
|
|
|
|
*
|
|
|
|
* However, this is very difficult to achieve in practice since
|
2008-01-25 15:08:02 -05:00
|
|
|
* get_online_cpus() not an api which is called all that often.
|
2008-01-25 15:08:01 -05:00
|
|
|
*
|
|
|
|
*/
|
|
|
|
static void cpu_hotplug_begin(void)
|
|
|
|
{
|
|
|
|
cpu_hotplug.active_writer = current;
|
2008-04-29 04:00:29 -04:00
|
|
|
|
|
|
|
for (;;) {
|
|
|
|
mutex_lock(&cpu_hotplug.lock);
|
|
|
|
if (likely(!cpu_hotplug.refcount))
|
|
|
|
break;
|
|
|
|
__set_current_state(TASK_UNINTERRUPTIBLE);
|
2008-01-25 15:08:01 -05:00
|
|
|
mutex_unlock(&cpu_hotplug.lock);
|
|
|
|
schedule();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void cpu_hotplug_done(void)
|
|
|
|
{
|
|
|
|
cpu_hotplug.active_writer = NULL;
|
|
|
|
mutex_unlock(&cpu_hotplug.lock);
|
|
|
|
}
|
2005-04-16 18:20:36 -04:00
|
|
|
/* Need to know about CPUs going up/down? */
|
2008-04-29 03:58:51 -04:00
|
|
|
int __ref register_cpu_notifier(struct notifier_block *nb)
|
2005-04-16 18:20:36 -04:00
|
|
|
{
|
2006-10-17 03:10:35 -04:00
|
|
|
int ret;
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2006-10-17 03:10:35 -04:00
|
|
|
ret = raw_notifier_chain_register(&cpu_chain, nb);
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2006-10-17 03:10:35 -04:00
|
|
|
return ret;
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
2006-06-27 05:54:08 -04:00
|
|
|
|
|
|
|
#ifdef CONFIG_HOTPLUG_CPU
|
|
|
|
|
2005-04-16 18:20:36 -04:00
|
|
|
EXPORT_SYMBOL(register_cpu_notifier);
|
|
|
|
|
2008-04-29 03:58:48 -04:00
|
|
|
void __ref unregister_cpu_notifier(struct notifier_block *nb)
|
2005-04-16 18:20:36 -04:00
|
|
|
{
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2006-10-17 03:10:35 -04:00
|
|
|
raw_notifier_chain_unregister(&cpu_chain, nb);
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
|
|
|
EXPORT_SYMBOL(unregister_cpu_notifier);
|
|
|
|
|
|
|
|
static inline void check_for_tasks(int cpu)
|
|
|
|
{
|
|
|
|
struct task_struct *p;
|
|
|
|
|
|
|
|
write_lock_irq(&tasklist_lock);
|
|
|
|
for_each_process(p) {
|
|
|
|
if (task_cpu(p) == cpu &&
|
|
|
|
(!cputime_eq(p->utime, cputime_zero) ||
|
|
|
|
!cputime_eq(p->stime, cputime_zero)))
|
|
|
|
printk(KERN_WARNING "Task %s (pid = %d) is on cpu %d\
|
2007-05-09 05:34:04 -04:00
|
|
|
(state = %ld, flags = %x) \n",
|
2007-10-19 02:40:40 -04:00
|
|
|
p->comm, task_pid_nr(p), cpu,
|
|
|
|
p->state, p->flags);
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
|
|
|
write_unlock_irq(&tasklist_lock);
|
|
|
|
}
|
|
|
|
|
2007-05-24 05:23:10 -04:00
|
|
|
struct take_cpu_down_param {
|
|
|
|
unsigned long mod;
|
|
|
|
void *hcpu;
|
|
|
|
};
|
|
|
|
|
2005-04-16 18:20:36 -04:00
|
|
|
/* Take this CPU down. */
|
2008-04-29 03:58:50 -04:00
|
|
|
static int __ref take_cpu_down(void *_param)
|
2005-04-16 18:20:36 -04:00
|
|
|
{
|
2007-05-24 05:23:10 -04:00
|
|
|
struct take_cpu_down_param *param = _param;
|
2005-04-16 18:20:36 -04:00
|
|
|
int err;
|
|
|
|
|
|
|
|
/* Ensure this CPU doesn't handle any more interrupts. */
|
|
|
|
err = __cpu_disable();
|
|
|
|
if (err < 0)
|
2005-06-25 17:54:50 -04:00
|
|
|
return err;
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2008-08-31 13:58:49 -04:00
|
|
|
raw_notifier_call_chain(&cpu_chain, CPU_DYING | param->mod,
|
|
|
|
param->hcpu);
|
|
|
|
|
2005-06-25 17:54:50 -04:00
|
|
|
/* Force idle task to run as soon as we yield: it should
|
|
|
|
immediately notice cpu is offline and die quickly. */
|
|
|
|
sched_idle_next();
|
|
|
|
return 0;
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
/* Requires cpu_add_remove_lock to be held */
|
2008-04-29 03:58:50 -04:00
|
|
|
static int __ref _cpu_down(unsigned int cpu, int tasks_frozen)
|
2005-04-16 18:20:36 -04:00
|
|
|
{
|
2007-05-09 05:34:04 -04:00
|
|
|
int err, nr_calls = 0;
|
2005-04-16 18:20:36 -04:00
|
|
|
cpumask_t old_allowed, tmp;
|
2007-05-09 05:34:04 -04:00
|
|
|
void *hcpu = (void *)(long)cpu;
|
2007-05-09 05:35:10 -04:00
|
|
|
unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0;
|
2007-05-24 05:23:10 -04:00
|
|
|
struct take_cpu_down_param tcd_param = {
|
|
|
|
.mod = mod,
|
|
|
|
.hcpu = hcpu,
|
|
|
|
};
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
if (num_online_cpus() == 1)
|
|
|
|
return -EBUSY;
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
if (!cpu_online(cpu))
|
|
|
|
return -EINVAL;
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_hotplug_begin();
|
2007-05-09 05:35:10 -04:00
|
|
|
err = __raw_notifier_call_chain(&cpu_chain, CPU_DOWN_PREPARE | mod,
|
2007-05-09 05:34:04 -04:00
|
|
|
hcpu, -1, &nr_calls);
|
2005-04-16 18:20:36 -04:00
|
|
|
if (err == NOTIFY_BAD) {
|
2007-10-18 06:05:12 -04:00
|
|
|
nr_calls--;
|
2007-05-09 05:35:10 -04:00
|
|
|
__raw_notifier_call_chain(&cpu_chain, CPU_DOWN_FAILED | mod,
|
|
|
|
hcpu, nr_calls, NULL);
|
2005-04-16 18:20:36 -04:00
|
|
|
printk("%s: attempt to take down CPU %u failed\n",
|
2008-04-30 03:55:08 -04:00
|
|
|
__func__, cpu);
|
2007-05-09 05:34:03 -04:00
|
|
|
err = -EINVAL;
|
|
|
|
goto out_release;
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
|
|
|
|
|
|
|
/* Ensure that we are not runnable on dying cpu */
|
|
|
|
old_allowed = current->cpus_allowed;
|
2008-04-04 21:11:06 -04:00
|
|
|
cpus_setall(tmp);
|
2005-04-16 18:20:36 -04:00
|
|
|
cpu_clear(cpu, tmp);
|
2008-04-04 21:11:06 -04:00
|
|
|
set_cpus_allowed_ptr(current, &tmp);
|
2008-07-28 13:16:30 -04:00
|
|
|
tmp = cpumask_of_cpu(cpu);
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2008-07-28 13:16:30 -04:00
|
|
|
err = __stop_machine(take_cpu_down, &tcd_param, &tmp);
|
2008-07-28 13:16:29 -04:00
|
|
|
if (err) {
|
2005-04-16 18:20:36 -04:00
|
|
|
/* CPU didn't die: tell everyone. Can't complain. */
|
2007-05-09 05:35:10 -04:00
|
|
|
if (raw_notifier_call_chain(&cpu_chain, CPU_DOWN_FAILED | mod,
|
2007-05-09 05:34:04 -04:00
|
|
|
hcpu) == NOTIFY_BAD)
|
2005-04-16 18:20:36 -04:00
|
|
|
BUG();
|
|
|
|
|
2008-07-28 13:16:28 -04:00
|
|
|
goto out_allowed;
|
2006-10-28 13:38:57 -04:00
|
|
|
}
|
2008-07-28 13:16:29 -04:00
|
|
|
BUG_ON(cpu_online(cpu));
|
2005-04-16 18:20:36 -04:00
|
|
|
|
|
|
|
/* Wait for it to sleep (leaving idle task). */
|
|
|
|
while (!idle_cpu(cpu))
|
|
|
|
yield();
|
|
|
|
|
|
|
|
/* This actually kills the CPU. */
|
|
|
|
__cpu_die(cpu);
|
|
|
|
|
|
|
|
/* CPU is completely dead: tell everyone. Too late to complain. */
|
2007-05-09 05:35:10 -04:00
|
|
|
if (raw_notifier_call_chain(&cpu_chain, CPU_DEAD | mod,
|
|
|
|
hcpu) == NOTIFY_BAD)
|
2005-04-16 18:20:36 -04:00
|
|
|
BUG();
|
|
|
|
|
|
|
|
check_for_tasks(cpu);
|
|
|
|
|
|
|
|
out_allowed:
|
2008-04-04 21:11:06 -04:00
|
|
|
set_cpus_allowed_ptr(current, &old_allowed);
|
2007-05-09 05:34:03 -04:00
|
|
|
out_release:
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_hotplug_done();
|
2008-07-25 04:47:50 -04:00
|
|
|
if (!err) {
|
|
|
|
if (raw_notifier_call_chain(&cpu_chain, CPU_POST_DEAD | mod,
|
|
|
|
hcpu) == NOTIFY_BAD)
|
|
|
|
BUG();
|
|
|
|
}
|
2006-09-26 02:32:48 -04:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2008-04-29 03:58:50 -04:00
|
|
|
int __ref cpu_down(unsigned int cpu)
|
2006-09-26 02:32:48 -04:00
|
|
|
{
|
|
|
|
int err = 0;
|
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2008-07-15 07:43:49 -04:00
|
|
|
|
|
|
|
if (cpu_hotplug_disabled) {
|
2006-09-26 02:32:48 -04:00
|
|
|
err = -EBUSY;
|
2008-07-15 07:43:49 -04:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
cpu_clear(cpu, cpu_active_map);
|
|
|
|
|
2008-07-15 23:56:26 -04:00
|
|
|
/*
|
|
|
|
* Make sure the all cpus did the reschedule and are not
|
|
|
|
* using stale version of the cpu_active_map.
|
|
|
|
* This is not strictly necessary becuase stop_machine()
|
|
|
|
* that we run down the line already provides the required
|
|
|
|
* synchronization. But it's really a side effect and we do not
|
|
|
|
* want to depend on the innards of the stop_machine here.
|
|
|
|
*/
|
|
|
|
synchronize_sched();
|
2006-09-26 02:32:48 -04:00
|
|
|
|
2008-07-15 07:43:49 -04:00
|
|
|
err = _cpu_down(cpu, 0);
|
2006-09-26 02:32:48 -04:00
|
|
|
|
2008-07-15 07:43:49 -04:00
|
|
|
if (cpu_online(cpu))
|
|
|
|
cpu_set(cpu, cpu_active_map);
|
|
|
|
|
|
|
|
out:
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2005-04-16 18:20:36 -04:00
|
|
|
return err;
|
|
|
|
}
|
2008-04-29 02:35:56 -04:00
|
|
|
EXPORT_SYMBOL(cpu_down);
|
2005-04-16 18:20:36 -04:00
|
|
|
#endif /*CONFIG_HOTPLUG_CPU*/
|
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
/* Requires cpu_add_remove_lock to be held */
|
2007-05-09 05:35:10 -04:00
|
|
|
static int __cpuinit _cpu_up(unsigned int cpu, int tasks_frozen)
|
2005-04-16 18:20:36 -04:00
|
|
|
{
|
2007-05-09 05:34:03 -04:00
|
|
|
int ret, nr_calls = 0;
|
2005-04-16 18:20:36 -04:00
|
|
|
void *hcpu = (void *)(long)cpu;
|
2007-05-09 05:35:10 -04:00
|
|
|
unsigned long mod = tasks_frozen ? CPU_TASKS_FROZEN : 0;
|
2005-04-16 18:20:36 -04:00
|
|
|
|
2006-09-26 02:32:48 -04:00
|
|
|
if (cpu_online(cpu) || !cpu_present(cpu))
|
|
|
|
return -EINVAL;
|
2005-11-09 00:34:24 -05:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_hotplug_begin();
|
2007-05-09 05:35:10 -04:00
|
|
|
ret = __raw_notifier_call_chain(&cpu_chain, CPU_UP_PREPARE | mod, hcpu,
|
2007-05-09 05:34:03 -04:00
|
|
|
-1, &nr_calls);
|
2005-04-16 18:20:36 -04:00
|
|
|
if (ret == NOTIFY_BAD) {
|
2007-10-18 06:05:12 -04:00
|
|
|
nr_calls--;
|
2005-04-16 18:20:36 -04:00
|
|
|
printk("%s: attempt to bring up CPU %u failed\n",
|
2008-04-30 03:55:08 -04:00
|
|
|
__func__, cpu);
|
2005-04-16 18:20:36 -04:00
|
|
|
ret = -EINVAL;
|
|
|
|
goto out_notify;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Arch-specific enabling code. */
|
|
|
|
ret = __cpu_up(cpu);
|
|
|
|
if (ret != 0)
|
|
|
|
goto out_notify;
|
2006-03-24 12:45:21 -05:00
|
|
|
BUG_ON(!cpu_online(cpu));
|
2005-04-16 18:20:36 -04:00
|
|
|
|
sched, cpu hotplug: fix set_cpus_allowed() use in hotplug callbacks
Mark Langsdorf reported:
> One of my co-workers noticed that the powernow-k8
> driver no longer restarts when a CPU core is
> hot-disabled and then hot-enabled on AMD quad-core
> systems.
>
> The following comands work fine on 2.6.26 and fail
> on 2.6.27-rc1:
>
> echo 0 > /sys/devices/system/cpu/cpu3/online
> echo 1 > /sys/devices/system/cpu/cpu3/online
> find /sys -name cpufreq
>
> For 2.6.26, the find will return a cpufreq
> directory for each processor. In 2.6.27-rc1,
> the cpu3 directory is missing.
>
> After digging through the code, the following
> logic is failing when the core is hot-enabled
> at runtime. The code works during the boot
> sequence.
>
> cpumask_t = current->cpus_allowed;
> set_cpus_allowed_ptr(current, &cpumask_of_cpu(cpu));
> if (smp_processor_id() != cpu)
> return -ENODEV;
So set the CPU active before calling the CPU_ONLINE notifier chain,
there are a handful of notifiers that use set_cpus_allowed().
This fix also solves the problem with x86-microcode. I've sent
alternative patches for microcode, but as this "rely on
set_cpus_allowed_ptr() being workable in cpu-hotplug(CPU_ONLINE, ...)"
assumption seems to be more broad than what we thought, perhaps this fix
should be applied.
With this patch we define that by the moment CPU_ONLINE is being sent,
a 'cpu' is online and ready for tasks to be migrated onto it.
Signed-off-by: Dmitry Adamushko <dmitry.adamushko@gmail.com>
Reported-by: Mark Langsdorf <mark.langsdorf@amd.com>
Tested-by: Mark Langsdorf <mark.langsdorf@amd.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-30 06:34:04 -04:00
|
|
|
cpu_set(cpu, cpu_active_map);
|
|
|
|
|
2005-04-16 18:20:36 -04:00
|
|
|
/* Now call notifier in preparation. */
|
2007-05-09 05:35:10 -04:00
|
|
|
raw_notifier_call_chain(&cpu_chain, CPU_ONLINE | mod, hcpu);
|
2005-04-16 18:20:36 -04:00
|
|
|
|
|
|
|
out_notify:
|
|
|
|
if (ret != 0)
|
2007-05-09 05:34:03 -04:00
|
|
|
__raw_notifier_call_chain(&cpu_chain,
|
2007-05-09 05:35:10 -04:00
|
|
|
CPU_UP_CANCELED | mod, hcpu, nr_calls, NULL);
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_hotplug_done();
|
2006-09-26 02:32:48 -04:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2007-01-11 02:15:34 -05:00
|
|
|
int __cpuinit cpu_up(unsigned int cpu)
|
2006-09-26 02:32:48 -04:00
|
|
|
{
|
|
|
|
int err = 0;
|
2007-10-19 02:40:47 -04:00
|
|
|
if (!cpu_isset(cpu, cpu_possible_map)) {
|
|
|
|
printk(KERN_ERR "can't online cpu %d because it is not "
|
|
|
|
"configured as may-hotadd at boot time\n", cpu);
|
2008-08-12 18:08:40 -04:00
|
|
|
#if defined(CONFIG_IA64) || defined(CONFIG_X86_64)
|
2007-10-19 02:40:47 -04:00
|
|
|
printk(KERN_ERR "please check additional_cpus= boot "
|
|
|
|
"parameter\n");
|
|
|
|
#endif
|
|
|
|
return -EINVAL;
|
|
|
|
}
|
2006-09-26 02:32:48 -04:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2008-07-15 07:43:49 -04:00
|
|
|
|
|
|
|
if (cpu_hotplug_disabled) {
|
2006-09-26 02:32:48 -04:00
|
|
|
err = -EBUSY;
|
2008-07-15 07:43:49 -04:00
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
err = _cpu_up(cpu, 0);
|
|
|
|
|
|
|
|
out:
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2006-09-26 02:32:48 -04:00
|
|
|
return err;
|
|
|
|
}
|
|
|
|
|
2007-08-31 02:56:29 -04:00
|
|
|
#ifdef CONFIG_PM_SLEEP_SMP
|
2006-09-26 02:32:48 -04:00
|
|
|
static cpumask_t frozen_cpus;
|
|
|
|
|
|
|
|
int disable_nonboot_cpus(void)
|
|
|
|
{
|
2006-12-23 10:55:29 -05:00
|
|
|
int cpu, first_cpu, error = 0;
|
2006-09-26 02:32:48 -04:00
|
|
|
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2007-04-02 02:49:49 -04:00
|
|
|
first_cpu = first_cpu(cpu_online_map);
|
2006-09-26 02:32:48 -04:00
|
|
|
/* We take down all of the non-boot CPUs in one shot to avoid races
|
|
|
|
* with the userspace trying to use the CPU hotplug at the same time
|
|
|
|
*/
|
|
|
|
cpus_clear(frozen_cpus);
|
|
|
|
printk("Disabling non-boot CPUs ...\n");
|
|
|
|
for_each_online_cpu(cpu) {
|
|
|
|
if (cpu == first_cpu)
|
|
|
|
continue;
|
2007-05-09 05:35:10 -04:00
|
|
|
error = _cpu_down(cpu, 1);
|
2006-09-26 02:32:48 -04:00
|
|
|
if (!error) {
|
|
|
|
cpu_set(cpu, frozen_cpus);
|
|
|
|
printk("CPU%d is down\n", cpu);
|
|
|
|
} else {
|
|
|
|
printk(KERN_ERR "Error taking CPU%d down: %d\n",
|
|
|
|
cpu, error);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (!error) {
|
|
|
|
BUG_ON(num_online_cpus() > 1);
|
|
|
|
/* Make sure the CPUs won't be enabled by someone else */
|
|
|
|
cpu_hotplug_disabled = 1;
|
|
|
|
} else {
|
2006-12-23 10:55:29 -05:00
|
|
|
printk(KERN_ERR "Non-boot CPUs are not disabled\n");
|
2006-09-26 02:32:48 -04:00
|
|
|
}
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2006-09-26 02:32:48 -04:00
|
|
|
return error;
|
|
|
|
}
|
|
|
|
|
2008-02-08 07:21:55 -05:00
|
|
|
void __ref enable_nonboot_cpus(void)
|
2006-09-26 02:32:48 -04:00
|
|
|
{
|
|
|
|
int cpu, error;
|
|
|
|
|
|
|
|
/* Allow everyone to use the CPU hotplug again */
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_begin();
|
2006-09-26 02:32:48 -04:00
|
|
|
cpu_hotplug_disabled = 0;
|
2007-02-10 04:43:32 -05:00
|
|
|
if (cpus_empty(frozen_cpus))
|
2007-04-02 02:49:49 -04:00
|
|
|
goto out;
|
2006-09-26 02:32:48 -04:00
|
|
|
|
|
|
|
printk("Enabling non-boot CPUs ...\n");
|
2008-05-12 15:21:13 -04:00
|
|
|
for_each_cpu_mask_nr(cpu, frozen_cpus) {
|
2007-05-09 05:35:10 -04:00
|
|
|
error = _cpu_up(cpu, 1);
|
2006-09-26 02:32:48 -04:00
|
|
|
if (!error) {
|
|
|
|
printk("CPU%d is up\n", cpu);
|
|
|
|
continue;
|
|
|
|
}
|
2007-04-02 02:49:49 -04:00
|
|
|
printk(KERN_WARNING "Error taking CPU%d up: %d\n", cpu, error);
|
2006-09-26 02:32:48 -04:00
|
|
|
}
|
|
|
|
cpus_clear(frozen_cpus);
|
2007-04-02 02:49:49 -04:00
|
|
|
out:
|
2008-01-25 15:08:01 -05:00
|
|
|
cpu_maps_update_done();
|
2005-04-16 18:20:36 -04:00
|
|
|
}
|
2007-08-31 02:56:29 -04:00
|
|
|
#endif /* CONFIG_PM_SLEEP_SMP */
|
2008-05-29 14:17:02 -04:00
|
|
|
|
2008-09-07 10:57:22 -04:00
|
|
|
/**
|
|
|
|
* notify_cpu_starting(cpu) - call the CPU_STARTING notifiers
|
|
|
|
* @cpu: cpu that just started
|
|
|
|
*
|
|
|
|
* This function calls the cpu_chain notifiers with CPU_STARTING.
|
|
|
|
* It must be called by the arch code on the new cpu, before the new cpu
|
|
|
|
* enables interrupts and before the "boot" cpu returns from __cpu_up().
|
|
|
|
*/
|
|
|
|
void notify_cpu_starting(unsigned int cpu)
|
|
|
|
{
|
|
|
|
unsigned long val = CPU_STARTING;
|
|
|
|
|
|
|
|
#ifdef CONFIG_PM_SLEEP_SMP
|
|
|
|
if (cpu_isset(cpu, frozen_cpus))
|
|
|
|
val = CPU_STARTING_FROZEN;
|
|
|
|
#endif /* CONFIG_PM_SLEEP_SMP */
|
|
|
|
raw_notifier_call_chain(&cpu_chain, val, (void *)(long)cpu);
|
|
|
|
}
|
|
|
|
|
2008-05-29 14:17:02 -04:00
|
|
|
#endif /* CONFIG_SMP */
|
2008-07-24 21:21:29 -04:00
|
|
|
|
cpu masks: optimize and clean up cpumask_of_cpu()
Clean up and optimize cpumask_of_cpu(), by sharing all the zero words.
Instead of stupidly generating all possible i=0...NR_CPUS 2^i patterns
creating a huge array of constant bitmasks, realize that the zero words
can be shared.
In other words, on a 64-bit architecture, we only ever need 64 of these
arrays - with a different bit set in one single world (with enough zero
words around it so that we can create any bitmask by just offsetting in
that big array). And then we just put enough zeroes around it that we
can point every single cpumask to be one of those things.
So when we have 4k CPU's, instead of having 4k arrays (of 4k bits each,
with one bit set in each array - 2MB memory total), we have exactly 64
arrays instead, each 8k bits in size (64kB total).
And then we just point cpumask(n) to the right position (which we can
calculate dynamically). Once we have the right arrays, getting
"cpumask(n)" ends up being:
static inline const cpumask_t *get_cpu_mask(unsigned int cpu)
{
const unsigned long *p = cpu_bit_bitmap[1 + cpu % BITS_PER_LONG];
p -= cpu / BITS_PER_LONG;
return (const cpumask_t *)p;
}
This brings other advantages and simplifications as well:
- we are not wasting memory that is just filled with a single bit in
various different places
- we don't need all those games to re-create the arrays in some dense
format, because they're already going to be dense enough.
if we compile a kernel for up to 4k CPU's, "wasting" that 64kB of memory
is a non-issue (especially since by doing this "overlapping" trick we
probably get better cache behaviour anyway).
[ mingo@elte.hu:
Converted Linus's mails into a commit. See:
http://lkml.org/lkml/2008/7/27/156
http://lkml.org/lkml/2008/7/28/320
Also applied a family filter - which also has the side-effect of leaving
out the bits where Linus calls me an idio... Oh, never mind ;-)
]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-28 14:32:33 -04:00
|
|
|
/*
|
|
|
|
* cpu_bit_bitmap[] is a special, "compressed" data structure that
|
|
|
|
* represents all NR_CPUS bits binary values of 1<<nr.
|
|
|
|
*
|
|
|
|
* It is used by cpumask_of_cpu() to get a constant address to a CPU
|
|
|
|
* mask value that has a single bit set only.
|
|
|
|
*/
|
2008-07-24 21:21:29 -04:00
|
|
|
|
cpu masks: optimize and clean up cpumask_of_cpu()
Clean up and optimize cpumask_of_cpu(), by sharing all the zero words.
Instead of stupidly generating all possible i=0...NR_CPUS 2^i patterns
creating a huge array of constant bitmasks, realize that the zero words
can be shared.
In other words, on a 64-bit architecture, we only ever need 64 of these
arrays - with a different bit set in one single world (with enough zero
words around it so that we can create any bitmask by just offsetting in
that big array). And then we just put enough zeroes around it that we
can point every single cpumask to be one of those things.
So when we have 4k CPU's, instead of having 4k arrays (of 4k bits each,
with one bit set in each array - 2MB memory total), we have exactly 64
arrays instead, each 8k bits in size (64kB total).
And then we just point cpumask(n) to the right position (which we can
calculate dynamically). Once we have the right arrays, getting
"cpumask(n)" ends up being:
static inline const cpumask_t *get_cpu_mask(unsigned int cpu)
{
const unsigned long *p = cpu_bit_bitmap[1 + cpu % BITS_PER_LONG];
p -= cpu / BITS_PER_LONG;
return (const cpumask_t *)p;
}
This brings other advantages and simplifications as well:
- we are not wasting memory that is just filled with a single bit in
various different places
- we don't need all those games to re-create the arrays in some dense
format, because they're already going to be dense enough.
if we compile a kernel for up to 4k CPU's, "wasting" that 64kB of memory
is a non-issue (especially since by doing this "overlapping" trick we
probably get better cache behaviour anyway).
[ mingo@elte.hu:
Converted Linus's mails into a commit. See:
http://lkml.org/lkml/2008/7/27/156
http://lkml.org/lkml/2008/7/28/320
Also applied a family filter - which also has the side-effect of leaving
out the bits where Linus calls me an idio... Oh, never mind ;-)
]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-28 14:32:33 -04:00
|
|
|
/* cpu_bit_bitmap[0] is empty - so we can back into it */
|
|
|
|
#define MASK_DECLARE_1(x) [x+1][0] = 1UL << (x)
|
|
|
|
#define MASK_DECLARE_2(x) MASK_DECLARE_1(x), MASK_DECLARE_1(x+1)
|
|
|
|
#define MASK_DECLARE_4(x) MASK_DECLARE_2(x), MASK_DECLARE_2(x+2)
|
|
|
|
#define MASK_DECLARE_8(x) MASK_DECLARE_4(x), MASK_DECLARE_4(x+4)
|
2008-07-24 21:21:29 -04:00
|
|
|
|
cpu masks: optimize and clean up cpumask_of_cpu()
Clean up and optimize cpumask_of_cpu(), by sharing all the zero words.
Instead of stupidly generating all possible i=0...NR_CPUS 2^i patterns
creating a huge array of constant bitmasks, realize that the zero words
can be shared.
In other words, on a 64-bit architecture, we only ever need 64 of these
arrays - with a different bit set in one single world (with enough zero
words around it so that we can create any bitmask by just offsetting in
that big array). And then we just put enough zeroes around it that we
can point every single cpumask to be one of those things.
So when we have 4k CPU's, instead of having 4k arrays (of 4k bits each,
with one bit set in each array - 2MB memory total), we have exactly 64
arrays instead, each 8k bits in size (64kB total).
And then we just point cpumask(n) to the right position (which we can
calculate dynamically). Once we have the right arrays, getting
"cpumask(n)" ends up being:
static inline const cpumask_t *get_cpu_mask(unsigned int cpu)
{
const unsigned long *p = cpu_bit_bitmap[1 + cpu % BITS_PER_LONG];
p -= cpu / BITS_PER_LONG;
return (const cpumask_t *)p;
}
This brings other advantages and simplifications as well:
- we are not wasting memory that is just filled with a single bit in
various different places
- we don't need all those games to re-create the arrays in some dense
format, because they're already going to be dense enough.
if we compile a kernel for up to 4k CPU's, "wasting" that 64kB of memory
is a non-issue (especially since by doing this "overlapping" trick we
probably get better cache behaviour anyway).
[ mingo@elte.hu:
Converted Linus's mails into a commit. See:
http://lkml.org/lkml/2008/7/27/156
http://lkml.org/lkml/2008/7/28/320
Also applied a family filter - which also has the side-effect of leaving
out the bits where Linus calls me an idio... Oh, never mind ;-)
]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-28 14:32:33 -04:00
|
|
|
const unsigned long cpu_bit_bitmap[BITS_PER_LONG+1][BITS_TO_LONGS(NR_CPUS)] = {
|
|
|
|
|
|
|
|
MASK_DECLARE_8(0), MASK_DECLARE_8(8),
|
|
|
|
MASK_DECLARE_8(16), MASK_DECLARE_8(24),
|
|
|
|
#if BITS_PER_LONG > 32
|
|
|
|
MASK_DECLARE_8(32), MASK_DECLARE_8(40),
|
|
|
|
MASK_DECLARE_8(48), MASK_DECLARE_8(56),
|
2008-07-24 21:21:29 -04:00
|
|
|
#endif
|
|
|
|
};
|
cpu masks: optimize and clean up cpumask_of_cpu()
Clean up and optimize cpumask_of_cpu(), by sharing all the zero words.
Instead of stupidly generating all possible i=0...NR_CPUS 2^i patterns
creating a huge array of constant bitmasks, realize that the zero words
can be shared.
In other words, on a 64-bit architecture, we only ever need 64 of these
arrays - with a different bit set in one single world (with enough zero
words around it so that we can create any bitmask by just offsetting in
that big array). And then we just put enough zeroes around it that we
can point every single cpumask to be one of those things.
So when we have 4k CPU's, instead of having 4k arrays (of 4k bits each,
with one bit set in each array - 2MB memory total), we have exactly 64
arrays instead, each 8k bits in size (64kB total).
And then we just point cpumask(n) to the right position (which we can
calculate dynamically). Once we have the right arrays, getting
"cpumask(n)" ends up being:
static inline const cpumask_t *get_cpu_mask(unsigned int cpu)
{
const unsigned long *p = cpu_bit_bitmap[1 + cpu % BITS_PER_LONG];
p -= cpu / BITS_PER_LONG;
return (const cpumask_t *)p;
}
This brings other advantages and simplifications as well:
- we are not wasting memory that is just filled with a single bit in
various different places
- we don't need all those games to re-create the arrays in some dense
format, because they're already going to be dense enough.
if we compile a kernel for up to 4k CPU's, "wasting" that 64kB of memory
is a non-issue (especially since by doing this "overlapping" trick we
probably get better cache behaviour anyway).
[ mingo@elte.hu:
Converted Linus's mails into a commit. See:
http://lkml.org/lkml/2008/7/27/156
http://lkml.org/lkml/2008/7/28/320
Also applied a family filter - which also has the side-effect of leaving
out the bits where Linus calls me an idio... Oh, never mind ;-)
]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Mike Travis <travis@sgi.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-07-28 14:32:33 -04:00
|
|
|
EXPORT_SYMBOL_GPL(cpu_bit_bitmap);
|