[CPUFREQ] fix (utter) cpufreq_add_dev mess
OK, I've tried to clean it up the best I could, but please test this with
concurrent cpu hotplug and cpufreq add/remove in loops. I'm sure we will make
other interesting findings.
This is step one of fixing the overall locking dependency mess in cpufreq.
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
CC: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
CC: rjw@sisk.pl
CC: mingo@elte.hu
CC: Shaohua Li <shaohua.li@intel.com>
CC: Pekka Enberg <penberg@cs.helsinki.fi>
CC: Dave Young <hidave.darkstar@gmail.com>
CC: "Rafael J. Wysocki" <rjw@sisk.pl>
CC: Rusty Russell <rusty@rustcorp.com.au>
CC: sven.wegener@stealer.net
CC: cpufreq@vger.kernel.org
CC: Thomas Renninger <trenn@suse.de>
Signed-off-by: Dave Jones <davej@redhat.com>
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c
index c7fe16e..c668ac8 100644
--- a/drivers/cpufreq/cpufreq.c
+++ b/drivers/cpufreq/cpufreq.c
@@ -761,6 +761,10 @@
* cpufreq_add_dev - add a CPU device
*
* Adds the cpufreq interface for a CPU device.
+ *
+ * The Oracle says: try running cpufreq registration/unregistration concurrently
+ * with with cpu hotplugging and all hell will break loose. Tried to clean this
+ * mess up, but more thorough testing is needed. - Mathieu
*/
static int cpufreq_add_dev(struct sys_device *sys_dev)
{
@@ -804,15 +808,12 @@
goto nomem_out;
}
if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL)) {
- kfree(policy);
ret = -ENOMEM;
- goto nomem_out;
+ goto err_free_policy;
}
if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL)) {
- free_cpumask_var(policy->cpus);
- kfree(policy);
ret = -ENOMEM;
- goto nomem_out;
+ goto err_free_cpumask;
}
policy->cpu = cpu;
@@ -820,7 +821,8 @@
/* Initially set CPU itself as the policy_cpu */
per_cpu(policy_cpu, cpu) = cpu;
- lock_policy_rwsem_write(cpu);
+ ret = (lock_policy_rwsem_write(cpu) < 0);
+ WARN_ON(ret);
init_completion(&policy->kobj_unregister);
INIT_WORK(&policy->update, handle_update);
@@ -833,7 +835,7 @@
ret = cpufreq_driver->init(policy);
if (ret) {
dprintk("initialization failed\n");
- goto err_out;
+ goto err_unlock_policy;
}
policy->user_policy.min = policy->min;
policy->user_policy.max = policy->max;
@@ -858,15 +860,21 @@
/* Check for existing affected CPUs.
* They may not be aware of it due to CPU Hotplug.
*/
- managed_policy = cpufreq_cpu_get(j); /* FIXME: Where is this released? What about error paths? */
+ managed_policy = cpufreq_cpu_get(j);
if (unlikely(managed_policy)) {
/* Set proper policy_cpu */
unlock_policy_rwsem_write(cpu);
per_cpu(policy_cpu, cpu) = managed_policy->cpu;
- if (lock_policy_rwsem_write(cpu) < 0)
- goto err_out_driver_exit;
+ if (lock_policy_rwsem_write(cpu) < 0) {
+ /* Should not go through policy unlock path */
+ if (cpufreq_driver->exit)
+ cpufreq_driver->exit(policy);
+ ret = -EBUSY;
+ cpufreq_cpu_put(managed_policy);
+ goto err_free_cpumask;
+ }
spin_lock_irqsave(&cpufreq_driver_lock, flags);
cpumask_copy(managed_policy->cpus, policy->cpus);
@@ -877,12 +885,14 @@
ret = sysfs_create_link(&sys_dev->kobj,
&managed_policy->kobj,
"cpufreq");
- if (ret)
- goto err_out_driver_exit;
-
- cpufreq_debug_enable_ratelimit();
- ret = 0;
- goto err_out_driver_exit; /* call driver->exit() */
+ if (!ret)
+ cpufreq_cpu_put(managed_policy);
+ /*
+ * Success. We only needed to be added to the mask.
+ * Call driver->exit() because only the cpu parent of
+ * the kobj needed to call init().
+ */
+ goto out_driver_exit; /* call driver->exit() */
}
}
#endif
@@ -892,25 +902,25 @@
ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq, &sys_dev->kobj,
"cpufreq");
if (ret)
- goto err_out_driver_exit;
+ goto out_driver_exit;
/* set up files for this cpu device */
drv_attr = cpufreq_driver->attr;
while ((drv_attr) && (*drv_attr)) {
ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
drv_attr++;
}
if (cpufreq_driver->get) {
ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
}
if (cpufreq_driver->target) {
ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
if (ret)
- goto err_out_driver_exit;
+ goto err_out_kobj_put;
}
spin_lock_irqsave(&cpufreq_driver_lock, flags);
@@ -928,12 +938,14 @@
continue;
dprintk("CPU %u already managed, adding link\n", j);
- cpufreq_cpu_get(cpu);
+ managed_policy = cpufreq_cpu_get(cpu);
cpu_sys_dev = get_cpu_sysdev(j);
ret = sysfs_create_link(&cpu_sys_dev->kobj, &policy->kobj,
"cpufreq");
- if (ret)
+ if (ret) {
+ cpufreq_cpu_put(managed_policy);
goto err_out_unregister;
+ }
}
policy->governor = NULL; /* to assure that the starting sequence is
@@ -965,17 +977,20 @@
per_cpu(cpufreq_cpu_data, j) = NULL;
spin_unlock_irqrestore(&cpufreq_driver_lock, flags);
+err_out_kobj_put:
kobject_put(&policy->kobj);
wait_for_completion(&policy->kobj_unregister);
-err_out_driver_exit:
+out_driver_exit:
if (cpufreq_driver->exit)
cpufreq_driver->exit(policy);
-err_out:
+err_unlock_policy:
unlock_policy_rwsem_write(cpu);
+err_free_cpumask:
+ free_cpumask_var(policy->cpus);
+err_free_policy:
kfree(policy);
-
nomem_out:
module_put(cpufreq_driver->owner);
module_out: