[PATCH v3] cpufreq: s3c: fix unbalances of cpufreq policy refcount

[PATCH v3] cpufreq: s3c: fix unbalances of cpufreq policy refcount

[qiwuchen55]

From: chenqiwu <chenqiwu@xiaomi.com>

The cpufreq_reboot_notifier_evt() call cpufreq_cpu_get() to get the
cpufreq policy of cpu0, meanwhile, it also increments the kobject
reference count to mark it busy. However, a corresponding call of
cpufreq_cpu_put() is ignored to decrement the kobject reference count
back, which may lead to a potential stuck risk that the cpuhp thread
deadly waits for dropping of kobject refcount when cpufreq policy free.

With this patch, the cpuhp thread can be easily exercised by attempting
to force an unbind of the CPUfreq driver.

Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>
---
changes in v3:
 - Rewrite title and commit message.
---
 drivers/cpufreq/s3c2416-cpufreq.c | 12 +++++++++++-
 drivers/cpufreq/s5pv210-cpufreq.c | 11 ++++++++++-
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/drivers/cpufreq/s3c2416-cpufreq.c b/drivers/cpufreq/s3c2416-cpufreq.c
index 1069103..5c221bc 100644
--- a/drivers/cpufreq/s3c2416-cpufreq.c
+++ b/drivers/cpufreq/s3c2416-cpufreq.c
@@ -304,6 +304,7 @@ static int s3c2416_cpufreq_reboot_notifier_evt(struct notifier_block *this,
 {
 	struct s3c2416_data *s3c_freq = &s3c2416_cpufreq;
 	int ret;
+	struct cpufreq_policy *policy;
 
 	mutex_lock(&cpufreq_lock);
 
@@ -318,7 +319,16 @@ static int s3c2416_cpufreq_reboot_notifier_evt(struct notifier_block *this,
 	 */
 	if (s3c_freq->is_dvs) {
 		pr_debug("cpufreq: leave dvs on reboot\n");
-		ret = cpufreq_driver_target(cpufreq_cpu_get(0), FREQ_SLEEP, 0);
+
+		policy = cpufreq_cpu_get(0);
+		if (!policy) {
+			pr_debug("cpufreq: get no policy for cpu0\n");
+			return NOTIFY_BAD;
+		}
+
+		ret = cpufreq_driver_target(policy, FREQ_SLEEP, 0);
+		cpufreq_cpu_put(policy);
+
 		if (ret < 0)
 			return NOTIFY_BAD;
 	}
diff --git a/drivers/cpufreq/s5pv210-cpufreq.c b/drivers/cpufreq/s5pv210-cpufreq.c
index 5d10030..e84281e 100644
--- a/drivers/cpufreq/s5pv210-cpufreq.c
+++ b/drivers/cpufreq/s5pv210-cpufreq.c
@@ -555,8 +555,17 @@ static int s5pv210_cpufreq_reboot_notifier_event(struct notifier_block *this,
 						 unsigned long event, void *ptr)
 {
 	int ret;
+	struct cpufreq_policy *policy;
+
+	policy = cpufreq_cpu_get(0);
+	if (!policy) {
+		pr_debug("cpufreq: get no policy for cpu0\n");
+		return NOTIFY_BAD;
+	}
+
+	ret = cpufreq_driver_target(policy, SLEEP_FREQ, 0);
+	cpufreq_cpu_put(policy);
 
-	ret = cpufreq_driver_target(cpufreq_cpu_get(0), SLEEP_FREQ, 0);
 	if (ret < 0)
 		return NOTIFY_BAD;
 
-- 
1.9.1

Re: [PATCH v3] cpufreq: s3c: fix unbalances of cpufreq policy refcount

[Viresh Kumar]

On 07-01-20, 15:53, qiwuchen55@gmail.com wrote:
> From: chenqiwu <chenqiwu@xiaomi.com>
> 
> The cpufreq_reboot_notifier_evt() call cpufreq_cpu_get() to get the
> cpufreq policy of cpu0, meanwhile, it also increments the kobject
> reference count to mark it busy. However, a corresponding call of
> cpufreq_cpu_put() is ignored to decrement the kobject reference count
> back, which may lead to a potential stuck risk that the cpuhp thread
> deadly waits for dropping of kobject refcount when cpufreq policy free.
> 
> With this patch, the cpuhp thread can be easily exercised by attempting
> to force an unbind of the CPUfreq driver.
> 
> Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>
> ---
> changes in v3:
>  - Rewrite title and commit message.
> ---
>  drivers/cpufreq/s3c2416-cpufreq.c | 12 +++++++++++-
>  drivers/cpufreq/s5pv210-cpufreq.c | 11 ++++++++++-
>  2 files changed, 21 insertions(+), 2 deletions(-)

Applied. Thanks.

-- 
viresh

Re: [PATCH v3] cpufreq: s3c: fix unbalances of cpufreq policy refcount

[Viresh Kumar]

On 19-01-20, 10:25, qiwu chen wrote:
> Hi
> Any progress about this patch?

I already applied it in my tree, it will be part for v5.5-rc1.

-- 
viresh

Re: [PATCH v3] cpufreq: s3c: fix unbalances of cpufreq policy refcount

[chenqiwu]

On Mon, Jan 20, 2020 at 10:58:21AM +0530, Viresh Kumar wrote:
> On 19-01-20, 10:25, qiwu chen wrote:
> > Hi
> > Any progress about this patch?
> 
> I already applied it in my tree, it will be part for v5.5-rc1.
> 
> -- 
> viresh

OK, Thanks a lot!

Qiwu