| Message ID | f67b4b395c777a3e933ab51eafcd76b1299bf1fb.1464776797.git.viresh.kumar@linaro.org |
|---|---|
| State | New |
| Headers | show |
On 02-06-16, 21:37, Michael Ellerman wrote: > On Thu, 2016-06-02 at 16:52 +0530, Viresh Kumar wrote: > > On 02-06-16, 21:08, Michael Ellerman wrote: > > > On Wed, 2016-06-01 at 16:04 +0530, Viresh Kumar wrote: > > > > > > > The policy is copied (unnecessarily) and is never freed. Fix it by just > > > > getting a reference to the existing policy structure and putting it > > > > back. > > > > > > > > Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> > > > > > > When was it broken, always? > > > > > > Cc: stable ? > > > > Its a small memory leak and its not that we will fail on something. So > > didn't bother to add those details, but in case they are required: > > > > Cc: <stable@vger.kernel.org> # 4.3+ > > Fixes: 227942809b52 ("cpufreq: powernv: Restore cpu frequency to policy->cur on unthrottling") > > OK. I can't actually see where the copy is? > > But if we are leaking even a small amount of memory in a loop like that, in a > function that's run semi-regularly, then it's going to add up eventually. Urg, it wasn't a memory leak actually. I misread. I somehow thought that cpufreq_get_policy() is also allocating memory for the policy, but it just memcpy's it into the callers buffer. So, that's not a problem really. This patch should be just dropped. Sorry for the noise. -- viresh
diff --git a/drivers/cpufreq/powernv-cpufreq.c b/drivers/cpufreq/powernv-cpufreq.c index 54c45368e3f1..96bb4acd366e 100644 --- a/drivers/cpufreq/powernv-cpufreq.c +++ b/drivers/cpufreq/powernv-cpufreq.c @@ -756,15 +756,18 @@ void powernv_cpufreq_work_fn(struct work_struct *work) chip->restore = false; for_each_cpu(cpu, &mask) { + struct cpufreq_policy *policy = cpufreq_cpu_get(cpu) int index; - struct cpufreq_policy policy; - cpufreq_get_policy(&policy, cpu); - cpufreq_frequency_table_target(&policy, policy.freq_table, - policy.cur, + if (!policy) + continue; + + cpufreq_frequency_table_target(policy, policy->freq_table, + policy->cur, CPUFREQ_RELATION_C, &index); - powernv_cpufreq_target_index(&policy, index); - cpumask_andnot(&mask, &mask, policy.cpus); + powernv_cpufreq_target_index(policy, index); + cpumask_andnot(&mask, &mask, policy->cpus); + cpufreq_cpu_put(policy); } out: put_online_cpus();
The policy is copied (unnecessarily) and is never freed. Fix it by just getting a reference to the existing policy structure and putting it back. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> --- drivers/cpufreq/powernv-cpufreq.c | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) -- 2.7.1.410.g6faf27b