On Wed, 2015-07-08 at 11:23 +0530, Viresh Kumar wrote:
When all CPUs of a policy are hot-unplugged, we EXIT the governor but don't mark policy->governor as NULL. This was done in order to keep last used governor's information intact in sysfs, while the CPUs are offline.
We also missed marking policy->governor as NULL while restoring the policy. Because of that, we call __cpufreq_governor(CPUFREQ_GOV_LIMITS) for an uninitialized policy. Which eventually returns -EBUSY.
Fix this by setting policy->governor to NULL while restoring the policy.
Reported-by: Pi-Cheng Chen pi-cheng.chen@linaro.org Reported-by: "Jon Medhurst (Tixy)" tixy@linaro.org Fixes: 18bf3a124ef8 ("cpufreq: Mark policy->governor = NULL for inactive policies") Signed-off-by: Viresh Kumar viresh.kumar@linaro.org
Tested-by: Jon Medhurst tixy@linaro.org
Thanks for that.
I believe this also fixes the other issue I mentioned (nullptr deref in in arm_big_little driver). To test that, after applying this patch, I modified the code to force __cpufreq_governor to still return an error when a cpu is hotpluged back in. Now the arm_big_little driver doesn't get called when I manually poke scaling_setspeed, presumably because policy->governor==NULL prevents that from reaching the driver?
For 4.2-rc
drivers/cpufreq/cpufreq.c | 1 + 1 file changed, 1 insertion(+)
diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index b612411655f9..2c22e3902e72 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1132,6 +1132,7 @@ static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu) down_write(&policy->rwsem); policy->cpu = cpu;
policy->governor = NULL;