From 9c8f1ee40b6368e6b2775c9c9f816e2a5dca3c07 Mon Sep 17 00:00:00 2001 From: Viresh Kumar Date: Thu, 12 Sep 2013 17:06:33 +0530 Subject: cpufreq: Clear policy->cpus bits in __cpufreq_remove_dev_finish() This broke after a recent change "cedb70a cpufreq: Split __cpufreq_remove_dev() into two parts" from Srivatsa. Consider a scenario where we have two CPUs in a policy (0 & 1) and we are removing CPU 1. On the call to __cpufreq_remove_dev_prepare() we have cleared 1 from policy->cpus and now on a call to __cpufreq_remove_dev_finish() we read cpumask_weight of policy->cpus, which will come as 1 and this code will behave as if we are removing the last CPU from policy :) Fix it by clearing the CPU mask in __cpufreq_remove_dev_finish() instead of __cpufreq_remove_dev_prepare(). Tested-by: Stephen Warren Reviewed-by: Srivatsa S. Bhat Signed-off-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 43c24aa756f..dbfe219667d 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1125,7 +1125,7 @@ static int cpufreq_nominate_new_policy_cpu(struct cpufreq_policy *policy, int ret; /* first sibling now owns the new sysfs dir */ - cpu_dev = get_cpu_device(cpumask_first(policy->cpus)); + cpu_dev = get_cpu_device(cpumask_any_but(policy->cpus, old_cpu)); /* Don't touch sysfs files during light-weight tear-down */ if (frozen) @@ -1189,12 +1189,9 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, policy->governor->name, CPUFREQ_NAME_LEN); #endif - WARN_ON(lock_policy_rwsem_write(cpu)); + lock_policy_rwsem_read(cpu); cpus = cpumask_weight(policy->cpus); - - if (cpus > 1) - cpumask_clear_cpu(cpu, policy->cpus); - unlock_policy_rwsem_write(cpu); + unlock_policy_rwsem_read(cpu); if (cpu != policy->cpu) { if (!frozen) @@ -1237,9 +1234,12 @@ static int __cpufreq_remove_dev_finish(struct device *dev, return -EINVAL; } - lock_policy_rwsem_read(cpu); + WARN_ON(lock_policy_rwsem_write(cpu)); cpus = cpumask_weight(policy->cpus); - unlock_policy_rwsem_read(cpu); + + if (cpus > 1) + cpumask_clear_cpu(cpu, policy->cpus); + unlock_policy_rwsem_write(cpu); /* If cpu is last user of policy, free policy */ if (cpus == 1) { -- cgit v1.2.3-70-g09d2 From 8efd57657d8ef666810b55e609da72de92314dc4 Mon Sep 17 00:00:00 2001 From: Viresh Kumar Date: Tue, 17 Sep 2013 10:22:11 +0530 Subject: cpufreq: unlock correct rwsem while updating policy->cpu Current code looks like this: WARN_ON(lock_policy_rwsem_write(cpu)); update_policy_cpu(policy, new_cpu); unlock_policy_rwsem_write(cpu); {lock|unlock}_policy_rwsem_write(cpu) takes/releases policy->cpu's rwsem. Because cpu is changing with the call to update_policy_cpu(), the unlock_policy_rwsem_write() will release the incorrect lock. The right solution would be to release the same lock as was taken earlier. Also update_policy_cpu() was also called from cpufreq_add_dev() without any locks and so its better if we move this locking to inside update_policy_cpu(). This patch fixes a regression introduced in 3.12 by commit f9ba680d23 (cpufreq: Extract the handover of policy cpu to a helper function). Reported-and-tested-by: Jon Medhurst Signed-off-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq.c | 13 +++++++++++-- 1 file changed, 11 insertions(+), 2 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index dbfe219667d..82ecbe39dfb 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -952,9 +952,20 @@ static void update_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu) if (cpu == policy->cpu) return; + /* + * Take direct locks as lock_policy_rwsem_write wouldn't work here. + * Also lock for last cpu is enough here as contention will happen only + * after policy->cpu is changed and after it is changed, other threads + * will try to acquire lock for new cpu. And policy is already updated + * by then. + */ + down_write(&per_cpu(cpu_policy_rwsem, policy->cpu)); + policy->last_cpu = policy->cpu; policy->cpu = cpu; + up_write(&per_cpu(cpu_policy_rwsem, policy->last_cpu)); + #ifdef CONFIG_CPU_FREQ_TABLE cpufreq_frequency_table_update_policy_cpu(policy); #endif @@ -1200,9 +1211,7 @@ static int __cpufreq_remove_dev_prepare(struct device *dev, new_cpu = cpufreq_nominate_new_policy_cpu(policy, cpu, frozen); if (new_cpu >= 0) { - WARN_ON(lock_policy_rwsem_write(cpu)); update_policy_cpu(policy, new_cpu); - unlock_policy_rwsem_write(cpu); if (!frozen) { pr_debug("%s: policy Kobject moved to cpu: %d " -- cgit v1.2.3-70-g09d2 From 4dea5806d332f91d640d99943db99a5539e832c3 Mon Sep 17 00:00:00 2001 From: Yinghai Lu Date: Wed, 18 Sep 2013 21:05:20 -0700 Subject: cpufreq: return EEXIST instead of EBUSY for second registering On systems that support intel_pstate, acpi_cpufreq fails to load, and udev keeps trying until trace gets filled up and kernel crashes. The root cause is driver return ret from cpufreq_register_driver(), because when some other driver takes over before, it will return EBUSY and then udev will keep trying ... cpufreq_register_driver() should return EEXIST instead so that the system can boot without appending intel_pstate=disable and still use intel_pstate. Signed-off-by: Yinghai Lu Acked-by: Viresh Kumar Signed-off-by: Rafael J. Wysocki --- drivers/cpufreq/cpufreq.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 82ecbe39dfb..89b3c52cd5c 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -2104,7 +2104,7 @@ int cpufreq_register_driver(struct cpufreq_driver *driver_data) write_lock_irqsave(&cpufreq_driver_lock, flags); if (cpufreq_driver) { write_unlock_irqrestore(&cpufreq_driver_lock, flags); - return -EBUSY; + return -EEXIST; } cpufreq_driver = driver_data; write_unlock_irqrestore(&cpufreq_driver_lock, flags); -- cgit v1.2.3-70-g09d2