From 7d26e2d5e2da37e92c6c7644b26b294dedd8c982 Mon Sep 17 00:00:00 2001 From: "venkatesh.pallipadi@intel.com" Date: Thu, 2 Jul 2009 17:08:30 -0700 Subject: [CPUFREQ] Eliminate the recent lockdep warnings in cpufreq Commit b14893a62c73af0eca414cfed505b8c09efc613c although it was very much needed to properly cleanup ondemand timer, opened-up a can of worms related to locking dependencies in cpufreq. Patch here defines the need for dbs_mutex and cleans up its usage in ondemand governor. This also resolves the lockdep warnings reported here http://lkml.indiana.edu/hypermail/linux/kernel/0906.1/01925.html http://lkml.indiana.edu/hypermail/linux/kernel/0907.0/00820.html and few others.. Signed-off-by: Venkatesh Pallipadi Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 6e2ec0b18948..c7fe16e0474b 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1070,8 +1070,6 @@ static int __cpufreq_remove_dev(struct sys_device *sys_dev) spin_unlock_irqrestore(&cpufreq_driver_lock, flags); #endif - unlock_policy_rwsem_write(cpu); - if (cpufreq_driver->target) __cpufreq_governor(data, CPUFREQ_GOV_STOP); @@ -1088,6 +1086,8 @@ static int __cpufreq_remove_dev(struct sys_device *sys_dev) if (cpufreq_driver->exit) cpufreq_driver->exit(data); + unlock_policy_rwsem_write(cpu); + free_cpumask_var(data->related_cpus); free_cpumask_var(data->cpus); kfree(data); -- cgit v1.2.3 From 3f4a782b5ce2698b1870b5a7b573cd721d4fce33 Mon Sep 17 00:00:00 2001 From: Mathieu Desnoyers Date: Fri, 3 Jul 2009 11:25:16 -0400 Subject: [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 CC: Venkatesh Pallipadi CC: rjw@sisk.pl CC: mingo@elte.hu CC: Shaohua Li CC: Pekka Enberg CC: Dave Young CC: "Rafael J. Wysocki" CC: Rusty Russell CC: sven.wegener@stealer.net CC: cpufreq@vger.kernel.org CC: Thomas Renninger Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 65 +++++++++++++++++++++++++++++------------------ 1 file changed, 40 insertions(+), 25 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index c7fe16e0474b..c668ac855f71 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -761,6 +761,10 @@ static struct kobj_type ktype_cpufreq = { * 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) /* 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) /* 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) 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 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) 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 @@ err_out_unregister: 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: -- cgit v1.2.3 From 5e1596f75395e7a402e1059c518e633d2732dcf8 Mon Sep 17 00:00:00 2001 From: Dave Jones Date: Wed, 8 Jul 2009 16:14:23 -0400 Subject: [CPUFREQ] Fix compile failure in cpufreq.c managed_policy is out of scope for the non-smp case. Declare it locally where used (twice) Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index c668ac855f71..b90eda8b3440 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -776,9 +776,6 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) struct sys_device *cpu_sys_dev; unsigned long flags; unsigned int j; -#ifdef CONFIG_SMP - struct cpufreq_policy *managed_policy; -#endif if (cpu_is_offline(cpu)) return 0; @@ -854,6 +851,8 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) #endif for_each_cpu(j, policy->cpus) { + struct cpufreq_policy *managed_policy; + if (cpu == j) continue; @@ -932,6 +931,8 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) /* symlink affected CPUs */ for_each_cpu(j, policy->cpus) { + struct cpufreq_policy *managed_policy; + if (j == cpu) continue; if (!cpu_online(j)) -- cgit v1.2.3 From 42c74b84c64633dd3badbfc2abd2ef1728b64b30 Mon Sep 17 00:00:00 2001 From: Prarit Bhargava Date: Mon, 3 Aug 2009 10:58:11 -0400 Subject: [CPUFREQ] Do not set policy for offline cpus Suspend/Resume fails on multi socket, multi core systems because the cpufreq code erroneously sets the per_cpu policy_cpu value when a logical cpu is offline. This most notably results in missing sysfs files that are used to set the cpu frequencies of the various cpus. Signed-off-by: Prarit Bhargava Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index b90eda8b3440..120d236c0ffb 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -924,6 +924,8 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) spin_lock_irqsave(&cpufreq_driver_lock, flags); for_each_cpu(j, policy->cpus) { + if (!cpu_online(j)) + continue; per_cpu(cpufreq_cpu_data, j) = policy; per_cpu(policy_cpu, j) = policy->cpu; } -- cgit v1.2.3 From d5194decd0a6f792b2789eebd4ddf022a248f655 Mon Sep 17 00:00:00 2001 From: Thomas Renninger Date: Wed, 29 Jul 2009 11:26:20 +0200 Subject: [CPUFREQ] Fix a kobject reference bug related to managed CPUs The first offline/online cycle is successful, the second not. Doing: echo 0 >cpu1/online echo 1 >cpu1/online echo 0 >cpu1/online The last command will trigger: Jul 22 14:39:50 linux kernel: [ 593.210125] ------------[ cut here ]------------ Jul 22 14:39:50 linux kernel: [ 593.210139] WARNING: at lib/kref.c:43 kref_get+0x23/0x2b() Jul 22 14:39:50 linux kernel: [ 593.210144] Hardware name: To Be Filled By O.E.M. Jul 22 14:39:50 linux kernel: [ 593.210148] Modules linked in: powernow_k8 Jul 22 14:39:50 linux kernel: [ 593.210158] Pid: 378, comm: kondemand/2 Tainted: G W 2.6.31-rc2 #38 Jul 22 14:39:50 linux kernel: [ 593.210163] Call Trace: Jul 22 14:39:50 linux kernel: [ 593.210171] [] ? kref_get+0x23/0x2b Jul 22 14:39:50 linux kernel: [ 593.210181] [] warn_slowpath_common+0x77/0xa4 Jul 22 14:39:50 linux kernel: [ 593.210190] [] warn_slowpath_null+0xf/0x11 Jul 22 14:39:50 linux kernel: [ 593.210198] [] kref_get+0x23/0x2b Jul 22 14:39:50 linux kernel: [ 593.210206] [] kobject_get+0x1a/0x22 Jul 22 14:39:50 linux kernel: [ 593.210214] [] cpufreq_cpu_get+0x8a/0xcb Jul 22 14:39:50 linux kernel: [ 593.210222] [] __cpufreq_driver_getavg+0x1d/0x67 Jul 22 14:39:50 linux kernel: [ 593.210231] [] do_dbs_timer+0x158/0x27f Jul 22 14:39:50 linux kernel: [ 593.210240] [] worker_thread+0x200/0x313 ... The output continues on every do_dbs_timer ondemand freq checking poll. This regression was introduced by git commit: 3f4a782b5ce2698b1870b5a7b573cd721d4fce33 The policy is released when the cpufreq device is removed in: __cpufreq_remove_dev(): /* if this isn't the CPU which is the parent of the kobj, we * only need to unlink, put and exit */ Not creating the symlink is not sever at all. As long as: sysfs_remove_link(&sys_dev->kobj, "cpufreq"); handles it gracefully that the symlink did not exist. Possibly no error should be returned at all, because ondemand governor would still provide the same functionality. Userspace in userspace gov case might be confused if the link is missing. Resolves http://bugzilla.kernel.org/show_bug.cgi?id=13903 CC: Mathieu Desnoyers CC: Venkatesh Pallipadi Signed-off-by: Thomas Renninger Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index 120d236c0ffb..bd74a0b12176 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -858,6 +858,8 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) /* Check for existing affected CPUs. * They may not be aware of it due to CPU Hotplug. + * cpufreq_cpu_put is called when the device is removed + * in __cpufreq_remove_dev() */ managed_policy = cpufreq_cpu_get(j); if (unlikely(managed_policy)) { @@ -884,7 +886,7 @@ static int cpufreq_add_dev(struct sys_device *sys_dev) ret = sysfs_create_link(&sys_dev->kobj, &managed_policy->kobj, "cpufreq"); - if (!ret) + if (ret) cpufreq_cpu_put(managed_policy); /* * Success. We only needed to be added to the mask. -- cgit v1.2.3 From 4bc5d34135039566b8d6efa2de7515b2be505da8 Mon Sep 17 00:00:00 2001 From: Dave Jones Date: Tue, 4 Aug 2009 14:03:25 -0400 Subject: [CPUFREQ] Make cpufreq suspend code conditional on powerpc. The suspend code runs with interrupts disabled, and the powerpc workaround we do in the cpufreq suspend hook calls the drivers ->get method. powernow-k8's ->get does an smp_call_function_single which needs interrupts enabled cpufreq's suspend/resume code was added in 42d4dc3f4e1e to work around a hardware problem on ppc powerbooks. If we make all this code conditional on powerpc, we avoid the issue above. Signed-off-by: Dave Jones --- drivers/cpufreq/cpufreq.c | 21 +++++++++++++++++++-- 1 file changed, 19 insertions(+), 2 deletions(-) (limited to 'drivers/cpufreq/cpufreq.c') diff --git a/drivers/cpufreq/cpufreq.c b/drivers/cpufreq/cpufreq.c index bd74a0b12176..fd69086d08d5 100644 --- a/drivers/cpufreq/cpufreq.c +++ b/drivers/cpufreq/cpufreq.c @@ -1248,13 +1248,22 @@ EXPORT_SYMBOL(cpufreq_get); static int cpufreq_suspend(struct sys_device *sysdev, pm_message_t pmsg) { - int cpu = sysdev->id; int ret = 0; + +#ifdef __powerpc__ + int cpu = sysdev->id; unsigned int cur_freq = 0; struct cpufreq_policy *cpu_policy; dprintk("suspending cpu %u\n", cpu); + /* + * This whole bogosity is here because Powerbooks are made of fail. + * No sane platform should need any of the code below to be run. + * (it's entirely the wrong thing to do, as driver->get may + * reenable interrupts on some architectures). + */ + if (!cpu_online(cpu)) return 0; @@ -1313,6 +1322,7 @@ static int cpufreq_suspend(struct sys_device *sysdev, pm_message_t pmsg) out: cpufreq_cpu_put(cpu_policy); +#endif /* __powerpc__ */ return ret; } @@ -1326,12 +1336,18 @@ out: */ static int cpufreq_resume(struct sys_device *sysdev) { - int cpu = sysdev->id; int ret = 0; + +#ifdef __powerpc__ + int cpu = sysdev->id; struct cpufreq_policy *cpu_policy; dprintk("resuming cpu %u\n", cpu); + /* As with the ->suspend method, all the code below is + * only necessary because Powerbooks suck. + * See commit 42d4dc3f4e1e for jokes. */ + if (!cpu_online(cpu)) return 0; @@ -1395,6 +1411,7 @@ out: schedule_work(&cpu_policy->update); fail: cpufreq_cpu_put(cpu_policy); +#endif /* __powerpc__ */ return ret; } -- cgit v1.2.3