kernel_optimize_test/drivers/cpufreq/cpufreq.c

2162 lines
55 KiB
C
Raw Normal View History

/*
* linux/drivers/cpufreq/cpufreq.c
*
* Copyright (C) 2001 Russell King
* (C) 2002 - 2003 Dominik Brodowski <linux@brodo.de>
* (C) 2013 Viresh Kumar <viresh.kumar@linaro.org>
*
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
* Oct 2005 - Ashok Raj <ashok.raj@intel.com>
* Added handling for CPU hotplug
* Feb 2006 - Jacob Shin <jacob.shin@amd.com>
* Fix handling for CPU hotplug -- affected CPUs
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
#include <linux/cpu.h>
#include <linux/cpufreq.h>
#include <linux/delay.h>
#include <linux/device.h>
#include <linux/init.h>
#include <linux/kernel_stat.h>
#include <linux/module.h>
#include <linux/mutex.h>
#include <linux/slab.h>
#include <linux/syscore_ops.h>
#include <linux/tick.h>
#include <trace/events/power.h>
/**
* The "cpufreq driver" - the arch- or hardware-dependent low
* level driver of CPUFreq support, and its spinlock. This lock
* also protects the cpufreq_cpu_data array.
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
static struct cpufreq_driver *cpufreq_driver;
static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data);
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
static DEFINE_PER_CPU(struct cpufreq_policy *, cpufreq_cpu_data_fallback);
static DEFINE_RWLOCK(cpufreq_driver_lock);
static DEFINE_MUTEX(cpufreq_governor_lock);
static LIST_HEAD(cpufreq_policy_list);
#ifdef CONFIG_HOTPLUG_CPU
/* This one keeps track of the previously set governor of a removed CPU */
static DEFINE_PER_CPU(char[CPUFREQ_NAME_LEN], cpufreq_cpu_governor);
#endif
/*
* cpu_policy_rwsem is a per CPU reader-writer semaphore designed to cure
* all cpufreq/hotplug/workqueue/etc related lock issues.
*
* The rules for this semaphore:
* - Any routine that wants to read from the policy structure will
* do a down_read on this semaphore.
* - Any routine that will write to the policy structure and/or may take away
* the policy altogether (eg. CPU hotplug), will hold this lock in write
* mode before doing so.
*
* Additional rules:
* - Governor routines that can be called in cpufreq hotplug path should not
* take this sem as top level hotplug notifier handler takes this.
* - Lock should not be held across
* __cpufreq_governor(data, CPUFREQ_GOV_STOP);
*/
static DEFINE_PER_CPU(int, cpufreq_policy_cpu);
static DEFINE_PER_CPU(struct rw_semaphore, cpu_policy_rwsem);
#define lock_policy_rwsem(mode, cpu) \
static int lock_policy_rwsem_##mode(int cpu) \
{ \
int policy_cpu = per_cpu(cpufreq_policy_cpu, cpu); \
BUG_ON(policy_cpu == -1); \
down_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu)); \
\
return 0; \
}
lock_policy_rwsem(read, cpu);
lock_policy_rwsem(write, cpu);
#define unlock_policy_rwsem(mode, cpu) \
static void unlock_policy_rwsem_##mode(int cpu) \
{ \
int policy_cpu = per_cpu(cpufreq_policy_cpu, cpu); \
BUG_ON(policy_cpu == -1); \
up_##mode(&per_cpu(cpu_policy_rwsem, policy_cpu)); \
}
unlock_policy_rwsem(read, cpu);
unlock_policy_rwsem(write, cpu);
/* internal prototypes */
static int __cpufreq_governor(struct cpufreq_policy *policy,
unsigned int event);
static unsigned int __cpufreq_get(unsigned int cpu);
2006-11-22 22:55:48 +08:00
static void handle_update(struct work_struct *work);
/**
* Two notifier lists: the "policy" list is involved in the
* validation process for a new CPU frequency policy; the
* "transition" list for kernel code that needs to handle
* changes to devices when the CPU clock speed changes.
* The mutex locks both lists.
*/
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
static BLOCKING_NOTIFIER_HEAD(cpufreq_policy_notifier_list);
static struct srcu_notifier_head cpufreq_transition_notifier_list;
static bool init_cpufreq_transition_notifier_list_called;
static int __init init_cpufreq_transition_notifier_list(void)
{
srcu_init_notifier_head(&cpufreq_transition_notifier_list);
init_cpufreq_transition_notifier_list_called = true;
return 0;
}
pure_initcall(init_cpufreq_transition_notifier_list);
static int off __read_mostly;
static int cpufreq_disabled(void)
{
return off;
}
void disable_cpufreq(void)
{
off = 1;
}
static LIST_HEAD(cpufreq_governor_list);
static DEFINE_MUTEX(cpufreq_governor_mutex);
bool have_governor_per_policy(void)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
return cpufreq_driver->have_governor_per_policy;
}
EXPORT_SYMBOL_GPL(have_governor_per_policy);
struct kobject *get_governor_parent_kobj(struct cpufreq_policy *policy)
{
if (have_governor_per_policy())
return &policy->kobj;
else
return cpufreq_global_kobject;
}
EXPORT_SYMBOL_GPL(get_governor_parent_kobj);
static inline u64 get_cpu_idle_time_jiffy(unsigned int cpu, u64 *wall)
{
u64 idle_time;
u64 cur_wall_time;
u64 busy_time;
cur_wall_time = jiffies64_to_cputime64(get_jiffies_64());
busy_time = kcpustat_cpu(cpu).cpustat[CPUTIME_USER];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SYSTEM];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_IRQ];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_SOFTIRQ];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_STEAL];
busy_time += kcpustat_cpu(cpu).cpustat[CPUTIME_NICE];
idle_time = cur_wall_time - busy_time;
if (wall)
*wall = cputime_to_usecs(cur_wall_time);
return cputime_to_usecs(idle_time);
}
u64 get_cpu_idle_time(unsigned int cpu, u64 *wall, int io_busy)
{
u64 idle_time = get_cpu_idle_time_us(cpu, io_busy ? wall : NULL);
if (idle_time == -1ULL)
return get_cpu_idle_time_jiffy(cpu, wall);
else if (!io_busy)
idle_time += get_cpu_iowait_time_us(cpu, wall);
return idle_time;
}
EXPORT_SYMBOL_GPL(get_cpu_idle_time);
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
static struct cpufreq_policy *__cpufreq_cpu_get(unsigned int cpu, bool sysfs)
{
struct cpufreq_policy *policy;
unsigned long flags;
if (cpu >= nr_cpu_ids)
goto err_out;
/* get the cpufreq driver */
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
read_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver)
goto err_out_unlock;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!try_module_get(cpufreq_driver->owner))
goto err_out_unlock;
/* get the CPU */
policy = per_cpu(cpufreq_cpu_data, cpu);
if (!policy)
goto err_out_put_module;
if (!sysfs && !kobject_get(&policy->kobj))
goto err_out_put_module;
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
return policy;
err_out_put_module:
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
module_put(cpufreq_driver->owner);
err_out_unlock:
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
err_out:
return NULL;
}
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
struct cpufreq_policy *cpufreq_cpu_get(unsigned int cpu)
{
if (cpufreq_disabled())
return NULL;
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
return __cpufreq_cpu_get(cpu, false);
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_get);
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
static struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu)
{
return __cpufreq_cpu_get(cpu, true);
}
static void __cpufreq_cpu_put(struct cpufreq_policy *policy, bool sysfs)
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
{
if (!sysfs)
kobject_put(&policy->kobj);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
module_put(cpufreq_driver->owner);
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
}
void cpufreq_cpu_put(struct cpufreq_policy *policy)
{
if (cpufreq_disabled())
return;
__cpufreq_cpu_put(policy, false);
}
EXPORT_SYMBOL_GPL(cpufreq_cpu_put);
static void cpufreq_cpu_put_sysfs(struct cpufreq_policy *policy)
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
{
__cpufreq_cpu_put(policy, true);
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
}
/*********************************************************************
* EXTERNALLY AFFECTING FREQUENCY CHANGES *
*********************************************************************/
/**
* adjust_jiffies - adjust the system "loops_per_jiffy"
*
* This function alters the system "loops_per_jiffy" for the clock
* speed change. Note that loops_per_jiffy cannot be updated on SMP
* systems as each CPU might be scaled differently. So, use the arch
* per-CPU loops_per_jiffy value wherever possible.
*/
#ifndef CONFIG_SMP
static unsigned long l_p_j_ref;
static unsigned int l_p_j_ref_freq;
static void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
if (ci->flags & CPUFREQ_CONST_LOOPS)
return;
if (!l_p_j_ref_freq) {
l_p_j_ref = loops_per_jiffy;
l_p_j_ref_freq = ci->old;
pr_debug("saving %lu as reference value for loops_per_jiffy; "
"freq is %u kHz\n", l_p_j_ref, l_p_j_ref_freq);
}
if ((val == CPUFREQ_POSTCHANGE && ci->old != ci->new) ||
(val == CPUFREQ_RESUMECHANGE || val == CPUFREQ_SUSPENDCHANGE)) {
loops_per_jiffy = cpufreq_scale(l_p_j_ref, l_p_j_ref_freq,
ci->new);
pr_debug("scaling loops_per_jiffy to %lu "
"for frequency %u kHz\n", loops_per_jiffy, ci->new);
}
}
#else
static inline void adjust_jiffies(unsigned long val, struct cpufreq_freqs *ci)
{
return;
}
#endif
static void __cpufreq_notify_transition(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs, unsigned int state)
{
BUG_ON(irqs_disabled());
if (cpufreq_disabled())
return;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
freqs->flags = cpufreq_driver->flags;
pr_debug("notification %u of frequency transition to %u kHz\n",
state, freqs->new);
switch (state) {
case CPUFREQ_PRECHANGE:
if (WARN(policy->transition_ongoing ==
cpumask_weight(policy->cpus),
"In middle of another frequency transition\n"))
return;
policy->transition_ongoing++;
/* detect if the driver reported a value as "old frequency"
* which is not equal to what the cpufreq core thinks is
* "old frequency".
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
if ((policy) && (policy->cpu == freqs->cpu) &&
(policy->cur) && (policy->cur != freqs->old)) {
pr_debug("Warning: CPU frequency is"
" %u, cpufreq assumed %u kHz.\n",
freqs->old, policy->cur);
freqs->old = policy->cur;
}
}
srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
CPUFREQ_PRECHANGE, freqs);
adjust_jiffies(CPUFREQ_PRECHANGE, freqs);
break;
case CPUFREQ_POSTCHANGE:
if (WARN(!policy->transition_ongoing,
"No frequency transition in progress\n"))
return;
policy->transition_ongoing--;
adjust_jiffies(CPUFREQ_POSTCHANGE, freqs);
pr_debug("FREQ: %lu - CPU: %lu", (unsigned long)freqs->new,
(unsigned long)freqs->cpu);
trace_cpu_frequency(freqs->new, freqs->cpu);
srcu_notifier_call_chain(&cpufreq_transition_notifier_list,
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
CPUFREQ_POSTCHANGE, freqs);
if (likely(policy) && likely(policy->cpu == freqs->cpu))
policy->cur = freqs->new;
break;
}
}
/**
* cpufreq_notify_transition - call notifier chain and adjust_jiffies
* on frequency transition.
*
* This function calls the transition notifiers and the "adjust_jiffies"
* function. It is called twice on all CPU frequency changes that have
* external effects.
*/
void cpufreq_notify_transition(struct cpufreq_policy *policy,
struct cpufreq_freqs *freqs, unsigned int state)
{
for_each_cpu(freqs->cpu, policy->cpus)
__cpufreq_notify_transition(policy, freqs, state);
}
EXPORT_SYMBOL_GPL(cpufreq_notify_transition);
/*********************************************************************
* SYSFS INTERFACE *
*********************************************************************/
static struct cpufreq_governor *__find_governor(const char *str_governor)
{
struct cpufreq_governor *t;
list_for_each_entry(t, &cpufreq_governor_list, governor_list)
if (!strnicmp(str_governor, t->name, CPUFREQ_NAME_LEN))
return t;
return NULL;
}
/**
* cpufreq_parse_governor - parse a governor string
*/
static int cpufreq_parse_governor(char *str_governor, unsigned int *policy,
struct cpufreq_governor **governor)
{
int err = -EINVAL;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver)
goto out;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->setpolicy) {
if (!strnicmp(str_governor, "performance", CPUFREQ_NAME_LEN)) {
*policy = CPUFREQ_POLICY_PERFORMANCE;
err = 0;
} else if (!strnicmp(str_governor, "powersave",
CPUFREQ_NAME_LEN)) {
*policy = CPUFREQ_POLICY_POWERSAVE;
err = 0;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
} else if (cpufreq_driver->target) {
struct cpufreq_governor *t;
mutex_lock(&cpufreq_governor_mutex);
t = __find_governor(str_governor);
if (t == NULL) {
int ret;
mutex_unlock(&cpufreq_governor_mutex);
ret = request_module("cpufreq_%s", str_governor);
mutex_lock(&cpufreq_governor_mutex);
if (ret == 0)
t = __find_governor(str_governor);
}
if (t != NULL) {
*governor = t;
err = 0;
}
mutex_unlock(&cpufreq_governor_mutex);
}
out:
return err;
}
/**
* cpufreq_per_cpu_attr_read() / show_##file_name() -
* print out cpufreq information
*
* Write out information from cpufreq_driver->policy[cpu]; object must be
* "unsigned int".
*/
#define show_one(file_name, object) \
static ssize_t show_##file_name \
(struct cpufreq_policy *policy, char *buf) \
{ \
return sprintf(buf, "%u\n", policy->object); \
}
show_one(cpuinfo_min_freq, cpuinfo.min_freq);
show_one(cpuinfo_max_freq, cpuinfo.max_freq);
show_one(cpuinfo_transition_latency, cpuinfo.transition_latency);
show_one(scaling_min_freq, min);
show_one(scaling_max_freq, max);
show_one(scaling_cur_freq, cur);
static int __cpufreq_set_policy(struct cpufreq_policy *policy,
struct cpufreq_policy *new_policy);
[CPUFREQ] If max_freq got reduced (e.g. by _PPC) a write to sysfs scaling_governor let cpufreq core stuck at low max_freq for ever The previous patch had bugs (locking and refcount). This one could also be related to the latest DELL reports. But they only slip into this if a user prog (e.g. powersave daemon does when AC got (un) plugged due to a scheme change) echos something to /sys/../cpufreq/scaling_governor while the frequencies got limited by BIOS. This one works: Subject: Max freq stucks at low freq if reduced by _PPC and sysfs gov access The problem is reproducable by(if machine is limiting freqs via BIOS): - Unplugging AC -> max freq gets limited - echo ${governor} >/sys/.../cpufreq/scaling_governor (policy->user_data.max gets overridden with policy->max and will never come up again.) This patch exchanged the cpufreq_set_policy call to __cpufreq_set_policy and duplicated it's functionality but did not override user_data.max. The same happens with overridding min/max values. If freqs are limited and you override the min freq value, the max freq global value will also get stuck to the limited freq, even if BIOS allows all freqs again. Last scenario does only happen if BIOS does not reduce the frequency to the lowest value (should never happen, just for correctness...) drivers/cpufreq/cpufreq.c | 17 +++++++++++++++-- 1 files changed, 15 insertions(+), 2 deletions(-) Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-04-13 21:14:04 +08:00
/**
* cpufreq_per_cpu_attr_write() / store_##file_name() - sysfs write access
*/
#define store_one(file_name, object) \
static ssize_t store_##file_name \
(struct cpufreq_policy *policy, const char *buf, size_t count) \
{ \
unsigned int ret; \
struct cpufreq_policy new_policy; \
\
ret = cpufreq_get_policy(&new_policy, policy->cpu); \
if (ret) \
return -EINVAL; \
\
ret = sscanf(buf, "%u", &new_policy.object); \
if (ret != 1) \
return -EINVAL; \
\
[CPUFREQ] If max_freq got reduced (e.g. by _PPC) a write to sysfs scaling_governor let cpufreq core stuck at low max_freq for ever The previous patch had bugs (locking and refcount). This one could also be related to the latest DELL reports. But they only slip into this if a user prog (e.g. powersave daemon does when AC got (un) plugged due to a scheme change) echos something to /sys/../cpufreq/scaling_governor while the frequencies got limited by BIOS. This one works: Subject: Max freq stucks at low freq if reduced by _PPC and sysfs gov access The problem is reproducable by(if machine is limiting freqs via BIOS): - Unplugging AC -> max freq gets limited - echo ${governor} >/sys/.../cpufreq/scaling_governor (policy->user_data.max gets overridden with policy->max and will never come up again.) This patch exchanged the cpufreq_set_policy call to __cpufreq_set_policy and duplicated it's functionality but did not override user_data.max. The same happens with overridding min/max values. If freqs are limited and you override the min freq value, the max freq global value will also get stuck to the limited freq, even if BIOS allows all freqs again. Last scenario does only happen if BIOS does not reduce the frequency to the lowest value (should never happen, just for correctness...) drivers/cpufreq/cpufreq.c | 17 +++++++++++++++-- 1 files changed, 15 insertions(+), 2 deletions(-) Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-04-13 21:14:04 +08:00
ret = __cpufreq_set_policy(policy, &new_policy); \
policy->user_policy.object = policy->object; \
\
return ret ? ret : count; \
}
store_one(scaling_min_freq, min);
store_one(scaling_max_freq, max);
/**
* show_cpuinfo_cur_freq - current CPU frequency as detected by hardware
*/
static ssize_t show_cpuinfo_cur_freq(struct cpufreq_policy *policy,
char *buf)
{
unsigned int cur_freq = __cpufreq_get(policy->cpu);
if (!cur_freq)
return sprintf(buf, "<unknown>");
return sprintf(buf, "%u\n", cur_freq);
}
/**
* show_scaling_governor - show the current policy for the specified CPU
*/
static ssize_t show_scaling_governor(struct cpufreq_policy *policy, char *buf)
{
if (policy->policy == CPUFREQ_POLICY_POWERSAVE)
return sprintf(buf, "powersave\n");
else if (policy->policy == CPUFREQ_POLICY_PERFORMANCE)
return sprintf(buf, "performance\n");
else if (policy->governor)
return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n",
policy->governor->name);
return -EINVAL;
}
/**
* store_scaling_governor - store policy for the specified CPU
*/
static ssize_t store_scaling_governor(struct cpufreq_policy *policy,
const char *buf, size_t count)
{
unsigned int ret;
char str_governor[16];
struct cpufreq_policy new_policy;
ret = cpufreq_get_policy(&new_policy, policy->cpu);
if (ret)
return ret;
ret = sscanf(buf, "%15s", str_governor);
if (ret != 1)
return -EINVAL;
if (cpufreq_parse_governor(str_governor, &new_policy.policy,
&new_policy.governor))
return -EINVAL;
/*
* Do not use cpufreq_set_policy here or the user_policy.max
* will be wrongly overridden
*/
[CPUFREQ] If max_freq got reduced (e.g. by _PPC) a write to sysfs scaling_governor let cpufreq core stuck at low max_freq for ever The previous patch had bugs (locking and refcount). This one could also be related to the latest DELL reports. But they only slip into this if a user prog (e.g. powersave daemon does when AC got (un) plugged due to a scheme change) echos something to /sys/../cpufreq/scaling_governor while the frequencies got limited by BIOS. This one works: Subject: Max freq stucks at low freq if reduced by _PPC and sysfs gov access The problem is reproducable by(if machine is limiting freqs via BIOS): - Unplugging AC -> max freq gets limited - echo ${governor} >/sys/.../cpufreq/scaling_governor (policy->user_data.max gets overridden with policy->max and will never come up again.) This patch exchanged the cpufreq_set_policy call to __cpufreq_set_policy and duplicated it's functionality but did not override user_data.max. The same happens with overridding min/max values. If freqs are limited and you override the min freq value, the max freq global value will also get stuck to the limited freq, even if BIOS allows all freqs again. Last scenario does only happen if BIOS does not reduce the frequency to the lowest value (should never happen, just for correctness...) drivers/cpufreq/cpufreq.c | 17 +++++++++++++++-- 1 files changed, 15 insertions(+), 2 deletions(-) Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2006-04-13 21:14:04 +08:00
ret = __cpufreq_set_policy(policy, &new_policy);
policy->user_policy.policy = policy->policy;
policy->user_policy.governor = policy->governor;
if (ret)
return ret;
else
return count;
}
/**
* show_scaling_driver - show the cpufreq driver currently loaded
*/
static ssize_t show_scaling_driver(struct cpufreq_policy *policy, char *buf)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
return scnprintf(buf, CPUFREQ_NAME_PLEN, "%s\n", cpufreq_driver->name);
}
/**
* show_scaling_available_governors - show the available CPUfreq governors
*/
static ssize_t show_scaling_available_governors(struct cpufreq_policy *policy,
char *buf)
{
ssize_t i = 0;
struct cpufreq_governor *t;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver->target) {
i += sprintf(buf, "performance powersave");
goto out;
}
list_for_each_entry(t, &cpufreq_governor_list, governor_list) {
if (i >= (ssize_t) ((PAGE_SIZE / sizeof(char))
- (CPUFREQ_NAME_LEN + 2)))
goto out;
i += scnprintf(&buf[i], CPUFREQ_NAME_PLEN, "%s ", t->name);
}
out:
i += sprintf(&buf[i], "\n");
return i;
}
ssize_t cpufreq_show_cpus(const struct cpumask *mask, char *buf)
{
ssize_t i = 0;
unsigned int cpu;
for_each_cpu(cpu, mask) {
if (i)
i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), " ");
i += scnprintf(&buf[i], (PAGE_SIZE - i - 2), "%u", cpu);
if (i >= (PAGE_SIZE - 5))
break;
}
i += sprintf(&buf[i], "\n");
return i;
}
EXPORT_SYMBOL_GPL(cpufreq_show_cpus);
/**
* show_related_cpus - show the CPUs affected by each transition even if
* hw coordination is in use
*/
static ssize_t show_related_cpus(struct cpufreq_policy *policy, char *buf)
{
return cpufreq_show_cpus(policy->related_cpus, buf);
}
/**
* show_affected_cpus - show the CPUs affected by each transition
*/
static ssize_t show_affected_cpus(struct cpufreq_policy *policy, char *buf)
{
return cpufreq_show_cpus(policy->cpus, buf);
}
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-27 01:18:21 +08:00
static ssize_t store_scaling_setspeed(struct cpufreq_policy *policy,
const char *buf, size_t count)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-27 01:18:21 +08:00
{
unsigned int freq = 0;
unsigned int ret;
if (!policy->governor || !policy->governor->store_setspeed)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-27 01:18:21 +08:00
return -EINVAL;
ret = sscanf(buf, "%u", &freq);
if (ret != 1)
return -EINVAL;
policy->governor->store_setspeed(policy, freq);
return count;
}
static ssize_t show_scaling_setspeed(struct cpufreq_policy *policy, char *buf)
{
if (!policy->governor || !policy->governor->show_setspeed)
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-27 01:18:21 +08:00
return sprintf(buf, "<unsupported>\n");
return policy->governor->show_setspeed(policy, buf);
}
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 19:31:01 +08:00
/**
* show_bios_limit - show the current cpufreq HW/BIOS limitation
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 19:31:01 +08:00
*/
static ssize_t show_bios_limit(struct cpufreq_policy *policy, char *buf)
{
unsigned int limit;
int ret;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->bios_limit) {
ret = cpufreq_driver->bios_limit(policy->cpu, &limit);
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 19:31:01 +08:00
if (!ret)
return sprintf(buf, "%u\n", limit);
}
return sprintf(buf, "%u\n", policy->cpuinfo.max_freq);
}
cpufreq_freq_attr_ro_perm(cpuinfo_cur_freq, 0400);
cpufreq_freq_attr_ro(cpuinfo_min_freq);
cpufreq_freq_attr_ro(cpuinfo_max_freq);
cpufreq_freq_attr_ro(cpuinfo_transition_latency);
cpufreq_freq_attr_ro(scaling_available_governors);
cpufreq_freq_attr_ro(scaling_driver);
cpufreq_freq_attr_ro(scaling_cur_freq);
cpufreq_freq_attr_ro(bios_limit);
cpufreq_freq_attr_ro(related_cpus);
cpufreq_freq_attr_ro(affected_cpus);
cpufreq_freq_attr_rw(scaling_min_freq);
cpufreq_freq_attr_rw(scaling_max_freq);
cpufreq_freq_attr_rw(scaling_governor);
cpufreq_freq_attr_rw(scaling_setspeed);
static struct attribute *default_attrs[] = {
&cpuinfo_min_freq.attr,
&cpuinfo_max_freq.attr,
&cpuinfo_transition_latency.attr,
&scaling_min_freq.attr,
&scaling_max_freq.attr,
&affected_cpus.attr,
&related_cpus.attr,
&scaling_governor.attr,
&scaling_driver.attr,
&scaling_available_governors.attr,
[CPUFREQ] Eliminate cpufreq_userspace scaling_setspeed deadlock Eliminate cpufreq_userspace scaling_setspeed deadlock. Luming Yu recently uncovered yet another cpufreq related deadlock. One thread that continuously switches the governors and the other thread that repeatedly cats the contents of cpufreq directory causes both these threads to go into a deadlock. Detailed examination of the deadlock showed the exact flow before the deadlock as: Thread 1 Thread 2 ________ ________ cats files under /sys/devices/.../cpufreq/ Set governor to userspace Adds a new sysfs entry for scaling_setspeed cats files under /sys/devices/.../cpufreq/ Set governor to performance Holds cpufreq_rw_sem in write mode Sends a STOP notify to userspace governor cat /sys/devices/.../cpufreq/scaling_setspeed Gets a handle on the above sysfs entry with sysfs_get_active Blocks while trying to get cpufreq_rw_sem in read mode Remove a sysfs entry for scaling_setspeed Blocks on sysfs_deactivate while waiting for earlier get_active (on other thread) to drain At this point both threads go into deadlock and any other thread that tries to do anything with sysfs cpufreq will also block. There seems to be no easy way to avoid this deadlock as long as cpufreq_userspace adds/removes the sysfs entry under same kobject as cpufreq. Below patch moves scaling_setspeed to cpufreq.c, keeping it always and calling back the governor on read/write. This is the cleanest fix I could think of, even though adding two callbacks in governor structure just for this seems unnecessary. Note that the change makes scaling_setspeed under /sys/.../cpufreq permanent and returns <unsupported> when governor is not userspace. Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Dave Jones <davej@redhat.com>
2007-10-27 01:18:21 +08:00
&scaling_setspeed.attr,
NULL
};
#define to_policy(k) container_of(k, struct cpufreq_policy, kobj)
#define to_attr(a) container_of(a, struct freq_attr, attr)
static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
{
struct cpufreq_policy *policy = to_policy(kobj);
struct freq_attr *fattr = to_attr(attr);
ssize_t ret = -EINVAL;
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
policy = cpufreq_cpu_get_sysfs(policy->cpu);
if (!policy)
goto no_policy;
if (lock_policy_rwsem_read(policy->cpu) < 0)
goto fail;
if (fattr->show)
ret = fattr->show(policy, buf);
else
ret = -EIO;
unlock_policy_rwsem_read(policy->cpu);
fail:
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
cpufreq_cpu_put_sysfs(policy);
no_policy:
return ret;
}
static ssize_t store(struct kobject *kobj, struct attribute *attr,
const char *buf, size_t count)
{
struct cpufreq_policy *policy = to_policy(kobj);
struct freq_attr *fattr = to_attr(attr);
ssize_t ret = -EINVAL;
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
policy = cpufreq_cpu_get_sysfs(policy->cpu);
if (!policy)
goto no_policy;
if (lock_policy_rwsem_write(policy->cpu) < 0)
goto fail;
if (fattr->store)
ret = fattr->store(policy, buf, count);
else
ret = -EIO;
unlock_policy_rwsem_write(policy->cpu);
fail:
cpufreq: Fix sysfs deadlock with concurrent hotplug/frequency switch Running one program that continuously hotplugs and replugs a cpu concurrently with another program that continuously writes to the scaling_setspeed node eventually deadlocks with: ============================================= [ INFO: possible recursive locking detected ] 3.4.0 #37 Tainted: G W --------------------------------------------- filemonkey/122 is trying to acquire lock: (s_active#13){++++.+}, at: [<c01a3d28>] sysfs_remove_dir+0x9c/0xb4 but task is already holding lock: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(s_active#13); lock(s_active#13); *** DEADLOCK *** May be due to missing lock nesting notation 2 locks held by filemonkey/122: #0: (&buffer->mutex){+.+.+.}, at: [<c01a2230>] sysfs_write_file+0x28/0x140 #1: (s_active#13){++++.+}, at: [<c01a22f0>] sysfs_write_file+0xe8/0x140 stack backtrace: [<c0014fcc>] (unwind_backtrace+0x0/0x120) from [<c00ca600>] (validate_chain+0x6f8/0x1054) [<c00ca600>] (validate_chain+0x6f8/0x1054) from [<c00cb778>] (__lock_acquire+0x81c/0x8d8) [<c00cb778>] (__lock_acquire+0x81c/0x8d8) from [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) [<c00cb9c0>] (lock_acquire+0x18c/0x1e8) from [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) [<c01a3ba8>] (sysfs_addrm_finish+0xd0/0x180) from [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) [<c01a3d28>] (sysfs_remove_dir+0x9c/0xb4) from [<c02d0e5c>] (kobject_del+0x10/0x38) [<c02d0e5c>] (kobject_del+0x10/0x38) from [<c02d0f74>] (kobject_release+0xf0/0x194) [<c02d0f74>] (kobject_release+0xf0/0x194) from [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) [<c0565a98>] (cpufreq_cpu_put+0xc/0x24) from [<c05683f0>] (store+0x6c/0x74) [<c05683f0>] (store+0x6c/0x74) from [<c01a2314>] (sysfs_write_file+0x10c/0x140) [<c01a2314>] (sysfs_write_file+0x10c/0x140) from [<c014af44>] (vfs_write+0xb0/0x128) [<c014af44>] (vfs_write+0xb0/0x128) from [<c014b06c>] (sys_write+0x3c/0x68) [<c014b06c>] (sys_write+0x3c/0x68) from [<c000e0e0>] (ret_fast_syscall+0x0/0x3c) This is because store() in cpufreq.c indirectly calls kobject_get() via cpufreq_cpu_get() and is the last one to call kobject_put() via cpufreq_cpu_put(). Sysfs code should not call kobject_get() or kobject_put() directly (see the comment around sysfs_schedule_callback() for more information). Fix this deadlock by introducing two new functions: struct cpufreq_policy *cpufreq_cpu_get_sysfs(unsigned int cpu) void cpufreq_cpu_put_sysfs(struct cpufreq_policy *data) which do the same thing as cpufreq_cpu_{get,put}() but don't call kobject functions. To easily trigger this deadlock you can insert an msleep() with a reasonably large value right after the fail label at the bottom of the store() function in cpufreq.c and then write scaling_setspeed in one task and offline the cpu in another. The first task will hang and be detected by the hung task detector. Signed-off-by: Stephen Boyd <sboyd@codeaurora.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
2012-07-21 02:14:38 +08:00
cpufreq_cpu_put_sysfs(policy);
no_policy:
return ret;
}
static void cpufreq_sysfs_release(struct kobject *kobj)
{
struct cpufreq_policy *policy = to_policy(kobj);
pr_debug("last reference is dropped\n");
complete(&policy->kobj_unregister);
}
static const struct sysfs_ops sysfs_ops = {
.show = show,
.store = store,
};
static struct kobj_type ktype_cpufreq = {
.sysfs_ops = &sysfs_ops,
.default_attrs = default_attrs,
.release = cpufreq_sysfs_release,
};
struct kobject *cpufreq_global_kobject;
EXPORT_SYMBOL(cpufreq_global_kobject);
static int cpufreq_global_kobject_usage;
int cpufreq_get_global_kobject(void)
{
if (!cpufreq_global_kobject_usage++)
return kobject_add(cpufreq_global_kobject,
&cpu_subsys.dev_root->kobj, "%s", "cpufreq");
return 0;
}
EXPORT_SYMBOL(cpufreq_get_global_kobject);
void cpufreq_put_global_kobject(void)
{
if (!--cpufreq_global_kobject_usage)
kobject_del(cpufreq_global_kobject);
}
EXPORT_SYMBOL(cpufreq_put_global_kobject);
int cpufreq_sysfs_create_file(const struct attribute *attr)
{
int ret = cpufreq_get_global_kobject();
if (!ret) {
ret = sysfs_create_file(cpufreq_global_kobject, attr);
if (ret)
cpufreq_put_global_kobject();
}
return ret;
}
EXPORT_SYMBOL(cpufreq_sysfs_create_file);
void cpufreq_sysfs_remove_file(const struct attribute *attr)
{
sysfs_remove_file(cpufreq_global_kobject, attr);
cpufreq_put_global_kobject();
}
EXPORT_SYMBOL(cpufreq_sysfs_remove_file);
/* symlink affected CPUs */
static int cpufreq_add_dev_symlink(struct cpufreq_policy *policy)
{
unsigned int j;
int ret = 0;
for_each_cpu(j, policy->cpus) {
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
struct device *cpu_dev;
if (j == policy->cpu)
continue;
pr_debug("Adding link for CPU: %u\n", j);
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
cpu_dev = get_cpu_device(j);
ret = sysfs_create_link(&cpu_dev->kobj, &policy->kobj,
"cpufreq");
cpufreq: Do not hold driver module references for additional policy CPUs The cpufreq core is a little inconsistent in the way it uses the driver module refcount. Namely, if __cpufreq_add_dev() is called for a CPU that doesn't share the policy object with any other CPUs, the driver module refcount it grabs to start with will be dropped by it before returning and will be equal to whatever it had been before that function was invoked. However, if the given CPU does share the policy object with other CPUs, either cpufreq_add_policy_cpu() is called to link the new CPU to the existing policy, or cpufreq_add_dev_symlink() is used to link the other CPUs sharing the policy with it to the just created policy object. In that case, because both cpufreq_add_policy_cpu() and cpufreq_add_dev_symlink() call cpufreq_cpu_get() for the given policy (the latter possibly many times) without the balancing cpufreq_cpu_put() (unless there is an error), the driver module refcount will be left by __cpufreq_add_dev() with a nonzero value (different from the initial one). To remove that inconsistency make cpufreq_add_policy_cpu() execute cpufreq_cpu_put() for the given policy before returning, which decrements the driver module refcount so that it will be equal to its initial value after __cpufreq_add_dev() returns. Also remove the cpufreq_cpu_get() call from cpufreq_add_dev_symlink(), since both the policy refcount and the driver module refcount are nonzero when it is called and they don't need to be bumped up by it. Accordingly, drop the cpufreq_cpu_put() from __cpufreq_remove_dev(), since it is only necessary to balance the cpufreq_cpu_get() called by cpufreq_add_policy_cpu() or cpufreq_add_dev_symlink(). Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2013-08-04 07:19:34 +08:00
if (ret)
break;
}
return ret;
}
static int cpufreq_add_dev_interface(struct cpufreq_policy *policy,
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
struct device *dev)
{
struct freq_attr **drv_attr;
int ret = 0;
/* prepare interface data */
ret = kobject_init_and_add(&policy->kobj, &ktype_cpufreq,
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
&dev->kobj, "cpufreq");
if (ret)
return ret;
/* set up files for this cpu device */
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
drv_attr = cpufreq_driver->attr;
while ((drv_attr) && (*drv_attr)) {
ret = sysfs_create_file(&policy->kobj, &((*drv_attr)->attr));
if (ret)
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
goto err_out_kobj_put;
drv_attr++;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->get) {
ret = sysfs_create_file(&policy->kobj, &cpuinfo_cur_freq.attr);
if (ret)
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
goto err_out_kobj_put;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->target) {
ret = sysfs_create_file(&policy->kobj, &scaling_cur_freq.attr);
if (ret)
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
goto err_out_kobj_put;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->bios_limit) {
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 19:31:01 +08:00
ret = sysfs_create_file(&policy->kobj, &bios_limit.attr);
if (ret)
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
goto err_out_kobj_put;
[ACPI/CPUFREQ] Introduce bios_limit per cpu cpufreq sysfs interface This interface is mainly intended (and implemented) for ACPI _PPC BIOS frequency limitations, but other cpufreq drivers can also use it for similar use-cases. Why is this needed: Currently it's not obvious why cpufreq got limited. People see cpufreq/scaling_max_freq reduced, but this could have happened by: - any userspace prog writing to scaling_max_freq - thermal limitations - hardware (_PPC in ACPI case) limitiations Therefore export bios_limit (in kHz) to: - Point the user that it's the BIOS (broken or intended) which limits frequency - Export it as a sysfs interface for userspace progs. While this was a rarely used feature on laptops, there will appear more and more server implemenations providing "Green IT" features like allowing the service processor to limit the frequency. People want to know about HW/BIOS frequency limitations. All ACPI P-state driven cpufreq drivers are covered with this patch: - powernow-k8 - powernow-k7 - acpi-cpufreq Tested with a patched DSDT which limits the first two cores (_PPC returns 1) via _PPC, exposed by bios_limit: # echo 2200000 >cpu2/cpufreq/scaling_max_freq # cat cpu*/cpufreq/scaling_max_freq 2600000 2600000 2200000 2200000 # #scaling_max_freq shows general user/thermal/BIOS limitations # cat cpu*/cpufreq/bios_limit 2600000 2600000 2800000 2800000 # #bios_limit only shows the HW/BIOS limitation CC: Pallipadi Venkatesh <venkatesh.pallipadi@intel.com> CC: Len Brown <lenb@kernel.org> CC: davej@codemonkey.org.uk CC: linux@dominikbrodowski.net Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-11-19 19:31:01 +08:00
}
ret = cpufreq_add_dev_symlink(policy);
if (ret)
goto err_out_kobj_put;
return ret;
err_out_kobj_put:
kobject_put(&policy->kobj);
wait_for_completion(&policy->kobj_unregister);
return ret;
}
static void cpufreq_init_policy(struct cpufreq_policy *policy)
{
struct cpufreq_policy new_policy;
int ret = 0;
memcpy(&new_policy, policy, sizeof(*policy));
/* assure that the starting sequence is run in __cpufreq_set_policy */
policy->governor = NULL;
/* set default policy */
ret = __cpufreq_set_policy(policy, &new_policy);
policy->user_policy.policy = policy->policy;
policy->user_policy.governor = policy->governor;
if (ret) {
pr_debug("setting policy failed\n");
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->exit)
cpufreq_driver->exit(policy);
}
}
#ifdef CONFIG_HOTPLUG_CPU
static int cpufreq_add_policy_cpu(struct cpufreq_policy *policy,
unsigned int cpu, struct device *dev,
bool frozen)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
int ret = 0, has_target = !!cpufreq_driver->target;
unsigned long flags;
if (has_target)
__cpufreq_governor(policy, CPUFREQ_GOV_STOP);
lock_policy_rwsem_write(policy->cpu);
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpumask_set_cpu(cpu, policy->cpus);
per_cpu(cpufreq_policy_cpu, cpu) = policy->cpu;
per_cpu(cpufreq_cpu_data, cpu) = policy;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
unlock_policy_rwsem_write(policy->cpu);
if (has_target) {
__cpufreq_governor(policy, CPUFREQ_GOV_START);
__cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
}
/* Don't touch sysfs links during light-weight init */
cpufreq: Do not hold driver module references for additional policy CPUs The cpufreq core is a little inconsistent in the way it uses the driver module refcount. Namely, if __cpufreq_add_dev() is called for a CPU that doesn't share the policy object with any other CPUs, the driver module refcount it grabs to start with will be dropped by it before returning and will be equal to whatever it had been before that function was invoked. However, if the given CPU does share the policy object with other CPUs, either cpufreq_add_policy_cpu() is called to link the new CPU to the existing policy, or cpufreq_add_dev_symlink() is used to link the other CPUs sharing the policy with it to the just created policy object. In that case, because both cpufreq_add_policy_cpu() and cpufreq_add_dev_symlink() call cpufreq_cpu_get() for the given policy (the latter possibly many times) without the balancing cpufreq_cpu_put() (unless there is an error), the driver module refcount will be left by __cpufreq_add_dev() with a nonzero value (different from the initial one). To remove that inconsistency make cpufreq_add_policy_cpu() execute cpufreq_cpu_put() for the given policy before returning, which decrements the driver module refcount so that it will be equal to its initial value after __cpufreq_add_dev() returns. Also remove the cpufreq_cpu_get() call from cpufreq_add_dev_symlink(), since both the policy refcount and the driver module refcount are nonzero when it is called and they don't need to be bumped up by it. Accordingly, drop the cpufreq_cpu_put() from __cpufreq_remove_dev(), since it is only necessary to balance the cpufreq_cpu_get() called by cpufreq_add_policy_cpu() or cpufreq_add_dev_symlink(). Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
2013-08-04 07:19:34 +08:00
if (!frozen)
ret = sysfs_create_link(&dev->kobj, &policy->kobj, "cpufreq");
return ret;
}
#endif
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
static struct cpufreq_policy *cpufreq_policy_restore(unsigned int cpu)
{
struct cpufreq_policy *policy;
unsigned long flags;
write_lock_irqsave(&cpufreq_driver_lock, flags);
policy = per_cpu(cpufreq_cpu_data_fallback, cpu);
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
return policy;
}
static struct cpufreq_policy *cpufreq_policy_alloc(void)
{
struct cpufreq_policy *policy;
policy = kzalloc(sizeof(*policy), GFP_KERNEL);
if (!policy)
return NULL;
if (!alloc_cpumask_var(&policy->cpus, GFP_KERNEL))
goto err_free_policy;
if (!zalloc_cpumask_var(&policy->related_cpus, GFP_KERNEL))
goto err_free_cpumask;
INIT_LIST_HEAD(&policy->policy_list);
return policy;
err_free_cpumask:
free_cpumask_var(policy->cpus);
err_free_policy:
kfree(policy);
return NULL;
}
static void cpufreq_policy_free(struct cpufreq_policy *policy)
{
unsigned long flags;
write_lock_irqsave(&cpufreq_driver_lock, flags);
list_del(&policy->policy_list);
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
free_cpumask_var(policy->related_cpus);
free_cpumask_var(policy->cpus);
kfree(policy);
}
static int __cpufreq_add_dev(struct device *dev, struct subsys_interface *sif,
bool frozen)
{
unsigned int j, cpu = dev->id;
int ret = -ENOMEM;
struct cpufreq_policy *policy;
unsigned long flags;
#ifdef CONFIG_HOTPLUG_CPU
struct cpufreq_policy *tpolicy;
struct cpufreq_governor *gov;
#endif
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
if (cpu_is_offline(cpu))
return 0;
pr_debug("adding CPU %u\n", cpu);
#ifdef CONFIG_SMP
/* check whether a different CPU already registered this
* CPU because it is in the same boat. */
policy = cpufreq_cpu_get(cpu);
if (unlikely(policy)) {
cpufreq_cpu_put(policy);
return 0;
}
#ifdef CONFIG_HOTPLUG_CPU
/* Check if this cpu was hot-unplugged earlier and has siblings */
read_lock_irqsave(&cpufreq_driver_lock, flags);
list_for_each_entry(tpolicy, &cpufreq_policy_list, policy_list) {
if (cpumask_test_cpu(cpu, tpolicy->related_cpus)) {
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
return cpufreq_add_policy_cpu(tpolicy, cpu, dev,
frozen);
}
}
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
#endif
#endif
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!try_module_get(cpufreq_driver->owner)) {
ret = -EINVAL;
goto module_out;
}
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
if (frozen)
/* Restore the saved policy when doing light-weight init */
policy = cpufreq_policy_restore(cpu);
else
policy = cpufreq_policy_alloc();
if (!policy)
goto nomem_out;
policy->cpu = cpu;
policy->governor = CPUFREQ_DEFAULT_GOVERNOR;
cpumask_copy(policy->cpus, cpumask_of(cpu));
/* Initially set CPU itself as the policy_cpu */
per_cpu(cpufreq_policy_cpu, cpu) = cpu;
init_completion(&policy->kobj_unregister);
2006-11-22 22:55:48 +08:00
INIT_WORK(&policy->update, handle_update);
/* call driver. From then on the cpufreq must be able
* to accept all calls to ->verify and ->setpolicy for this CPU
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
ret = cpufreq_driver->init(policy);
if (ret) {
pr_debug("initialization failed\n");
goto err_set_policy_cpu;
}
/* related cpus should atleast have policy->cpus */
cpumask_or(policy->related_cpus, policy->related_cpus, policy->cpus);
/*
* affected cpus must always be the one, which are online. We aren't
* managing offline cpus here.
*/
cpumask_and(policy->cpus, policy->cpus, cpu_online_mask);
policy->user_policy.min = policy->min;
policy->user_policy.max = policy->max;
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_START, policy);
#ifdef CONFIG_HOTPLUG_CPU
gov = __find_governor(per_cpu(cpufreq_cpu_governor, cpu));
if (gov) {
policy->governor = gov;
pr_debug("Restoring governor %s for cpu %d\n",
policy->governor->name, cpu);
[CPUFREQ] Bail out of cpufreq_add_dev if the link for a managed CPU got created Doing: echo 0 >cpu1/online echo 1 >cpu1/online on a managed CPU will result in: Jul 22 15:15:37 linux kernel: [ 80.013864] WARNING: at fs/sysfs/dir.c:487 sysfs_add_one+0xcf/0xe6() Jul 22 15:15:37 linux kernel: [ 80.013866] Hardware name: To Be Filled By O.E.M. Jul 22 15:15:37 linux kernel: [ 80.013868] sysfs: cannot create duplicate filename '/devices/system/cpu/cpu1/cpufreq' Jul 22 15:15:37 linux kernel: [ 80.013870] Modules linked in: powernow_k8 Jul 22 15:15:37 linux kernel: [ 80.013874] Pid: 5750, comm: bash Not tainted 2.6.31-rc2 #40 Jul 22 15:15:37 linux kernel: [ 80.013876] Call Trace: Jul 22 15:15:37 linux kernel: [ 80.013879] [<ffffffff8112ebda>] ? sysfs_add_one+0xcf/0xe6 Jul 22 15:15:37 linux kernel: [ 80.013884] [<ffffffff81041926>] warn_slowpath_common+0x77/0xa4 Jul 22 15:15:37 linux kernel: [ 80.013888] [<ffffffff810419a0>] warn_slowpath_fmt+0x3c/0x3e Jul 22 15:15:37 linux kernel: [ 80.013891] [<ffffffff8112ebda>] sysfs_add_one+0xcf/0xe6 Jul 22 15:15:37 linux kernel: [ 80.013894] [<ffffffff8112f213>] create_dir+0x58/0x87 Jul 22 15:15:37 linux kernel: [ 80.013898] [<ffffffff8112f27a>] sysfs_create_dir+0x38/0x4f Jul 22 15:15:37 linux kernel: [ 80.013902] [<ffffffff811ffb8a>] kobject_add_internal+0x11f/0x1de Jul 22 15:15:37 linux kernel: [ 80.013905] [<ffffffff811ffd21>] kobject_add_varg+0x41/0x4e Jul 22 15:15:37 linux kernel: [ 80.013908] [<ffffffff811ffd7a>] kobject_init_and_add+0x4c/0x57 Jul 22 15:15:37 linux kernel: [ 80.013913] [<ffffffff810667bc>] ? mark_lock+0x22/0x228 Jul 22 15:15:37 linux kernel: [ 80.013918] [<ffffffff813e8a3b>] cpufreq_add_dev_interface+0x40/0x1e4 ... This bug slipped in by git commit: 150b06f7f223cfd0f808737a5243cceca8ea47fa When splitting up cpufreq_add_dev, the whole cpufreq_add_dev function is not left anymore, only cpufreq_add_dev_policy. This patch should reconstruct the identical functionality again as it was before the split. CC: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Thomas Renninger <trenn@suse.de> Signed-off-by: Dave Jones <davej@redhat.com>
2009-07-24 21:25:03 +08:00
}
#endif
write_lock_irqsave(&cpufreq_driver_lock, flags);
for_each_cpu(j, policy->cpus) {
per_cpu(cpufreq_cpu_data, j) = policy;
per_cpu(cpufreq_policy_cpu, j) = policy->cpu;
}
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
if (!frozen) {
ret = cpufreq_add_dev_interface(policy, dev);
if (ret)
goto err_out_unregister;
write_lock_irqsave(&cpufreq_driver_lock, flags);
list_add(&policy->policy_list, &cpufreq_policy_list);
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
}
cpufreq_init_policy(policy);
kobject_uevent(&policy->kobj, KOBJ_ADD);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
module_put(cpufreq_driver->owner);
pr_debug("initialization complete\n");
return 0;
err_out_unregister:
write_lock_irqsave(&cpufreq_driver_lock, flags);
for_each_cpu(j, policy->cpus) {
per_cpu(cpufreq_cpu_data, j) = NULL;
if (j != cpu)
per_cpu(cpufreq_policy_cpu, j) = -1;
}
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
err_set_policy_cpu:
per_cpu(cpufreq_policy_cpu, cpu) = -1;
cpufreq_policy_free(policy);
nomem_out:
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
module_put(cpufreq_driver->owner);
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
module_out:
return ret;
}
/**
* 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 device *dev, struct subsys_interface *sif)
{
return __cpufreq_add_dev(dev, sif, false);
}
static void update_policy_cpu(struct cpufreq_policy *policy, unsigned int cpu)
{
int j;
policy->last_cpu = policy->cpu;
policy->cpu = cpu;
for_each_cpu(j, policy->cpus)
per_cpu(cpufreq_policy_cpu, j) = cpu;
#ifdef CONFIG_CPU_FREQ_TABLE
cpufreq_frequency_table_update_policy_cpu(policy);
#endif
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_UPDATE_POLICY_CPU, policy);
}
static int cpufreq_nominate_new_policy_cpu(struct cpufreq_policy *policy,
unsigned int old_cpu, bool frozen)
{
struct device *cpu_dev;
unsigned long flags;
int ret;
/* first sibling now owns the new sysfs dir */
cpu_dev = get_cpu_device(cpumask_first(policy->cpus));
/* Don't touch sysfs files during light-weight tear-down */
if (frozen)
return cpu_dev->id;
sysfs_remove_link(&cpu_dev->kobj, "cpufreq");
ret = kobject_move(&policy->kobj, &cpu_dev->kobj);
if (ret) {
pr_err("%s: Failed to move kobj: %d", __func__, ret);
WARN_ON(lock_policy_rwsem_write(old_cpu));
cpumask_set_cpu(old_cpu, policy->cpus);
write_lock_irqsave(&cpufreq_driver_lock, flags);
per_cpu(cpufreq_cpu_data, old_cpu) = policy;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
unlock_policy_rwsem_write(old_cpu);
ret = sysfs_create_link(&cpu_dev->kobj, &policy->kobj,
"cpufreq");
return -EINVAL;
}
return cpu_dev->id;
}
/**
* __cpufreq_remove_dev - remove a CPU device
*
* Removes the cpufreq interface for a CPU device.
* Caller should already have policy_rwsem in write mode for this CPU.
* This routine frees the rwsem before returning.
*/
static int __cpufreq_remove_dev(struct device *dev,
struct subsys_interface *sif, bool frozen)
{
unsigned int cpu = dev->id, cpus;
int new_cpu;
unsigned long flags;
struct cpufreq_policy *policy;
struct kobject *kobj;
struct completion *cmp;
pr_debug("%s: unregistering CPU %u\n", __func__, cpu);
write_lock_irqsave(&cpufreq_driver_lock, flags);
policy = per_cpu(cpufreq_cpu_data, cpu);
per_cpu(cpufreq_cpu_data, cpu) = NULL;
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
/* Save the policy somewhere when doing a light-weight tear-down */
if (frozen)
per_cpu(cpufreq_cpu_data_fallback, cpu) = policy;
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
if (!policy) {
pr_debug("%s: No cpu_data found\n", __func__);
return -EINVAL;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->target)
__cpufreq_governor(policy, CPUFREQ_GOV_STOP);
#ifdef CONFIG_HOTPLUG_CPU
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver->setpolicy)
strncpy(per_cpu(cpufreq_cpu_governor, cpu),
policy->governor->name, CPUFREQ_NAME_LEN);
#endif
WARN_ON(lock_policy_rwsem_write(cpu));
cpus = cpumask_weight(policy->cpus);
if (cpus > 1)
cpumask_clear_cpu(cpu, policy->cpus);
unlock_policy_rwsem_write(cpu);
if (cpu != policy->cpu && !frozen) {
sysfs_remove_link(&dev->kobj, "cpufreq");
} else if (cpus > 1) {
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 "
"from: %d\n",__func__, new_cpu, cpu);
}
}
}
/* If cpu is last user of policy, free policy */
if (cpus == 1) {
cpufreq: Fix cpufreq driver module refcount balance after suspend/resume Since cpufreq_cpu_put() called by __cpufreq_remove_dev() drops the driver module refcount, __cpufreq_remove_dev() causes that refcount to become negative for the cpufreq driver after a suspend/resume cycle. This is not the only bad thing that happens there, however, because kobject_put() should only be called for the policy kobject at this point if the CPU is not the last one for that policy. Namely, if the given CPU is the last one for that policy, the policy kobject's refcount should be 1 at this point, as set by cpufreq_add_dev_interface(), and only needs to be dropped once for the kobject to go away. This actually happens under the cpu == 1 check, so it need not be done before by cpufreq_cpu_put(). On the other hand, if the given CPU is not the last one for that policy, this means that cpufreq_add_policy_cpu() has been called at least once for that policy and cpufreq_cpu_get() has been called for it too. To balance that cpufreq_cpu_get(), we need to call cpufreq_cpu_put() in that case. Thus, to fix the described problem and keep the reference counters balanced in both cases, move the cpufreq_cpu_get() call in __cpufreq_remove_dev() to the code path executed only for CPUs that share the policy with other CPUs. Reported-and-tested-by: Toralf Förster <toralf.foerster@gmx.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Cc: 3.10+ <stable@vger.kernel.org>
2013-07-30 06:32:00 +08:00
if (cpufreq_driver->target)
__cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT);
cpufreq: Fix cpufreq driver module refcount balance after suspend/resume Since cpufreq_cpu_put() called by __cpufreq_remove_dev() drops the driver module refcount, __cpufreq_remove_dev() causes that refcount to become negative for the cpufreq driver after a suspend/resume cycle. This is not the only bad thing that happens there, however, because kobject_put() should only be called for the policy kobject at this point if the CPU is not the last one for that policy. Namely, if the given CPU is the last one for that policy, the policy kobject's refcount should be 1 at this point, as set by cpufreq_add_dev_interface(), and only needs to be dropped once for the kobject to go away. This actually happens under the cpu == 1 check, so it need not be done before by cpufreq_cpu_put(). On the other hand, if the given CPU is not the last one for that policy, this means that cpufreq_add_policy_cpu() has been called at least once for that policy and cpufreq_cpu_get() has been called for it too. To balance that cpufreq_cpu_get(), we need to call cpufreq_cpu_put() in that case. Thus, to fix the described problem and keep the reference counters balanced in both cases, move the cpufreq_cpu_get() call in __cpufreq_remove_dev() to the code path executed only for CPUs that share the policy with other CPUs. Reported-and-tested-by: Toralf Förster <toralf.foerster@gmx.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Cc: 3.10+ <stable@vger.kernel.org>
2013-07-30 06:32:00 +08:00
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
if (!frozen) {
lock_policy_rwsem_read(cpu);
kobj = &policy->kobj;
cmp = &policy->kobj_unregister;
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
unlock_policy_rwsem_read(cpu);
kobject_put(kobj);
/*
* We need to make sure that the underlying kobj is
* actually not referenced anymore by anybody before we
* proceed with unloading.
*/
pr_debug("waiting for dropping of refcount\n");
wait_for_completion(cmp);
pr_debug("wait complete\n");
}
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
/*
* Perform the ->exit() even during light-weight tear-down,
* since this is a core component, and is essential for the
* subsequent light-weight ->init() to succeed.
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->exit)
cpufreq_driver->exit(policy);
cpufreq: Preserve policy structure across suspend/resume To perform light-weight cpu-init and teardown in the cpufreq subsystem during suspend/resume, we need to separate out the 2 main functionalities of the cpufreq CPU hotplug callbacks, as outlined below: 1. Init/tear-down of core cpufreq and CPU-specific components, which are critical to the correct functioning of the cpufreq subsystem. 2. Init/tear-down of cpufreq sysfs files during suspend/resume. The first part requires accurate updates to the policy structure such as its ->cpus and ->related_cpus masks, whereas the second part requires that the policy->kobj structure is not released or re-initialized during suspend/resume. To handle both these requirements, we need to allow updates to the policy structure throughout suspend/resume, but prevent the structure from getting freed up. Also, we must have a mechanism by which the cpu-up callbacks can restore the policy structure, without allocating things afresh. (That also helps avoid memory leaks). To achieve this, we use 2 schemes: a. Use a fallback per-cpu storage area for preserving the policy structures during suspend, so that they can be restored during resume appropriately. b. Use the 'frozen' flag to determine when to free or allocate the policy structure vs when to restore the policy from the saved fallback storage. Thus we can successfully preserve the structure across suspend/resume. Effectively, this helps us complete the separation of the 'light-weight' and the 'full' init/tear-down sequences in the cpufreq subsystem, so that this can be made use of in the suspend/resume scenario. Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-07-30 06:55:10 +08:00
if (!frozen)
cpufreq_policy_free(policy);
cpufreq: Fix cpufreq driver module refcount balance after suspend/resume Since cpufreq_cpu_put() called by __cpufreq_remove_dev() drops the driver module refcount, __cpufreq_remove_dev() causes that refcount to become negative for the cpufreq driver after a suspend/resume cycle. This is not the only bad thing that happens there, however, because kobject_put() should only be called for the policy kobject at this point if the CPU is not the last one for that policy. Namely, if the given CPU is the last one for that policy, the policy kobject's refcount should be 1 at this point, as set by cpufreq_add_dev_interface(), and only needs to be dropped once for the kobject to go away. This actually happens under the cpu == 1 check, so it need not be done before by cpufreq_cpu_put(). On the other hand, if the given CPU is not the last one for that policy, this means that cpufreq_add_policy_cpu() has been called at least once for that policy and cpufreq_cpu_get() has been called for it too. To balance that cpufreq_cpu_get(), we need to call cpufreq_cpu_put() in that case. Thus, to fix the described problem and keep the reference counters balanced in both cases, move the cpufreq_cpu_get() call in __cpufreq_remove_dev() to the code path executed only for CPUs that share the policy with other CPUs. Reported-and-tested-by: Toralf Förster <toralf.foerster@gmx.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Cc: 3.10+ <stable@vger.kernel.org>
2013-07-30 06:32:00 +08:00
} else {
if (cpufreq_driver->target) {
__cpufreq_governor(policy, CPUFREQ_GOV_START);
__cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
cpufreq: Fix cpufreq driver module refcount balance after suspend/resume Since cpufreq_cpu_put() called by __cpufreq_remove_dev() drops the driver module refcount, __cpufreq_remove_dev() causes that refcount to become negative for the cpufreq driver after a suspend/resume cycle. This is not the only bad thing that happens there, however, because kobject_put() should only be called for the policy kobject at this point if the CPU is not the last one for that policy. Namely, if the given CPU is the last one for that policy, the policy kobject's refcount should be 1 at this point, as set by cpufreq_add_dev_interface(), and only needs to be dropped once for the kobject to go away. This actually happens under the cpu == 1 check, so it need not be done before by cpufreq_cpu_put(). On the other hand, if the given CPU is not the last one for that policy, this means that cpufreq_add_policy_cpu() has been called at least once for that policy and cpufreq_cpu_get() has been called for it too. To balance that cpufreq_cpu_get(), we need to call cpufreq_cpu_put() in that case. Thus, to fix the described problem and keep the reference counters balanced in both cases, move the cpufreq_cpu_get() call in __cpufreq_remove_dev() to the code path executed only for CPUs that share the policy with other CPUs. Reported-and-tested-by: Toralf Förster <toralf.foerster@gmx.de> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com> Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com> Cc: 3.10+ <stable@vger.kernel.org>
2013-07-30 06:32:00 +08:00
}
}
per_cpu(cpufreq_policy_cpu, cpu) = -1;
return 0;
}
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
static int cpufreq_remove_dev(struct device *dev, struct subsys_interface *sif)
{
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
unsigned int cpu = dev->id;
int retval;
if (cpu_is_offline(cpu))
return 0;
retval = __cpufreq_remove_dev(dev, sif, false);
return retval;
}
2006-11-22 22:55:48 +08:00
static void handle_update(struct work_struct *work)
{
2006-11-22 22:55:48 +08:00
struct cpufreq_policy *policy =
container_of(work, struct cpufreq_policy, update);
unsigned int cpu = policy->cpu;
pr_debug("handle_update for cpu %u called\n", cpu);
cpufreq_update_policy(cpu);
}
/**
* cpufreq_out_of_sync - If actual and saved CPU frequency differs, we're
* in deep trouble.
* @cpu: cpu number
* @old_freq: CPU frequency the kernel thinks the CPU runs at
* @new_freq: CPU frequency the CPU actually runs at
*
* We adjust to current frequency first, and need to clean up later.
* So either call to cpufreq_update_policy() or schedule handle_update()).
*/
static void cpufreq_out_of_sync(unsigned int cpu, unsigned int old_freq,
unsigned int new_freq)
{
struct cpufreq_policy *policy;
struct cpufreq_freqs freqs;
unsigned long flags;
pr_debug("Warning: CPU frequency out of sync: cpufreq and timing "
"core thinks of %u, is %u kHz.\n", old_freq, new_freq);
freqs.old = old_freq;
freqs.new = new_freq;
read_lock_irqsave(&cpufreq_driver_lock, flags);
policy = per_cpu(cpufreq_cpu_data, cpu);
read_unlock_irqrestore(&cpufreq_driver_lock, flags);
cpufreq_notify_transition(policy, &freqs, CPUFREQ_PRECHANGE);
cpufreq_notify_transition(policy, &freqs, CPUFREQ_POSTCHANGE);
}
/**
* cpufreq_quick_get - get the CPU frequency (in kHz) from policy->cur
* @cpu: CPU number
*
* This is the last known freq, without actually getting it from the driver.
* Return value will be same as what is shown in scaling_cur_freq in sysfs.
*/
unsigned int cpufreq_quick_get(unsigned int cpu)
{
struct cpufreq_policy *policy;
unsigned int ret_freq = 0;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver && cpufreq_driver->setpolicy && cpufreq_driver->get)
return cpufreq_driver->get(cpu);
policy = cpufreq_cpu_get(cpu);
if (policy) {
ret_freq = policy->cur;
cpufreq_cpu_put(policy);
}
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get);
/**
* cpufreq_quick_get_max - get the max reported CPU frequency for this CPU
* @cpu: CPU number
*
* Just return the max possible frequency for a given CPU.
*/
unsigned int cpufreq_quick_get_max(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
unsigned int ret_freq = 0;
if (policy) {
ret_freq = policy->max;
cpufreq_cpu_put(policy);
}
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_quick_get_max);
static unsigned int __cpufreq_get(unsigned int cpu)
{
struct cpufreq_policy *policy = per_cpu(cpufreq_cpu_data, cpu);
unsigned int ret_freq = 0;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver->get)
return ret_freq;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
ret_freq = cpufreq_driver->get(cpu);
if (ret_freq && policy->cur &&
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
!(cpufreq_driver->flags & CPUFREQ_CONST_LOOPS)) {
/* verify no discrepancy between actual and
saved value exists */
if (unlikely(ret_freq != policy->cur)) {
cpufreq_out_of_sync(cpu, policy->cur, ret_freq);
schedule_work(&policy->update);
}
}
return ret_freq;
}
/**
* cpufreq_get - get the current CPU frequency (in kHz)
* @cpu: CPU number
*
* Get the CPU current (static) CPU frequency
*/
unsigned int cpufreq_get(unsigned int cpu)
{
unsigned int ret_freq = 0;
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
if (!policy)
goto out;
if (unlikely(lock_policy_rwsem_read(cpu)))
goto out_policy;
ret_freq = __cpufreq_get(cpu);
unlock_policy_rwsem_read(cpu);
out_policy:
cpufreq_cpu_put(policy);
out:
return ret_freq;
}
EXPORT_SYMBOL(cpufreq_get);
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
static struct subsys_interface cpufreq_interface = {
.name = "cpufreq",
.subsys = &cpu_subsys,
.add_dev = cpufreq_add_dev,
.remove_dev = cpufreq_remove_dev,
};
/**
* cpufreq_bp_suspend - Prepare the boot CPU for system suspend.
*
* This function is only executed for the boot processor. The other CPUs
* have been put offline by means of CPU hotplug.
*/
static int cpufreq_bp_suspend(void)
{
int ret = 0;
int cpu = smp_processor_id();
struct cpufreq_policy *policy;
pr_debug("suspending cpu %u\n", cpu);
/* If there's no policy for the boot CPU, we have nothing to do. */
policy = cpufreq_cpu_get(cpu);
if (!policy)
return 0;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->suspend) {
ret = cpufreq_driver->suspend(policy);
[CPUFREQ] Re-enable cpufreq suspend and resume code Commit 4bc5d3413503 is broken and causes regressions: (1) cpufreq_driver->resume() and ->suspend() were only called on __powerpc__, but you could set them on all architectures. In fact, ->resume() was defined and used before the PPC-related commit 42d4dc3f4e1e complained about in 4bc5d3413503. (2) Therfore, the resume functions in acpi_cpufreq and speedstep-smi would never be called. (3) This means speedstep-smi would be unusuable after suspend or resume. The _real_ problem was calling cpufreq_driver->get() with interrupts off, but it re-enabling interrupts on some platforms. Why is ->get() necessary? Some systems like to change the CPU frequency behind our back, especially during BIOS-intensive operations like suspend or resume. If such systems also use a CPU frequency-dependant timing loop, delays might be off by large factors. Therefore, we need to ascertain as soon as possible that the CPU frequency is indeed at the speed we think it is. You can do this two ways: either setting it anew, or trying to get it. The latter is what was done, the former also has the same IRQ issue. So, let's try something different: defer the checking to after interrupts are re-enabled, by calling cpufreq_update_policy() (via schedule_work()). Timings may be off until this later stage, so let's watch out for resume regressions caused by the deferred handling of frequency changes behind the kernel's back. Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Dave Jones <davej@redhat.com>
2009-08-08 04:58:51 +08:00
if (ret)
printk(KERN_ERR "cpufreq: suspend failed in ->suspend "
"step on CPU %u\n", policy->cpu);
}
cpufreq_cpu_put(policy);
return ret;
}
/**
* cpufreq_bp_resume - Restore proper frequency handling of the boot CPU.
*
* 1.) resume CPUfreq hardware support (cpufreq_driver->resume())
[CPUFREQ] Re-enable cpufreq suspend and resume code Commit 4bc5d3413503 is broken and causes regressions: (1) cpufreq_driver->resume() and ->suspend() were only called on __powerpc__, but you could set them on all architectures. In fact, ->resume() was defined and used before the PPC-related commit 42d4dc3f4e1e complained about in 4bc5d3413503. (2) Therfore, the resume functions in acpi_cpufreq and speedstep-smi would never be called. (3) This means speedstep-smi would be unusuable after suspend or resume. The _real_ problem was calling cpufreq_driver->get() with interrupts off, but it re-enabling interrupts on some platforms. Why is ->get() necessary? Some systems like to change the CPU frequency behind our back, especially during BIOS-intensive operations like suspend or resume. If such systems also use a CPU frequency-dependant timing loop, delays might be off by large factors. Therefore, we need to ascertain as soon as possible that the CPU frequency is indeed at the speed we think it is. You can do this two ways: either setting it anew, or trying to get it. The latter is what was done, the former also has the same IRQ issue. So, let's try something different: defer the checking to after interrupts are re-enabled, by calling cpufreq_update_policy() (via schedule_work()). Timings may be off until this later stage, so let's watch out for resume regressions caused by the deferred handling of frequency changes behind the kernel's back. Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Dave Jones <davej@redhat.com>
2009-08-08 04:58:51 +08:00
* 2.) schedule call cpufreq_update_policy() ASAP as interrupts are
* restored. It will verify that the current freq is in sync with
* what we believe it to be. This is a bit later than when it
* should be, but nonethteless it's better than calling
* cpufreq_driver->get() here which might re-enable interrupts...
*
* This function is only executed for the boot CPU. The other CPUs have not
* been turned on yet.
*/
static void cpufreq_bp_resume(void)
{
int ret = 0;
int cpu = smp_processor_id();
struct cpufreq_policy *policy;
pr_debug("resuming cpu %u\n", cpu);
/* If there's no policy for the boot CPU, we have nothing to do. */
policy = cpufreq_cpu_get(cpu);
if (!policy)
return;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->resume) {
ret = cpufreq_driver->resume(policy);
if (ret) {
printk(KERN_ERR "cpufreq: resume failed in ->resume "
"step on CPU %u\n", policy->cpu);
goto fail;
}
}
schedule_work(&policy->update);
[CPUFREQ] Re-enable cpufreq suspend and resume code Commit 4bc5d3413503 is broken and causes regressions: (1) cpufreq_driver->resume() and ->suspend() were only called on __powerpc__, but you could set them on all architectures. In fact, ->resume() was defined and used before the PPC-related commit 42d4dc3f4e1e complained about in 4bc5d3413503. (2) Therfore, the resume functions in acpi_cpufreq and speedstep-smi would never be called. (3) This means speedstep-smi would be unusuable after suspend or resume. The _real_ problem was calling cpufreq_driver->get() with interrupts off, but it re-enabling interrupts on some platforms. Why is ->get() necessary? Some systems like to change the CPU frequency behind our back, especially during BIOS-intensive operations like suspend or resume. If such systems also use a CPU frequency-dependant timing loop, delays might be off by large factors. Therefore, we need to ascertain as soon as possible that the CPU frequency is indeed at the speed we think it is. You can do this two ways: either setting it anew, or trying to get it. The latter is what was done, the former also has the same IRQ issue. So, let's try something different: defer the checking to after interrupts are re-enabled, by calling cpufreq_update_policy() (via schedule_work()). Timings may be off until this later stage, so let's watch out for resume regressions caused by the deferred handling of frequency changes behind the kernel's back. Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net> Signed-off-by: Dave Jones <davej@redhat.com>
2009-08-08 04:58:51 +08:00
fail:
cpufreq_cpu_put(policy);
}
static struct syscore_ops cpufreq_syscore_ops = {
.suspend = cpufreq_bp_suspend,
.resume = cpufreq_bp_resume,
};
/**
* cpufreq_get_current_driver - return current driver's name
*
* Return the name string of the currently loaded cpufreq driver
* or NULL, if none.
*/
const char *cpufreq_get_current_driver(void)
{
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver)
return cpufreq_driver->name;
return NULL;
}
EXPORT_SYMBOL_GPL(cpufreq_get_current_driver);
/*********************************************************************
* NOTIFIER LISTS INTERFACE *
*********************************************************************/
/**
* cpufreq_register_notifier - register a driver with cpufreq
* @nb: notifier function to register
* @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
*
* Add a driver to one of two lists: either a list of drivers that
* are notified about clock rate changes (once before and once after
* the transition), or a list of drivers that are notified about
* changes in cpufreq policy.
*
* This function may sleep, and has the same return conditions as
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
* blocking_notifier_chain_register.
*/
int cpufreq_register_notifier(struct notifier_block *nb, unsigned int list)
{
int ret;
if (cpufreq_disabled())
return -EINVAL;
WARN_ON(!init_cpufreq_transition_notifier_list_called);
switch (list) {
case CPUFREQ_TRANSITION_NOTIFIER:
ret = srcu_notifier_chain_register(
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
&cpufreq_transition_notifier_list, nb);
break;
case CPUFREQ_POLICY_NOTIFIER:
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
ret = blocking_notifier_chain_register(
&cpufreq_policy_notifier_list, nb);
break;
default:
ret = -EINVAL;
}
return ret;
}
EXPORT_SYMBOL(cpufreq_register_notifier);
/**
* cpufreq_unregister_notifier - unregister a driver with cpufreq
* @nb: notifier block to be unregistered
* @list: CPUFREQ_TRANSITION_NOTIFIER or CPUFREQ_POLICY_NOTIFIER
*
* Remove a driver from the CPU frequency notifier list.
*
* This function may sleep, and has the same return conditions as
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
* blocking_notifier_chain_unregister.
*/
int cpufreq_unregister_notifier(struct notifier_block *nb, unsigned int list)
{
int ret;
if (cpufreq_disabled())
return -EINVAL;
switch (list) {
case CPUFREQ_TRANSITION_NOTIFIER:
ret = srcu_notifier_chain_unregister(
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
&cpufreq_transition_notifier_list, nb);
break;
case CPUFREQ_POLICY_NOTIFIER:
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
ret = blocking_notifier_chain_unregister(
&cpufreq_policy_notifier_list, nb);
break;
default:
ret = -EINVAL;
}
return ret;
}
EXPORT_SYMBOL(cpufreq_unregister_notifier);
/*********************************************************************
* GOVERNORS *
*********************************************************************/
int __cpufreq_driver_target(struct cpufreq_policy *policy,
unsigned int target_freq,
unsigned int relation)
{
int retval = -EINVAL;
unsigned int old_target_freq = target_freq;
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
if (cpufreq_disabled())
return -ENODEV;
if (policy->transition_ongoing)
return -EBUSY;
/* Make sure that target_freq is within supported range */
if (target_freq > policy->max)
target_freq = policy->max;
if (target_freq < policy->min)
target_freq = policy->min;
pr_debug("target for CPU %u: %u kHz, relation %u, requested %u kHz\n",
policy->cpu, target_freq, relation, old_target_freq);
if (target_freq == policy->cur)
return 0;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->target)
retval = cpufreq_driver->target(policy, target_freq, relation);
return retval;
}
EXPORT_SYMBOL_GPL(__cpufreq_driver_target);
int cpufreq_driver_target(struct cpufreq_policy *policy,
unsigned int target_freq,
unsigned int relation)
{
int ret = -EINVAL;
if (unlikely(lock_policy_rwsem_write(policy->cpu)))
goto fail;
ret = __cpufreq_driver_target(policy, target_freq, relation);
unlock_policy_rwsem_write(policy->cpu);
fail:
return ret;
}
EXPORT_SYMBOL_GPL(cpufreq_driver_target);
/*
* when "event" is CPUFREQ_GOV_LIMITS
*/
static int __cpufreq_governor(struct cpufreq_policy *policy,
unsigned int event)
{
int ret;
/* Only must be defined when default governor is known to have latency
restrictions, like e.g. conservative or ondemand.
That this is the case is already ensured in Kconfig
*/
#ifdef CONFIG_CPU_FREQ_GOV_PERFORMANCE
struct cpufreq_governor *gov = &cpufreq_gov_performance;
#else
struct cpufreq_governor *gov = NULL;
#endif
if (policy->governor->max_transition_latency &&
policy->cpuinfo.transition_latency >
policy->governor->max_transition_latency) {
if (!gov)
return -EINVAL;
else {
printk(KERN_WARNING "%s governor failed, too long"
" transition latency of HW, fallback"
" to %s governor\n",
policy->governor->name,
gov->name);
policy->governor = gov;
}
}
if (!try_module_get(policy->governor->owner))
return -EINVAL;
pr_debug("__cpufreq_governor for CPU %u, event %u\n",
policy->cpu, event);
cpufreq: Fix governor start/stop race condition Cpufreq governors' stop and start operations should be carried out in sequence. Otherwise, there will be unexpected behavior, like in the example below. Suppose there are 4 CPUs and policy->cpu=CPU0, CPU1/2/3 are linked to CPU0. The normal sequence is: 1) Current governor is userspace. An application tries to set the governor to ondemand. It will call __cpufreq_set_policy() in which it will stop the userspace governor and then start the ondemand governor. 2) Current governor is userspace. The online of CPU3 runs on CPU0. It will call cpufreq_add_policy_cpu() in which it will first stop the userspace governor, and then start it again. If the sequence of the above two cases interleaves, it becomes: 1) Application stops userspace governor 2) Hotplug stops userspace governor which is a problem, because the governor shouldn't be stopped twice in a row. What happens next is: 3) Application starts ondemand governor 4) Hotplug starts a governor In step 4, the hotplug is supposed to start the userspace governor, but now the governor has been changed by the application to ondemand, so the ondemand governor is started once again, which is incorrect. The solution is to prevent policy governors from being stopped multiple times in a row. A governor should only be stopped once for one policy. After it has been stopped, no more governor stop operations should be executed. Also add a mutex to serialize governor operations. [rjw: Changelog. And you owe me a beverage of my choice.] Signed-off-by: Xiaoguang Chen <chenxg@marvell.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-06-19 15:00:07 +08:00
mutex_lock(&cpufreq_governor_lock);
if ((!policy->governor_enabled && (event == CPUFREQ_GOV_STOP)) ||
(policy->governor_enabled && (event == CPUFREQ_GOV_START))) {
mutex_unlock(&cpufreq_governor_lock);
return -EBUSY;
}
if (event == CPUFREQ_GOV_STOP)
policy->governor_enabled = false;
else if (event == CPUFREQ_GOV_START)
policy->governor_enabled = true;
mutex_unlock(&cpufreq_governor_lock);
ret = policy->governor->governor(policy, event);
if (!ret) {
if (event == CPUFREQ_GOV_POLICY_INIT)
policy->governor->initialized++;
else if (event == CPUFREQ_GOV_POLICY_EXIT)
policy->governor->initialized--;
cpufreq: Fix governor start/stop race condition Cpufreq governors' stop and start operations should be carried out in sequence. Otherwise, there will be unexpected behavior, like in the example below. Suppose there are 4 CPUs and policy->cpu=CPU0, CPU1/2/3 are linked to CPU0. The normal sequence is: 1) Current governor is userspace. An application tries to set the governor to ondemand. It will call __cpufreq_set_policy() in which it will stop the userspace governor and then start the ondemand governor. 2) Current governor is userspace. The online of CPU3 runs on CPU0. It will call cpufreq_add_policy_cpu() in which it will first stop the userspace governor, and then start it again. If the sequence of the above two cases interleaves, it becomes: 1) Application stops userspace governor 2) Hotplug stops userspace governor which is a problem, because the governor shouldn't be stopped twice in a row. What happens next is: 3) Application starts ondemand governor 4) Hotplug starts a governor In step 4, the hotplug is supposed to start the userspace governor, but now the governor has been changed by the application to ondemand, so the ondemand governor is started once again, which is incorrect. The solution is to prevent policy governors from being stopped multiple times in a row. A governor should only be stopped once for one policy. After it has been stopped, no more governor stop operations should be executed. Also add a mutex to serialize governor operations. [rjw: Changelog. And you owe me a beverage of my choice.] Signed-off-by: Xiaoguang Chen <chenxg@marvell.com> Acked-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-06-19 15:00:07 +08:00
} else {
/* Restore original values */
mutex_lock(&cpufreq_governor_lock);
if (event == CPUFREQ_GOV_STOP)
policy->governor_enabled = true;
else if (event == CPUFREQ_GOV_START)
policy->governor_enabled = false;
mutex_unlock(&cpufreq_governor_lock);
}
/* we keep one module reference alive for
each CPU governed by this CPU */
if ((event != CPUFREQ_GOV_START) || ret)
module_put(policy->governor->owner);
if ((event == CPUFREQ_GOV_STOP) && !ret)
module_put(policy->governor->owner);
return ret;
}
int cpufreq_register_governor(struct cpufreq_governor *governor)
{
int err;
if (!governor)
return -EINVAL;
if (cpufreq_disabled())
return -ENODEV;
mutex_lock(&cpufreq_governor_mutex);
governor->initialized = 0;
err = -EBUSY;
if (__find_governor(governor->name) == NULL) {
err = 0;
list_add(&governor->governor_list, &cpufreq_governor_list);
}
mutex_unlock(&cpufreq_governor_mutex);
return err;
}
EXPORT_SYMBOL_GPL(cpufreq_register_governor);
void cpufreq_unregister_governor(struct cpufreq_governor *governor)
{
#ifdef CONFIG_HOTPLUG_CPU
int cpu;
#endif
if (!governor)
return;
if (cpufreq_disabled())
return;
#ifdef CONFIG_HOTPLUG_CPU
for_each_present_cpu(cpu) {
if (cpu_online(cpu))
continue;
if (!strcmp(per_cpu(cpufreq_cpu_governor, cpu), governor->name))
strcpy(per_cpu(cpufreq_cpu_governor, cpu), "\0");
}
#endif
mutex_lock(&cpufreq_governor_mutex);
list_del(&governor->governor_list);
mutex_unlock(&cpufreq_governor_mutex);
return;
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_governor);
/*********************************************************************
* POLICY INTERFACE *
*********************************************************************/
/**
* cpufreq_get_policy - get the current cpufreq_policy
* @policy: struct cpufreq_policy into which the current cpufreq_policy
* is written
*
* Reads the current cpufreq policy.
*/
int cpufreq_get_policy(struct cpufreq_policy *policy, unsigned int cpu)
{
struct cpufreq_policy *cpu_policy;
if (!policy)
return -EINVAL;
cpu_policy = cpufreq_cpu_get(cpu);
if (!cpu_policy)
return -EINVAL;
memcpy(policy, cpu_policy, sizeof(*policy));
cpufreq_cpu_put(cpu_policy);
return 0;
}
EXPORT_SYMBOL(cpufreq_get_policy);
/*
* data : current policy.
* policy : policy to be set.
*/
static int __cpufreq_set_policy(struct cpufreq_policy *policy,
struct cpufreq_policy *new_policy)
{
int ret = 0, failed = 1;
pr_debug("setting new policy for CPU %u: %u - %u kHz\n", new_policy->cpu,
new_policy->min, new_policy->max);
memcpy(&new_policy->cpuinfo, &policy->cpuinfo, sizeof(policy->cpuinfo));
if (new_policy->min > policy->max || new_policy->max < policy->min) {
ret = -EINVAL;
goto error_out;
}
/* verify the cpu speed can be set within this limit */
ret = cpufreq_driver->verify(new_policy);
if (ret)
goto error_out;
/* adjust if necessary - all reasons */
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_ADJUST, new_policy);
/* adjust if necessary - hardware incompatibility*/
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_INCOMPATIBLE, new_policy);
/*
* verify the cpu speed can be set within this limit, which might be
* different to the first one
*/
ret = cpufreq_driver->verify(new_policy);
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
if (ret)
goto error_out;
/* notification of the new policy */
[PATCH] Notifier chain update: API changes The kernel's implementation of notifier chains is unsafe. There is no protection against entries being added to or removed from a chain while the chain is in use. The issues were discussed in this thread: http://marc.theaimsgroup.com/?l=linux-kernel&m=113018709002036&w=2 We noticed that notifier chains in the kernel fall into two basic usage classes: "Blocking" chains are always called from a process context and the callout routines are allowed to sleep; "Atomic" chains can be called from an atomic context and the callout routines are not allowed to sleep. We decided to codify this distinction and make it part of the API. Therefore this set of patches introduces three new, parallel APIs: one for blocking notifiers, one for atomic notifiers, and one for "raw" notifiers (which is really just the old API under a new name). New kinds of data structures are used for the heads of the chains, and new routines are defined for registration, unregistration, and calling a chain. The three APIs are explained in include/linux/notifier.h and their implementation is in kernel/sys.c. With atomic and blocking chains, the implementation guarantees that the chain links will not be corrupted and that chain callers will not get messed up by entries being added or removed. For raw chains the implementation provides no guarantees at all; users of this API must provide their own protections. (The idea was that situations may come up where the assumptions of the atomic and blocking APIs are not appropriate, so it should be possible for users to handle these things in their own way.) There are some limitations, which should not be too hard to live with. For atomic/blocking chains, registration and unregistration must always be done in a process context since the chain is protected by a mutex/rwsem. Also, a callout routine for a non-raw chain must not try to register or unregister entries on its own chain. (This did happen in a couple of places and the code had to be changed to avoid it.) Since atomic chains may be called from within an NMI handler, they cannot use spinlocks for synchronization. Instead we use RCU. The overhead falls almost entirely in the unregister routine, which is okay since unregistration is much less frequent that calling a chain. Here is the list of chains that we adjusted and their classifications. None of them use the raw API, so for the moment it is only a placeholder. ATOMIC CHAINS ------------- arch/i386/kernel/traps.c: i386die_chain arch/ia64/kernel/traps.c: ia64die_chain arch/powerpc/kernel/traps.c: powerpc_die_chain arch/sparc64/kernel/traps.c: sparc64die_chain arch/x86_64/kernel/traps.c: die_chain drivers/char/ipmi/ipmi_si_intf.c: xaction_notifier_list kernel/panic.c: panic_notifier_list kernel/profile.c: task_free_notifier net/bluetooth/hci_core.c: hci_notifier net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_chain net/ipv4/netfilter/ip_conntrack_core.c: ip_conntrack_expect_chain net/ipv6/addrconf.c: inet6addr_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_chain net/netfilter/nf_conntrack_core.c: nf_conntrack_expect_chain net/netlink/af_netlink.c: netlink_chain BLOCKING CHAINS --------------- arch/powerpc/platforms/pseries/reconfig.c: pSeries_reconfig_chain arch/s390/kernel/process.c: idle_chain arch/x86_64/kernel/process.c idle_notifier drivers/base/memory.c: memory_chain drivers/cpufreq/cpufreq.c cpufreq_policy_notifier_list drivers/cpufreq/cpufreq.c cpufreq_transition_notifier_list drivers/macintosh/adb.c: adb_client_list drivers/macintosh/via-pmu.c sleep_notifier_list drivers/macintosh/via-pmu68k.c sleep_notifier_list drivers/macintosh/windfarm_core.c wf_client_list drivers/usb/core/notify.c usb_notifier_list drivers/video/fbmem.c fb_notifier_list kernel/cpu.c cpu_chain kernel/module.c module_notify_list kernel/profile.c munmap_notifier kernel/profile.c task_exit_notifier kernel/sys.c reboot_notifier_list net/core/dev.c netdev_chain net/decnet/dn_dev.c: dnaddr_chain net/ipv4/devinet.c: inetaddr_chain It's possible that some of these classifications are wrong. If they are, please let us know or submit a patch to fix them. Note that any chain that gets called very frequently should be atomic, because the rwsem read-locking used for blocking chains is very likely to incur cache misses on SMP systems. (However, if the chain's callout routines may sleep then the chain cannot be atomic.) The patch set was written by Alan Stern and Chandra Seetharaman, incorporating material written by Keith Owens and suggestions from Paul McKenney and Andrew Morton. [jes@sgi.com: restructure the notifier chain initialization macros] Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com> Signed-off-by: Jes Sorensen <jes@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-27 17:16:30 +08:00
blocking_notifier_call_chain(&cpufreq_policy_notifier_list,
CPUFREQ_NOTIFY, new_policy);
policy->min = new_policy->min;
policy->max = new_policy->max;
pr_debug("new min and max freqs are %u - %u kHz\n",
policy->min, policy->max);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->setpolicy) {
policy->policy = new_policy->policy;
pr_debug("setting range\n");
ret = cpufreq_driver->setpolicy(new_policy);
} else {
if (new_policy->governor != policy->governor) {
/* save old, working values */
struct cpufreq_governor *old_gov = policy->governor;
pr_debug("governor switch\n");
/* end old governor */
if (policy->governor) {
__cpufreq_governor(policy, CPUFREQ_GOV_STOP);
unlock_policy_rwsem_write(new_policy->cpu);
__cpufreq_governor(policy,
CPUFREQ_GOV_POLICY_EXIT);
lock_policy_rwsem_write(new_policy->cpu);
}
/* start new governor */
policy->governor = new_policy->governor;
if (!__cpufreq_governor(policy, CPUFREQ_GOV_POLICY_INIT)) {
if (!__cpufreq_governor(policy, CPUFREQ_GOV_START)) {
failed = 0;
cpufreq: Drop rwsem lock around CPUFREQ_GOV_POLICY_EXIT With the rwsem lock around __cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT), we get circular dependency when we call sysfs_remove_group(). ====================================================== [ INFO: possible circular locking dependency detected ] 3.9.0-rc7+ #15 Not tainted ------------------------------------------------------- cat/2387 is trying to acquire lock: (&per_cpu(cpu_policy_rwsem, cpu)){+++++.}, at: [<c02f6179>] lock_policy_rwsem_read+0x25/0x34 but task is already holding lock: (s_active#41){++++.+}, at: [<c00f9bf7>] sysfs_read_file+0x4f/0xcc which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (s_active#41){++++.+}: [<c0055a79>] lock_acquire+0x61/0xbc [<c00fabf1>] sysfs_addrm_finish+0xc1/0x128 [<c00f9819>] sysfs_hash_and_remove+0x35/0x64 [<c00fbe6f>] remove_files.isra.0+0x1b/0x24 [<c00fbea5>] sysfs_remove_group+0x2d/0xa8 [<c02f9a0b>] cpufreq_governor_interactive+0x13b/0x35c [<c02f61df>] __cpufreq_governor+0x2b/0x8c [<c02f6579>] __cpufreq_set_policy+0xa9/0xf8 [<c02f6b75>] store_scaling_governor+0x61/0x100 [<c02f6f4d>] store+0x39/0x60 [<c00f9b81>] sysfs_write_file+0xed/0x114 [<c00b3fd1>] vfs_write+0x65/0xd8 [<c00b424b>] sys_write+0x2f/0x50 [<c000cdc1>] ret_fast_syscall+0x1/0x52 -> #0 (&per_cpu(cpu_policy_rwsem, cpu)){+++++.}: [<c0055253>] __lock_acquire+0xef3/0x13dc [<c0055a79>] lock_acquire+0x61/0xbc [<c03ee1f5>] down_read+0x25/0x30 [<c02f6179>] lock_policy_rwsem_read+0x25/0x34 [<c02f6edd>] show+0x21/0x58 [<c00f9c0f>] sysfs_read_file+0x67/0xcc [<c00b40a7>] vfs_read+0x63/0xd8 [<c00b41fb>] sys_read+0x2f/0x50 [<c000cdc1>] ret_fast_syscall+0x1/0x52 other info that might help us debug this: Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(s_active#41); lock(&per_cpu(cpu_policy_rwsem, cpu)); lock(s_active#41); lock(&per_cpu(cpu_policy_rwsem, cpu)); *** DEADLOCK *** 2 locks held by cat/2387: #0: (&buffer->mutex){+.+.+.}, at: [<c00f9bcd>] sysfs_read_file+0x25/0xcc #1: (s_active#41){++++.+}, at: [<c00f9bf7>] sysfs_read_file+0x4f/0xcc stack backtrace: [<c0011d55>] (unwind_backtrace+0x1/0x9c) from [<c03e9a09>] (print_circular_bug+0x19d/0x1e8) [<c03e9a09>] (print_circular_bug+0x19d/0x1e8) from [<c0055253>] (__lock_acquire+0xef3/0x13dc) [<c0055253>] (__lock_acquire+0xef3/0x13dc) from [<c0055a79>] (lock_acquire+0x61/0xbc) [<c0055a79>] (lock_acquire+0x61/0xbc) from [<c03ee1f5>] (down_read+0x25/0x30) [<c03ee1f5>] (down_read+0x25/0x30) from [<c02f6179>] (lock_policy_rwsem_read+0x25/0x34) [<c02f6179>] (lock_policy_rwsem_read+0x25/0x34) from [<c02f6edd>] (show+0x21/0x58) [<c02f6edd>] (show+0x21/0x58) from [<c00f9c0f>] (sysfs_read_file+0x67/0xcc) [<c00f9c0f>] (sysfs_read_file+0x67/0xcc) from [<c00b40a7>] (vfs_read+0x63/0xd8) [<c00b40a7>] (vfs_read+0x63/0xd8) from [<c00b41fb>] (sys_read+0x2f/0x50) [<c00b41fb>] (sys_read+0x2f/0x50) from [<c000cdc1>] (ret_fast_syscall+0x1/0x52) This lock isn't required while calling __cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT). Remove it. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-05-16 13:09:58 +08:00
} else {
unlock_policy_rwsem_write(new_policy->cpu);
__cpufreq_governor(policy,
CPUFREQ_GOV_POLICY_EXIT);
lock_policy_rwsem_write(new_policy->cpu);
cpufreq: Drop rwsem lock around CPUFREQ_GOV_POLICY_EXIT With the rwsem lock around __cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT), we get circular dependency when we call sysfs_remove_group(). ====================================================== [ INFO: possible circular locking dependency detected ] 3.9.0-rc7+ #15 Not tainted ------------------------------------------------------- cat/2387 is trying to acquire lock: (&per_cpu(cpu_policy_rwsem, cpu)){+++++.}, at: [<c02f6179>] lock_policy_rwsem_read+0x25/0x34 but task is already holding lock: (s_active#41){++++.+}, at: [<c00f9bf7>] sysfs_read_file+0x4f/0xcc which lock already depends on the new lock. the existing dependency chain (in reverse order) is: -> #1 (s_active#41){++++.+}: [<c0055a79>] lock_acquire+0x61/0xbc [<c00fabf1>] sysfs_addrm_finish+0xc1/0x128 [<c00f9819>] sysfs_hash_and_remove+0x35/0x64 [<c00fbe6f>] remove_files.isra.0+0x1b/0x24 [<c00fbea5>] sysfs_remove_group+0x2d/0xa8 [<c02f9a0b>] cpufreq_governor_interactive+0x13b/0x35c [<c02f61df>] __cpufreq_governor+0x2b/0x8c [<c02f6579>] __cpufreq_set_policy+0xa9/0xf8 [<c02f6b75>] store_scaling_governor+0x61/0x100 [<c02f6f4d>] store+0x39/0x60 [<c00f9b81>] sysfs_write_file+0xed/0x114 [<c00b3fd1>] vfs_write+0x65/0xd8 [<c00b424b>] sys_write+0x2f/0x50 [<c000cdc1>] ret_fast_syscall+0x1/0x52 -> #0 (&per_cpu(cpu_policy_rwsem, cpu)){+++++.}: [<c0055253>] __lock_acquire+0xef3/0x13dc [<c0055a79>] lock_acquire+0x61/0xbc [<c03ee1f5>] down_read+0x25/0x30 [<c02f6179>] lock_policy_rwsem_read+0x25/0x34 [<c02f6edd>] show+0x21/0x58 [<c00f9c0f>] sysfs_read_file+0x67/0xcc [<c00b40a7>] vfs_read+0x63/0xd8 [<c00b41fb>] sys_read+0x2f/0x50 [<c000cdc1>] ret_fast_syscall+0x1/0x52 other info that might help us debug this: Possible unsafe locking scenario: CPU0 CPU1 ---- ---- lock(s_active#41); lock(&per_cpu(cpu_policy_rwsem, cpu)); lock(s_active#41); lock(&per_cpu(cpu_policy_rwsem, cpu)); *** DEADLOCK *** 2 locks held by cat/2387: #0: (&buffer->mutex){+.+.+.}, at: [<c00f9bcd>] sysfs_read_file+0x25/0xcc #1: (s_active#41){++++.+}, at: [<c00f9bf7>] sysfs_read_file+0x4f/0xcc stack backtrace: [<c0011d55>] (unwind_backtrace+0x1/0x9c) from [<c03e9a09>] (print_circular_bug+0x19d/0x1e8) [<c03e9a09>] (print_circular_bug+0x19d/0x1e8) from [<c0055253>] (__lock_acquire+0xef3/0x13dc) [<c0055253>] (__lock_acquire+0xef3/0x13dc) from [<c0055a79>] (lock_acquire+0x61/0xbc) [<c0055a79>] (lock_acquire+0x61/0xbc) from [<c03ee1f5>] (down_read+0x25/0x30) [<c03ee1f5>] (down_read+0x25/0x30) from [<c02f6179>] (lock_policy_rwsem_read+0x25/0x34) [<c02f6179>] (lock_policy_rwsem_read+0x25/0x34) from [<c02f6edd>] (show+0x21/0x58) [<c02f6edd>] (show+0x21/0x58) from [<c00f9c0f>] (sysfs_read_file+0x67/0xcc) [<c00f9c0f>] (sysfs_read_file+0x67/0xcc) from [<c00b40a7>] (vfs_read+0x63/0xd8) [<c00b40a7>] (vfs_read+0x63/0xd8) from [<c00b41fb>] (sys_read+0x2f/0x50) [<c00b41fb>] (sys_read+0x2f/0x50) from [<c000cdc1>] (ret_fast_syscall+0x1/0x52) This lock isn't required while calling __cpufreq_governor(policy, CPUFREQ_GOV_POLICY_EXIT). Remove it. Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-05-16 13:09:58 +08:00
}
}
if (failed) {
/* new governor failed, so re-start old one */
pr_debug("starting governor %s failed\n",
policy->governor->name);
if (old_gov) {
policy->governor = old_gov;
__cpufreq_governor(policy,
CPUFREQ_GOV_POLICY_INIT);
__cpufreq_governor(policy,
CPUFREQ_GOV_START);
}
ret = -EINVAL;
goto error_out;
}
/* might be a policy change, too, so fall through */
}
pr_debug("governor: change or update limits\n");
__cpufreq_governor(policy, CPUFREQ_GOV_LIMITS);
}
error_out:
return ret;
}
/**
* cpufreq_update_policy - re-evaluate an existing cpufreq policy
* @cpu: CPU which shall be re-evaluated
*
* Useful for policy notifiers which have different necessities
* at different times.
*/
int cpufreq_update_policy(unsigned int cpu)
{
struct cpufreq_policy *policy = cpufreq_cpu_get(cpu);
struct cpufreq_policy new_policy;
int ret;
if (!policy) {
ret = -ENODEV;
goto no_policy;
}
if (unlikely(lock_policy_rwsem_write(cpu))) {
ret = -EINVAL;
goto fail;
}
pr_debug("updating policy for CPU %u\n", cpu);
memcpy(&new_policy, policy, sizeof(*policy));
new_policy.min = policy->user_policy.min;
new_policy.max = policy->user_policy.max;
new_policy.policy = policy->user_policy.policy;
new_policy.governor = policy->user_policy.governor;
/*
* BIOS might change freq behind our back
* -> ask driver for current freq and notify governors about a change
*/
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver->get) {
new_policy.cur = cpufreq_driver->get(cpu);
if (!policy->cur) {
pr_debug("Driver did not initialize current freq");
policy->cur = new_policy.cur;
} else {
if (policy->cur != new_policy.cur && cpufreq_driver->target)
cpufreq_out_of_sync(cpu, policy->cur,
new_policy.cur);
}
}
ret = __cpufreq_set_policy(policy, &new_policy);
unlock_policy_rwsem_write(cpu);
fail:
cpufreq_cpu_put(policy);
no_policy:
return ret;
}
EXPORT_SYMBOL(cpufreq_update_policy);
static int cpufreq_cpu_callback(struct notifier_block *nfb,
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
unsigned long action, void *hcpu)
{
unsigned int cpu = (unsigned long)hcpu;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
struct device *dev;
bool frozen = false;
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
dev = get_cpu_device(cpu);
if (dev) {
if (action & CPU_TASKS_FROZEN)
frozen = true;
switch (action & ~CPU_TASKS_FROZEN) {
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
case CPU_ONLINE:
__cpufreq_add_dev(dev, NULL, frozen);
cpufreq_update_policy(cpu);
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
break;
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
case CPU_DOWN_PREPARE:
__cpufreq_remove_dev(dev, NULL, frozen);
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
break;
case CPU_DOWN_FAILED:
__cpufreq_add_dev(dev, NULL, frozen);
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
break;
}
}
return NOTIFY_OK;
}
static struct notifier_block __refdata cpufreq_cpu_notifier = {
.notifier_call = cpufreq_cpu_callback,
[PATCH] create and destroy cpufreq sysfs entries based on cpu notifiers cpufreq entries in sysfs should only be populated when CPU is online state. When we either boot with maxcpus=x and then boot the other cpus by echoing to sysfs online file, these entries should be created and destroyed when CPU_DEAD is notified. Same treatement as cache entries under sysfs. We place the processor in the lowest frequency, so hw managed P-State transitions can still work on the other threads to save power. Primary goal was to just make these directories appear/disapper dynamically. There is one in this patch i had to do, which i really dont like myself but probably best if someone handling the cpufreq infrastructure could give this code right treatment if this is not acceptable. I guess its probably good for the first cut. - Converting lock_cpu_hotplug()/unlock_cpu_hotplug() to disable/enable preempt. The locking was smack in the middle of the notification path, when the hotplug is already holding the lock. I tried another solution to avoid this so avoid taking locks if we know we are from notification path. The solution was getting very ugly and i decided this was probably good for this iteration until someone who understands cpufreq could do a better job than me. (akpm: export cpucontrol to GPL modules: drivers/cpufreq/cpufreq_stats.c now does lock_cpu_hotplug()) Signed-off-by: Ashok Raj <ashok.raj@intel.com> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Cc: Dave Jones <davej@codemonkey.org.uk> Cc: Zwane Mwaikambo <zwane@holomorphy.com> Cc: Greg KH <greg@kroah.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-31 06:59:54 +08:00
};
/*********************************************************************
* REGISTER / UNREGISTER CPUFREQ DRIVER *
*********************************************************************/
/**
* cpufreq_register_driver - register a CPU Frequency driver
* @driver_data: A struct cpufreq_driver containing the values#
* submitted by the CPU Frequency driver.
*
* Registers a CPU Frequency driver to this core code. This code
* returns zero on success, -EBUSY when another driver got here first
* (and isn't unregistered in the meantime).
*
*/
int cpufreq_register_driver(struct cpufreq_driver *driver_data)
{
unsigned long flags;
int ret;
if (cpufreq_disabled())
return -ENODEV;
if (!driver_data || !driver_data->verify || !driver_data->init ||
((!driver_data->setpolicy) && (!driver_data->target)))
return -EINVAL;
pr_debug("trying to register driver %s\n", driver_data->name);
if (driver_data->setpolicy)
driver_data->flags |= CPUFREQ_CONST_LOOPS;
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (cpufreq_driver) {
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
return -EBUSY;
}
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
cpufreq_driver = driver_data;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
ret = subsys_interface_register(&cpufreq_interface);
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-02 00:41:10 +08:00
if (ret)
goto err_null_driver;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!(cpufreq_driver->flags & CPUFREQ_STICKY)) {
int i;
ret = -ENODEV;
/* check for at least one working CPU */
for (i = 0; i < nr_cpu_ids; i++)
if (cpu_possible(i) && per_cpu(cpufreq_cpu_data, i)) {
ret = 0;
break;
}
/* if all ->init() calls failed, unregister */
if (ret) {
pr_debug("no CPU initialized for driver %s\n",
driver_data->name);
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
goto err_if_unreg;
}
}
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-02 00:41:10 +08:00
register_hotcpu_notifier(&cpufreq_cpu_notifier);
pr_debug("driver %s up and running\n", driver_data->name);
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-02 00:41:10 +08:00
return 0;
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
err_if_unreg:
subsys_interface_unregister(&cpufreq_interface);
[CPUFREQ] fix BUG on cpufreq policy init failure cpufreq_register_driver sets cpufreq_driver to a structure owned (and placed) in the caller's memory. If cpufreq policy fails in its ->init function, sysdev_driver_register returns nonzero in cpufreq_register_driver. Now, cpufreq_register_driver returns an error without setting cpufreq_driver back to NULL. Usually cpufreq policy modules are unloaded because they propagate the error to the module init function and return that. So a later access to any member of cpufreq_driver causes bugs like: BUG: unable to handle kernel paging request at ffffffffa00270a0 IP: [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 PGD 1805067 PUD 1809063 PMD 1c3f90067 PTE 0 Oops: 0000 [#1] SMP last sysfs file: /sys/devices/virtual/net/tun0/statistics/collisions CPU 0 Modules linked in: ... Pid: 5677, comm: thunderbird-bin Tainted: G W 2.6.38-rc4-mm1_64+ #1389 To be filled by O.E.M./To Be Filled By O.E.M. RIP: 0010:[<ffffffff8145eca3>] [<ffffffff8145eca3>] cpufreq_cpu_get+0x53/0xe0 RSP: 0018:ffff8801aec37d98 EFLAGS: 00010086 RAX: 0000000000000202 RBX: 0000000000000000 RCX: 0000000000000001 RDX: ffffffffa00270a0 RSI: 0000000000001000 RDI: ffffffff8199ece8 ... Call Trace: [<ffffffff8145f490>] cpufreq_quick_get+0x10/0x30 [<ffffffff8103f12b>] show_cpuinfo+0x2ab/0x300 [<ffffffff81136292>] seq_read+0xf2/0x3f0 [<ffffffff8126c5d3>] ? __strncpy_from_user+0x33/0x60 [<ffffffff8116850d>] proc_reg_read+0x6d/0xa0 [<ffffffff81116e53>] vfs_read+0xc3/0x180 [<ffffffff81116f5c>] sys_read+0x4c/0x90 [<ffffffff81030dbb>] system_call_fastpath+0x16/0x1b ... It's all cause by weird fail path handling in cpufreq_register_driver. To fix that, shuffle the code to do proper handling with gotos. Signed-off-by: Jiri Slaby <jslaby@suse.cz> Signed-off-by: Dave Jones <davej@redhat.com>
2011-03-02 00:41:10 +08:00
err_null_driver:
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
cpufreq_driver = NULL;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
return ret;
}
EXPORT_SYMBOL_GPL(cpufreq_register_driver);
/**
* cpufreq_unregister_driver - unregister the current CPUFreq driver
*
* Unregister the current CPUFreq driver. Only call this if you have
* the right to do so, i.e. if you have succeeded in initialising before!
* Returns zero if successful, and -EINVAL if the cpufreq_driver is
* currently not initialised.
*/
int cpufreq_unregister_driver(struct cpufreq_driver *driver)
{
unsigned long flags;
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
if (!cpufreq_driver || (driver != cpufreq_driver))
return -EINVAL;
pr_debug("unregistering driver %s\n", driver->name);
cpu: convert 'cpu' and 'machinecheck' sysdev_class to a regular subsystem This moves the 'cpu sysdev_class' over to a regular 'cpu' subsystem and converts the devices to regular devices. The sysdev drivers are implemented as subsystem interfaces now. After all sysdev classes are ported to regular driver core entities, the sysdev implementation will be entirely removed from the kernel. Userspace relies on events and generic sysfs subsystem infrastructure from sysdev devices, which are made available with this conversion. Cc: Haavard Skinnemoen <hskinnemoen@gmail.com> Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no> Cc: Tony Luck <tony.luck@intel.com> Cc: Fenghua Yu <fenghua.yu@intel.com> Cc: Arnd Bergmann <arnd@arndb.de> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Cc: Chris Metcalf <cmetcalf@tilera.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Borislav Petkov <bp@amd64.org> Cc: Tigran Aivazian <tigran@aivazian.fsnet.co.uk> Cc: Len Brown <lenb@kernel.org> Cc: Zhang Rui <rui.zhang@intel.com> Cc: Dave Jones <davej@redhat.com> Cc: Peter Zijlstra <peterz@infradead.org> Cc: Russell King <rmk+kernel@arm.linux.org.uk> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Arjan van de Ven <arjan@linux.intel.com> Cc: "Rafael J. Wysocki" <rjw@sisk.pl> Cc: "Srivatsa S. Bhat" <srivatsa.bhat@linux.vnet.ibm.com> Signed-off-by: Kay Sievers <kay.sievers@vrfy.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
2011-12-22 06:29:42 +08:00
subsys_interface_unregister(&cpufreq_interface);
unregister_hotcpu_notifier(&cpufreq_cpu_notifier);
write_lock_irqsave(&cpufreq_driver_lock, flags);
cpufreq: Revert incorrect commit 5800043 Commit 5800043 (cpufreq: convert cpufreq_driver to using RCU) causes the following call trace to be spit on boot: BUG: sleeping function called from invalid context at /scratch/rafael/work/linux-pm/mm/slab.c:3179 in_atomic(): 0, irqs_disabled(): 0, pid: 292, name: systemd-udevd 2 locks held by systemd-udevd/292: #0: (subsys mutex){+.+.+.}, at: [<ffffffff8146851a>] subsys_interface_register+0x4a/0xe0 #1: (rcu_read_lock){.+.+.+}, at: [<ffffffff81538210>] cpufreq_add_dev_interface+0x60/0x5e0 Pid: 292, comm: systemd-udevd Not tainted 3.9.0-rc8+ #323 Call Trace: [<ffffffff81072c90>] __might_sleep+0x140/0x1f0 [<ffffffff811581c2>] kmem_cache_alloc+0x42/0x2b0 [<ffffffff811e7179>] sysfs_new_dirent+0x59/0x130 [<ffffffff811e63cb>] sysfs_add_file_mode+0x6b/0x110 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff811e647d>] sysfs_add_file+0xd/0x10 [<ffffffff811e6541>] sysfs_create_file+0x21/0x30 [<ffffffff81538280>] cpufreq_add_dev_interface+0xd0/0x5e0 [<ffffffff81538210>] ? cpufreq_add_dev_interface+0x60/0x5e0 [<ffffffffa000337f>] ? acpi_processor_get_platform_limit+0x32/0xbb [processor] [<ffffffffa022f540>] ? do_drv_write+0x70/0x70 [acpi_cpufreq] [<ffffffff810a3254>] ? __lock_is_held+0x54/0x80 [<ffffffff8106c97e>] ? up_read+0x1e/0x40 [<ffffffff8106e632>] ? __blocking_notifier_call_chain+0x72/0xc0 [<ffffffff81538dbd>] cpufreq_add_dev+0x62d/0xae0 [<ffffffff815389b8>] ? cpufreq_add_dev+0x228/0xae0 [<ffffffff81468569>] subsys_interface_register+0x99/0xe0 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffff81535d5d>] cpufreq_register_driver+0x9d/0x200 [<ffffffffa014d000>] ? 0xffffffffa014cfff [<ffffffffa014d0e9>] acpi_cpufreq_init+0xe9/0x1000 [acpi_cpufreq] [<ffffffff810002fa>] do_one_initcall+0x11a/0x170 [<ffffffff810b4b87>] load_module+0x1cf7/0x2920 [<ffffffff81322580>] ? ddebug_proc_open+0xb0/0xb0 [<ffffffff816baee0>] ? retint_restore_args+0xe/0xe [<ffffffff810b5887>] sys_init_module+0xd7/0x120 [<ffffffff816bb6d2>] system_call_fastpath+0x16/0x1b which is quite obvious, because that commit put (multiple instances of) sysfs_create_file() under rcu_read_lock()/rcu_read_unlock(), although sysfs_create_file() may cause memory to be allocated with GFP_KERNEL and that may sleep, which is not permitted in RCU read critical section. Revert the buggy commit altogether along with some changes on top of it. Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
2013-04-29 06:08:16 +08:00
cpufreq_driver = NULL;
write_unlock_irqrestore(&cpufreq_driver_lock, flags);
return 0;
}
EXPORT_SYMBOL_GPL(cpufreq_unregister_driver);
static int __init cpufreq_core_init(void)
{
int cpu;
if (cpufreq_disabled())
return -ENODEV;
for_each_possible_cpu(cpu) {
per_cpu(cpufreq_policy_cpu, cpu) = -1;
init_rwsem(&per_cpu(cpu_policy_rwsem, cpu));
}
cpufreq_global_kobject = kobject_create();
BUG_ON(!cpufreq_global_kobject);
register_syscore_ops(&cpufreq_syscore_ops);
return 0;
}
core_initcall(cpufreq_core_init);