From e40b17208b6805be50ffe891878662b6076206b9 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 5 Feb 2010 21:47:03 -0500 Subject: x86: Move notify_die from nmi.c to traps.c In order to handle a new nmi_watchdog approach, I need to move the notify_die() routine out of nmi_watchdog_tick() and into default_do_nmi(). This lets me easily swap out the old nmi_watchdog with the new one with just a config change. The change probably makes sense from a high level perspective because the nmi_watchdog shouldn't be handling notify_die routines anyway. However, this move does change the semantics a little bit. Instead of checking on every nmi interrupt if the cpus are stuck, only check them on the nmi_watchdog interrupts. v2: Move notify_die call into #idef block Signed-off-by: Don Zickus Cc: Linus Torvalds Cc: Andrew Morton Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: peterz@infradead.org LKML-Reference: <1265424425-31562-2-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/kernel/apic/nmi.c b/arch/x86/kernel/apic/nmi.c index 0159a69..5d47682 100644 --- a/arch/x86/kernel/apic/nmi.c +++ b/arch/x86/kernel/apic/nmi.c @@ -400,13 +400,6 @@ nmi_watchdog_tick(struct pt_regs *regs, unsigned reason) int cpu = smp_processor_id(); int rc = 0; - /* check for other users first */ - if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) - == NOTIFY_STOP) { - rc = 1; - touched = 1; - } - sum = get_timer_irqs(cpu); if (__get_cpu_var(nmi_touch)) { diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index 1168e44..51ef893f 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -400,7 +400,12 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT) == NOTIFY_STOP) return; + #ifdef CONFIG_X86_LOCAL_APIC + if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) + == NOTIFY_STOP) + return; + /* * Ok, so this is none of the documented NMI sources, * so it must be the NMI watchdog. -- cgit v0.10.2 From 1fb9d6ad2766a1dd70d167552988375049a97f21 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 5 Feb 2010 21:47:04 -0500 Subject: nmi_watchdog: Add new, generic implementation, using perf events This is a new generic nmi_watchdog implementation using the perf events infrastructure as suggested by Ingo. The implementation is simple, just create an in-kernel perf event and register an overflow handler to check for cpu lockups. I created a generic implementation that lives in kernel/ and the hardware specific part that for now lives in arch/x86. This approach has a number of advantages: - It simplifies the x86 PMU implementation in the long run, in that it removes the hardcoded low-level PMU implementation that was the NMI watchdog before. - It allows new NMI watchdog features to be added in a central place. - It allows other architectures to enable the NMI watchdog, as long as they have perf events (that provide NMIs) implemented. - It also allows for more graceful co-existence of existing perf events apps and the NMI watchdog - before these changes the relationship was exclusive. (The NMI watchdog will 'spend' a perf event when enabled. In later iterations we might be able to piggyback from an existing NMI event without having to allocate a hardware event for the NMI watchdog - turning this into a no-hardware-cost feature.) As for compatibility, we'll keep the old NMI watchdog code as well until the new one can 100% replace it on all CPUs, old and new alike. That might take some time as the NMI watchdog has been ported to many CPU models. I have done light testing to make sure the framework works correctly and it does. v2: Set the correct timeout values based on the old nmi watchdog Signed-off-by: Don Zickus Cc: Linus Torvalds Cc: Andrew Morton Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: peterz@infradead.org LKML-Reference: <1265424425-31562-3-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c new file mode 100644 index 0000000..8c0e6a4 --- /dev/null +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -0,0 +1,114 @@ +/* + * HW NMI watchdog support + * + * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc. + * + * Arch specific calls to support NMI watchdog + * + * Bits copied from original nmi.c file + * + */ + +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +/* For reliability, we're prepared to waste bits here. */ +static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; + +static DEFINE_PER_CPU(unsigned, last_irq_sum); + +/* + * Take the local apic timer and PIT/HPET into account. We don't + * know which one is active, when we have highres/dyntick on + */ +static inline unsigned int get_timer_irqs(int cpu) +{ + return per_cpu(irq_stat, cpu).apic_timer_irqs + + per_cpu(irq_stat, cpu).irq0_irqs; +} + +static inline int mce_in_progress(void) +{ +#if defined(CONFIG_X86_MCE) + return atomic_read(&mce_entry) > 0; +#endif + return 0; +} + +int hw_nmi_is_cpu_stuck(struct pt_regs *regs) +{ + unsigned int sum; + int cpu = smp_processor_id(); + + /* FIXME: cheap hack for this check, probably should get its own + * die_notifier handler + */ + if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { + static DEFINE_SPINLOCK(lock); /* Serialise the printks */ + + spin_lock(&lock); + printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); + show_regs(regs); + dump_stack(); + spin_unlock(&lock); + cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); + } + + /* if we are doing an mce, just assume the cpu is not stuck */ + /* Could check oops_in_progress here too, but it's safer not to */ + if (mce_in_progress()) + return 0; + + /* We determine if the cpu is stuck by checking whether any + * interrupts have happened since we last checked. Of course + * an nmi storm could create false positives, but the higher + * level logic should account for that + */ + sum = get_timer_irqs(cpu); + if (__get_cpu_var(last_irq_sum) == sum) { + return 1; + } else { + __get_cpu_var(last_irq_sum) = sum; + return 0; + } +} + +void arch_trigger_all_cpu_backtrace(void) +{ + int i; + + cpumask_copy(to_cpumask(backtrace_mask), cpu_online_mask); + + printk(KERN_INFO "sending NMI to all CPUs:\n"); + apic->send_IPI_all(NMI_VECTOR); + + /* Wait for up to 10 seconds for all CPUs to do the backtrace */ + for (i = 0; i < 10 * 1000; i++) { + if (cpumask_empty(to_cpumask(backtrace_mask))) + break; + mdelay(1); + } +} + +/* STUB calls to mimic old nmi_watchdog behaviour */ +unsigned int nmi_watchdog = NMI_NONE; +EXPORT_SYMBOL(nmi_watchdog); +atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ +EXPORT_SYMBOL(nmi_active); +int nmi_watchdog_enabled; +int unknown_nmi_panic; +void cpu_nmi_set_wd_enabled(void) { return; } +void acpi_nmi_enable(void) { return; } +void acpi_nmi_disable(void) { return; } +void stop_apic_nmi_watchdog(void *unused) { return; } +void setup_apic_nmi_watchdog(void *unused) { return; } +int __init check_nmi_watchdog(void) { return 0; } diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c new file mode 100644 index 0000000..36817b2 --- /dev/null +++ b/kernel/nmi_watchdog.c @@ -0,0 +1,191 @@ +/* + * Detect Hard Lockups using the NMI + * + * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc. + * + * this code detects hard lockups: incidents in where on a CPU + * the kernel does not respond to anything except NMI. + * + * Note: Most of this code is borrowed heavily from softlockup.c, + * so thanks to Ingo for the initial implementation. + * Some chunks also taken from arch/x86/kernel/apic/nmi.c, thanks + * to those contributors as well. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +static DEFINE_PER_CPU(struct perf_event *, nmi_watchdog_ev); +static DEFINE_PER_CPU(int, nmi_watchdog_touch); +static DEFINE_PER_CPU(long, alert_counter); + +void touch_nmi_watchdog(void) +{ + __raw_get_cpu_var(nmi_watchdog_touch) = 1; + touch_softlockup_watchdog(); +} +EXPORT_SYMBOL(touch_nmi_watchdog); + +void touch_all_nmi_watchdog(void) +{ + int cpu; + + for_each_online_cpu(cpu) + per_cpu(nmi_watchdog_touch, cpu) = 1; + touch_softlockup_watchdog(); +} + +#ifdef CONFIG_SYSCTL +/* + * proc handler for /proc/sys/kernel/nmi_watchdog + */ +int proc_nmi_enabled(struct ctl_table *table, int write, + void __user *buffer, size_t *length, loff_t *ppos) +{ + int cpu; + + if (per_cpu(nmi_watchdog_ev, smp_processor_id()) == NULL) + nmi_watchdog_enabled = 0; + else + nmi_watchdog_enabled = 1; + + touch_all_nmi_watchdog(); + proc_dointvec(table, write, buffer, length, ppos); + if (nmi_watchdog_enabled) + for_each_online_cpu(cpu) + perf_event_enable(per_cpu(nmi_watchdog_ev, cpu)); + else + for_each_online_cpu(cpu) + perf_event_disable(per_cpu(nmi_watchdog_ev, cpu)); + return 0; +} + +#endif /* CONFIG_SYSCTL */ + +struct perf_event_attr wd_attr = { + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, +}; + +static int panic_on_timeout; + +void wd_overflow(struct perf_event *event, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) +{ + int cpu = smp_processor_id(); + int touched = 0; + + if (__get_cpu_var(nmi_watchdog_touch)) { + per_cpu(nmi_watchdog_touch, cpu) = 0; + touched = 1; + } + + /* check to see if the cpu is doing anything */ + if (!touched && hw_nmi_is_cpu_stuck(regs)) { + /* + * Ayiee, looks like this CPU is stuck ... + * wait a few IRQs (5 seconds) before doing the oops ... + */ + per_cpu(alert_counter,cpu) += 1; + if (per_cpu(alert_counter,cpu) == 5) { + /* + * die_nmi will return ONLY if NOTIFY_STOP happens.. + */ + die_nmi("BUG: NMI Watchdog detected LOCKUP", + regs, panic_on_timeout); + } + } else { + per_cpu(alert_counter,cpu) = 0; + } + + return; +} + +/* + * Create/destroy watchdog threads as CPUs come and go: + */ +static int __cpuinit +cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) +{ + int hotcpu = (unsigned long)hcpu; + struct perf_event *event; + + switch (action) { + case CPU_UP_PREPARE: + case CPU_UP_PREPARE_FROZEN: + per_cpu(nmi_watchdog_touch, hotcpu) = 0; + break; + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + /* originally wanted the below chunk to be in CPU_UP_PREPARE, but caps is unpriv for non-CPU0 */ + wd_attr.sample_period = cpu_khz * 1000; + event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); + if (IS_ERR(event)) { + printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", hotcpu, event); + return NOTIFY_BAD; + } + per_cpu(nmi_watchdog_ev, hotcpu) = event; + perf_event_enable(per_cpu(nmi_watchdog_ev, hotcpu)); + break; +#ifdef CONFIG_HOTPLUG_CPU + case CPU_UP_CANCELED: + case CPU_UP_CANCELED_FROZEN: + perf_event_disable(per_cpu(nmi_watchdog_ev, hotcpu)); + case CPU_DEAD: + case CPU_DEAD_FROZEN: + event = per_cpu(nmi_watchdog_ev, hotcpu); + per_cpu(nmi_watchdog_ev, hotcpu) = NULL; + perf_event_release_kernel(event); + break; +#endif /* CONFIG_HOTPLUG_CPU */ + } + return NOTIFY_OK; +} + +static struct notifier_block __cpuinitdata cpu_nfb = { + .notifier_call = cpu_callback +}; + +static int __initdata nonmi_watchdog; + +static int __init nonmi_watchdog_setup(char *str) +{ + nonmi_watchdog = 1; + return 1; +} +__setup("nonmi_watchdog", nonmi_watchdog_setup); + +static int __init spawn_nmi_watchdog_task(void) +{ + void *cpu = (void *)(long)smp_processor_id(); + int err; + + if (nonmi_watchdog) + return 0; + + err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); + if (err == NOTIFY_BAD) { + BUG(); + return 1; + } + cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); + register_cpu_notifier(&cpu_nfb); + + return 0; +} +early_initcall(spawn_nmi_watchdog_task); -- cgit v0.10.2 From 84e478c6f1eb9c4bfa1fff2f8108e9a061b46428 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 5 Feb 2010 21:47:05 -0500 Subject: nmi_watchdog: Config option to enable new nmi_watchdog These are the bits that enable the new nmi_watchdog and safely isolate the old nmi_watchdog. Only one or the other can run, not both at the same time. Signed-off-by: Don Zickus Cc: Linus Torvalds Cc: Andrew Morton Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: peterz@infradead.org LKML-Reference: <1265424425-31562-4-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/kernel/apic/Makefile b/arch/x86/kernel/apic/Makefile index 565c1bf..1a4512e 100644 --- a/arch/x86/kernel/apic/Makefile +++ b/arch/x86/kernel/apic/Makefile @@ -2,7 +2,12 @@ # Makefile for local APIC drivers and for the IO-APIC code # -obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o nmi.o +obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o +ifneq ($(CONFIG_NMI_WATCHDOG),y) +obj-$(CONFIG_X86_LOCAL_APIC) += nmi.o +endif +obj-$(CONFIG_NMI_WATCHDOG) += hw_nmi.o + obj-$(CONFIG_X86_IO_APIC) += io_apic.o obj-$(CONFIG_SMP) += ipi.o diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index 51ef893f..973cbc4 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -406,6 +406,7 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) == NOTIFY_STOP) return; +#ifndef CONFIG_NMI_WATCHDOG /* * Ok, so this is none of the documented NMI sources, * so it must be the NMI watchdog. @@ -413,6 +414,7 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) if (nmi_watchdog_tick(regs, reason)) return; if (!do_nmi_callback(regs, cpu)) +#endif /* !CONFIG_NMI_WATCHDOG */ unknown_nmi_error(reason, regs); #else unknown_nmi_error(reason, regs); diff --git a/include/linux/nmi.h b/include/linux/nmi.h index b752e80..a42ff0b 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -47,4 +47,8 @@ static inline bool trigger_all_cpu_backtrace(void) } #endif +#ifdef CONFIG_NMI_WATCHDOG +int hw_nmi_is_cpu_stuck(struct pt_regs *); +#endif + #endif diff --git a/kernel/Makefile b/kernel/Makefile index 864ff75..8a5abe5 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -76,6 +76,7 @@ obj-$(CONFIG_AUDIT_TREE) += audit_tree.o obj-$(CONFIG_KPROBES) += kprobes.o obj-$(CONFIG_KGDB) += kgdb.o obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o +obj-$(CONFIG_NMI_WATCHDOG) += nmi_watchdog.o obj-$(CONFIG_DETECT_HUNG_TASK) += hung_task.o obj-$(CONFIG_GENERIC_HARDIRQS) += irq/ obj-$(CONFIG_SECCOMP) += seccomp.o diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 25c3ed5..f80b67e 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -170,6 +170,19 @@ config DETECT_SOFTLOCKUP can be detected via the NMI-watchdog, on platforms that support it.) +config NMI_WATCHDOG + bool "Detect Hard Lockups with an NMI Watchdog" + depends on DEBUG_KERNEL && PERF_EVENTS + default y + help + Say Y here to enable the kernel to use the NMI as a watchdog + to detect hard lockups. This is useful when a cpu hangs for no + reason but can still respond to NMIs. A backtrace is displayed + for reviewing and reporting. + + The overhead should be minimal, just an extra NMI every few + seconds. + config BOOTPARAM_SOFTLOCKUP_PANIC bool "Panic (Reboot) On Soft Lockups" depends on DETECT_SOFTLOCKUP -- cgit v0.10.2 From 8e7672cdb413af859086ffceaed68f7e1e8ea4c2 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Tue, 9 Feb 2010 06:11:00 +0100 Subject: nmi_watchdog: Only enable on x86 for now It wont even build on other platforms just yet - so restrict it to x86 for now. Cc: Don Zickus Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: peterz@infradead.org LKML-Reference: <1265424425-31562-4-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index f80b67e..acef882 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -173,6 +173,7 @@ config DETECT_SOFTLOCKUP config NMI_WATCHDOG bool "Detect Hard Lockups with an NMI Watchdog" depends on DEBUG_KERNEL && PERF_EVENTS + depends on X86 default y help Say Y here to enable the kernel to use the NMI as a watchdog -- cgit v0.10.2 From c3128fb6ad39b0edda6675d20585a64846cf89ea Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 12 Feb 2010 17:19:18 -0500 Subject: nmi_watchdog: Use a boolean config flag for compiling Determines if an arch has setup arch specific perf_events and nmi_watchdog code. This should restrict compiles to only those arches ready. Signed-off-by: Don Zickus Cc: peterz@infradead.org Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266013161-31197-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index cbcbfde..4f9685f 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -52,6 +52,7 @@ config X86 select HAVE_KERNEL_LZO select HAVE_HW_BREAKPOINT select PERF_EVENTS + select PERF_EVENTS_NMI select ANON_INODES select HAVE_ARCH_KMEMCHECK select HAVE_USER_RETURN_NOTIFIER diff --git a/init/Kconfig b/init/Kconfig index ada4844..7331a16d 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -946,6 +946,11 @@ config PERF_USE_VMALLOC help See tools/perf/design.txt for details +config PERF_EVENTS_NMI + bool + help + Arch has support for nmi_watchdog + menu "Kernel Performance Events And Counters" config PERF_EVENTS diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index acef882..01a4d85 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -172,8 +172,7 @@ config DETECT_SOFTLOCKUP config NMI_WATCHDOG bool "Detect Hard Lockups with an NMI Watchdog" - depends on DEBUG_KERNEL && PERF_EVENTS - depends on X86 + depends on DEBUG_KERNEL && PERF_EVENTS && PERF_EVENTS_NMI default y help Say Y here to enable the kernel to use the NMI as a watchdog -- cgit v0.10.2 From 504d7cf10ee42bb76b9556859f23d4121dee0a77 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 12 Feb 2010 17:19:19 -0500 Subject: nmi_watchdog: Compile and portability fixes The original patch was x86_64 centric. Changed the code to make it less so. ested by building and running on a powerpc. Signed-off-by: Don Zickus Cc: peterz@infradead.org Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266013161-31197-2-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h index 93da9c3..5b41b0f 100644 --- a/arch/x86/include/asm/nmi.h +++ b/arch/x86/include/asm/nmi.h @@ -17,7 +17,9 @@ int do_nmi_callback(struct pt_regs *regs, int cpu); extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern int check_nmi_watchdog(void); +#if !defined(CONFIG_NMI_WATCHDOG) extern int nmi_watchdog_enabled; +#endif extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); extern int reserve_perfctr_nmi(unsigned int); extern void release_perfctr_nmi(unsigned int); diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 8c0e6a4..312d772 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -32,8 +32,13 @@ static DEFINE_PER_CPU(unsigned, last_irq_sum); */ static inline unsigned int get_timer_irqs(int cpu) { - return per_cpu(irq_stat, cpu).apic_timer_irqs + - per_cpu(irq_stat, cpu).irq0_irqs; + unsigned int irqs = per_cpu(irq_stat, cpu).irq0_irqs; + +#if defined(CONFIG_X86_LOCAL_APIC) + irqs += per_cpu(irq_stat, cpu).apic_timer_irqs; +#endif + + return irqs; } static inline int mce_in_progress(void) @@ -82,6 +87,11 @@ int hw_nmi_is_cpu_stuck(struct pt_regs *regs) } } +u64 hw_nmi_get_sample_period(void) +{ + return cpu_khz * 1000; +} + void arch_trigger_all_cpu_backtrace(void) { int i; @@ -100,15 +110,16 @@ void arch_trigger_all_cpu_backtrace(void) } /* STUB calls to mimic old nmi_watchdog behaviour */ +#if defined(CONFIG_X86_LOCAL_APIC) unsigned int nmi_watchdog = NMI_NONE; EXPORT_SYMBOL(nmi_watchdog); +void acpi_nmi_enable(void) { return; } +void acpi_nmi_disable(void) { return; } +#endif atomic_t nmi_active = ATOMIC_INIT(0); /* oprofile uses this */ EXPORT_SYMBOL(nmi_active); -int nmi_watchdog_enabled; int unknown_nmi_panic; void cpu_nmi_set_wd_enabled(void) { return; } -void acpi_nmi_enable(void) { return; } -void acpi_nmi_disable(void) { return; } void stop_apic_nmi_watchdog(void *unused) { return; } void setup_apic_nmi_watchdog(void *unused) { return; } int __init check_nmi_watchdog(void) { return 0; } diff --git a/include/linux/nmi.h b/include/linux/nmi.h index a42ff0b..794e735 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -20,10 +20,14 @@ extern void touch_nmi_watchdog(void); extern void acpi_nmi_disable(void); extern void acpi_nmi_enable(void); #else +#ifndef CONFIG_NMI_WATCHDOG static inline void touch_nmi_watchdog(void) { touch_softlockup_watchdog(); } +#else +extern void touch_nmi_watchdog(void); +#endif static inline void acpi_nmi_disable(void) { } static inline void acpi_nmi_enable(void) { } #endif @@ -49,6 +53,11 @@ static inline bool trigger_all_cpu_backtrace(void) #ifdef CONFIG_NMI_WATCHDOG int hw_nmi_is_cpu_stuck(struct pt_regs *); +u64 hw_nmi_get_sample_period(void); +extern int nmi_watchdog_enabled; +struct ctl_table; +extern int proc_nmi_enabled(struct ctl_table *, int , + void __user *, size_t *, loff_t *); #endif #endif diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 36817b2..73c1954 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -30,6 +30,8 @@ static DEFINE_PER_CPU(struct perf_event *, nmi_watchdog_ev); static DEFINE_PER_CPU(int, nmi_watchdog_touch); static DEFINE_PER_CPU(long, alert_counter); +static int panic_on_timeout; + void touch_nmi_watchdog(void) { __raw_get_cpu_var(nmi_watchdog_touch) = 1; @@ -46,19 +48,49 @@ void touch_all_nmi_watchdog(void) touch_softlockup_watchdog(); } +static int __init setup_nmi_watchdog(char *str) +{ + if (!strncmp(str, "panic", 5)) { + panic_on_timeout = 1; + str = strchr(str, ','); + if (!str) + return 1; + ++str; + } + return 1; +} +__setup("nmi_watchdog=", setup_nmi_watchdog); + #ifdef CONFIG_SYSCTL /* * proc handler for /proc/sys/kernel/nmi_watchdog */ +int nmi_watchdog_enabled; + int proc_nmi_enabled(struct ctl_table *table, int write, void __user *buffer, size_t *length, loff_t *ppos) { int cpu; - if (per_cpu(nmi_watchdog_ev, smp_processor_id()) == NULL) + if (!write) { + struct perf_event *event; + for_each_online_cpu(cpu) { + event = per_cpu(nmi_watchdog_ev, cpu); + if (event->state > PERF_EVENT_STATE_OFF) { + nmi_watchdog_enabled = 1; + break; + } + } + proc_dointvec(table, write, buffer, length, ppos); + return 0; + } + + if (per_cpu(nmi_watchdog_ev, smp_processor_id()) == NULL) { nmi_watchdog_enabled = 0; - else - nmi_watchdog_enabled = 1; + proc_dointvec(table, write, buffer, length, ppos); + printk("NMI watchdog failed configuration, can not be enabled\n"); + return 0; + } touch_all_nmi_watchdog(); proc_dointvec(table, write, buffer, length, ppos); @@ -81,8 +113,6 @@ struct perf_event_attr wd_attr = { .disabled = 1, }; -static int panic_on_timeout; - void wd_overflow(struct perf_event *event, int nmi, struct perf_sample_data *data, struct pt_regs *regs) @@ -103,11 +133,11 @@ void wd_overflow(struct perf_event *event, int nmi, */ per_cpu(alert_counter,cpu) += 1; if (per_cpu(alert_counter,cpu) == 5) { - /* - * die_nmi will return ONLY if NOTIFY_STOP happens.. - */ - die_nmi("BUG: NMI Watchdog detected LOCKUP", - regs, panic_on_timeout); + if (panic_on_timeout) { + panic("NMI Watchdog detected LOCKUP on cpu %d", cpu); + } else { + WARN(1, "NMI Watchdog detected LOCKUP on cpu %d", cpu); + } } } else { per_cpu(alert_counter,cpu) = 0; @@ -133,7 +163,7 @@ cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) case CPU_ONLINE: case CPU_ONLINE_FROZEN: /* originally wanted the below chunk to be in CPU_UP_PREPARE, but caps is unpriv for non-CPU0 */ - wd_attr.sample_period = cpu_khz * 1000; + wd_attr.sample_period = hw_nmi_get_sample_period(); event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); if (IS_ERR(event)) { printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", hotcpu, event); diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 8a68b24..ac72c9e 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -60,6 +60,10 @@ #include #endif +#ifdef CONFIG_NMI_WATCHDOG +#include +#endif + #if defined(CONFIG_SYSCTL) @@ -692,7 +696,16 @@ static struct ctl_table kern_table[] = { .mode = 0444, .proc_handler = proc_dointvec, }, -#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) +#if defined(CONFIG_NMI_WATCHDOG) + { + .procname = "nmi_watchdog", + .data = &nmi_watchdog_enabled, + .maxlen = sizeof (int), + .mode = 0644, + .proc_handler = proc_nmi_enabled, + }, +#endif +#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_NMI_WATCHDOG) { .procname = "unknown_nmi_panic", .data = &unknown_nmi_panic, -- cgit v0.10.2 From cf454aecb31741a0438ed1201b3dd153c7c7b19a Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 12 Feb 2010 17:19:20 -0500 Subject: nmi_watchdog: Fallback to software events when no hardware pmu detected Not all arches have a PMU or have perf_event support for their PMU. The nmi_watchdog will fail in those cases. Fallback to using software events to generate nmi_watchdog traffic with local apic interrupts. Tested on a Pentium4 and it worked as expected, excepting for detecting cpu lockups. The problem with using software events as a cpu lock up detector is the nmi_watchdog uses the logic that if local apic interrupts stop incrementing then the cpu is probably locked up. But with software events we use the local apic to trigger the nmi_watchdog callback to see if local apic interrupts are still firing, which obviously they are otherwise we wouldn't have been triggered. The algorithm to detect cpu lock ups is the same as the old nmi_watchdog. Perhaps we need to find a better way to detect lock ups? Signed-off-by: Don Zickus Cc: peterz@infradead.org Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266013161-31197-3-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 73c1954..4f23505 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -166,8 +166,12 @@ cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) wd_attr.sample_period = hw_nmi_get_sample_period(); event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); if (IS_ERR(event)) { - printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", hotcpu, event); - return NOTIFY_BAD; + wd_attr.type = PERF_TYPE_SOFTWARE; + event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); + if (IS_ERR(event)) { + printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", hotcpu, event); + return NOTIFY_BAD; + } } per_cpu(nmi_watchdog_ev, hotcpu) = event; perf_event_enable(per_cpu(nmi_watchdog_ev, hotcpu)); -- cgit v0.10.2 From 6081b6cd9702967889de34fe5da1f96bb96d0ab8 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Tue, 16 Feb 2010 17:04:52 -0500 Subject: nmi_watchdog: support for oprofile Re-arrange the code so that when someone disables nmi_watchdog with: echo 0 > /proc/sys/kernel/nmi_watchdog it releases the hardware reservation on the PMUs. This allows the oprofile module to grab those PMUs and do its thing. Otherwise oprofile fails to load because the hardware is reserved by the perf_events subsystem. Tested using: oprofile --vm-linux --start and watched it failed when nmi_watchdog is enabled and succeed when: oprofile --deinit && echo 0 > /proc/sys/kernel/nmi_watchdog is run. Note: this has the side quirk of having the nmi_watchdog latch onto the software events instead of hardware events if oprofile has already reserved the hardware first. User beware! :-) Signed-off-by: Don Zickus Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: eranian@google.com LKML-Reference: <1266357892-30504-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 4f23505..633b230 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -61,50 +61,6 @@ static int __init setup_nmi_watchdog(char *str) } __setup("nmi_watchdog=", setup_nmi_watchdog); -#ifdef CONFIG_SYSCTL -/* - * proc handler for /proc/sys/kernel/nmi_watchdog - */ -int nmi_watchdog_enabled; - -int proc_nmi_enabled(struct ctl_table *table, int write, - void __user *buffer, size_t *length, loff_t *ppos) -{ - int cpu; - - if (!write) { - struct perf_event *event; - for_each_online_cpu(cpu) { - event = per_cpu(nmi_watchdog_ev, cpu); - if (event->state > PERF_EVENT_STATE_OFF) { - nmi_watchdog_enabled = 1; - break; - } - } - proc_dointvec(table, write, buffer, length, ppos); - return 0; - } - - if (per_cpu(nmi_watchdog_ev, smp_processor_id()) == NULL) { - nmi_watchdog_enabled = 0; - proc_dointvec(table, write, buffer, length, ppos); - printk("NMI watchdog failed configuration, can not be enabled\n"); - return 0; - } - - touch_all_nmi_watchdog(); - proc_dointvec(table, write, buffer, length, ppos); - if (nmi_watchdog_enabled) - for_each_online_cpu(cpu) - perf_event_enable(per_cpu(nmi_watchdog_ev, cpu)); - else - for_each_online_cpu(cpu) - perf_event_disable(per_cpu(nmi_watchdog_ev, cpu)); - return 0; -} - -#endif /* CONFIG_SYSCTL */ - struct perf_event_attr wd_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, @@ -146,6 +102,85 @@ void wd_overflow(struct perf_event *event, int nmi, return; } +static int enable_nmi_watchdog(int cpu) +{ + struct perf_event *event; + + event = per_cpu(nmi_watchdog_ev, cpu); + if (event && event->state > PERF_EVENT_STATE_OFF) + return 0; + + if (event == NULL) { + /* Try to register using hardware perf events first */ + wd_attr.sample_period = hw_nmi_get_sample_period(); + event = perf_event_create_kernel_counter(&wd_attr, cpu, -1, wd_overflow); + if (IS_ERR(event)) { + wd_attr.type = PERF_TYPE_SOFTWARE; + event = perf_event_create_kernel_counter(&wd_attr, cpu, -1, wd_overflow); + if (IS_ERR(event)) { + printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", cpu, event); + return -1; + } + } + per_cpu(nmi_watchdog_ev, cpu) = event; + } + perf_event_enable(per_cpu(nmi_watchdog_ev, cpu)); + return 0; +} + +static void disable_nmi_watchdog(int cpu) +{ + struct perf_event *event; + + event = per_cpu(nmi_watchdog_ev, cpu); + if (event) { + perf_event_disable(per_cpu(nmi_watchdog_ev, cpu)); + per_cpu(nmi_watchdog_ev, cpu) = NULL; + perf_event_release_kernel(event); + } +} + +#ifdef CONFIG_SYSCTL +/* + * proc handler for /proc/sys/kernel/nmi_watchdog + */ +int nmi_watchdog_enabled; + +int proc_nmi_enabled(struct ctl_table *table, int write, + void __user *buffer, size_t *length, loff_t *ppos) +{ + int cpu; + + if (!write) { + struct perf_event *event; + for_each_online_cpu(cpu) { + event = per_cpu(nmi_watchdog_ev, cpu); + if (event && event->state > PERF_EVENT_STATE_OFF) { + nmi_watchdog_enabled = 1; + break; + } + } + proc_dointvec(table, write, buffer, length, ppos); + return 0; + } + + touch_all_nmi_watchdog(); + proc_dointvec(table, write, buffer, length, ppos); + if (nmi_watchdog_enabled) { + for_each_online_cpu(cpu) + if (enable_nmi_watchdog(cpu)) { + printk("NMI watchdog failed configuration, " + " can not be enabled\n"); + } + } else { + for_each_online_cpu(cpu) + disable_nmi_watchdog(cpu); + } + return 0; +} + +#endif /* CONFIG_SYSCTL */ + /* * Create/destroy watchdog threads as CPUs come and go: */ @@ -153,7 +188,6 @@ static int __cpuinit cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) { int hotcpu = (unsigned long)hcpu; - struct perf_event *event; switch (action) { case CPU_UP_PREPARE: @@ -162,29 +196,15 @@ cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) break; case CPU_ONLINE: case CPU_ONLINE_FROZEN: - /* originally wanted the below chunk to be in CPU_UP_PREPARE, but caps is unpriv for non-CPU0 */ - wd_attr.sample_period = hw_nmi_get_sample_period(); - event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); - if (IS_ERR(event)) { - wd_attr.type = PERF_TYPE_SOFTWARE; - event = perf_event_create_kernel_counter(&wd_attr, hotcpu, -1, wd_overflow); - if (IS_ERR(event)) { - printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", hotcpu, event); - return NOTIFY_BAD; - } - } - per_cpu(nmi_watchdog_ev, hotcpu) = event; - perf_event_enable(per_cpu(nmi_watchdog_ev, hotcpu)); + if (enable_nmi_watchdog(hotcpu)) + return NOTIFY_BAD; break; #ifdef CONFIG_HOTPLUG_CPU case CPU_UP_CANCELED: case CPU_UP_CANCELED_FROZEN: - perf_event_disable(per_cpu(nmi_watchdog_ev, hotcpu)); + disable_nmi_watchdog(hotcpu); case CPU_DEAD: case CPU_DEAD_FROZEN: - event = per_cpu(nmi_watchdog_ev, hotcpu); - per_cpu(nmi_watchdog_ev, hotcpu) = NULL; - perf_event_release_kernel(event); break; #endif /* CONFIG_HOTPLUG_CPU */ } -- cgit v0.10.2 From 96ca4028aca0638d0e11dcbfabc4283054072933 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Tue, 16 Feb 2010 17:02:25 -0500 Subject: nmi_watchdog: Properly configure for software events Paul Mackerras brought up a good point that when fallbacking to software events, I may have been lucky in my configuration. Modified the code to explicit provide a new configuration for software events. Suggested-by: Paul Mackerras Signed-off-by: Don Zickus Cc: gorcunov@gmail.com Cc: aris@redhat.com Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker LKML-Reference: <1266357745-26671-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 633b230..3c75cbf 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -61,7 +61,7 @@ static int __init setup_nmi_watchdog(char *str) } __setup("nmi_watchdog=", setup_nmi_watchdog); -struct perf_event_attr wd_attr = { +struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, .size = sizeof(struct perf_event_attr), @@ -69,6 +69,14 @@ struct perf_event_attr wd_attr = { .disabled = 1, }; +struct perf_event_attr wd_sw_attr = { + .type = PERF_TYPE_SOFTWARE, + .config = PERF_COUNT_SW_CPU_CLOCK, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, +}; + void wd_overflow(struct perf_event *event, int nmi, struct perf_sample_data *data, struct pt_regs *regs) @@ -105,6 +113,7 @@ void wd_overflow(struct perf_event *event, int nmi, static int enable_nmi_watchdog(int cpu) { struct perf_event *event; + struct perf_event_attr *wd_attr; event = per_cpu(nmi_watchdog_ev, cpu); if (event && event->state > PERF_EVENT_STATE_OFF) @@ -112,11 +121,15 @@ static int enable_nmi_watchdog(int cpu) if (event == NULL) { /* Try to register using hardware perf events first */ - wd_attr.sample_period = hw_nmi_get_sample_period(); - event = perf_event_create_kernel_counter(&wd_attr, cpu, -1, wd_overflow); + wd_attr = &wd_hw_attr; + wd_attr->sample_period = hw_nmi_get_sample_period(); + event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); if (IS_ERR(event)) { - wd_attr.type = PERF_TYPE_SOFTWARE; - event = perf_event_create_kernel_counter(&wd_attr, cpu, -1, wd_overflow); + /* hardware doesn't exist or not supported, fallback to software events */ + printk("nmi_watchdog: hardware not available, trying software events\n"); + wd_attr = &wd_sw_attr; + wd_attr->sample_period = NSEC_PER_SEC; + event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); if (IS_ERR(event)) { printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", cpu, event); return -1; -- cgit v0.10.2 From 2cc4452bc31fc1cde6f0b64a4eb13269f982787d Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Thu, 18 Feb 2010 21:56:52 -0500 Subject: nmi_watchdog: Fix undefined 'apic' build bug Ingo provided me a config that fails to compile with: arch/x86/built-in.o: In function `arch_trigger_all_cpu_backtrace': (.text+0x17e78): undefined reference to `apic' make: *** [.tmp_vmlinux1] Error 1 I realized I changed the compile behaviour of the nmi code by not wrapping it with CONFIG_LOCAL_APIC. To fix this I add a compile check for ARCH_HAS_NMI_WATCHDOG around arch_trigger_all_cpu_backtrace. Signed-off-by: Don Zickus Cc: a.p.zijlstra@chello.nl Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266548212-24243-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 312d772..0b4d205 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -92,6 +92,7 @@ u64 hw_nmi_get_sample_period(void) return cpu_khz * 1000; } +#ifdef ARCH_HAS_NMI_WATCHDOG void arch_trigger_all_cpu_backtrace(void) { int i; @@ -108,6 +109,7 @@ void arch_trigger_all_cpu_backtrace(void) mdelay(1); } } +#endif /* STUB calls to mimic old nmi_watchdog behaviour */ #if defined(CONFIG_X86_LOCAL_APIC) -- cgit v0.10.2 From 47195d57636604ff6048b0d7aa3e4ed9643f6073 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Mon, 22 Feb 2010 18:09:03 -0500 Subject: nmi_watchdog: Clean up various small details Mostly copy/paste whitespace damage with a couple of nitpicks by the checkpatch script. Fix the struct definition as requested by Ingo too. Signed-off-by: Don Zickus Cc: peterz@infradead.org Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266880143-24943-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar -- arch/x86/kernel/apic/hw_nmi.c | 14 +++++------ arch/x86/kernel/traps.c | 6 ++-- include/linux/nmi.h | 2 - kernel/nmi_watchdog.c | 51 ++++++++++++++++++++---------------------- 4 files changed, 36 insertions(+), 37 deletions(-) diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 0b4d205..e8b78a0 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -38,15 +38,15 @@ static inline unsigned int get_timer_irqs(int cpu) irqs += per_cpu(irq_stat, cpu).apic_timer_irqs; #endif - return irqs; + return irqs; } static inline int mce_in_progress(void) { #if defined(CONFIG_X86_MCE) - return atomic_read(&mce_entry) > 0; + return atomic_read(&mce_entry) > 0; #endif - return 0; + return 0; } int hw_nmi_is_cpu_stuck(struct pt_regs *regs) @@ -69,9 +69,9 @@ int hw_nmi_is_cpu_stuck(struct pt_regs *regs) } /* if we are doing an mce, just assume the cpu is not stuck */ - /* Could check oops_in_progress here too, but it's safer not to */ - if (mce_in_progress()) - return 0; + /* Could check oops_in_progress here too, but it's safer not to */ + if (mce_in_progress()) + return 0; /* We determine if the cpu is stuck by checking whether any * interrupts have happened since we last checked. Of course @@ -89,7 +89,7 @@ int hw_nmi_is_cpu_stuck(struct pt_regs *regs) u64 hw_nmi_get_sample_period(void) { - return cpu_khz * 1000; + return cpu_khz * 1000; } #ifdef ARCH_HAS_NMI_WATCHDOG diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index 973cbc4..bdc7fab 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -402,9 +402,9 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) return; #ifdef CONFIG_X86_LOCAL_APIC - if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) - == NOTIFY_STOP) - return; + if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) + == NOTIFY_STOP) + return; #ifndef CONFIG_NMI_WATCHDOG /* diff --git a/include/linux/nmi.h b/include/linux/nmi.h index 794e735..22cc796 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -57,7 +57,7 @@ u64 hw_nmi_get_sample_period(void); extern int nmi_watchdog_enabled; struct ctl_table; extern int proc_nmi_enabled(struct ctl_table *, int , - void __user *, size_t *, loff_t *); + void __user *, size_t *, loff_t *); #endif #endif diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 3c75cbf..0a6f57f 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -50,31 +50,31 @@ void touch_all_nmi_watchdog(void) static int __init setup_nmi_watchdog(char *str) { - if (!strncmp(str, "panic", 5)) { - panic_on_timeout = 1; - str = strchr(str, ','); - if (!str) - return 1; - ++str; - } - return 1; + if (!strncmp(str, "panic", 5)) { + panic_on_timeout = 1; + str = strchr(str, ','); + if (!str) + return 1; + ++str; + } + return 1; } __setup("nmi_watchdog=", setup_nmi_watchdog); struct perf_event_attr wd_hw_attr = { - .type = PERF_TYPE_HARDWARE, - .config = PERF_COUNT_HW_CPU_CYCLES, - .size = sizeof(struct perf_event_attr), - .pinned = 1, - .disabled = 1, + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, }; struct perf_event_attr wd_sw_attr = { - .type = PERF_TYPE_SOFTWARE, - .config = PERF_COUNT_SW_CPU_CLOCK, - .size = sizeof(struct perf_event_attr), - .pinned = 1, - .disabled = 1, + .type = PERF_TYPE_SOFTWARE, + .config = PERF_COUNT_SW_CPU_CLOCK, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, }; void wd_overflow(struct perf_event *event, int nmi, @@ -95,16 +95,15 @@ void wd_overflow(struct perf_event *event, int nmi, * Ayiee, looks like this CPU is stuck ... * wait a few IRQs (5 seconds) before doing the oops ... */ - per_cpu(alert_counter,cpu) += 1; - if (per_cpu(alert_counter,cpu) == 5) { - if (panic_on_timeout) { + per_cpu(alert_counter, cpu) += 1; + if (per_cpu(alert_counter, cpu) == 5) { + if (panic_on_timeout) panic("NMI Watchdog detected LOCKUP on cpu %d", cpu); - } else { + else WARN(1, "NMI Watchdog detected LOCKUP on cpu %d", cpu); - } } } else { - per_cpu(alert_counter,cpu) = 0; + per_cpu(alert_counter, cpu) = 0; } return; @@ -126,7 +125,7 @@ static int enable_nmi_watchdog(int cpu) event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); if (IS_ERR(event)) { /* hardware doesn't exist or not supported, fallback to software events */ - printk("nmi_watchdog: hardware not available, trying software events\n"); + printk(KERN_INFO "nmi_watchdog: hardware not available, trying software events\n"); wd_attr = &wd_sw_attr; wd_attr->sample_period = NSEC_PER_SEC; event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); @@ -182,7 +181,7 @@ int proc_nmi_enabled(struct ctl_table *table, int write, if (nmi_watchdog_enabled) { for_each_online_cpu(cpu) if (enable_nmi_watchdog(cpu)) { - printk("NMI watchdog failed configuration, " + printk(KERN_ERR "NMI watchdog failed configuration, " " can not be enabled\n"); } } else { -- cgit v0.10.2 From c99c30feadf664ccd8590b96259cb9f3954bd246 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Sun, 28 Feb 2010 20:49:00 +0100 Subject: nmi_watchdog: Turn it off by default It was nice to enable it by default for testing - but before we push it upstream we want it to be off - so that people can opt-in gradually. Cc: Don Zickus Cc: peterz@infradead.org Cc: gorcunov@gmail.com Cc: aris@redhat.com LKML-Reference: <1266880143-24943-1-git-send-email-dzickus@redhat.com> Signed-off-by: Ingo Molnar diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 01a4d85..e2e73cc 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -173,7 +173,6 @@ config DETECT_SOFTLOCKUP config NMI_WATCHDOG bool "Detect Hard Lockups with an NMI Watchdog" depends on DEBUG_KERNEL && PERF_EVENTS && PERF_EVENTS_NMI - default y help Say Y here to enable the kernel to use the NMI as a watchdog to detect hard lockups. This is useful when a cpu hangs for no -- cgit v0.10.2 From 5671a10e2bc7f99d9157c6044faf8be2ef302361 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Tue, 2 Mar 2010 14:20:14 +0100 Subject: nmi_watchdog: Tell the world we're active Because I was wondering why perf stat wasn't working as expected.. Signed-off-by: Peter Zijlstra Cc: Don Zickus LKML-Reference: Signed-off-by: Ingo Molnar diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c index 0a6f57f..a79d211 100644 --- a/kernel/nmi_watchdog.c +++ b/kernel/nmi_watchdog.c @@ -244,6 +244,8 @@ static int __init spawn_nmi_watchdog_task(void) if (nonmi_watchdog) return 0; + printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n"); + err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); if (err == NOTIFY_BAD) { BUG(); -- cgit v0.10.2 From 58687acba59266735adb8ccd9b5b9aa2c7cd205b Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:44 -0400 Subject: lockup_detector: Combine nmi_watchdog and softlockup detector The new nmi_watchdog (which uses the perf event subsystem) is very similar in structure to the softlockup detector. Using Ingo's suggestion, I combined the two functionalities into one file: kernel/watchdog.c. Now both the nmi_watchdog (or hardlockup detector) and softlockup detector sit on top of the perf event subsystem, which is run every 60 seconds or so to see if there are any lockups. To detect hardlockups, cpus not responding to interrupts, I implemented an hrtimer that runs 5 times for every perf event overflow event. If that stops counting on a cpu, then the cpu is most likely in trouble. To detect softlockups, tasks not yielding to the scheduler, I used the previous kthread idea that now gets kicked every time the hrtimer fires. If the kthread isn't being scheduled neither is anyone else and the warning is printed to the console. I tested this on x86_64 and both the softlockup and hardlockup paths work. V2: - cleaned up the Kconfig and softlockup combination - surrounded hardlockup cases with #ifdef CONFIG_PERF_EVENTS_NMI - seperated out the softlockup case from perf event subsystem - re-arranged the enabling/disabling nmi watchdog from proc space - added cpumasks for hardlockup failure cases - removed fallback to soft events if no PMU exists for hard events V3: - comment cleanups - drop support for older softlockup code - per_cpu cleanups - completely remove software clock base hardlockup detector - use per_cpu masking on hard/soft lockup detection - #ifdef cleanups - rename config option NMI_WATCHDOG to LOCKUP_DETECTOR - documentation additions V4: - documentation fixes - convert per_cpu to __get_cpu_var - powerpc compile fixes V5: - split apart warn flags for hard and soft lockups TODO: - figure out how to make an arch-agnostic clock2cycles call (if possible) to feed into perf events as a sample period [fweisbec: merged conflict patch] Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-2-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/Documentation/kernel-parameters.txt b/Documentation/kernel-parameters.txt index 839b21b..dfe8d1c 100644 --- a/Documentation/kernel-parameters.txt +++ b/Documentation/kernel-parameters.txt @@ -1777,6 +1777,8 @@ and is between 256 and 4096 characters. It is defined in the file nousb [USB] Disable the USB subsystem + nowatchdog [KNL] Disable the lockup detector. + nowb [ARM] nox2apic [X86-64,APIC] Do not enable x2APIC mode. diff --git a/arch/x86/include/asm/nmi.h b/arch/x86/include/asm/nmi.h index 5b41b0f..932f0f8 100644 --- a/arch/x86/include/asm/nmi.h +++ b/arch/x86/include/asm/nmi.h @@ -17,7 +17,7 @@ int do_nmi_callback(struct pt_regs *regs, int cpu); extern void die_nmi(char *str, struct pt_regs *regs, int do_panic); extern int check_nmi_watchdog(void); -#if !defined(CONFIG_NMI_WATCHDOG) +#if !defined(CONFIG_LOCKUP_DETECTOR) extern int nmi_watchdog_enabled; #endif extern int avail_to_resrv_perfctr_nmi_bit(unsigned int); diff --git a/arch/x86/kernel/apic/Makefile b/arch/x86/kernel/apic/Makefile index 1a4512e..52f32e0 100644 --- a/arch/x86/kernel/apic/Makefile +++ b/arch/x86/kernel/apic/Makefile @@ -3,10 +3,10 @@ # obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o -ifneq ($(CONFIG_NMI_WATCHDOG),y) +ifneq ($(CONFIG_LOCKUP_DETECTOR),y) obj-$(CONFIG_X86_LOCAL_APIC) += nmi.o endif -obj-$(CONFIG_NMI_WATCHDOG) += hw_nmi.o +obj-$(CONFIG_LOCKUP_DETECTOR) += hw_nmi.o obj-$(CONFIG_X86_IO_APIC) += io_apic.o obj-$(CONFIG_SMP) += ipi.o diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index e8b78a0..79425f9 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -89,7 +89,7 @@ int hw_nmi_is_cpu_stuck(struct pt_regs *regs) u64 hw_nmi_get_sample_period(void) { - return cpu_khz * 1000; + return (u64)(cpu_khz) * 1000 * 60; } #ifdef ARCH_HAS_NMI_WATCHDOG diff --git a/arch/x86/kernel/traps.c b/arch/x86/kernel/traps.c index bdc7fab..bd347c2 100644 --- a/arch/x86/kernel/traps.c +++ b/arch/x86/kernel/traps.c @@ -406,7 +406,7 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) == NOTIFY_STOP) return; -#ifndef CONFIG_NMI_WATCHDOG +#ifndef CONFIG_LOCKUP_DETECTOR /* * Ok, so this is none of the documented NMI sources, * so it must be the NMI watchdog. @@ -414,7 +414,7 @@ static notrace __kprobes void default_do_nmi(struct pt_regs *regs) if (nmi_watchdog_tick(regs, reason)) return; if (!do_nmi_callback(regs, cpu)) -#endif /* !CONFIG_NMI_WATCHDOG */ +#endif /* !CONFIG_LOCKUP_DETECTOR */ unknown_nmi_error(reason, regs); #else unknown_nmi_error(reason, regs); diff --git a/include/linux/nmi.h b/include/linux/nmi.h index 22cc796..abd48aa 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -20,7 +20,7 @@ extern void touch_nmi_watchdog(void); extern void acpi_nmi_disable(void); extern void acpi_nmi_enable(void); #else -#ifndef CONFIG_NMI_WATCHDOG +#ifndef CONFIG_LOCKUP_DETECTOR static inline void touch_nmi_watchdog(void) { touch_softlockup_watchdog(); @@ -51,12 +51,12 @@ static inline bool trigger_all_cpu_backtrace(void) } #endif -#ifdef CONFIG_NMI_WATCHDOG +#ifdef CONFIG_LOCKUP_DETECTOR int hw_nmi_is_cpu_stuck(struct pt_regs *); u64 hw_nmi_get_sample_period(void); -extern int nmi_watchdog_enabled; +extern int watchdog_enabled; struct ctl_table; -extern int proc_nmi_enabled(struct ctl_table *, int , +extern int proc_dowatchdog_enabled(struct ctl_table *, int , void __user *, size_t *, loff_t *); #endif diff --git a/include/linux/sched.h b/include/linux/sched.h index dad7f66..37efe8f 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -346,6 +346,12 @@ extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, size_t *lenp, loff_t *ppos); #endif +#ifdef CONFIG_LOCKUP_DETECTOR +extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, + void __user *buffer, + size_t *lenp, loff_t *ppos); +#endif + /* Attach to any functions which should be ignored in wchan output. */ #define __sched __attribute__((__section__(".sched.text"))) diff --git a/init/Kconfig b/init/Kconfig index c6c8903..e44e254 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -944,8 +944,11 @@ config PERF_USE_VMALLOC config PERF_EVENTS_NMI bool + depends on PERF_EVENTS help - Arch has support for nmi_watchdog + System hardware can generate an NMI using the perf event + subsystem. Also has support for calculating CPU cycle events + to determine how many clock cycles in a given period. menu "Kernel Performance Events And Counters" diff --git a/kernel/Makefile b/kernel/Makefile index d5c3006..6adeafc 100644 --- a/kernel/Makefile +++ b/kernel/Makefile @@ -76,9 +76,8 @@ obj-$(CONFIG_GCOV_KERNEL) += gcov/ obj-$(CONFIG_AUDIT_TREE) += audit_tree.o obj-$(CONFIG_KPROBES) += kprobes.o obj-$(CONFIG_KGDB) += kgdb.o -obj-$(CONFIG_DETECT_SOFTLOCKUP) += softlockup.o -obj-$(CONFIG_NMI_WATCHDOG) += nmi_watchdog.o obj-$(CONFIG_DETECT_HUNG_TASK) += hung_task.o +obj-$(CONFIG_LOCKUP_DETECTOR) += watchdog.o obj-$(CONFIG_GENERIC_HARDIRQS) += irq/ obj-$(CONFIG_SECCOMP) += seccomp.o obj-$(CONFIG_RCU_TORTURE_TEST) += rcutorture.o diff --git a/kernel/sysctl.c b/kernel/sysctl.c index a38af43..0f9adda 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -74,7 +74,7 @@ #include #endif -#ifdef CONFIG_NMI_WATCHDOG +#ifdef CONFIG_LOCKUP_DETECTOR #include #endif @@ -686,16 +686,25 @@ static struct ctl_table kern_table[] = { .mode = 0444, .proc_handler = proc_dointvec, }, -#if defined(CONFIG_NMI_WATCHDOG) +#if defined(CONFIG_LOCKUP_DETECTOR) { - .procname = "nmi_watchdog", - .data = &nmi_watchdog_enabled, + .procname = "watchdog", + .data = &watchdog_enabled, .maxlen = sizeof (int), .mode = 0644, - .proc_handler = proc_nmi_enabled, + .proc_handler = proc_dowatchdog_enabled, + }, + { + .procname = "watchdog_thresh", + .data = &softlockup_thresh, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dowatchdog_thresh, + .extra1 = &neg_one, + .extra2 = &sixty, }, #endif -#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_NMI_WATCHDOG) +#if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR) { .procname = "unknown_nmi_panic", .data = &unknown_nmi_panic, diff --git a/kernel/watchdog.c b/kernel/watchdog.c new file mode 100644 index 0000000..6b7fad8 --- /dev/null +++ b/kernel/watchdog.c @@ -0,0 +1,592 @@ +/* + * Detect hard and soft lockups on a system + * + * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc. + * + * this code detects hard lockups: incidents in where on a CPU + * the kernel does not respond to anything except NMI. + * + * Note: Most of this code is borrowed heavily from softlockup.c, + * so thanks to Ingo for the initial implementation. + * Some chunks also taken from arch/x86/kernel/apic/nmi.c, thanks + * to those contributors as well. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include +#include + +int watchdog_enabled; +int __read_mostly softlockup_thresh = 60; + +static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts); +static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog); +static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer); +static DEFINE_PER_CPU(bool, softlockup_touch_sync); +static DEFINE_PER_CPU(bool, hard_watchdog_warn); +static DEFINE_PER_CPU(bool, soft_watchdog_warn); +#ifdef CONFIG_PERF_EVENTS_NMI +static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts); +static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved); +static DEFINE_PER_CPU(struct perf_event *, watchdog_ev); +#endif + +static int __read_mostly did_panic; +static int __initdata no_watchdog; + + +/* boot commands */ +/* + * Should we panic when a soft-lockup or hard-lockup occurs: + */ +#ifdef CONFIG_PERF_EVENTS_NMI +static int hardlockup_panic; + +static int __init hardlockup_panic_setup(char *str) +{ + if (!strncmp(str, "panic", 5)) + hardlockup_panic = 1; + return 1; +} +__setup("nmi_watchdog=", hardlockup_panic_setup); +#endif + +unsigned int __read_mostly softlockup_panic = + CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE; + +static int __init softlockup_panic_setup(char *str) +{ + softlockup_panic = simple_strtoul(str, NULL, 0); + + return 1; +} +__setup("softlockup_panic=", softlockup_panic_setup); + +static int __init nowatchdog_setup(char *str) +{ + no_watchdog = 1; + return 1; +} +__setup("nowatchdog", nowatchdog_setup); + +/* deprecated */ +static int __init nosoftlockup_setup(char *str) +{ + no_watchdog = 1; + return 1; +} +__setup("nosoftlockup", nosoftlockup_setup); +/* */ + + +/* + * Returns seconds, approximately. We don't need nanosecond + * resolution, and we don't need to waste time with a big divide when + * 2^30ns == 1.074s. + */ +static unsigned long get_timestamp(int this_cpu) +{ + return cpu_clock(this_cpu) >> 30LL; /* 2^30 ~= 10^9 */ +} + +static unsigned long get_sample_period(void) +{ + /* + * convert softlockup_thresh from seconds to ns + * the divide by 5 is to give hrtimer 5 chances to + * increment before the hardlockup detector generates + * a warning + */ + return softlockup_thresh / 5 * NSEC_PER_SEC; +} + +/* Commands for resetting the watchdog */ +static void __touch_watchdog(void) +{ + int this_cpu = raw_smp_processor_id(); + + __get_cpu_var(watchdog_touch_ts) = get_timestamp(this_cpu); +} + +void touch_watchdog(void) +{ + __get_cpu_var(watchdog_touch_ts) = 0; +} +EXPORT_SYMBOL(touch_watchdog); + +void touch_all_watchdog(void) +{ + int cpu; + + /* + * this is done lockless + * do we care if a 0 races with a timestamp? + * all it means is the softlock check starts one cycle later + */ + for_each_online_cpu(cpu) + per_cpu(watchdog_touch_ts, cpu) = 0; +} + +void touch_nmi_watchdog(void) +{ + touch_watchdog(); +} +EXPORT_SYMBOL(touch_nmi_watchdog); + +void touch_all_nmi_watchdog(void) +{ + touch_all_watchdog(); +} + +void touch_softlockup_watchdog(void) +{ + touch_watchdog(); +} + +void touch_all_softlockup_watchdogs(void) +{ + touch_all_watchdog(); +} + +void touch_softlockup_watchdog_sync(void) +{ + __raw_get_cpu_var(softlockup_touch_sync) = true; + __raw_get_cpu_var(watchdog_touch_ts) = 0; +} + +void softlockup_tick(void) +{ +} + +#ifdef CONFIG_PERF_EVENTS_NMI +/* watchdog detector functions */ +static int is_hardlockup(int cpu) +{ + unsigned long hrint = per_cpu(hrtimer_interrupts, cpu); + + if (per_cpu(hrtimer_interrupts_saved, cpu) == hrint) + return 1; + + per_cpu(hrtimer_interrupts_saved, cpu) = hrint; + return 0; +} +#endif + +static int is_softlockup(unsigned long touch_ts, int cpu) +{ + unsigned long now = get_timestamp(cpu); + + /* Warn about unreasonable delays: */ + if (time_after(now, touch_ts + softlockup_thresh)) + return now - touch_ts; + + return 0; +} + +static int +watchdog_panic(struct notifier_block *this, unsigned long event, void *ptr) +{ + did_panic = 1; + + return NOTIFY_DONE; +} + +static struct notifier_block panic_block = { + .notifier_call = watchdog_panic, +}; + +#ifdef CONFIG_PERF_EVENTS_NMI +static struct perf_event_attr wd_hw_attr = { + .type = PERF_TYPE_HARDWARE, + .config = PERF_COUNT_HW_CPU_CYCLES, + .size = sizeof(struct perf_event_attr), + .pinned = 1, + .disabled = 1, +}; + +/* Callback function for perf event subsystem */ +void watchdog_overflow_callback(struct perf_event *event, int nmi, + struct perf_sample_data *data, + struct pt_regs *regs) +{ + int this_cpu = smp_processor_id(); + unsigned long touch_ts = per_cpu(watchdog_touch_ts, this_cpu); + + if (touch_ts == 0) { + __touch_watchdog(); + return; + } + + /* check for a hardlockup + * This is done by making sure our timer interrupt + * is incrementing. The timer interrupt should have + * fired multiple times before we overflow'd. If it hasn't + * then this is a good indication the cpu is stuck + */ + if (is_hardlockup(this_cpu)) { + /* only print hardlockups once */ + if (__get_cpu_var(hard_watchdog_warn) == true) + return; + + if (hardlockup_panic) + panic("Watchdog detected hard LOCKUP on cpu %d", this_cpu); + else + WARN(1, "Watchdog detected hard LOCKUP on cpu %d", this_cpu); + + __get_cpu_var(hard_watchdog_warn) = true; + return; + } + + __get_cpu_var(hard_watchdog_warn) = false; + return; +} +static void watchdog_interrupt_count(void) +{ + __get_cpu_var(hrtimer_interrupts)++; +} +#else +static inline void watchdog_interrupt_count(void) { return; } +#endif /* CONFIG_PERF_EVENTS_NMI */ + +/* watchdog kicker functions */ +static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) +{ + int this_cpu = smp_processor_id(); + unsigned long touch_ts = __get_cpu_var(watchdog_touch_ts); + struct pt_regs *regs = get_irq_regs(); + int duration; + + /* kick the hardlockup detector */ + watchdog_interrupt_count(); + + /* kick the softlockup detector */ + wake_up_process(__get_cpu_var(softlockup_watchdog)); + + /* .. and repeat */ + hrtimer_forward_now(hrtimer, ns_to_ktime(get_sample_period())); + + if (touch_ts == 0) { + if (unlikely(per_cpu(softlockup_touch_sync, this_cpu))) { + /* + * If the time stamp was touched atomically + * make sure the scheduler tick is up to date. + */ + per_cpu(softlockup_touch_sync, this_cpu) = false; + sched_clock_tick(); + } + __touch_watchdog(); + return HRTIMER_RESTART; + } + + /* check for a softlockup + * This is done by making sure a high priority task is + * being scheduled. The task touches the watchdog to + * indicate it is getting cpu time. If it hasn't then + * this is a good indication some task is hogging the cpu + */ + duration = is_softlockup(touch_ts, this_cpu); + if (unlikely(duration)) { + /* only warn once */ + if (__get_cpu_var(soft_watchdog_warn) == true) + return HRTIMER_RESTART; + + printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n", + this_cpu, duration, + current->comm, task_pid_nr(current)); + print_modules(); + print_irqtrace_events(current); + if (regs) + show_regs(regs); + else + dump_stack(); + + if (softlockup_panic) + panic("softlockup: hung tasks"); + __get_cpu_var(soft_watchdog_warn) = true; + } else + __get_cpu_var(soft_watchdog_warn) = false; + + return HRTIMER_RESTART; +} + + +/* + * The watchdog thread - touches the timestamp. + */ +static int watchdog(void *__bind_cpu) +{ + struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; + struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, (unsigned long)__bind_cpu); + + sched_setscheduler(current, SCHED_FIFO, ¶m); + + /* initialize timestamp */ + __touch_watchdog(); + + /* kick off the timer for the hardlockup detector */ + /* done here because hrtimer_start can only pin to smp_processor_id() */ + hrtimer_start(hrtimer, ns_to_ktime(get_sample_period()), + HRTIMER_MODE_REL_PINNED); + + set_current_state(TASK_INTERRUPTIBLE); + /* + * Run briefly once per second to reset the softlockup timestamp. + * If this gets delayed for more than 60 seconds then the + * debug-printout triggers in softlockup_tick(). + */ + while (!kthread_should_stop()) { + __touch_watchdog(); + schedule(); + + if (kthread_should_stop()) + break; + + set_current_state(TASK_INTERRUPTIBLE); + } + __set_current_state(TASK_RUNNING); + + return 0; +} + + +#ifdef CONFIG_PERF_EVENTS_NMI +static int watchdog_nmi_enable(int cpu) +{ + struct perf_event_attr *wd_attr; + struct perf_event *event = per_cpu(watchdog_ev, cpu); + + /* is it already setup and enabled? */ + if (event && event->state > PERF_EVENT_STATE_OFF) + goto out; + + /* it is setup but not enabled */ + if (event != NULL) + goto out_enable; + + /* Try to register using hardware perf events */ + wd_attr = &wd_hw_attr; + wd_attr->sample_period = hw_nmi_get_sample_period(); + event = perf_event_create_kernel_counter(wd_attr, cpu, -1, watchdog_overflow_callback); + if (!IS_ERR(event)) { + printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n"); + goto out_save; + } + + printk(KERN_ERR "NMI watchdog failed to create perf event on cpu%i: %p\n", cpu, event); + return -1; + + /* success path */ +out_save: + per_cpu(watchdog_ev, cpu) = event; +out_enable: + perf_event_enable(per_cpu(watchdog_ev, cpu)); +out: + return 0; +} + +static void watchdog_nmi_disable(int cpu) +{ + struct perf_event *event = per_cpu(watchdog_ev, cpu); + + if (event) { + perf_event_disable(event); + per_cpu(watchdog_ev, cpu) = NULL; + + /* should be in cleanup, but blocks oprofile */ + perf_event_release_kernel(event); + } + return; +} +#else +static int watchdog_nmi_enable(int cpu) { return 0; } +static void watchdog_nmi_disable(int cpu) { return; } +#endif /* CONFIG_PERF_EVENTS_NMI */ + +/* prepare/enable/disable routines */ +static int watchdog_prepare_cpu(int cpu) +{ + struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu); + + WARN_ON(per_cpu(softlockup_watchdog, cpu)); + hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL); + hrtimer->function = watchdog_timer_fn; + + return 0; +} + +static int watchdog_enable(int cpu) +{ + struct task_struct *p = per_cpu(softlockup_watchdog, cpu); + + /* enable the perf event */ + if (watchdog_nmi_enable(cpu) != 0) + return -1; + + /* create the watchdog thread */ + if (!p) { + p = kthread_create(watchdog, (void *)(unsigned long)cpu, "watchdog/%d", cpu); + if (IS_ERR(p)) { + printk(KERN_ERR "softlockup watchdog for %i failed\n", cpu); + return -1; + } + kthread_bind(p, cpu); + per_cpu(watchdog_touch_ts, cpu) = 0; + per_cpu(softlockup_watchdog, cpu) = p; + wake_up_process(p); + } + + return 0; +} + +static void watchdog_disable(int cpu) +{ + struct task_struct *p = per_cpu(softlockup_watchdog, cpu); + struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, cpu); + + /* + * cancel the timer first to stop incrementing the stats + * and waking up the kthread + */ + hrtimer_cancel(hrtimer); + + /* disable the perf event */ + watchdog_nmi_disable(cpu); + + /* stop the watchdog thread */ + if (p) { + per_cpu(softlockup_watchdog, cpu) = NULL; + kthread_stop(p); + } + + /* if any cpu succeeds, watchdog is considered enabled for the system */ + watchdog_enabled = 1; +} + +static void watchdog_enable_all_cpus(void) +{ + int cpu; + int result; + + for_each_online_cpu(cpu) + result += watchdog_enable(cpu); + + if (result) + printk(KERN_ERR "watchdog: failed to be enabled on some cpus\n"); + +} + +static void watchdog_disable_all_cpus(void) +{ + int cpu; + + for_each_online_cpu(cpu) + watchdog_disable(cpu); + + /* if all watchdogs are disabled, then they are disabled for the system */ + watchdog_enabled = 0; +} + + +/* sysctl functions */ +#ifdef CONFIG_SYSCTL +/* + * proc handler for /proc/sys/kernel/nmi_watchdog + */ + +int proc_dowatchdog_enabled(struct ctl_table *table, int write, + void __user *buffer, size_t *length, loff_t *ppos) +{ + proc_dointvec(table, write, buffer, length, ppos); + + if (watchdog_enabled) + watchdog_enable_all_cpus(); + else + watchdog_disable_all_cpus(); + return 0; +} + +int proc_dowatchdog_thresh(struct ctl_table *table, int write, + void __user *buffer, + size_t *lenp, loff_t *ppos) +{ + return proc_dointvec_minmax(table, write, buffer, lenp, ppos); +} + +/* stub functions */ +int proc_dosoftlockup_thresh(struct ctl_table *table, int write, + void __user *buffer, + size_t *lenp, loff_t *ppos) +{ + return proc_dowatchdog_thresh(table, write, buffer, lenp, ppos); +} +/* end of stub functions */ +#endif /* CONFIG_SYSCTL */ + + +/* + * Create/destroy watchdog threads as CPUs come and go: + */ +static int __cpuinit +cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) +{ + int hotcpu = (unsigned long)hcpu; + + switch (action) { + case CPU_UP_PREPARE: + case CPU_UP_PREPARE_FROZEN: + if (watchdog_prepare_cpu(hotcpu)) + return NOTIFY_BAD; + break; + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + if (watchdog_enable(hotcpu)) + return NOTIFY_BAD; + break; +#ifdef CONFIG_HOTPLUG_CPU + case CPU_UP_CANCELED: + case CPU_UP_CANCELED_FROZEN: + watchdog_disable(hotcpu); + break; + case CPU_DEAD: + case CPU_DEAD_FROZEN: + watchdog_disable(hotcpu); + break; +#endif /* CONFIG_HOTPLUG_CPU */ + } + return NOTIFY_OK; +} + +static struct notifier_block __cpuinitdata cpu_nfb = { + .notifier_call = cpu_callback +}; + +static int __init spawn_watchdog_task(void) +{ + void *cpu = (void *)(long)smp_processor_id(); + int err; + + if (no_watchdog) + return 0; + + err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); + WARN_ON(err == NOTIFY_BAD); + + cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); + register_cpu_notifier(&cpu_nfb); + + atomic_notifier_chain_register(&panic_notifier_list, &panic_block); + + return 0; +} +early_initcall(spawn_watchdog_task); diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 220ae60..49e285d 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -153,7 +153,7 @@ config DEBUG_SHIRQ points; some don't and need to be caught. config DETECT_SOFTLOCKUP - bool "Detect Soft Lockups" + bool depends on DEBUG_KERNEL && !S390 default y help @@ -171,17 +171,27 @@ config DETECT_SOFTLOCKUP can be detected via the NMI-watchdog, on platforms that support it.) -config NMI_WATCHDOG - bool "Detect Hard Lockups with an NMI Watchdog" - depends on DEBUG_KERNEL && PERF_EVENTS && PERF_EVENTS_NMI +config LOCKUP_DETECTOR + bool "Detect Hard and Soft Lockups" + depends on DEBUG_KERNEL + default DETECT_SOFTLOCKUP help - Say Y here to enable the kernel to use the NMI as a watchdog - to detect hard lockups. This is useful when a cpu hangs for no - reason but can still respond to NMIs. A backtrace is displayed - for reviewing and reporting. + Say Y here to enable the kernel to act as a watchdog to detect + hard and soft lockups. + + Softlockups are bugs that cause the kernel to loop in kernel + mode for more than 60 seconds, without giving other tasks a + chance to run. The current stack trace is displayed upon + detection and the system will stay locked up. + + Hardlockups are bugs that cause the CPU to loop in kernel mode + for more than 60 seconds, without letting other interrupts have a + chance to run. The current stack trace is displayed upon detection + and the system will stay locked up. - The overhead should be minimal, just an extra NMI every few - seconds. + The overhead should be minimal. A periodic hrtimer runs to + generate interrupts and kick the watchdog task every 10-12 seconds. + An NMI is generated every 60 seconds or so to check for hardlockups. config BOOTPARAM_SOFTLOCKUP_PANIC bool "Panic (Reboot) On Soft Lockups" -- cgit v0.10.2 From 332fbdbca3f7716c5620970755ae054d213bcc4e Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:45 -0400 Subject: lockup_detector: Touch_softlockup cleanups and softlockup_tick removal Just some code cleanup to make touch_softlockup clearer and remove the softlockup_tick function as it is no longer needed. Also remove the /proc softlockup_thres call as it has been changed to watchdog_thres. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-3-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/include/linux/sched.h b/include/linux/sched.h index 37efe8f..33f9b2a 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -312,19 +312,15 @@ extern void scheduler_tick(void); extern void sched_show_task(struct task_struct *p); #ifdef CONFIG_DETECT_SOFTLOCKUP -extern void softlockup_tick(void); extern void touch_softlockup_watchdog(void); extern void touch_softlockup_watchdog_sync(void); extern void touch_all_softlockup_watchdogs(void); -extern int proc_dosoftlockup_thresh(struct ctl_table *table, int write, - void __user *buffer, - size_t *lenp, loff_t *ppos); +extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, + void __user *buffer, + size_t *lenp, loff_t *ppos); extern unsigned int softlockup_panic; extern int softlockup_thresh; #else -static inline void softlockup_tick(void) -{ -} static inline void touch_softlockup_watchdog(void) { } @@ -346,12 +342,6 @@ extern int proc_dohung_task_timeout_secs(struct ctl_table *table, int write, size_t *lenp, loff_t *ppos); #endif -#ifdef CONFIG_LOCKUP_DETECTOR -extern int proc_dowatchdog_thresh(struct ctl_table *table, int write, - void __user *buffer, - size_t *lenp, loff_t *ppos); -#endif - /* Attach to any functions which should be ignored in wchan output. */ #define __sched __attribute__((__section__(".sched.text"))) diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 0f9adda..999bc3f 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -817,15 +817,6 @@ static struct ctl_table kern_table[] = { .extra1 = &zero, .extra2 = &one, }, - { - .procname = "softlockup_thresh", - .data = &softlockup_thresh, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = proc_dosoftlockup_thresh, - .extra1 = &neg_one, - .extra2 = &sixty, - }, #endif #ifdef CONFIG_DETECT_HUNG_TASK { diff --git a/kernel/timer.c b/kernel/timer.c index aeb6a54..e8de5eb 100644 --- a/kernel/timer.c +++ b/kernel/timer.c @@ -1225,7 +1225,6 @@ void run_local_timers(void) { hrtimer_run_queues(); raise_softirq(TIMER_SOFTIRQ); - softlockup_tick(); } /* diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 6b7fad8..f1541b7 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -119,13 +119,12 @@ static void __touch_watchdog(void) __get_cpu_var(watchdog_touch_ts) = get_timestamp(this_cpu); } -void touch_watchdog(void) +void touch_softlockup_watchdog(void) { __get_cpu_var(watchdog_touch_ts) = 0; } -EXPORT_SYMBOL(touch_watchdog); -void touch_all_watchdog(void) +void touch_all_softlockup_watchdogs(void) { int cpu; @@ -140,35 +139,16 @@ void touch_all_watchdog(void) void touch_nmi_watchdog(void) { - touch_watchdog(); + touch_softlockup_watchdog(); } EXPORT_SYMBOL(touch_nmi_watchdog); -void touch_all_nmi_watchdog(void) -{ - touch_all_watchdog(); -} - -void touch_softlockup_watchdog(void) -{ - touch_watchdog(); -} - -void touch_all_softlockup_watchdogs(void) -{ - touch_all_watchdog(); -} - void touch_softlockup_watchdog_sync(void) { __raw_get_cpu_var(softlockup_touch_sync) = true; __raw_get_cpu_var(watchdog_touch_ts) = 0; } -void softlockup_tick(void) -{ -} - #ifdef CONFIG_PERF_EVENTS_NMI /* watchdog detector functions */ static int is_hardlockup(int cpu) @@ -522,15 +502,6 @@ int proc_dowatchdog_thresh(struct ctl_table *table, int write, { return proc_dointvec_minmax(table, write, buffer, lenp, ppos); } - -/* stub functions */ -int proc_dosoftlockup_thresh(struct ctl_table *table, int write, - void __user *buffer, - size_t *lenp, loff_t *ppos) -{ - return proc_dowatchdog_thresh(table, write, buffer, lenp, ppos); -} -/* end of stub functions */ #endif /* CONFIG_SYSCTL */ -- cgit v0.10.2 From 2508ce1845a3b256798532b2c6b7997c2dc6533b Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:46 -0400 Subject: lockup_detector: Remove old softlockup code Now that is no longer compiled or used, just remove it. Also move some of the code wrapped with DETECT_SOFTLOCKUP to the LOCKUP_DETECTOR wrappers because that is the code that uses it now. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-4-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/kernel/softlockup.c b/kernel/softlockup.c deleted file mode 100644 index 4b493f6..0000000 --- a/kernel/softlockup.c +++ /dev/null @@ -1,293 +0,0 @@ -/* - * Detect Soft Lockups - * - * started by Ingo Molnar, Copyright (C) 2005, 2006 Red Hat, Inc. - * - * this code detects soft lockups: incidents in where on a CPU - * the kernel does not reschedule for 10 seconds or more. - */ -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include - -static DEFINE_SPINLOCK(print_lock); - -static DEFINE_PER_CPU(unsigned long, softlockup_touch_ts); /* touch timestamp */ -static DEFINE_PER_CPU(unsigned long, softlockup_print_ts); /* print timestamp */ -static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog); -static DEFINE_PER_CPU(bool, softlock_touch_sync); - -static int __read_mostly did_panic; -int __read_mostly softlockup_thresh = 60; - -/* - * Should we panic (and reboot, if panic_timeout= is set) when a - * soft-lockup occurs: - */ -unsigned int __read_mostly softlockup_panic = - CONFIG_BOOTPARAM_SOFTLOCKUP_PANIC_VALUE; - -static int __init softlockup_panic_setup(char *str) -{ - softlockup_panic = simple_strtoul(str, NULL, 0); - - return 1; -} -__setup("softlockup_panic=", softlockup_panic_setup); - -static int -softlock_panic(struct notifier_block *this, unsigned long event, void *ptr) -{ - did_panic = 1; - - return NOTIFY_DONE; -} - -static struct notifier_block panic_block = { - .notifier_call = softlock_panic, -}; - -/* - * Returns seconds, approximately. We don't need nanosecond - * resolution, and we don't need to waste time with a big divide when - * 2^30ns == 1.074s. - */ -static unsigned long get_timestamp(int this_cpu) -{ - return cpu_clock(this_cpu) >> 30LL; /* 2^30 ~= 10^9 */ -} - -static void __touch_softlockup_watchdog(void) -{ - int this_cpu = raw_smp_processor_id(); - - __raw_get_cpu_var(softlockup_touch_ts) = get_timestamp(this_cpu); -} - -void touch_softlockup_watchdog(void) -{ - __raw_get_cpu_var(softlockup_touch_ts) = 0; -} -EXPORT_SYMBOL(touch_softlockup_watchdog); - -void touch_softlockup_watchdog_sync(void) -{ - __raw_get_cpu_var(softlock_touch_sync) = true; - __raw_get_cpu_var(softlockup_touch_ts) = 0; -} - -void touch_all_softlockup_watchdogs(void) -{ - int cpu; - - /* Cause each CPU to re-update its timestamp rather than complain */ - for_each_online_cpu(cpu) - per_cpu(softlockup_touch_ts, cpu) = 0; -} -EXPORT_SYMBOL(touch_all_softlockup_watchdogs); - -int proc_dosoftlockup_thresh(struct ctl_table *table, int write, - void __user *buffer, - size_t *lenp, loff_t *ppos) -{ - touch_all_softlockup_watchdogs(); - return proc_dointvec_minmax(table, write, buffer, lenp, ppos); -} - -/* - * This callback runs from the timer interrupt, and checks - * whether the watchdog thread has hung or not: - */ -void softlockup_tick(void) -{ - int this_cpu = smp_processor_id(); - unsigned long touch_ts = per_cpu(softlockup_touch_ts, this_cpu); - unsigned long print_ts; - struct pt_regs *regs = get_irq_regs(); - unsigned long now; - - /* Is detection switched off? */ - if (!per_cpu(softlockup_watchdog, this_cpu) || softlockup_thresh <= 0) { - /* Be sure we don't false trigger if switched back on */ - if (touch_ts) - per_cpu(softlockup_touch_ts, this_cpu) = 0; - return; - } - - if (touch_ts == 0) { - if (unlikely(per_cpu(softlock_touch_sync, this_cpu))) { - /* - * If the time stamp was touched atomically - * make sure the scheduler tick is up to date. - */ - per_cpu(softlock_touch_sync, this_cpu) = false; - sched_clock_tick(); - } - __touch_softlockup_watchdog(); - return; - } - - print_ts = per_cpu(softlockup_print_ts, this_cpu); - - /* report at most once a second */ - if (print_ts == touch_ts || did_panic) - return; - - /* do not print during early bootup: */ - if (unlikely(system_state != SYSTEM_RUNNING)) { - __touch_softlockup_watchdog(); - return; - } - - now = get_timestamp(this_cpu); - - /* - * Wake up the high-prio watchdog task twice per - * threshold timespan. - */ - if (time_after(now - softlockup_thresh/2, touch_ts)) - wake_up_process(per_cpu(softlockup_watchdog, this_cpu)); - - /* Warn about unreasonable delays: */ - if (time_before_eq(now - softlockup_thresh, touch_ts)) - return; - - per_cpu(softlockup_print_ts, this_cpu) = touch_ts; - - spin_lock(&print_lock); - printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %lus! [%s:%d]\n", - this_cpu, now - touch_ts, - current->comm, task_pid_nr(current)); - print_modules(); - print_irqtrace_events(current); - if (regs) - show_regs(regs); - else - dump_stack(); - spin_unlock(&print_lock); - - if (softlockup_panic) - panic("softlockup: hung tasks"); -} - -/* - * The watchdog thread - runs every second and touches the timestamp. - */ -static int watchdog(void *__bind_cpu) -{ - struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; - - sched_setscheduler(current, SCHED_FIFO, ¶m); - - /* initialize timestamp */ - __touch_softlockup_watchdog(); - - set_current_state(TASK_INTERRUPTIBLE); - /* - * Run briefly once per second to reset the softlockup timestamp. - * If this gets delayed for more than 60 seconds then the - * debug-printout triggers in softlockup_tick(). - */ - while (!kthread_should_stop()) { - __touch_softlockup_watchdog(); - schedule(); - - if (kthread_should_stop()) - break; - - set_current_state(TASK_INTERRUPTIBLE); - } - __set_current_state(TASK_RUNNING); - - return 0; -} - -/* - * Create/destroy watchdog threads as CPUs come and go: - */ -static int __cpuinit -cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) -{ - int hotcpu = (unsigned long)hcpu; - struct task_struct *p; - - switch (action) { - case CPU_UP_PREPARE: - case CPU_UP_PREPARE_FROZEN: - BUG_ON(per_cpu(softlockup_watchdog, hotcpu)); - p = kthread_create(watchdog, hcpu, "watchdog/%d", hotcpu); - if (IS_ERR(p)) { - printk(KERN_ERR "watchdog for %i failed\n", hotcpu); - return NOTIFY_BAD; - } - per_cpu(softlockup_touch_ts, hotcpu) = 0; - per_cpu(softlockup_watchdog, hotcpu) = p; - kthread_bind(p, hotcpu); - break; - case CPU_ONLINE: - case CPU_ONLINE_FROZEN: - wake_up_process(per_cpu(softlockup_watchdog, hotcpu)); - break; -#ifdef CONFIG_HOTPLUG_CPU - case CPU_UP_CANCELED: - case CPU_UP_CANCELED_FROZEN: - if (!per_cpu(softlockup_watchdog, hotcpu)) - break; - /* Unbind so it can run. Fall thru. */ - kthread_bind(per_cpu(softlockup_watchdog, hotcpu), - cpumask_any(cpu_online_mask)); - case CPU_DEAD: - case CPU_DEAD_FROZEN: - p = per_cpu(softlockup_watchdog, hotcpu); - per_cpu(softlockup_watchdog, hotcpu) = NULL; - kthread_stop(p); - break; -#endif /* CONFIG_HOTPLUG_CPU */ - } - return NOTIFY_OK; -} - -static struct notifier_block __cpuinitdata cpu_nfb = { - .notifier_call = cpu_callback -}; - -static int __initdata nosoftlockup; - -static int __init nosoftlockup_setup(char *str) -{ - nosoftlockup = 1; - return 1; -} -__setup("nosoftlockup", nosoftlockup_setup); - -static int __init spawn_softlockup_task(void) -{ - void *cpu = (void *)(long)smp_processor_id(); - int err; - - if (nosoftlockup) - return 0; - - err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); - if (err == NOTIFY_BAD) { - BUG(); - return 1; - } - cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); - register_cpu_notifier(&cpu_nfb); - - atomic_notifier_chain_register(&panic_notifier_list, &panic_block); - - return 0; -} -early_initcall(spawn_softlockup_task); diff --git a/kernel/sysctl.c b/kernel/sysctl.c index 999bc3f..04bcd8a 100644 --- a/kernel/sysctl.c +++ b/kernel/sysctl.c @@ -108,7 +108,7 @@ extern int blk_iopoll_enabled; #endif /* Constants used for minimum and maximum */ -#ifdef CONFIG_DETECT_SOFTLOCKUP +#ifdef CONFIG_LOCKUP_DETECTOR static int sixty = 60; static int neg_one = -1; #endif @@ -703,6 +703,15 @@ static struct ctl_table kern_table[] = { .extra1 = &neg_one, .extra2 = &sixty, }, + { + .procname = "softlockup_panic", + .data = &softlockup_panic, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec_minmax, + .extra1 = &zero, + .extra2 = &one, + }, #endif #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR) { @@ -807,17 +816,6 @@ static struct ctl_table kern_table[] = { .proc_handler = proc_dointvec, }, #endif -#ifdef CONFIG_DETECT_SOFTLOCKUP - { - .procname = "softlockup_panic", - .data = &softlockup_panic, - .maxlen = sizeof(int), - .mode = 0644, - .proc_handler = proc_dointvec_minmax, - .extra1 = &zero, - .extra2 = &one, - }, -#endif #ifdef CONFIG_DETECT_HUNG_TASK { .procname = "hung_task_panic", -- cgit v0.10.2 From f69bcf60c3f17aa367e16eef7bc6ab001ea6d58a Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:47 -0400 Subject: lockup_detector: Remove nmi_watchdog.c file This file migrated to kernel/watchdog.c and then combined with kernel/softlockup.c. As a result kernel/nmi_watchdog.c is no longer needed. Just remove it. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-5-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/kernel/nmi_watchdog.c b/kernel/nmi_watchdog.c deleted file mode 100644 index a79d211..0000000 --- a/kernel/nmi_watchdog.c +++ /dev/null @@ -1,259 +0,0 @@ -/* - * Detect Hard Lockups using the NMI - * - * started by Don Zickus, Copyright (C) 2010 Red Hat, Inc. - * - * this code detects hard lockups: incidents in where on a CPU - * the kernel does not respond to anything except NMI. - * - * Note: Most of this code is borrowed heavily from softlockup.c, - * so thanks to Ingo for the initial implementation. - * Some chunks also taken from arch/x86/kernel/apic/nmi.c, thanks - * to those contributors as well. - */ - -#include -#include -#include -#include -#include -#include -#include -#include -#include -#include - -#include -#include - -static DEFINE_PER_CPU(struct perf_event *, nmi_watchdog_ev); -static DEFINE_PER_CPU(int, nmi_watchdog_touch); -static DEFINE_PER_CPU(long, alert_counter); - -static int panic_on_timeout; - -void touch_nmi_watchdog(void) -{ - __raw_get_cpu_var(nmi_watchdog_touch) = 1; - touch_softlockup_watchdog(); -} -EXPORT_SYMBOL(touch_nmi_watchdog); - -void touch_all_nmi_watchdog(void) -{ - int cpu; - - for_each_online_cpu(cpu) - per_cpu(nmi_watchdog_touch, cpu) = 1; - touch_softlockup_watchdog(); -} - -static int __init setup_nmi_watchdog(char *str) -{ - if (!strncmp(str, "panic", 5)) { - panic_on_timeout = 1; - str = strchr(str, ','); - if (!str) - return 1; - ++str; - } - return 1; -} -__setup("nmi_watchdog=", setup_nmi_watchdog); - -struct perf_event_attr wd_hw_attr = { - .type = PERF_TYPE_HARDWARE, - .config = PERF_COUNT_HW_CPU_CYCLES, - .size = sizeof(struct perf_event_attr), - .pinned = 1, - .disabled = 1, -}; - -struct perf_event_attr wd_sw_attr = { - .type = PERF_TYPE_SOFTWARE, - .config = PERF_COUNT_SW_CPU_CLOCK, - .size = sizeof(struct perf_event_attr), - .pinned = 1, - .disabled = 1, -}; - -void wd_overflow(struct perf_event *event, int nmi, - struct perf_sample_data *data, - struct pt_regs *regs) -{ - int cpu = smp_processor_id(); - int touched = 0; - - if (__get_cpu_var(nmi_watchdog_touch)) { - per_cpu(nmi_watchdog_touch, cpu) = 0; - touched = 1; - } - - /* check to see if the cpu is doing anything */ - if (!touched && hw_nmi_is_cpu_stuck(regs)) { - /* - * Ayiee, looks like this CPU is stuck ... - * wait a few IRQs (5 seconds) before doing the oops ... - */ - per_cpu(alert_counter, cpu) += 1; - if (per_cpu(alert_counter, cpu) == 5) { - if (panic_on_timeout) - panic("NMI Watchdog detected LOCKUP on cpu %d", cpu); - else - WARN(1, "NMI Watchdog detected LOCKUP on cpu %d", cpu); - } - } else { - per_cpu(alert_counter, cpu) = 0; - } - - return; -} - -static int enable_nmi_watchdog(int cpu) -{ - struct perf_event *event; - struct perf_event_attr *wd_attr; - - event = per_cpu(nmi_watchdog_ev, cpu); - if (event && event->state > PERF_EVENT_STATE_OFF) - return 0; - - if (event == NULL) { - /* Try to register using hardware perf events first */ - wd_attr = &wd_hw_attr; - wd_attr->sample_period = hw_nmi_get_sample_period(); - event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); - if (IS_ERR(event)) { - /* hardware doesn't exist or not supported, fallback to software events */ - printk(KERN_INFO "nmi_watchdog: hardware not available, trying software events\n"); - wd_attr = &wd_sw_attr; - wd_attr->sample_period = NSEC_PER_SEC; - event = perf_event_create_kernel_counter(wd_attr, cpu, -1, wd_overflow); - if (IS_ERR(event)) { - printk(KERN_ERR "nmi watchdog failed to create perf event on %i: %p\n", cpu, event); - return -1; - } - } - per_cpu(nmi_watchdog_ev, cpu) = event; - } - perf_event_enable(per_cpu(nmi_watchdog_ev, cpu)); - return 0; -} - -static void disable_nmi_watchdog(int cpu) -{ - struct perf_event *event; - - event = per_cpu(nmi_watchdog_ev, cpu); - if (event) { - perf_event_disable(per_cpu(nmi_watchdog_ev, cpu)); - per_cpu(nmi_watchdog_ev, cpu) = NULL; - perf_event_release_kernel(event); - } -} - -#ifdef CONFIG_SYSCTL -/* - * proc handler for /proc/sys/kernel/nmi_watchdog - */ -int nmi_watchdog_enabled; - -int proc_nmi_enabled(struct ctl_table *table, int write, - void __user *buffer, size_t *length, loff_t *ppos) -{ - int cpu; - - if (!write) { - struct perf_event *event; - for_each_online_cpu(cpu) { - event = per_cpu(nmi_watchdog_ev, cpu); - if (event && event->state > PERF_EVENT_STATE_OFF) { - nmi_watchdog_enabled = 1; - break; - } - } - proc_dointvec(table, write, buffer, length, ppos); - return 0; - } - - touch_all_nmi_watchdog(); - proc_dointvec(table, write, buffer, length, ppos); - if (nmi_watchdog_enabled) { - for_each_online_cpu(cpu) - if (enable_nmi_watchdog(cpu)) { - printk(KERN_ERR "NMI watchdog failed configuration, " - " can not be enabled\n"); - } - } else { - for_each_online_cpu(cpu) - disable_nmi_watchdog(cpu); - } - return 0; -} - -#endif /* CONFIG_SYSCTL */ - -/* - * Create/destroy watchdog threads as CPUs come and go: - */ -static int __cpuinit -cpu_callback(struct notifier_block *nfb, unsigned long action, void *hcpu) -{ - int hotcpu = (unsigned long)hcpu; - - switch (action) { - case CPU_UP_PREPARE: - case CPU_UP_PREPARE_FROZEN: - per_cpu(nmi_watchdog_touch, hotcpu) = 0; - break; - case CPU_ONLINE: - case CPU_ONLINE_FROZEN: - if (enable_nmi_watchdog(hotcpu)) - return NOTIFY_BAD; - break; -#ifdef CONFIG_HOTPLUG_CPU - case CPU_UP_CANCELED: - case CPU_UP_CANCELED_FROZEN: - disable_nmi_watchdog(hotcpu); - case CPU_DEAD: - case CPU_DEAD_FROZEN: - break; -#endif /* CONFIG_HOTPLUG_CPU */ - } - return NOTIFY_OK; -} - -static struct notifier_block __cpuinitdata cpu_nfb = { - .notifier_call = cpu_callback -}; - -static int __initdata nonmi_watchdog; - -static int __init nonmi_watchdog_setup(char *str) -{ - nonmi_watchdog = 1; - return 1; -} -__setup("nonmi_watchdog", nonmi_watchdog_setup); - -static int __init spawn_nmi_watchdog_task(void) -{ - void *cpu = (void *)(long)smp_processor_id(); - int err; - - if (nonmi_watchdog) - return 0; - - printk(KERN_INFO "NMI watchdog enabled, takes one hw-pmu counter.\n"); - - err = cpu_callback(&cpu_nfb, CPU_UP_PREPARE, cpu); - if (err == NOTIFY_BAD) { - BUG(); - return 1; - } - cpu_callback(&cpu_nfb, CPU_ONLINE, cpu); - register_cpu_notifier(&cpu_nfb); - - return 0; -} -early_initcall(spawn_nmi_watchdog_task); -- cgit v0.10.2 From 7cbb7e7fa46f6e5229438ac9e4a5c72ec0d53e0b Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:48 -0400 Subject: x86: Move trigger_all_cpu_backtrace to its own die_notifier As part of the transition of the nmi watchdog to something more generic, the trigger_all_cpu_backtrace code is getting left behind. Put it in its own die_notifier so it can still be used. V2: - use arch_spin_locks Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-6-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 79425f9..8c3edfb 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -17,6 +17,10 @@ #include #include #include +#include +#include +#include + #include #include @@ -54,20 +58,6 @@ int hw_nmi_is_cpu_stuck(struct pt_regs *regs) unsigned int sum; int cpu = smp_processor_id(); - /* FIXME: cheap hack for this check, probably should get its own - * die_notifier handler - */ - if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { - static DEFINE_SPINLOCK(lock); /* Serialise the printks */ - - spin_lock(&lock); - printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); - show_regs(regs); - dump_stack(); - spin_unlock(&lock); - cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); - } - /* if we are doing an mce, just assume the cpu is not stuck */ /* Could check oops_in_progress here too, but it's safer not to */ if (mce_in_progress()) @@ -109,6 +99,53 @@ void arch_trigger_all_cpu_backtrace(void) mdelay(1); } } + +static int __kprobes +arch_trigger_all_cpu_backtrace_handler(struct notifier_block *self, + unsigned long cmd, void *__args) +{ + struct die_args *args = __args; + struct pt_regs *regs; + int cpu = smp_processor_id(); + + switch (cmd) { + case DIE_NMI: + case DIE_NMI_IPI: + break; + + default: + return NOTIFY_DONE; + } + + regs = args->regs; + + if (cpumask_test_cpu(cpu, to_cpumask(backtrace_mask))) { + static arch_spinlock_t lock = __ARCH_SPIN_LOCK_UNLOCKED; + + arch_spin_lock(&lock); + printk(KERN_WARNING "NMI backtrace for cpu %d\n", cpu); + show_regs(regs); + dump_stack(); + arch_spin_unlock(&lock); + cpumask_clear_cpu(cpu, to_cpumask(backtrace_mask)); + return NOTIFY_STOP; + } + + return NOTIFY_DONE; +} + +static __read_mostly struct notifier_block backtrace_notifier = { + .notifier_call = arch_trigger_all_cpu_backtrace_handler, + .next = NULL, + .priority = 1 +}; + +static int __init register_trigger_all_cpu_backtrace(void) +{ + register_die_notifier(&backtrace_notifier); + return 0; +} +early_initcall(register_trigger_all_cpu_backtrace); #endif /* STUB calls to mimic old nmi_watchdog behaviour */ -- cgit v0.10.2 From 10f9014912a2b1cb59c39cdea777e6d9afa8f17e Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:49 -0400 Subject: x86: Cleanup hw_nmi.c cruft The design of the hardlockup watchdog has changed and cruft was left behind in the hw_nmi.c file. Just remove the code that isn't used anymore. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-7-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 8c3edfb..3b40082 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -9,74 +9,16 @@ * */ -#include -#include #include -#include -#include -#include -#include -#include #include #include #include - - #include #include /* For reliability, we're prepared to waste bits here. */ static DECLARE_BITMAP(backtrace_mask, NR_CPUS) __read_mostly; -static DEFINE_PER_CPU(unsigned, last_irq_sum); - -/* - * Take the local apic timer and PIT/HPET into account. We don't - * know which one is active, when we have highres/dyntick on - */ -static inline unsigned int get_timer_irqs(int cpu) -{ - unsigned int irqs = per_cpu(irq_stat, cpu).irq0_irqs; - -#if defined(CONFIG_X86_LOCAL_APIC) - irqs += per_cpu(irq_stat, cpu).apic_timer_irqs; -#endif - - return irqs; -} - -static inline int mce_in_progress(void) -{ -#if defined(CONFIG_X86_MCE) - return atomic_read(&mce_entry) > 0; -#endif - return 0; -} - -int hw_nmi_is_cpu_stuck(struct pt_regs *regs) -{ - unsigned int sum; - int cpu = smp_processor_id(); - - /* if we are doing an mce, just assume the cpu is not stuck */ - /* Could check oops_in_progress here too, but it's safer not to */ - if (mce_in_progress()) - return 0; - - /* We determine if the cpu is stuck by checking whether any - * interrupts have happened since we last checked. Of course - * an nmi storm could create false positives, but the higher - * level logic should account for that - */ - sum = get_timer_irqs(cpu); - if (__get_cpu_var(last_irq_sum) == sum) { - return 1; - } else { - __get_cpu_var(last_irq_sum) = sum; - return 0; - } -} - u64 hw_nmi_get_sample_period(void) { return (u64)(cpu_khz) * 1000 * 60; -- cgit v0.10.2 From d7c547335fa6b0090fa09c46ea0e965ac273a27e Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 7 May 2010 17:11:51 -0400 Subject: lockup_detector: Separate touch_nmi_watchdog code path from touch_watchdog When I combined the nmi_watchdog (hardlockup) and softlockup code, I also combined the paths the touch_watchdog and touch_nmi_watchdog took. This may not be the best idea as pointed out by Frederic W., that the touch_watchdog case probably should not reset the hardlockup count. Therefore the patch below falls back to the previous idea of keeping the touch_nmi_watchdog a superset of the touch_watchdog case. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap LKML-Reference: <1273266711-18706-9-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/kernel/watchdog.c b/kernel/watchdog.c index f1541b7..57b8e2c 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -31,6 +31,7 @@ int watchdog_enabled; int __read_mostly softlockup_thresh = 60; static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts); +static DEFINE_PER_CPU(bool, watchdog_nmi_touch); static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog); static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer); static DEFINE_PER_CPU(bool, softlockup_touch_sync); @@ -139,6 +140,7 @@ void touch_all_softlockup_watchdogs(void) void touch_nmi_watchdog(void) { + __get_cpu_var(watchdog_nmi_touch) = true; touch_softlockup_watchdog(); } EXPORT_SYMBOL(touch_nmi_watchdog); @@ -201,10 +203,9 @@ void watchdog_overflow_callback(struct perf_event *event, int nmi, struct pt_regs *regs) { int this_cpu = smp_processor_id(); - unsigned long touch_ts = per_cpu(watchdog_touch_ts, this_cpu); - if (touch_ts == 0) { - __touch_watchdog(); + if (__get_cpu_var(watchdog_nmi_touch) == true) { + __get_cpu_var(watchdog_nmi_touch) = false; return; } -- cgit v0.10.2 From 89d7ce2a2178e7f562f608b466a18c8c2ece87af Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 May 2010 00:27:20 +0200 Subject: lockup_detector: Make BOOTPARAM_SOFTLOCKUP_PANIC depend on LOCKUP_DETECTOR Panic on softlockups was still depending on the softlockup detector. But the latter has been merged into the lockup detector now. Let's update this config dependency. Signed-off-by: Frederic Weisbecker Cc: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 49e285d..755713a 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -195,7 +195,7 @@ config LOCKUP_DETECTOR config BOOTPARAM_SOFTLOCKUP_PANIC bool "Panic (Reboot) On Soft Lockups" - depends on DETECT_SOFTLOCKUP + depends on LOCKUP_DETECTOR help Say Y here to enable the kernel to panic on "soft lockups", which are bugs that cause the kernel to loop in kernel -- cgit v0.10.2 From 19cc36c0f0457e5c6629ec24036fbbe8255c88ec Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Thu, 13 May 2010 02:30:49 +0200 Subject: lockup_detector: Fix forgotten config conversion Fix forgotten CONFIG_DETECT_SOFTLOCKUP -> CONFIG_LOCKUP_DETECTOR in sched.h Fixes: arch/x86/built-in.o: In function `touch_nmi_watchdog': (.text+0x1bd59): undefined reference to `touch_softlockup_watchdog' kernel/built-in.o: In function `show_state_filter': (.text+0x10d01): undefined reference to `touch_all_softlockup_watchdogs' kernel/built-in.o: In function `sched_clock_idle_wakeup_event': (.text+0x362f9): undefined reference to `touch_softlockup_watchdog' kernel/built-in.o: In function `timekeeping_resume': timekeeping.c:(.text+0x38757): undefined reference to `touch_softlockup_watchdog' kernel/built-in.o: In function `tick_nohz_handler': tick-sched.c:(.text+0x3e5b9): undefined reference to `touch_softlockup_watchdog' kernel/built-in.o: In function `tick_sched_timer': tick-sched.c:(.text+0x3e671): undefined reference to `touch_softlockup_watchdog' kernel/built-in.o: In function `tick_check_idle': (.text+0x3e90b): undefined reference to `touch_softlockup_watchdog' Signed-off-by: Frederic Weisbecker Cc: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov Cc: Eric Paris Cc: Randy Dunlap diff --git a/include/linux/sched.h b/include/linux/sched.h index 33f9b2a..3958e0c 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -311,7 +311,7 @@ extern void scheduler_tick(void); extern void sched_show_task(struct task_struct *p); -#ifdef CONFIG_DETECT_SOFTLOCKUP +#ifdef CONFIG_LOCKUP_DETECTOR extern void touch_softlockup_watchdog(void); extern void touch_softlockup_watchdog_sync(void); extern void touch_all_softlockup_watchdogs(void); -- cgit v0.10.2 From 0167c781907fcdc3e1f144ef5ce31d402c91eb94 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 13 May 2010 08:53:33 +0200 Subject: watchdog: Export touch_softlockup_watchdog There are modules that rely on it: ERROR: "touch_softlockup_watchdog" [drivers/video/nvidia/nvidiafb.ko] undefined! Cc: Frederic Weisbecker Cc: Don Zickus Cc: Peter Zijlstra Cc: Cyrill Gorcunov LKML-Reference: <1273713674-8434-1-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 57b8e2c..be5e74e 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -124,6 +124,7 @@ void touch_softlockup_watchdog(void) { __get_cpu_var(watchdog_touch_ts) = 0; } +EXPORT_SYMBOL(touch_softlockup_watchdog); void touch_all_softlockup_watchdogs(void) { -- cgit v0.10.2 From 5e85391b3badd3f0e50ebdd0cafe0202a979f73a Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Thu, 13 May 2010 09:12:39 +0200 Subject: x86, watchdog: Fix build error in hw_nmi.c On some configs the following build error triggers: arch/x86/kernel/apic/hw_nmi.c:35: error: 'apic' undeclared (first use in this function) arch/x86/kernel/apic/hw_nmi.c:35: error: (Each undeclared identifier is reported only once arch/x86/kernel/apic/hw_nmi.c:35: error: for each function it appears in.) Because asm/apic.h was only included implicitly. Include it explicitly. Cc: Frederic Weisbecker Cc: Don Zickus Cc: Peter Zijlstra Cc: Cyrill Gorcunov LKML-Reference: <1273713674-8434-1-git-send-regression-fweisbec@gmail.com> Signed-off-by: Ingo Molnar diff --git a/arch/x86/kernel/apic/hw_nmi.c b/arch/x86/kernel/apic/hw_nmi.c index 3b40082..cefd694 100644 --- a/arch/x86/kernel/apic/hw_nmi.c +++ b/arch/x86/kernel/apic/hw_nmi.c @@ -8,6 +8,7 @@ * Bits copied from original nmi.c file * */ +#include #include #include -- cgit v0.10.2 From e16bb1d7fe07609bc8b0e4c043eff2f47ada78d8 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 15 May 2010 22:30:22 +0200 Subject: lockup_detector: Update some config We kept CONFIG_DETECT_SOFTLOCKUP around for compatibility with older configs. But it was enabled by default if CONFIG_DEBUG_KERNEL. So if we want to enable CONFIG_LOCKUP_DETECTOR on configs that had CONFIG_DETECT_SOFTLOCKUP, all we need is to have the same enabling by default if CONFIG_DEBUG_KERNEL. We can then remove CONFIG_DETECT_SOFTLOCKUP directly. So tag CONFIG_LOCKUP_DETECTOR as default y. This is what we want for most serious kernel debugging anyway. And also forbid the lockup detector in S390 as it was for the previous softlockup detector, event though the true reason for that is not outlined. Signed-off-by: Frederic Weisbecker Cc: Ingo Molnar Cc: Don Zickus Cc: Peter Zijlstra Cc: Cyrill Gorcunov diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 755713a..3a18b0b 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -152,29 +152,10 @@ config DEBUG_SHIRQ Drivers ought to be able to handle interrupts coming in at those points; some don't and need to be caught. -config DETECT_SOFTLOCKUP - bool - depends on DEBUG_KERNEL && !S390 - default y - help - Say Y here to enable the kernel to detect "soft lockups", - which are bugs that cause the kernel to loop in kernel - mode for more than 60 seconds, without giving other tasks a - chance to run. - - When a soft-lockup is detected, the kernel will print the - current stack trace (which you should report), but the - system will stay locked up. This feature has negligible - overhead. - - (Note that "hard lockups" are separate type of bugs that - can be detected via the NMI-watchdog, on platforms that - support it.) - config LOCKUP_DETECTOR bool "Detect Hard and Soft Lockups" - depends on DEBUG_KERNEL - default DETECT_SOFTLOCKUP + depends on DEBUG_KERNEL && !S390 + default y help Say Y here to enable the kernel to act as a watchdog to detect hard and soft lockups. @@ -212,7 +193,7 @@ config BOOTPARAM_SOFTLOCKUP_PANIC config BOOTPARAM_SOFTLOCKUP_PANIC_VALUE int - depends on DETECT_SOFTLOCKUP + depends on LOCKUP_DETECTOR range 0 1 default 0 if !BOOTPARAM_SOFTLOCKUP_PANIC default 1 if BOOTPARAM_SOFTLOCKUP_PANIC -- cgit v0.10.2 From c01d4323309a90a298fd81cf3a059ee1b12be2e9 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 15 May 2010 22:57:48 +0200 Subject: lockup_detector: Adapt CONFIG_PERF_EVENT_NMI to other archs CONFIG_PERF_EVENT_NMI is something that need to be enabled from the arch. This is fine on x86 as PERF_EVENTS is builtin but if other archs select it, they will need to handle the PERF_EVENTS dependency. Instead, handle the dependency in the generic layer: - archs need to tell what they support through HAVE_PERF_EVENTS_NMI - Enable magically PERF_EVENTS_NMI if we have PERF_EVENTS and HAVE_PERF_EVENTS_NMI. Signed-off-by: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Don Zickus Cc: Cyrill Gorcunov diff --git a/arch/Kconfig b/arch/Kconfig index e5eb133..89b0efb 100644 --- a/arch/Kconfig +++ b/arch/Kconfig @@ -145,4 +145,7 @@ config HAVE_HW_BREAKPOINT config HAVE_USER_RETURN_NOTIFIER bool +config HAVE_PERF_EVENTS_NMI + bool + source "kernel/gcov/Kconfig" diff --git a/arch/x86/Kconfig b/arch/x86/Kconfig index 3cb28cd..3cb5bb0 100644 --- a/arch/x86/Kconfig +++ b/arch/x86/Kconfig @@ -54,7 +54,7 @@ config X86 select HAVE_KERNEL_LZO select HAVE_HW_BREAKPOINT select PERF_EVENTS - select PERF_EVENTS_NMI + select HAVE_PERF_EVENTS_NMI select ANON_INODES select HAVE_ARCH_KMEMCHECK select HAVE_USER_RETURN_NOTIFIER diff --git a/init/Kconfig b/init/Kconfig index e44e254..ab733c3 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -943,8 +943,7 @@ config PERF_USE_VMALLOC See tools/perf/design.txt for details config PERF_EVENTS_NMI - bool - depends on PERF_EVENTS + def_bool PERF_EVENTS && HAVE_PERF_EVENTS_NMI help System hardware can generate an NMI using the perf event subsystem. Also has support for calculating CPU cycle events -- cgit v0.10.2 From 23637d477c1f53acbb176a02c241d60a25888fae Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Sat, 15 May 2010 23:15:20 +0200 Subject: lockup_detector: Introduce CONFIG_HARDLOCKUP_DETECTOR This new config is deemed to simplify even more the lockup detector dependencies and can make it easier to bring a smooth sorting between archs that support the new generic lockup detector and those that still have their own, especially for those that are in the middle of this migration. Instead of checking whether we have CONFIG_LOCKUP_DETECTOR + CONFIG_PERF_EVENTS_NMI each time an arch wants to know if it needs to build its own lockup detector, take a shortcut with this new config. It is enabled only if the hardlockup detection part of the whole lockup detector is on. Signed-off-by: Frederic Weisbecker Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Don Zickus Cc: Cyrill Gorcunov diff --git a/arch/Kconfig b/arch/Kconfig index 89b0efb..35084f2 100644 --- a/arch/Kconfig +++ b/arch/Kconfig @@ -147,5 +147,9 @@ config HAVE_USER_RETURN_NOTIFIER config HAVE_PERF_EVENTS_NMI bool + help + System hardware can generate an NMI using the perf event + subsystem. Also has support for calculating CPU cycle events + to determine how many clock cycles in a given period. source "kernel/gcov/Kconfig" diff --git a/init/Kconfig b/init/Kconfig index ab733c3..eb77e8c 100644 --- a/init/Kconfig +++ b/init/Kconfig @@ -942,13 +942,6 @@ config PERF_USE_VMALLOC help See tools/perf/design.txt for details -config PERF_EVENTS_NMI - def_bool PERF_EVENTS && HAVE_PERF_EVENTS_NMI - help - System hardware can generate an NMI using the perf event - subsystem. Also has support for calculating CPU cycle events - to determine how many clock cycles in a given period. - menu "Kernel Performance Events And Counters" config PERF_EVENTS diff --git a/kernel/watchdog.c b/kernel/watchdog.c index be5e74e..83fb631 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -37,7 +37,7 @@ static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer); static DEFINE_PER_CPU(bool, softlockup_touch_sync); static DEFINE_PER_CPU(bool, hard_watchdog_warn); static DEFINE_PER_CPU(bool, soft_watchdog_warn); -#ifdef CONFIG_PERF_EVENTS_NMI +#ifdef CONFIG_HARDLOCKUP_DETECTOR static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts); static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved); static DEFINE_PER_CPU(struct perf_event *, watchdog_ev); @@ -51,7 +51,7 @@ static int __initdata no_watchdog; /* * Should we panic when a soft-lockup or hard-lockup occurs: */ -#ifdef CONFIG_PERF_EVENTS_NMI +#ifdef CONFIG_HARDLOCKUP_DETECTOR static int hardlockup_panic; static int __init hardlockup_panic_setup(char *str) @@ -152,7 +152,7 @@ void touch_softlockup_watchdog_sync(void) __raw_get_cpu_var(watchdog_touch_ts) = 0; } -#ifdef CONFIG_PERF_EVENTS_NMI +#ifdef CONFIG_HARDLOCKUP_DETECTOR /* watchdog detector functions */ static int is_hardlockup(int cpu) { @@ -189,7 +189,7 @@ static struct notifier_block panic_block = { .notifier_call = watchdog_panic, }; -#ifdef CONFIG_PERF_EVENTS_NMI +#ifdef CONFIG_HARDLOCKUP_DETECTOR static struct perf_event_attr wd_hw_attr = { .type = PERF_TYPE_HARDWARE, .config = PERF_COUNT_HW_CPU_CYCLES, @@ -239,7 +239,7 @@ static void watchdog_interrupt_count(void) } #else static inline void watchdog_interrupt_count(void) { return; } -#endif /* CONFIG_PERF_EVENTS_NMI */ +#endif /* CONFIG_HARDLOCKUP_DETECTOR */ /* watchdog kicker functions */ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) @@ -342,7 +342,7 @@ static int watchdog(void *__bind_cpu) } -#ifdef CONFIG_PERF_EVENTS_NMI +#ifdef CONFIG_HARDLOCKUP_DETECTOR static int watchdog_nmi_enable(int cpu) { struct perf_event_attr *wd_attr; @@ -393,7 +393,7 @@ static void watchdog_nmi_disable(int cpu) #else static int watchdog_nmi_enable(int cpu) { return 0; } static void watchdog_nmi_disable(int cpu) { return; } -#endif /* CONFIG_PERF_EVENTS_NMI */ +#endif /* CONFIG_HARDLOCKUP_DETECTOR */ /* prepare/enable/disable routines */ static int watchdog_prepare_cpu(int cpu) diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index 3a18b0b..e65e47d 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -174,6 +174,9 @@ config LOCKUP_DETECTOR generate interrupts and kick the watchdog task every 10-12 seconds. An NMI is generated every 60 seconds or so to check for hardlockups. +config HARDLOCKUP_DETECTOR + def_bool LOCKUP_DETECTOR && PERF_EVENTS && HAVE_PERF_EVENTS_NMI + config BOOTPARAM_SOFTLOCKUP_PANIC bool "Panic (Reboot) On Soft Lockups" depends on LOCKUP_DETECTOR -- cgit v0.10.2 From cafcd80d216bc2136b8edbb794327e495792c666 Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Fri, 14 May 2010 11:11:21 -0400 Subject: lockup_detector: Cross arch compile fixes Combining the softlockup and hardlockup code causes watchdog.c to build even without the hardlockup detection support. So if an arch, that has the previous and the new nmi watchdog implementations cohabiting, wants to know if the generic one is in use, CONFIG_LOCKUP_DETECTOR is not a reliable check. We need to use CONFIG_HARDLOCKUP_DETECTOR instead. Fixes: kernel/built-in.o: In function `touch_nmi_watchdog': (.text+0x449bc): multiple definition of `touch_nmi_watchdog' arch/sparc/kernel/built-in.o:(.text+0x11b28): first defined here Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Don Zickus Cc: Cyrill Gorcunov LKML-Reference: <20100514151121.GR15159@redhat.com> [ use CONFIG_HARDLOCKUP_DETECTOR instead of CONFIG_PERF_EVENTS_NMI] Signed-off-by: Frederic Weisbecker diff --git a/arch/x86/kernel/apic/Makefile b/arch/x86/kernel/apic/Makefile index 52f32e0..910f20b4 100644 --- a/arch/x86/kernel/apic/Makefile +++ b/arch/x86/kernel/apic/Makefile @@ -3,10 +3,10 @@ # obj-$(CONFIG_X86_LOCAL_APIC) += apic.o apic_noop.o probe_$(BITS).o ipi.o -ifneq ($(CONFIG_LOCKUP_DETECTOR),y) +ifneq ($(CONFIG_HARDLOCKUP_DETECTOR),y) obj-$(CONFIG_X86_LOCAL_APIC) += nmi.o endif -obj-$(CONFIG_LOCKUP_DETECTOR) += hw_nmi.o +obj-$(CONFIG_HARDLOCKUP_DETECTOR) += hw_nmi.o obj-$(CONFIG_X86_IO_APIC) += io_apic.o obj-$(CONFIG_SMP) += ipi.o diff --git a/include/linux/nmi.h b/include/linux/nmi.h index abd48aa..06aab5e 100644 --- a/include/linux/nmi.h +++ b/include/linux/nmi.h @@ -20,7 +20,7 @@ extern void touch_nmi_watchdog(void); extern void acpi_nmi_disable(void); extern void acpi_nmi_enable(void); #else -#ifndef CONFIG_LOCKUP_DETECTOR +#ifndef CONFIG_HARDLOCKUP_DETECTOR static inline void touch_nmi_watchdog(void) { touch_softlockup_watchdog(); diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 83fb631..e53622c 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -31,13 +31,13 @@ int watchdog_enabled; int __read_mostly softlockup_thresh = 60; static DEFINE_PER_CPU(unsigned long, watchdog_touch_ts); -static DEFINE_PER_CPU(bool, watchdog_nmi_touch); static DEFINE_PER_CPU(struct task_struct *, softlockup_watchdog); static DEFINE_PER_CPU(struct hrtimer, watchdog_hrtimer); static DEFINE_PER_CPU(bool, softlockup_touch_sync); -static DEFINE_PER_CPU(bool, hard_watchdog_warn); static DEFINE_PER_CPU(bool, soft_watchdog_warn); #ifdef CONFIG_HARDLOCKUP_DETECTOR +static DEFINE_PER_CPU(bool, hard_watchdog_warn); +static DEFINE_PER_CPU(bool, watchdog_nmi_touch); static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts); static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts_saved); static DEFINE_PER_CPU(struct perf_event *, watchdog_ev); @@ -139,6 +139,7 @@ void touch_all_softlockup_watchdogs(void) per_cpu(watchdog_touch_ts, cpu) = 0; } +#ifdef CONFIG_HARDLOCKUP_DETECTOR void touch_nmi_watchdog(void) { __get_cpu_var(watchdog_nmi_touch) = true; @@ -146,6 +147,8 @@ void touch_nmi_watchdog(void) } EXPORT_SYMBOL(touch_nmi_watchdog); +#endif + void touch_softlockup_watchdog_sync(void) { __raw_get_cpu_var(softlockup_touch_sync) = true; -- cgit v0.10.2 From 26e09c6eee14f4827b55137ba0eedc4e77cd50ab Mon Sep 17 00:00:00 2001 From: Don Zickus Date: Mon, 17 May 2010 18:06:04 -0400 Subject: lockup_detector: Convert per_cpu to __get_cpu_var for readability Just a bunch of conversions as suggested by Frederic W. __get_cpu_var() provides preemption disabled checks. Plus it gives more readability as it makes it obvious we are dealing locally now with these vars. Signed-off-by: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Don Zickus Cc: Cyrill Gorcunov LKML-Reference: <1274133966-18415-2-git-send-email-dzickus@redhat.com> Signed-off-by: Frederic Weisbecker diff --git a/kernel/watchdog.c b/kernel/watchdog.c index e53622c..91b0b26 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -115,7 +115,7 @@ static unsigned long get_sample_period(void) /* Commands for resetting the watchdog */ static void __touch_watchdog(void) { - int this_cpu = raw_smp_processor_id(); + int this_cpu = smp_processor_id(); __get_cpu_var(watchdog_touch_ts) = get_timestamp(this_cpu); } @@ -157,21 +157,21 @@ void touch_softlockup_watchdog_sync(void) #ifdef CONFIG_HARDLOCKUP_DETECTOR /* watchdog detector functions */ -static int is_hardlockup(int cpu) +static int is_hardlockup(void) { - unsigned long hrint = per_cpu(hrtimer_interrupts, cpu); + unsigned long hrint = __get_cpu_var(hrtimer_interrupts); - if (per_cpu(hrtimer_interrupts_saved, cpu) == hrint) + if (__get_cpu_var(hrtimer_interrupts_saved) == hrint) return 1; - per_cpu(hrtimer_interrupts_saved, cpu) = hrint; + __get_cpu_var(hrtimer_interrupts_saved) = hrint; return 0; } #endif -static int is_softlockup(unsigned long touch_ts, int cpu) +static int is_softlockup(unsigned long touch_ts) { - unsigned long now = get_timestamp(cpu); + unsigned long now = get_timestamp(smp_processor_id()); /* Warn about unreasonable delays: */ if (time_after(now, touch_ts + softlockup_thresh)) @@ -206,8 +206,6 @@ void watchdog_overflow_callback(struct perf_event *event, int nmi, struct perf_sample_data *data, struct pt_regs *regs) { - int this_cpu = smp_processor_id(); - if (__get_cpu_var(watchdog_nmi_touch) == true) { __get_cpu_var(watchdog_nmi_touch) = false; return; @@ -219,7 +217,9 @@ void watchdog_overflow_callback(struct perf_event *event, int nmi, * fired multiple times before we overflow'd. If it hasn't * then this is a good indication the cpu is stuck */ - if (is_hardlockup(this_cpu)) { + if (is_hardlockup()) { + int this_cpu = smp_processor_id(); + /* only print hardlockups once */ if (__get_cpu_var(hard_watchdog_warn) == true) return; @@ -247,7 +247,6 @@ static inline void watchdog_interrupt_count(void) { return; } /* watchdog kicker functions */ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) { - int this_cpu = smp_processor_id(); unsigned long touch_ts = __get_cpu_var(watchdog_touch_ts); struct pt_regs *regs = get_irq_regs(); int duration; @@ -262,12 +261,12 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) hrtimer_forward_now(hrtimer, ns_to_ktime(get_sample_period())); if (touch_ts == 0) { - if (unlikely(per_cpu(softlockup_touch_sync, this_cpu))) { + if (unlikely(__get_cpu_var(softlockup_touch_sync))) { /* * If the time stamp was touched atomically * make sure the scheduler tick is up to date. */ - per_cpu(softlockup_touch_sync, this_cpu) = false; + __get_cpu_var(softlockup_touch_sync) = false; sched_clock_tick(); } __touch_watchdog(); @@ -280,14 +279,14 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) * indicate it is getting cpu time. If it hasn't then * this is a good indication some task is hogging the cpu */ - duration = is_softlockup(touch_ts, this_cpu); + duration = is_softlockup(touch_ts); if (unlikely(duration)) { /* only warn once */ if (__get_cpu_var(soft_watchdog_warn) == true) return HRTIMER_RESTART; printk(KERN_ERR "BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n", - this_cpu, duration, + smp_processor_id(), duration, current->comm, task_pid_nr(current)); print_modules(); print_irqtrace_events(current); @@ -309,10 +308,10 @@ static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer) /* * The watchdog thread - touches the timestamp. */ -static int watchdog(void *__bind_cpu) +static int watchdog(void *unused) { struct sched_param param = { .sched_priority = MAX_RT_PRIO-1 }; - struct hrtimer *hrtimer = &per_cpu(watchdog_hrtimer, (unsigned long)__bind_cpu); + struct hrtimer *hrtimer = &__raw_get_cpu_var(watchdog_hrtimer); sched_setscheduler(current, SCHED_FIFO, ¶m); @@ -328,7 +327,7 @@ static int watchdog(void *__bind_cpu) /* * Run briefly once per second to reset the softlockup timestamp. * If this gets delayed for more than 60 seconds then the - * debug-printout triggers in softlockup_tick(). + * debug-printout triggers in watchdog_timer_fn(). */ while (!kthread_should_stop()) { __touch_watchdog(); -- cgit v0.10.2 From e35e7fb0e9ea557f7504ced6fe4ccf69e44b7f07 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 19 May 2010 11:36:49 +0200 Subject: lockup_detector: Don't enable the lockup detector by default The lockup detector is a new feature that now involves the nmi watchdog. Drop the default y and let the user choose. Signed-off-by: Frederic Weisbecker Cc: Don Zickus Cc: Ingo Molnar Cc: Peter Zijlstra Cc: Cyrill Gorcunov diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug index e65e47d..63968a9 100644 --- a/lib/Kconfig.debug +++ b/lib/Kconfig.debug @@ -155,7 +155,6 @@ config DEBUG_SHIRQ config LOCKUP_DETECTOR bool "Detect Hard and Soft Lockups" depends on DEBUG_KERNEL && !S390 - default y help Say Y here to enable the kernel to act as a watchdog to detect hard and soft lockups. -- cgit v0.10.2 From eb703f98191a505f78d0066712ad67d5dedc4c90 Mon Sep 17 00:00:00 2001 From: Kulikov Vasiliy Date: Mon, 5 Jul 2010 12:00:54 +0400 Subject: kernel/watchdog: Initialize 'result' Variable on the stack is not initialized to zero, do it explicitly. This bug was found by a compiler warning: kernel/watchdog.c:463: warning: 'result' may be used uninitialized in this function Signed-off-by: Kulikov Vasiliy Acked-by: Don Zickus Cc: Frederic Weisbecker Cc: Peter Zijlstra Cc: Arnaldo Carvalho de Melo Cc: Frederic Weisbecker Cc: Paul Mackerras Cc: Mike Galbraith Cc: Steven Rostedt LKML-Reference: <1278316854-28442-1-git-send-email-segooon@gmail.com> Signed-off-by: Ingo Molnar diff --git a/kernel/watchdog.c b/kernel/watchdog.c index 91b0b26..613bc1f 100644 --- a/kernel/watchdog.c +++ b/kernel/watchdog.c @@ -460,7 +460,7 @@ static void watchdog_disable(int cpu) static void watchdog_enable_all_cpus(void) { int cpu; - int result; + int result = 0; for_each_online_cpu(cpu) result += watchdog_enable(cpu); -- cgit v0.10.2