Linux：soft lockup 检测机制
linux-dash
A beautiful web dashboard for Linux
项目地址：https://gitcode.com/gh_mirrors/li/linux-dash
免费下载资源
·
1. 前言
限于作者能力水平，本文可能存在谬误，因此而给读者带来的损失，作者不做任何承诺。
2. 分析背景
本文分析基于 linux-4.14.132 内核代码分析，运行环境 Ubuntu 16.04.4 LTS + QEMU + ARM vexpress-a9 ，rootfs 基于 ubuntu-base-16.04-core-armhf.tar.gz 制作。
3. soft lockup 机制
3.1 什么是 soft lockup ?
soft lockup 是指某段代码一直占住当前 CPU ，其它进程无法在当前 CPU 上得到调度的情形。soft lockup 在不同的抢占模式下，有着不同的情形。
3.2 各种抢占配置下的 soft lockup
3.2.1 CONFIG_PREEMPT_NONE 配置下的 soft lockup
CONFIG_PREEMPT_NONE 不支持内核态抢占，致力于提高吞吐量，通常用于服务器。 。在 CONFIG_PREEMPT_NONE 配置下，包含死循环或等同情形的代码，可导致 soft lockup 。
3.2.2 CONFIG_PREEMPT 配置下的 soft lockup
CONFIG_PREEMPT 支持内核态抢占，用于低延迟的桌面系统。在CONFIG_PREEMPT 配置下，较长时间禁用抢占的进程，可导致 soft lockup 。
3.2.3 CONFIG_PREEMPT_VOLUNTARY 配置下的 soft lockup
CONFIG_PREEMPT_VOLUNTARY 不支持内核态抢占，通常用于桌面系统。相对于配置 CONFIG_PREEMPT_NONE 的情形，在一些可能导致睡眠的代码路径上，插入了一些调度点，以降低延迟。
/*
* include/linux/kernel.h
*/
...
#ifdef CONFIG_PREEMPT_VOLUNTARY
extern int _cond_resched(void);
# define might_resched() _cond_resched()
#else
# define might_resched() do { } while (0)
#endif
#ifdef CONFIG_DEBUG_ATOMIC_SLEEP
void ___might_sleep(const char *file, int line, int preempt_offset);
void __might_sleep(const char *file, int line, int preempt_offset);
/**
* might_sleep - annotation for functions that can sleep
*
* this macro will print a stack trace if it is executed in an atomic
* context (spinlock, irq-handler, ...).
*
* This is a useful debugging help to be able to catch problems early and not
* be bitten later when the calling function happens to sleep when it is not
* supposed to.
*/
# define might_sleep() \
do { __might_sleep(__FILE__, __LINE__, 0); might_resched(); } while (0)
# define sched_annotate_sleep() (current->task_state_change = 0)
#else
static inline void ___might_sleep(const char *file, int line,
int preempt_offset) { }
static inline void __might_sleep(const char *file, int line,
int preempt_offset) { }
# define might_sleep() do { might_resched(); } while (0)
# define sched_annotate_sleep() do { } while (0)
#endif
#define might_sleep_if(cond) do { if (cond) might_sleep(); } while (0)
...
在 CONFIG_PREEMPT_VOLUNTARY 配置下，包含死循环或等同情形的代码，可导致 soft lockup 。
3.3 soft lockup 的实现
3.3.1 创建 soft lockup watchdog 每 CPU 线程
为所有的 CPU 创建每 CPU 的 watchdog 内核线程，该内核线程每次被调度时（即定时器到期时）更新一次 watchdog时间戳，同时再次启动定时器；同时，定时器到期时，对比 watchdog时间戳 和 当前时间戳，如果（当前时间戳 - watchdog时间戳 >= 设定的soft lockup阈值），表示当前 CPU 上的当前进程已经占住 CPU 较长时间，如此报告一个 soft lockup 问题 。过程中涉及到内核线程创建细节部分，可参考 Linux: 内核线程简析。
void __init lockup_detector_init(void)
{
...
cpumask_copy(&watchdog_cpumask, cpu_possible_mask);
...
lockup_detector_setup();
}
static struct smp_hotplug_thread watchdog_threads = {
.store = &softlockup_watchdog,
.thread_should_run = watchdog_should_run,
.thread_fn = watchdog,
.thread_comm = "watchdog/%u",
.setup = watchdog_enable,
.cleanup = watchdog_cleanup,
.park = watchdog_disable,
.unpark = watchdog_enable,
};
static __init void lockup_detector_setup(void)
{
...
lockup_detector_update_enable();
...
/*
* 创建每 CPU 的 watchdog 内核线程：
* . 负责更新每 CPU 的 watchdog 时间戳；
* . 每次更新 watchdog 时间戳的同时启动一个定时器，
* 该定时器检测 watchdog 时间戳更新的时间间隔，如
* 果更新间隔超过设定的阈值，则报告 soft lockup 问题。
*/
ret = smpboot_register_percpu_thread_cpumask(&watchdog_threads,
&watchdog_allowed_mask);
if (ret) {
pr_err("Failed to initialize soft lockup detector threads\n");
return;
}
mutex_lock(&watchdog_mutex);
softlockup_threads_initialized = true;
lockup_detector_reconfigure();
mutex_unlock(&watchdog_mutex);
}
static void lockup_detector_update_enable(void)
{
watchdog_enabled = 0;
...
if (soft_watchdog_user_enabled)
watchdog_enabled |= SOFT_WATCHDOG_ENABLED;
}
int smpboot_register_percpu_thread_cpumask(struct smp_hotplug_thread *plug_thread,
const struct cpumask *cpumask)
{
...
for_each_online_cpu(cpu) {
ret = __smpboot_create_thread(plug_thread, cpu); /* 创建 @cpu 的内核线程 */
...
if (cpumask_test_cpu(cpu, cpumask))
smpboot_unpark_thread(plug_thread, cpu); /* 启动 @cpu 的内核线程，进入 smpboot_thread_fn() */
}
...
}
static int
__smpboot_create_thread(struct smp_hotplug_thread *ht, unsigned int cpu)
{
struct task_struct *tsk = *per_cpu_ptr(ht->store, cpu);
struct smpboot_thread_data *td;
td = kzalloc_node(sizeof(*td), GFP_KERNEL, cpu_to_node(cpu));
td->cpu = cpu;
td->ht = ht;
/*
* SMP 内核线程的公共入口为 smpboot_thread_fn() ，
* 通过 smpboot_thread_fn() 调用具体的线程入口。
*/
tsk = kthread_create_on_cpu(smpboot_thread_fn, td, cpu,
ht->thread_comm);
kthread_park(tsk); /* 暂停内核线程 */
get_task_struct(tsk);
*per_cpu_ptr(ht->store, cpu) = tsk;
...
return 0;
}
3.3.2 运行 soft lockup watchdog 内核线程
static int smpboot_thread_fn(void *data)
{
struct smpboot_thread_data *td = data;
struct smp_hotplug_thread *ht = td->ht; /* watchdog_threads */
while (1) {
set_current_state(TASK_INTERRUPTIBLE);
preempt_disable();
if (kthread_should_stop()) {
__set_current_state(TASK_RUNNING);
preempt_enable();
/* cleanup must mirror setup */
if (ht->cleanup && td->status != HP_THREAD_NONE)
ht->cleanup(td->cpu, cpu_online(td->cpu));
kfree(td);
return 0;
}
if (kthread_should_park()) {
__set_current_state(TASK_RUNNING);
preempt_enable();
if (ht->park && td->status == HP_THREAD_ACTIVE) {
BUG_ON(td->cpu != smp_processor_id());
ht->park(td->cpu);
td->status = HP_THREAD_PARKED;
}
kthread_parkme();
/* We might have been woken for stop */
continue;
}
BUG_ON(td->cpu != smp_processor_id());
/* Check for state change setup */
switch (td->status) {
case HP_THREAD_NONE:
__set_current_state(TASK_RUNNING);
preempt_enable();
if (ht->setup)
ht->setup(td->cpu); /* watchdog_enable(): 初始化 watchdog时间戳，以及时间戳更新间隔检测定时器 */
td->status = HP_THREAD_ACTIVE;
continue;
case HP_THREAD_PARKED:
__set_current_state(TASK_RUNNING);
preempt_enable();
if (ht->unpark)
ht->unpark(td->cpu);
td->status = HP_THREAD_ACTIVE;
continue;
}
if (!ht->thread_should_run(td->cpu)) {
preempt_enable_no_resched();
schedule();
} else {
__set_current_state(TASK_RUNNING);
preempt_enable();
/* 更新当前 CPU 的 watchdog时间戳 */
ht->thread_fn(td->cpu); /* kernel/watchdog.c: watchdog() */
}
}
}
static void watchdog_enable(unsigned int cpu)
{
/* 启动时间戳更新间隔检测定时器 */
hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
hrtimer->function = watchdog_timer_fn;
hrtimer_start(hrtimer, ns_to_ktime(sample_period),
HRTIMER_MODE_REL_PINNED);
__touch_watchdog(); /* 初始 watchdog 的时间戳 */
watchdog_set_prio(SCHED_FIFO, MAX_RT_PRIO - 1); /* 设置为 RT 调度类的 SCHED_FIFO */
}
static void watchdog(unsigned int cpu)
{
__this_cpu_write(soft_lockup_hrtimer_cnt,
__this_cpu_read(hrtimer_interrupts));
__touch_watchdog(); /* watchdog 线程负责时不时的更新当前CPU的watchdog时间戳 */
}
3.3.3 触发 soft lockup 问题报告
定时器到期后，进入 watchdog_timer_fn() ，触发可能的 soft lockup 问题报告：
/* watchdog kicker functions */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
unsigned long touch_ts = __this_cpu_read(watchdog_touch_ts); /* 读取当前CPU最近的更新的时间戳 */
...
/* kick the softlockup detector */
wake_up_process(__this_cpu_read(softlockup_watchdog)); /* 唤醒当前CPU时间戳更新线程 */
/* .. and repeat */
hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period));
if (touch_ts == 0) { /* 休眠重启等类似情形下，时间戳复位为0 */
...
__touch_watchdog(); /* 重新初始化时间戳 */
return HRTIMER_RESTART; /* 重启定时器 */
}
duration = is_softlockup(touch_ts); /* 上次更新的时间戳和当前时间戳的差值 */
if (unlikely(duration)) { /* 时间戳差值大于设定的阈值 */
...
/* 报告 sotf lockup 问题 */
pr_emerg("BUG: soft lockup - CPU#%d stuck for %us! [%s:%d]\n",
smp_processor_id(), duration,
current->comm, task_pid_nr(current));
__this_cpu_write(softlockup_task_ptr_saved, current);
print_modules();
print_irqtrace_events(current);
if (regs)
show_regs(regs);
else
dump_stack();
...
add_taint(TAINT_SOFTLOCKUP, LOCKDEP_STILL_OK);
if (softlockup_panic)
panic("softlockup: hung tasks");
__this_cpu_write(soft_watchdog_warn, true);
} else
__this_cpu_write(soft_watchdog_warn, false);
return HRTIMER_RESTART;
}
static int is_softlockup(unsigned long touch_ts)
{
unsigned long now = get_timestamp();
if ((watchdog_enabled & SOFT_WATCHDOG_ENABLED) && watchdog_thresh){
/* Warn about unreasonable delays. */
/* 更新时间戳的间隔超过了设定的阈值 */
if (time_after(now, touch_ts + get_softlockup_thresh()))
return now - touch_ts;
}
return 0;
}
思考一下，为什么 soft lockup 报告问题的时候，直指 current 进程？难道没有可能是 current 之前的某个进程占住 CPU 很长时间（超过设定的检查阈值），然后这个进程放开 CPU，紧接着 current 被调度上来，为之前那个干了坏事的进程背锅吗？为了防止这种情形，必须让 soft lockup 的 watchdog 内核线程拥有更高的被调度优先权，这样一旦出现调度机会，soft lockup 的 watchdog 内核线程总是比其它普通进程被先调度到 CPU 上执行。看代码片段：
static void watchdog_enable(unsigned int cpu)
{
...
watchdog_set_prio(SCHED_FIFO, MAX_RT_PRIO - 1); /* 设置为 RT 调度类的 SCHED_FIFO */
}
看到了吗？将 soft lockup 的 watchdog 内核线程的调度类切换为高优先级别的实时调度类，调度策略为 SCHED_FIFO 。因为更高优先级的调度类的进程，比其它低优先级调度类（如CFS）的进程拥有更高的优先调度权，如此保证了不会发生上述情形。但是， soft lockup 的 watchdog 内核线程总是比其它低优先调度类的进程先被调度，那其它低优先调度类的进程怎么办？啥时候轮到它们执行？soft lockup 的 watchdog 内核线程毕竟只是起到个监督作用，它也不需要时时刻运行，所以在不需要的时候，它们会主动让出 CPU 来，让其它低优先级调度类的进程有机会得到执行。看下面的代码：
static int smpboot_thread_fn(void *data)
{
while (1) {
set_current_state(TASK_INTERRUPTIBLE);
preempt_disable();
...
/*
* 当前 【定时器计数 hrtimer_interrupts】 和 【watchdog 时间戳更新次数
* soft_lockup_hrtimer_cnt】 同步(计数相同) ，表示当前处在【上次 watchdog
* 时间戳更新过后，检测定时器尚未到期之前】的一段时间内，这段期间无需更新
* watchdog 时间戳，主动让出 CPU 给其它进程，进入 TASK_INTERRUPTIBLE 睡眠
* 态，定时器到期后会唤醒它更新 watchdog 时间戳。
*/
if (!ht->thread_should_run(td->cpu)) { /* kernel/watchdog.c: watchdog_should_run() */
preempt_enable_no_resched();
schedule();
} else { /* 定时器到期了，唤醒 watchdog 内核线程更新时间戳 */
__set_current_state(TASK_RUNNING);
preempt_enable();
ht->thread_fn(td->cpu); /* kernel/watchdog.c: watchdog() */
}
}
}
/* 当然， hrtimer_interrupts 和 soft_lockup_hrtimer_cnt 都是每 CPU 的 */
static DEFINE_PER_CPU(unsigned long, hrtimer_interrupts);
static DEFINE_PER_CPU(unsigned long, soft_lockup_hrtimer_cnt);
/* 定时器到期了，watchdog 该起来更新时间戳 */
static int watchdog_should_run(unsigned int cpu)
{
return __this_cpu_read(hrtimer_interrupts) !=
__this_cpu_read(soft_lockup_hrtimer_cnt);
}
/*
* 每次更新 watchdog 时间戳的时候，
* 同步 hrtimer_interrupts (检测定时器触发的次数) 和 soft_lockup_hrtimer_cnt 计数。
*/
static void watchdog(unsigned int cpu)
{
/* 同步 hrtimer_interrupts (检测定时器触发的次数) 和 soft_lockup_hrtimer_cnt 计数 */
__this_cpu_write(soft_lockup_hrtimer_cnt,
__this_cpu_read(hrtimer_interrupts));
__touch_watchdog(); /* watchdog 线程负责时不时的更新时间戳 */
}
/* 定期检查定时器 */
static enum hrtimer_restart watchdog_timer_fn(struct hrtimer *hrtimer)
{
...
watchdog_interrupt_count(); /* /* 递增定时器触发次数 @hrtimer_interrupts */
wake_up_process(__this_cpu_read(softlockup_watchdog)); /* 唤醒当前CPU时间戳更新watchdog线程 */
hrtimer_forward_now(hrtimer, ns_to_ktime(sample_period)); /* 重启定时器 */
...
duration = is_softlockup(touch_ts);
if (unlikely(duration)) {
/* 报告 soft lockup */
}
...
}
看到这里，疑问又产生了，在 watchdog_timer_fn() 中，又是唤醒 watchdog 线程更新时间戳，又是检查 soft lockup 问题，这两个逻辑同时进行，不是互相冲突矛盾？先来解释下，为什么要对 watchdog 线程进行唤醒。前面我们看到，watchdog 线程会在一定条件下，主动让出 CPU 进入睡眠，所以需要在合适时机（就是定时器到期时）唤醒它，更新 watchdog 时间戳。在 watchdog_timer_fn() 中，唤醒 watchdog 线程更新时间戳 和 检查 soft lockup 问题 两个逻辑并行，并不矛盾。关键在于，定时器的周期，和判定 soft lockup 问题 的阈值并不相同。来看代码：
static void lockup_detector_reconfigure(void)
{
...
set_sample_period();
...
}
static void set_sample_period(void)
{
sample_period = get_softlockup_thresh() * ((u64)NSEC_PER_SEC / 5);
...
}
static void watchdog_enable(unsigned int cpu)
{
...
hrtimer_init(hrtimer, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
hrtimer->function = watchdog_timer_fn;
hrtimer_start(hrtimer, ns_to_ktime(sample_period),
HRTIMER_MODE_REL_PINNED);
...
}
原来，定时器的周期是检查阈值的的1/5，疑问得到了解答。
4. soft lockup 举例
#include <linux/module.h>
#include <linux/sched.h>
#include <linux/kthread.h>
#include <linux/delay.h>
static struct task_struct *softlockup_task;
static int softlockup_task_fn(void *ignored)
{
int ret = 0;
while (!kthread_should_stop()) {
#if (defined(CONFIG_PREEMPT_NONE) || defined(CONFIG_PREEMPT_VOLUNTARY))
asm("nop");
#else /* CONFIG_PREEMPT */
preempt_disable();
mdelay(30 * 1000);
preempt_enable();
#endif
}
return ret;
}
static int __init softlockup_task_demo_init(void)
{
int ret = 0;
softlockup_task = kthread_run(softlockup_task_fn, NULL, "softlockup_task");
if (IS_ERR(softlockup_task)) {
ret = PTR_ERR(softlockup_task);
printk(KERN_ERR "%s: Failed to create kernel thread, ret = [%d]\n", __func__, ret);
}
printk(KERN_INFO "soft lockup task example module loaded.\n");
return ret;
}
static void __exit softlockup_task_demo_exit(void)
{
if (softlockup_task) {
kthread_stop(softlockup_task);
softlockup_task = NULL;
}
printk(KERN_INFO "soft lockup task example module exited.\n");
}
module_init(softlockup_task_demo_init);
module_exit(softlockup_task_demo_exit);
MODULE_LICENSE("GPL");
#
# 配置
#
# 抢占模式 3 选一
CONFIG_PREEMPT_NONE=y
#CONFIG_PREEMPT_VOLUNTARY=y
#CONFIG_PREEMPT=y
CONFIG_DEBUG_KERNEL=y
CONFIG_LOCKUP_DETECTOR=y
CONFIG_SOFTLOCKUP_DETECTOR=y
CONFIG_SAMPLE_SOFTLOCKUP=m
cd linux-4.14.132
make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- -j8 O=output
sudo mount rootfs.img temp
cd linux-4.14.132
sudo make ARCH=arm CROSS_COMPILE=arm-linux-gnueabi- O=output INSTALL_MOD_PATH=/path/to/temp modules_install
cd -
sudo umount temp
然后用QEMU启动系统，加载测试模块：
sudo qemu-system-arm \
-M vexpress-a9 \
-smp 4 \
-m 512M \
-kernel /path/to/zImage \
-dtb /path/to/vexpress-v2p-ca9.dtb \
-nographic \
-append "root=/dev/mmcblk0 rw rootfstype=ext4 console=ttyAMA0" \
-sd rootfs.img
# 当前位于 QEMU 模拟器系统特权模式下
# ps -eo pid,class,rtprio,pri,psr,comm | grep watchdog
12 FF 99 139 0 watchdog/0
15 FF 99 139 1 watchdog/1
21 FF 99 139 2 watchdog/2
27 FF 99 139 3 watchdog/3
# modprobe softlockup_example
模块加载一段时间后，报 soft lockup BUG：
[ 157.356865] soft lockup task example module loaded.
[ 176.383362] INFO: rcu_sched self-detected stall on CPU
[ 176.384981] 1-...: (2113 ticks this GP) idle=a42/140000000000001/0 softirq=2326/2326 fqs=1019
[ 176.385513] (t=2100 jiffies g=903 c=902 q=636)
[ 176.387315] NMI backtrace for cpu 1
[ 176.387904] CPU: 1 PID: 939 Comm: softlockup_task Not tainted 4.14.132 #34
[ 176.387942] Hardware name: ARM-Versatile Express
[ 176.391912] [<8011149c>] (unwind_backtrace) from [<8010c330>] (show_stack+0x20/0x24)
[ 176.393289] [<8010c330>] (show_stack) from [<806ddfd8>] (dump_stack+0x8c/0xa0)
[ 176.393356] [<806ddfd8>] (dump_stack) from [<806e3dc4>] (nmi_cpu_backtrace+0xc0/0xc4)
[ 176.393417] [<806e3dc4>] (nmi_cpu_backtrace) from [<806e3eb0>] (nmi_trigger_cpumask_backtrace+0xe8/0x12c)
[ 176.393474] [<806e3eb0>] (nmi_trigger_cpumask_backtrace) from [<8010f490>] (arch_trigger_cpumask_backtrace+0x20/0x24)
[ 176.393534] [<8010f490>] (arch_trigger_cpumask_backtrace) from [<80182ea0>] (rcu_dump_cpu_stacks+0xac/0xd8)
[ 176.393672] [<80182ea0>] (rcu_dump_cpu_stacks) from [<80182478>] (rcu_check_callbacks+0x7f8/0x9f8)
[ 176.393725] [<80182478>] (rcu_check_callbacks) from [<80187f84>] (update_process_times+0x44/0x6c)
[ 176.393775] [<80187f84>] (update_process_times) from [<80197144>] (tick_periodic+0x4c/0xcc)
[ 176.393828] [<80197144>] (tick_periodic) from [<80197368>] (tick_handle_periodic+0x38/0x98)
[ 176.393877] [<80197368>] (tick_handle_periodic) from [<8010ffe4>] (twd_handler+0x40/0x50)
[ 176.393924] [<8010ffe4>] (twd_handler) from [<80172128>] (handle_percpu_devid_irq+0x98/0x24c)
[ 176.393979] [<80172128>] (handle_percpu_devid_irq) from [<8016c728>] (generic_handle_irq+0x34/0x44)
[ 176.394031] [<8016c728>] (generic_handle_irq) from [<8016cd3c>] (__handle_domain_irq+0x6c/0xc4)
[ 176.394125] [<8016cd3c>] (__handle_domain_irq) from [<80101508>] (gic_handle_irq+0x5c/0xa0)
[ 176.394166] [<80101508>] (gic_handle_irq) from [<8010d10c>] (__irq_svc+0x6c/0x90)
[ 176.394240] Exception stack(0x9e90df18 to 0x9e90df60)
[ 176.394533] df00: 00000000 9e983bc0
[ 176.394835] df20: 00000000 9e983bc0 9e983bc0 00000000 9f5fc940 9e90c000 00000000 9e983bdc
[ 176.394945] df40: 9e8fdd30 9e90df74 9e90df68 9e90df68 7f000020 7f000020 00000013 ffffffff
[ 176.396350] [<8010d10c>] (__irq_svc) from [<7f000020>] (softlockup_task_fn+0x20/0x30 [softlockup_example])
[ 176.396667] [<7f000020>] (softlockup_task_fn [softlockup_example]) from [<80143ae0>] (kthread+0x144/0x174)
[ 176.396718] [<80143ae0>] (kthread) from [<80107ee8>] (ret_from_fork+0x14/0x2c)
[ 204.214887] watchdog: BUG: soft lockup - CPU#1 stuck for 23s! [softlockup_task:939]
[ 204.215332] Modules linked in: softlockup_example
[ 204.215593] CPU: 1 PID: 939 Comm: softlockup_task Not tainted 4.14.132 #34
[ 204.215604] Hardware name: ARM-Versatile Express
[ 204.215648] task: 9f700c00 task.stack: 9e90c000
[ 204.215701] PC is at kthread_should_stop+0x30/0x54
[ 204.215736] LR is at softlockup_task_fn+0x20/0x30 [softlockup_example]
[ 204.215755] pc : [<80143524>] lr : [<7f000020>] psr: 00000013
[ 204.215770] sp : 9e90df50 ip : 9e90df68 fp : 9e90df64
[ 204.215785] r10: 9e8fdd30 r9 : 9e983bdc r8 : 00000000
[ 204.215804] r7 : 9e90c000 r6 : 9f5fc940 r5 : 00000000 r4 : 9f700c00
[ 204.215821] r3 : 00208040 r2 : 00000000 r1 : 9e983bc0 r0 : 00000000
[ 204.215875] Flags: nzcv IRQs on FIQs on Mode SVC_32 ISA ARM Segment none
[ 204.215899] Control: 10c5387d Table: 7e92806a DAC: 00000051
[ 204.215927] CPU: 1 PID: 939 Comm: softlockup_task Not tainted 4.14.132 #34
[ 204.215936] Hardware name: ARM-Versatile Express
[ 204.215979] [<8011149c>] (unwind_backtrace) from [<8010c330>] (show_stack+0x20/0x24)
[ 204.216011] [<8010c330>] (show_stack) from [<806ddfd8>] (dump_stack+0x8c/0xa0)
[ 204.216039] [<806ddfd8>] (dump_stack) from [<80108acc>] (show_regs+0x1c/0x20)
[ 204.216067] [<80108acc>] (show_regs) from [<801b5f08>] (watchdog_timer_fn+0x248/0x2c0)
[ 204.216095] [<801b5f08>] (watchdog_timer_fn) from [<80188f48>] (hrtimer_run_queues+0x1b8/0x370)
[ 204.216147] [<80188f48>] (hrtimer_run_queues) from [<80187f14>] (run_local_timers+0x24/0x50)
[ 204.216182] [<80187f14>] (run_local_timers) from [<80187f7c>] (update_process_times+0x3c/0x6c)
[ 204.216207] [<80187f7c>] (update_process_times) from [<80197144>] (tick_periodic+0x4c/0xcc)
[ 204.216232] [<80197144>] (tick_periodic) from [<80197368>] (tick_handle_periodic+0x38/0x98)
[ 204.216256] [<80197368>] (tick_handle_periodic) from [<8010ffe4>] (twd_handler+0x40/0x50)
[ 204.216282] [<8010ffe4>] (twd_handler) from [<80172128>] (handle_percpu_devid_irq+0x98/0x24c)
[ 204.216312] [<80172128>] (handle_percpu_devid_irq) from [<8016c728>] (generic_handle_irq+0x34/0x44)
[ 204.216338] [<8016c728>] (generic_handle_irq) from [<8016cd3c>] (__handle_domain_irq+0x6c/0xc4)
[ 204.216363] [<8016cd3c>] (__handle_domain_irq) from [<80101508>] (gic_handle_irq+0x5c/0xa0)
[ 204.216385] [<80101508>] (gic_handle_irq) from [<8010d10c>] (__irq_svc+0x6c/0x90)
[ 204.216399] Exception stack(0x9e90df00 to 0x9e90df48)
[ 204.216511] df00: 00000000 9e983bc0 00000000 00208040 9f700c00 00000000 9f5fc940 9e90c000
[ 204.216617] df20: 00000000 9e983bdc 9e8fdd30 9e90df64 9e90df68 9e90df50 7f000020 80143524
[ 204.216657] df40: 00000013 ffffffff
[ 204.216697] [<8010d10c>] (__irq_svc) from [<80143524>] (kthread_should_stop+0x30/0x54)
[ 204.216730] [<80143524>] (kthread_should_stop) from [<7f000020>] (softlockup_task_fn+0x20/0x30 [softlockup_example])
[ 204.216766] [<7f000020>] (softlockup_task_fn [softlockup_example]) from [<80143ae0>] (kthread+0x144/0x174)
[ 204.216791] [<80143ae0>] (kthread) from [<80107ee8>] (ret_from_fork+0x14/0x2c)
从上面信息可以看到，soft lockup 机制是依赖于中断的：当前 CPU 被占住无法调度的情形下，只有中断才能打断它，而用于检测的定时器接口，正是在中断上下文下执行。另外，只有内核代码问题才能导致 soft lockup ，因为用户空间代码无法关闭抢占，也总是会被抢占的。 一旦发生了 soft lockup ，可以肯定这是内核问题代码导致的。
5. soft lockup 用户空间接口
/proc/sys/kernel/soft_watchdog # 是否启用 softlockup 功能
/proc/sys/kernel/softlockup_panic # lockup是否导致kernel panic
/proc/sys/kernel/softlockup_all_cpu_backtrace # 是否不只输出lockup CPU调用栈，还输出其它CPU调用栈
/proc/sys/kernel/watchdog_thresh # softlockup 触发时间阈值
/proc/sys/kernel/watchdog_cpumask # 启用 softlockup 检测的CPU掩码
GitHub 加速计划 / li / linux-dash
10.39 K
1.2 K
下载
A beautiful web dashboard for Linux
最近提交(Master分支：22 天前 )
186a802e
added ecosystem file for PM2 4 年前
5def40a3
Add host customization support for the NodeJS version 4 年前
旨在为数千万中国开发者提供一个无缝且高效的云端环境，以支持学习、使用和贡献开源项目。
更多推荐
1
0- 0
已为社区贡献3条内容
所有评论(0)