关键词:kthread、irq、ksoftirqd、kworker、workqueues
在使用ps查看线程的时候,会有不少[...]名称的线程,这些有别于其它线程,都是内核线程。
其中多数内核线程从名称看,就知道其主要功能。
比如给中断线程化使用的irq内核线程,软中断使用的内核线程ksoftirqd,以及work使用的kworker内核线程。
本文首先概览一下Linux都有哪些内核线程,然后分析创建内核线程的API。
在介绍内核线程和普通线程都有哪些区别?
最后介绍主要内核线程(irq/ksoftirqd/kworker/)的创建过程及其作用。
1. ps下初步认识Linux内核线程
在ps -a会显示如下,可以看出内核线程都用[...]标注。
并且pid=1的init进程是所有用户空间进程的父进程;pid=2的kthreadd内核线程是所有内核线程的父线程。
内核线程分为几大类:softirq、kworker、irq及其他。
PID USER TIME COMMAND 1 0 0:01 {linuxrc} init 2 0 0:00 [kthreadd] 3 0 0:00 [ksoftirqd/0] 4 0 0:00 [kworker/0:0] 5 0 0:00 [kworker/0:0H] 6 0 0:00 [kworker/u8:0] 7 0 0:00 [rcu_sched] 8 0 0:00 [rcu_bh] 9 0 0:00 [migration/0] 10 0 0:00 [migration/1] 11 0 0:00 [ksoftirqd/1] 12 0 0:00 [kworker/1:0] 13 0 0:00 [kworker/1:0H] 14 0 0:00 [migration/2] 15 0 0:00 [ksoftirqd/2] 16 0 0:00 [kworker/2:0] 17 0 0:00 [kworker/2:0H] 18 0 0:00 [migration/3] 19 0 0:00 [ksoftirqd/3] 20 0 0:00 [kworker/3:0] 21 0 0:00 [kworker/3:0H] 22 0 0:00 [khelper] 23 0 0:00 [kdevtmpfs] 24 0 0:00 [perf] 25 0 0:00 [kworker/u8:1] 279 0 0:00 [khungtaskd] 280 0 0:00 [writeback] 281 0 0:00 [kintegrityd] 282 0 0:00 [kworker/0:1] 284 0 0:00 [bioset] 286 0 0:00 [kblockd] 294 0 0:00 [ata_sff] 408 0 0:00 [rpciod] 409 0 0:00 [kworker/2:1] 410 0 0:00 [kworker/1:1] 412 0 0:00 [kswapd0] 416 0 0:00 [fsnotify_mark] 429 0 0:00 [nfsiod] 449 0 0:00 [kworker/3:1] 527 0 0:00 [kpsmoused] 537 0 0:00 [kworker/1:2] 613 0 0:00 [deferwq]
2. kthreadd以及创建内核线程API
2.1 kthreadd:kthreadd内核线程的创建
内核其他线程的创立,要基于kthreadd。kthreadd线程是其他线程的父线程。
start_kernel-->rest_init如下:
static noinline void __init_refok rest_init(void) { int pid; rcu_scheduler_starting(); /* * We need to spawn init first so that it obtains pid 1, however * the init task will end up wanting to create kthreads, which, if * we schedule it before we create kthreadd, will OOPS. */ kernel_thread(kernel_init, NULL, CLONE_FS);--------------------------------创建第一个用户空间线程init numa_default_policy(); pid = kernel_thread(kthreadd, NULL, CLONE_FS | CLONE_FILES);---------------创建第一个内核线程kthreadd rcu_read_lock(); kthreadd_task = find_task_by_pid_ns(pid, &init_pid_ns);--------------------kthreadd_task指向kthreadd的task_strcut结构体 rcu_read_unlock(); complete(&kthreadd_done);--------------------------------------------------在init进程kernel_init-->kernel_init_freeable中等待kthreadd_done释放 /* * The boot idle thread must execute schedule() * at least once to get things moving: */ init_idle_bootup_task(current); schedule_preempt_disabled(); /* Call into cpu_idle with preempt disabled */ cpu_startup_entry(CPUHP_ONLINE); }
kernel_init在kthreadd之前启动,但是kernel_init的很多任务需要基于kthreadd。所以在kernel_init的开头等待reset_init的kthreadd_done完成量。
因为kernel_init-->kernel_init_freeable-->do_basic_setup-->do_initcalls中很多初始化需要kthread_create支援。
kernel_init-->kernel_init_freeable: static noinline void __init kernel_init_freeable(void) { /* * Wait until kthreadd is all set-up. */ wait_for_completion(&kthreadd_done);-------------------等待kthreadd_done完成量 ...
do_basic_setup();---------------------------------------很多初始化需要kthread_create支持
...
}
内核中有一个线程kthreadd_task负责创建其他内核线程,这个线程的函数为kthreadd()。
int kthreadd(void *unused) { struct task_struct *tsk = current; /* Setup a clean context for our children to inherit. */ set_task_comm(tsk, "kthreadd"); ignore_signals(tsk); set_cpus_allowed_ptr(tsk, cpu_all_mask); set_mems_allowed(node_states[N_MEMORY]); current->flags |= PF_NOFREEZE; for (;;) { set_current_state(TASK_INTERRUPTIBLE); if (list_empty(&kthread_create_list)) schedule();----------------------------------------------如果kthread_create_list为空,让出CPU,进入休眠状态。在kthread_create_on_node()中会将要创建进程节点加入到kthread_create_list中,然后唤醒此进程。 __set_current_state(TASK_RUNNING); spin_lock(&kthread_create_lock); while (!list_empty(&kthread_create_list)) {------------------只要kthread_create_list不为空,遍历kthread_create_list链表 struct kthread_create_info *create; create = list_entry(kthread_create_list.next, struct kthread_create_info, list); list_del_init(&create->list);----------------------------从kthread_create_list中摘除当前create spin_unlock(&kthread_create_lock); create_kthread(create);----------------------------------创建线程 spin_lock(&kthread_create_lock); } spin_unlock(&kthread_create_lock); } return 0; } static void create_kthread(struct kthread_create_info *create) { int pid; #ifdef CONFIG_NUMA current->pref_node_fork = create->node; #endif /* We want our own signal handler (we take no signals by default). */ pid = kernel_thread(kthread, create, CLONE_FS | CLONE_FILES | SIGCHLD);----调用do_fork()创建线程 if (pid < 0) { /* If user was SIGKILLed, I release the structure. */ struct completion *done = xchg(&create->done, NULL); if (!done) { kfree(create); return; } create->result = ERR_PTR(pid); complete(done);--------------------------------------------------------触发complete事件 } } pid_t kernel_thread(int (*fn)(void *), void *arg, unsigned long flags) { return do_fork(flags|CLONE_VM|CLONE_UNTRACED, (unsigned long)fn, (unsigned long)arg, NULL, NULL); }
2.2 创建内核线程接口:kthread_create等
kthread_create()是最常见的创建内核线程的接口。
kthread_create_on_cpu()相对于kthread_create多了个cpu,但都基于kthread_create_on_node()。
kthread_run基于kthreadd_create,所以这些函数都基于kthread_create_on_node。
#define kthread_create(threadfn, data, namefmt, arg...) kthread_create_on_node(threadfn, data, -1, namefmt, ##arg) struct task_struct *kthread_create_on_cpu(int (*threadfn)(void *data), void *data, unsigned int cpu, const char *namefmt); /** * kthread_run - create and wake a thread. * @threadfn: the function to run until signal_pending(current). * @data: data ptr for @threadfn. * @namefmt: printf-style name for the thread. * * Description: Convenient wrapper for kthread_create() followed by * wake_up_process(). Returns the kthread or ERR_PTR(-ENOMEM). */ #define kthread_run(threadfn, data, namefmt, ...) ({ struct task_struct *__k = kthread_create(threadfn, data, namefmt, ## __VA_ARGS__); if (!IS_ERR(__k)) --------------------------如果kthread_create()正确创建了一个进程,调用wake_up_process()唤醒它。 wake_up_process(__k); __k; })
kthread_create_on_node()负责创建一个线程,填充一个kthread_create_info结构体;然后将此结构体作为一个节点插入kthread_create_list队尾。
然后唤醒kthreadd_task进行处理,创建线程。
struct task_struct *kthread_create_on_node(int (*threadfn)(void *data), void *data, int node, const char namefmt[], ...) { DECLARE_COMPLETION_ONSTACK(done); struct task_struct *task; struct kthread_create_info *create = kmalloc(sizeof(*create), GFP_KERNEL);---------------------------------创建插入kthread_create_list的节点。 if (!create) return ERR_PTR(-ENOMEM); create->threadfn = threadfn; create->data = data; create->node = node; create->done = &done; spin_lock(&kthread_create_lock); list_add_tail(&create->list, &kthread_create_list);-------------------将填充的节点插入kthread_create_list中。 spin_unlock(&kthread_create_lock); wake_up_process(kthreadd_task);---------------------------------------唤醒kthread_task处理kthread_create_list链表,创建相应的线程。 /* * Wait for completion in killable state, for I might be chosen by * the OOM killer while kthreadd is trying to allocate memory for * new kernel thread. */ if (unlikely(wait_for_completion_killable(&done))) {------------------等待complete事件触发,在create_kthread()中触发。 /* * If I was SIGKILLed before kthreadd (or new kernel thread) * calls complete(), leave the cleanup of this structure to * that thread. */ if (xchg(&create->done, NULL)) return ERR_PTR(-EINTR); /* * kthreadd (or new kernel thread) will call complete() * shortly. */ wait_for_completion(&done);---------------------------------------等待complete事件触发。 } task = create->result;------------------------------------------------创建的结果为task_struct结构体。 if (!IS_ERR(task)) { static const struct sched_param param = { .sched_priority = 0 }; va_list args; va_start(args, namefmt); vsnprintf(task->comm, sizeof(task->comm), namefmt, args);---------配置进程名称。 va_end(args); /* * root may have changed our (kthreadd's) priority or CPU mask. * The kernel thread should not inherit these properties. */ sched_setscheduler_nocheck(task, SCHED_NORMAL, ¶m);-----------设置进程调度策略为NORMAL,优先级为0。 set_cpus_allowed_ptr(task, cpu_all_mask); } kfree(create);--------------------------------------------------------释放kthread_create_info。 return task; }
3. 内核线程和普通线程的区别
内核线程没有地址空间,所以task_struct->mm指针为NULL。内核线程没有用户上下文。
内核线程只工作在内核空间,不会切换至用户空间。但内核线程同样是可调度且可抢占的。
普通线程即可工作在内核空间,也可工作在用户空间。
内核线程只能访问3GB以上地址,而普通线程可访问所有4GB地址空间。
4. irq、softirq、woker内核线程
irq、softirq、worker都可能创建对应的内核线程,有线程就有优先级。
下面从优先来来看看它们的重要性。
可以看出中断内核线程优先级很高,为49,并且使用了实时调度策略。softirq和worker都是普通内核线程。
prio | policy | |
irq | 49 | SCHED_FIFO |
softirq | 120 | SCHED_NORMAL |
worker | 120 | SCHED_NORMAL |
init | 120 | SCHED_NORMAL |
kthreadd | 120 | SCHED_NORMAL |
cfinteractive | 0 | SCHED_FIFO |
其它特殊内核线程init优先级为120,kthreadd优先级为120.
cfinteractive优先级最高,主要处理CPU Frequency负载更新。
4.1 irq/xx-xx:创建处理线程化中断的线程
request_threaded_irq-->__setup_irq,可见如果设置了thread_fn,并且不允许中断嵌套,则创建一个类似"irq/中断号-终端名称"的线程。
线程函数是irq_thread,
/* * Internal function to register an irqaction - typically used to * allocate special interrupts that are part of the architecture. */ static int __setup_irq(unsigned int irq, struct irq_desc *desc, struct irqaction *new) { ... if (new->thread_fn && !nested) { struct task_struct *t; static const struct sched_param param = { .sched_priority = MAX_USER_RT_PRIO/2, }; t = kthread_create(irq_thread, new, "irq/%d-%s", irq,----------------在irq_thread中调用irq_thread_fn,进而调用action->thread_fn,request_threaded_irq参数thread_fn。 new->name); ... } ... }
request_irq是对request_threaded_irq的封装,创建中断线程的工作交给__setup_irq()。
static inline int __must_check request_irq(unsigned int irq, irq_handler_t handler, unsigned long flags, const char *name, void *dev) { return request_threaded_irq(irq, handler, NULL, flags, name, dev); }
更详细信息参考:《Linux中断管理 (1)Linux中断管理机制》中关于request_irq()介绍。
4.2 ksoftirqd/xx:创建处理软中断线程
软中断线程通过smpboot_register_percpu_thread注册softirq_threads创建。
static struct smp_hotplug_thread softirq_threads = { .store = &ksoftirqd, .thread_should_run = ksoftirqd_should_run, .thread_fn = run_ksoftirqd, .thread_comm = "ksoftirqd/%u", }; static __init int spawn_ksoftirqd(void) { register_cpu_notifier(&cpu_nfb); BUG_ON(smpboot_register_percpu_thread(&softirq_threads)); return 0; }
smpboot_register_percpu_thread-->__smpboot_create_thread,最终也还是调用kthread_create_on_cpu,创建了类似"ksoftirqd/xx"的内核线程,xx为cpuid号。
从ps -a中可以看出创建的结果如下,可以看出每个CPU创建了一个ksoftirqd内核线程。
3 0 0:03 [ksoftirqd/0] 11 0 0:03 [ksoftirqd/1] 15 0 0:00 [ksoftirqd/2] 19 0 0:00 [ksoftirqd/3]
更详细信息参考: 《Linux中断管理 (2)软中断和tasklet》
4.3 kworker:创建work的工作线程
kwoker线程是处理work的工作线程,详细参考《Linux中断管理 (3)workqueue工作队列》。
每个CPU都会创建自己的workqueue,用以集中处理内核kworker。
workquuue就是把一些任务(work)推迟到一个或一组内核线程中去执行,那个内核线程被称为worker_thread。
首先看看创建结果,可以看出在init_workqueues中创建了绑定CPU0的两个kworker,分别是nice=0和nice=-20。
apply_workqueue_attrs创建unbund worker,即kworker/u8:0。
然后在每个CPU_UP_PREPARE回调中创建两个不同nice的kworker。所以四个CPU一共9个内核线程。
PID USER TIME COMMAND 1 0 0:01 {linuxrc} init 2 0 0:00 [kthreadd] 3 0 0:00 [ksoftirqd/0] 4 0 0:00 [kworker/0:0] 5 0 0:00 [kworker/0:0H]---------------init_workqueues-->create_worker 6 0 0:00 [kworker/u8:0]---------------apply_workqueue_attrs-->alloc_unbound_pwq-->create_worker 7 0 0:00 [rcu_sched] 8 0 0:00 [rcu_bh] 9 0 0:00 [migration/0] 10 0 0:00 [migration/1] 11 0 0:00 [ksoftirqd/1] 12 0 0:00 [kworker/1:0]---------------workqueue_cpu_up_callback-->create_worker 13 0 0:00 [kworker/1:0H] 14 0 0:00 [migration/2] 15 0 0:00 [ksoftirqd/2] 16 0 0:00 [kworker/2:0] 17 0 0:00 [kworker/2:0H]--------------workqueue_cpu_up_callback-->create_worker 18 0 0:00 [migration/3] 19 0 0:00 [ksoftirqd/3] 20 0 0:00 [kworker/3:0] 21 0 0:00 [kworker/3:0H]--------------workqueue_cpu_up_callback-->create_worker 22 0 0:00 [khelper] 23 0 0:00 [kdevtmpfs] 24 0 0:00 [perf] 25 0 0:00 [kworker/u8:1]--------------worker_thread-->create_worker 279 0 0:00 [khungtaskd] 280 0 0:00 [writeback] 281 0 0:00 [kintegrityd] 282 0 0:00 [kworker/0:1]---------------worker_thread-->create_worker 284 0 0:00 [bioset] 286 0 0:00 [kblockd] 294 0 0:00 [ata_sff] 408 0 0:00 [rpciod] 409 0 0:00 [kworker/2:1]---------------worker_thread-->create_worker 410 0 0:00 [kworker/1:1]---------------worker_thread-->create_worker 412 0 0:00 [kswapd0] 416 0 0:00 [fsnotify_mark] 429 0 0:00 [nfsiod] 449 0 0:00 [kworker/3:1]---------------worker_thread-->create_worker 527 0 0:00 [kpsmoused] 537 0 0:00 [kworker/1:2]---------------worker_thread-->create_worker 613 0 0:00 [deferwq]
init_workqueues-->create_worker-->kthread_create_on_node,创建"kworker/xx:xxH"内核线程。
static int __init init_workqueues(void) { int std_nice[NR_STD_WORKER_POOLS] = { 0, HIGHPRI_NICE_LEVEL }; int i, cpu; ... /* create the initial worker */ for_each_online_cpu(cpu) {---------------------------------遍历CPU[0~3] struct worker_pool *pool; for_each_cpu_worker_pool(pool, cpu) {------------------NR_STD_WORKER_POOLS=2,所以每个CPU有两个pool pool->flags &= ~POOL_DISASSOCIATED; BUG_ON(!create_worker(pool)); } } ... system_wq = alloc_workqueue("events", 0, 0); system_highpri_wq = alloc_workqueue("events_highpri", WQ_HIGHPRI, 0); system_long_wq = alloc_workqueue("events_long", 0, 0); system_unbound_wq = alloc_workqueue("events_unbound", WQ_UNBOUND, WQ_UNBOUND_MAX_ACTIVE); system_freezable_wq = alloc_workqueue("events_freezable", WQ_FREEZABLE, 0); system_power_efficient_wq = alloc_workqueue("events_power_efficient", WQ_POWER_EFFICIENT, 0); system_freezable_power_efficient_wq = alloc_workqueue("events_freezable_power_efficient", WQ_FREEZABLE | WQ_POWER_EFFICIENT, 0); BUG_ON(!system_wq || !system_highpri_wq || !system_long_wq || !system_unbound_wq || !system_freezable_wq || !system_power_efficient_wq || !system_freezable_power_efficient_wq); return 0; }
create_worker()函数创建工作线程。
static struct worker *create_worker(struct worker_pool *pool) { ... if (pool->cpu >= 0) snprintf(id_buf, sizeof(id_buf), "%d:%d%s", pool->cpu, id,-------------cpuid和id,区分cpu和cpu内kworker。 pool->attrs->nice < 0 ? "H" : ""); else snprintf(id_buf, sizeof(id_buf), "u%d:%d", pool->id, id);--------------u表示不指定cpu。 worker->task = kthread_create_on_node(worker_thread, worker, pool->node, "kworker/%s", id_buf); ... }
更详细信息参考:《Linux中断管理 (3)workqueue工作队列》、《Linux workqueue工作原理》、《Concurrency Managed Workqueue之(一):workqueue的基本概念》
5. 其他内核线程
rcu_sched、rcu_bh
migration
khelper
kdevtmpfs
perf
writeback
kintegrityd
bioset
kblockd
ata_sff
rpciod
kswapd
nfsiod
kpsmpused
deferwq