简体繁体 English

Linux中的线程

[英]Threads in Linux

原文 2012-11-17 17:34:16 6 2 linux

I have gone through mostly all the questions on here regarding the topic of Pthreads in Linux but there is a basic doubt which remains unresolved for me: 我在这里主要讨论了关于Linux中Pthreads主题的所有问题，但是有一个基本的疑问对我来说仍然没有解决：

It is mentioned in various responses that when we create a POSIX thread on Linux, there is a 1:1 mapping between user thread and kernel thread. 在各种响应中提到，当我们在Linux上创建POSIX线程时，用户线程和内核线程之间存在1：1的映射。

My doubt is when we use pthread_create() in Linux, is there 1 user thread and a corresponding unique kernel thread created implicitly (ie a total of 2 threads, one of which is invisible to the user) ? 我怀疑的是，当我们在Linux中使用pthread_create()时，是否存在1个用户线程和隐式创建的相应唯一内核线程（即总共2个线程，其中一个对用户是不可见的） ？

OR 要么

There only one kernel thread created and there is nothing anymore in newer Linux kernels called a user thread? 只创建了一个内核线程，在称为用户线程的新Linux内核中什么也没有了？

2 个解决方案

The NPTL (Native POSIX Thread Library) and the older LinuxThreads both use a 1:1 model, where each threads (or process) created by the user corresponds with one schedulable entity in the kernel. NPTL（Native POSIX线程库）和较旧的LinuxThreads都使用1：1模型，其中用户创建的每个线程（或进程）对应于内核中的一个可调度实体。

However, you maybe confused by user-level threads, or fibers , those are threads of execution created via calls like makecontext() and swapcontext() that have N:1 model, the kernel doesn't know about user-level threads and their scheduling is done in user-space. 但是，您可能会对用户级线程或光纤感到困惑，这些线程是通过具有N：1模型的makecontext()和swapcontext()等调用创建的执行线程 ，内核不了解用户级线程及其调度在用户空间中完成。

pthread_create() internally calls fork(), and vfork()/fork() internally calls clone(). pthread_create（）在内部调用fork（），而vfork（）/ fork（）在内部调用clone（）。 So in most of the case it is 1:1 mapping. 所以在大多数情况下它是1：1映射。