文件/套接字描述符表

Question

我很想知道每进程文件/套接字描述符表是如何在Linux中实现的。具体来说，是使用哪种数据结构和算法来实现它并保持它的效率。

提前感谢!

Answer 1

进程打开的文件由struct files_struct管理，该struct位于进程的struct task_struct

struct task_struct {
    ...
    /* open file information */
    struct files_struct *files;

每个进程的文件描述符表(fdt)在struct files_struct

struct files_struct {
    ...
    struct fdtable __rcu *fdt;

当一个进程试图打开一个文件时，它发出一个打开的系统调用。它将调用sys_open。这基本上是代码流:

sys_open(filename, …)
    // 1)   copy filename from user space
    getname(filename)
            strncpy_from_user()
    // 2)  get first unused file descriptor (will be returned to process)
    int fd = get_unused_fd()
        struct files_struct *files = current->files
    // 3) get file from filesystem
    struct file *f = file_open(filename)
        open_namei
            // lookup operation for filesystem
            dentry = cached_lookup or real_lookup 
            // initializes file struct
            dentry_open 
    // 4) install file returned by filesystem into file descriptor table for the process
    fd_install
        current->files->fd[fd] = file

进程得到打开文件的文件描述符表的索引。