5

I'm slowly creating a mental model of how linux works, and I've tried to simplify the most, my model of how linux works. Let's assume that the kernel boots and inits the only software it's gonna run: the terminal. Let's assume that this terminal has the capabilities of appearing on the screen and rendering some text, and also, of course, getting input from the keyboard. Let's also assume that I type the name of an executable, and it knows where it's in memory. Now, how is the terminal able run this program? In my mental model, I thougth of the following:

The terminal is a program, which means it can do system calls. So it uses the fork() system call and creates a new process in the kernel. Then, it somehow makes this process run the code of my program. Now, how it printf() able to show text on my terminal live while the program runs?

jordanm
  • 42,678
Emilio Magrito
  • 51

1 Answers1

6

Your understanding is fairly accurate. The shell uses the clone() system call to create a new process. The manpage describes it's difference from fork():

Unlike fork(2), clone() allows the child process to share parts of its execution context with the calling process, such as the memory space, the table of file descriptors, and the table of signal handlers. (Note that on this manual page, "calling process" normally corresponds to "parent process".

It then uses an execve() system call to replace the current child process image with a new process image. This system call is what makes the process run the code of your program.

When a process forks the file descriptors of the parent are copied. From the fork(2) manual page:

The child inherits copies of the parent's set of open file descriptors.
Each file descriptor in the child refers to the same open file description (see open(2)) as the corresponding file descriptor in the parent. This means that the two descriptors share open file status flags, current file offset, and signal-driven I/O attributes (see the description of F_SETOWN and F_SETSIG in fcntl(2)).

This is why text is displayed to your terminal when a program writes to stdout. You can see this process happen using the strace program in Linux. Here are the main excerpts from running strace on a bash process in Linux and executing /bin/echo foo within the shell.

21:32:20 clone(child_stack=0, flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD, child_tidptr=0x7f3f419f19d0) = 32036
Process 32036 attached
[pid 32017] 21:32:20 wait4(-1,  <unfinished ...>
[pid 32036] 21:32:20 execve("/bin/echo", ["/bin/echo", "foo"], ["XDG_VTNR=8", "KDE_MULTIHEAD=false", "XDG_SESSION_ID=5512", "SSH_AGENT_PID=30259", "DM_CONTROL=/var/run/xdmctl", "TERM=xterm", "SHELL=/bin/bash", "XDM_MANAGED=method=classic", "XDG_SESSION_COOKIE=5c78dafb330601d94d7556bb52a6a2a6-1450467466.154128-547622992", "HISTSIZE=50000", "KONSOLE_DBUS_SERVICE=:1.160", "GTK2_RC_FILES=/etc/gtk-2.0/gtkrc:/home/jordan/.gtkrc-2.0:/home/jordan/.kde/share/config/gtkrc-2.0", "KONSOLE_PROFILE_NAME=Shell", "GTK_RC_FILES=/etc/gtk/gtkrc:/home/jordan/.gtkrc:/home/jordan/.kde/share/config/gtkrc", "GS_LIB=/home/jordan/.fonts", "WINDOWID=92274714", "SHELL_SESSION_ID=5b72a0038b0c4000a9299cae82f340a2", "KDE_FULL_SESSION=true", "USER=jordan", "SSH_AUTH_SOCK=/tmp/ssh-JEjo6RVmNhvR/agent.30205", "SESSION_MANAGER=local/tesla:@/tmp/.ICE-unix/30329,unix/tesla:/tmp/.ICE-unix/30329", "PATH=/home/jordan/.gem/ruby/1.9.1/bin:/home/jordan/.gem/ruby/1.9.1/bin:/home/jordan/bin:/home/jordan/local/packer:/home/jordan/local/bin:/usr/local/bin:/usr/bin:/bin:/usr/local/games:/usr/games:/sbin:/usr/local/sbin:/usr/sbin:/home/jordan/.rvm/bin:/home/jordan/prog/go/bin:/home/jordan/.rvm/bin:/home/jordan/prog/go/bin", "DESKTOP_SESSION=kde-plasma", "PWD=/home/jordan/games", "WORKING=/home/jordan/prog/greenspan", "KONSOLE_DBUS_WINDOW=/Windows/1", "EDITOR=emacs -nw", "LANG=en_US.UTF-8", "KDE_SESSION_UID=1000", "PS1=\\[\\033[01;32m\\]\\u@\\h\\[\\033[01;34m\\] \\w\\[\\033[1;31m\\]$(__git_ps1)\\[\\033[01;34m\\] \\$\\[\\033[00m\\] ", "KONSOLE_DBUS_SESSION=/Sessions/1", "SHLVL=2", "XDG_SEAT=seat0", "COLORFGBG=15;0", "HOME=/home/jordan", "LANGUAGE=", "KDE_SESSION_VERSION=4", "GOROOT=/home/jordan/local/go", "XCURSOR_THEME=oxy-zion", "LOGNAME=jordan", "DBUS_SESSION_BUS_ADDRESS=unix:abstract=/tmp/dbus-onouV6Cc66,guid=bcdceeabe7aa00a28d55899f5674608a", "XDG_DATA_DIRS=/usr/share:/usr/share:/usr/local/share", "GOPATH=/home/jordan/prog/go", "PROMPT_COMMAND=history -a", "WINDOWPATH=8", "DISPLAY=:0", "XDG_RUNTIME_DIR=/run/user/1000", "PROFILEHOME=", "QT_PLUGIN_PATH=/home/jordan/.kde/lib/kde4/plugins/:/usr/lib/kde4/plugins/", "XDG_CURRENT_DESKTOP=KDE", "HISTTIMEFORMAT=%F %T: ", "_=/bin/echo"]) = 0
[pid 32036] 21:32:20 write(1, "foo\n", 4) = 4
[pid 32036] 21:32:20 exit_group(0)      = ?
[pid 32036] 21:32:20 +++ exited with 0 +++
21:32:20 <... wait4 resumed> [{WIFEXITED(s) && WEXITSTATUS(s) == 0}], WSTOPPED|WCONTINUED, NULL) = 32036
jordanm
  • 42,678
  • I got a question too: does the output of the child pass through the shell itself? I think that since the child inherits the file descriptors from the shell, the kernel sends its output directly to the terminal, and the shell can't even know what the output is, but I don't know if I'm right. (unless, of course, the program is called through $(program), or anything like that) – fonini Dec 20 '15 at 05:47