Thread: Write a service in Linux

Originally Posted by MisterIO

Yes, I do "command > /dev/null &". Why would you want to have a daemon print its output to the console? Clearly, if you don't redirect the output, it screws things up.

Note that the CONSOLE is different from a TTY. CONSOLE should not close, _ever_ (keeping in mind that if you don't have one, it still NEVER closes). In fact, it even stays alive after the kernel has shut down (if doing a shutdown -h now), and can still display messages. Having a daemon print to the console is done quite often.

The reason that your jobs stop if you have started background jobs is that you started the jobs as a non-root user. If you start a job in background mode (either by & or by CTRL-Z and then bg) as root, init will adopt the job believing it to be a critical function.

Originally Posted by Kennedy

Note that the CONSOLE is different from a TTY. CONSOLE should not close, _ever_ (keeping in mind that if you don't have one, it still NEVER closes). In fact, it even stays alive after the kernel has shut down (if doing a shutdown -h now), and can still display messages. Having a daemon print to the console is done quite often.

The reason that your jobs stop if you have started background jobs is that you started the jobs as a non-root user. If you start a job in background mode (either by & or by CTRL-Z and then bg) as root, init will adopt the job believing it to be a critical function.

Yeah, I wasn't talking about consoles as in "CTRL+ALT+F1",etc., because those can never be closed and the post I was answering to was talking about closing the terminal.
But what do you mean that the console stays alive after the kernel has shut down?
Consoles stay alive even during the system is going to a halt, but when the kernel finally halts, no consoles remain.

Originally Posted by MisterIO

Yeah, I wasn't talking about consoles as in "CTRL+ALT+F1",etc., because those can never be closed and the post I was answering to was talking about closing the terminal.
But what do you mean that the console stays alive after the kernel has shut down?
Consoles stay alive even during the system is going to a halt, but when the kernel finally halts, no consoles remain.

ALT-F1 is a tty, as is ALT-F2 - ALT-F7 (F7 is mostly used for X<something>). THE console, however, is how the kernel talks to the world. There is a big difference between the console and a tty.

As far as CTRL+ALT+F1, I'm assuming you are referring to "shelling" out of X<something>. That is still not the console and YES these can be closed. It only requires you editing /etc/inittab and commenting out the shells that you don't want to run, then send init a hang-up (killall -HUP init).

Originally Posted by Kennedy

ALT-F1 is a tty, as is ALT-F2 - ALT-F7 (F7 is mostly used for X<something>). THE console, however, is how the kernel talks to the world. There is a big difference between the console and a tty.

As far as CTRL+ALT+F1, I'm assuming you are referring to "shelling" out of X<something>. That is still not the console and YES these can be closed. It only requires you editing /etc/inittab and commenting out the shells that you don't want to run, then send init a hang-up (killall -HUP init).

I know that I can change their number, but that's different from closing them. It seems as if by "The Console" you mean simply the interface that the kernel uses to do output, but how is that related to the much simpler discussion above about how to deamonize a program?

Originally Posted by MisterIO

I know that I can change their number, but that's different from closing them. It seems as if by "The Console" you mean simply the interface that the kernel uses to do output, but how is that related to the much simpler discussion above about how to deamonize a program?

Because you asked the question

Originally Posted by MisterIO

Why would you want to have a daemon print its output to the console?

And the answer is because you may want to send messages to the console after the controlling tty has died.

[EDIT]More info: check out the specs on /dev/console[/EDIT]

Originally Posted by Kennedy

And the answer is because you may want to send messages to the console after the controlling tty has died.

Thanks for this.

Thanks all!!! My daemon have to monitor other users and maybe logoff one of them,
remove some files, restart the system and something like these.
So I have to be root to do these. Am I wrong?!!! I have written this service for windows
with WIN API. I want just know that is there any API like WIN in LIN to do this?

Originally Posted by MK27

Since this is most easily done with a sequence of system commands, the normal and sane way to do this would be with a shell script, so there is no such API. Many linux daemons are (more and less complicated) bash shell (aka. the "command language interpreter") scripts that use system commands, which system commands are written in C. Indeed, the bash shell is derived from an older unix power tool called "the C shell", and many "system commands" are actually part of the shell -- they do not have their own executables.

If you do a bit of research into basic system commands you will realize you do not have to write any new ones for this*, so don't bother using C -- it will just be a string of system(), popen(), etc. I would almost bet that you could learn enough bash to do this in less time than it would take to actually write an awkward and probably inefficient (since you must invoke the shell implicitly to do all that) C program. If you don't use system/exec() calls, this will take you years and ten of thousands of lines of code to re-write commands that already exist.

Advanced Bash-Scripting Guide

You can even ask questions here about that; if no one else will answer them, I will...there are some goofy things about the syntax, like it is whitespace sensitive and heavy on braces. But it has regular expressions, functions, arrays, etc.

You're right, this is a root/sys admin task. So it's about the shell. The bash shell. You don't have to write anything in C, it's already written. I promise you, that is how the pros do it and that is what the shell is for.

*check users logged on: use "who", check the time: use "date", to log a user out: use "kill" to kill all the processes with that user ID (maybe you want to send a message first with "write" or "wall"). To delete files: "rm". To wait, "sleep". To fork, "&". To take the output of a such a SHELL command (who, date) and put into an array for further manipulation analysis, for example, is simple in a SHELL script; to do it in C, a real pain in the

I agree with you, unless he's doing that just for fun. In that case, he could simply take a look at those apps code.

Maybe you are right
I'm now reading about bash and linux command line. It seems that Bash is some thing more than it's name !!!!!

Thank you for all