Are there non-interactive login shells?

Question

I know what the difference between interactive/non-interactive and login/non-login shells are, but it seems like in practice there's never going to be a non-interactive login shell unless you have something like /bin/bash --login some-script.sh in a script (and even that seems a little odd). Is this correct or are they more common?

Answer 1

I assume you're talking about Bash's concept of login vs. non-login and interactive vs. non-interactive Bash shells as described in the Invocation section of the manpage. (This is different from the interpretation in James Youngman's answer of any arbitrary command used as the "shell" (or user command interpreter) in the passwd(5) file and whether or not that program accepts user input; some, such as /usr/sbin/nologin, obviously don't.)

You are correct that /bin/bash --login some-script.sh will produce a non-interactive login Bash invocation, and this is perhaps a pathological example. There is one case, perhaps uncommon but not truly weird, that produces a non-interactive login shell: ssh somehost < some-file. Here sshd will start Bash with argv[0] set to -bash because it's not been given a command to run, causing Bash to consider itself a login shell, but, because stdin is not connected to a terminal, Bash will not set itself to interactive mode ($- will not contain i).

(Personally, that case seems far more reasonable than the converse, ssh somehost somecommand, which is not considered a "login shell" even though it's a fresh login to somehost just as the above is.)

I have recently done what I should have done long ago and put together a table of Bash's modes and what init files are run. If you're finding it confusing, take heart in that at least I do as well. It mystifies me what their original aim was with the rules about when .bashrc is executed.

Answer 2

Most login shells by count on a newly installed system are non-interactive, actually:

$ awk -F:  '{print $7}' <  /etc/passwd | sort | uniq -c
  5 /bin/bash
 23 /bin/false
  1 /bin/sh
  1 /bin/sync
 17 /usr/sbin/nologin

Clearly /bin/bash and /bin/sh are traditional shells and they're interactive. But all the other items in that list are non-interactive. If you're reading the list and don't know what one of them does, you can just look up its manual page (with for example man nologin or man sync).

What the command actually does

Looking at the Captain Man's posting history I see that they're quite inexperienced with Unix. So maybe the question in the comment about not following related to the command line at the top of the answer, not simply the output. So I'll explain the command-line too, even though it's really off-topic for this question.

The command is a Unix pipeline. A pipeline is a chain of commands - you read it left-to-right - in which the output of the first command becomes the input of the second, the output of the second becomes the input of the third, and so on, until the end of the pipeline. The output of the last process is shown on the terminal (unless it's been redirected). See the Wikipedia entry on shell pipelines for more information.

If you don't understand what a pipeline is doing, you can simply run it in segments to see what's happening. You can also read the manual page for the commands which are being used (here, awk, sort and uniq). In fact, you should do that now. I'll wait.

Let's run the stages of the pipeline incrementally (you can safely do this on your own Unix system):

~$ awk -F:  '{print $7}' <  /etc/passwd | sed -e 's/^/    /'
/bin/bash
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/bin/sync
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/usr/sbin/nologin
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
/usr/sbin/nologin
[ ... and so on, I've left the rest out ... ]

The output above is simply the contents of the seventh field from the /etc/passwd file. That's the (flat-text-file) database which tells the system what everybody's login shell is. If you want to find out more about /etc/passwd just go read it (it's world-readable) and look at the manual page for it (man 5 passwd).

So reading the whole list there you can get an idea what the popular items are, but that's not a good format for an answer to this question, because the question was really about how common non-interactive shells are. Let's count them up. The simplest way to do that is to sort the items first:

~$ awk -F:  '{print $7}' <  /etc/passwd | sort
/bin/bash
/bin/bash
/bin/bash
/bin/bash
/bin/bash
/bin/false
/bin/false
/bin/false
/bin/false
/bin/false
[ ... and so on, I've left the rest out ... ]

We can use the program uniq to show us only the unique items:

~$ awk -F: '{print $7}' < /etc/passwd | sort | uniq | sed -e 's/^/ /' /bin/bash /bin/false /bin/sh /bin/sync /usr/sbin/nologin

But wait, that's no use, how many of each were there? Let's ask uniq (read the man page!):

~$ awk -F:  '{print $7}' <  /etc/passwd | sort | uniq -c
      5 /bin/bash
     23 /bin/false
      1 /bin/sh
      1 /bin/sync
     17 /usr/sbin/nologin

That's the output we saw at the top of the answer, of course. Let's sort it again to see the entries in order:

~$ awk -F:  '{print $7}' <  /etc/passwd | sort | uniq -c | sort
     17 /usr/sbin/nologin
      1 /bin/sh
      1 /bin/sync
     23 /bin/false
      5 /bin/bash

Wait, that can't be right, 17 comes before 1 and 5 after 23. The problem is that the items are being sorted lexicographically. Let's ask sort to sort them numerically, and in reverse order:

~$ awk -F:  '{print $7}' <  /etc/passwd | sort | uniq -c | sort -n -r
     23 /bin/false
     17 /usr/sbin/nologin
      5 /bin/bash
      1 /bin/sync
      1 /bin/sh

I think that explains everything in the original answer. If you're still unclear on the details of what those commands do, you can read the manual pages. If you're still unclear on the principles of what's going on, it might be better to start by reading a book (online or on paper) explaining Unix and Linux.