docs/51133: RSH environmental variable not described in rcmd(3)

[David Magda]

On Sun, May 04, 2003 at 02:46:50PM -0700, Murray Stokely wrote:
> Yes we should update these man pages.  Is there any chance the
> submitter can give us a patch to update the pages?  If not, I will
> work on this.

Don't know anything about groff (?) formatting so here's the text
and I'll let someone who knows it get the details right.

In the DESCRIPTION section, insert the following after the first
sentence in first paragraph:

	If the calling process is not setuid, the RSH environment
	variable is set, and inport is ``shell/tcp'', rcmdsh(3) is
	called instead with the value of RSH. Alternately, if the
	user is not the superuser, rcmd() will invoke rcmdsh(3) to
	run the command via rsh(1). While rcmd() can handle IPv4
	cases only, the rcmd_af() function can handle other cases as
	well.

Then start a new paragraph. Continue of with the rest of the text of
the man page.

This will basically make FreeBSD's man page match OpenBSD's 
rcmd(3) [1].


Next, create an ENVIRONMENT section and add the following text:

      RSH    When using the rcmd() function, this variable is used as
	     the program to run instead of rsh(1).

This part is taken from NetBSD's rcmd(3) [2]: the main change is that
Net has a program rcmd(1) so that users could call rcmd(3) without
having to be the super-user, and Net uses the variable RCMD_CMD.


The above ENVIRONMENT section should also be added to dump(8) and
restore(8) since they will run $RSH if it is defined. I sent an email
suggesting so to docs@ but haven't received a reply and nothing's
been committed to CVS. Doing a quick find/grep, rdist(1) also uses
rcmd(3) so would probably be effective by $RSH.


Also, just curious. Why does OpenBSD have the following code in
rcmd.c [3]:

        if (!issetugid() && (p = getenv("RSH")) && *p) {
                struct servent *sp = getservbyname("shell", "tcp");

While FreeBSD has the following [4]:

      if (!issetugid() && (p = getenv("RSH"))) {
		 struct servent *sp = getservbyname("shell", "tcp");

The difference being the "&& *p".

This difference was introduced by the OpenBSD (by millet) in their
rev. 1.37 [5] with the commit message of "Make sure $RSH is not the
empty string."

How beneficial would it be if this was added to FreeBSD's code?

[1] http://www.openbsd.org/cgi-bin/man.cgi?query=rcmd
[2] http://netbsd.gw.com/cgi-bin/man-cgi?rcmd+3
[3] http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/net/rcmd.c?rev=1.45
[4] http://www.freebsd.org/cgi/cvsweb.cgi/src/lib/libc/net/rcmd.c?rev=1.38
[5] http://www.openbsd.org/cgi-bin/cvsweb/src/lib/libc/net/rcmd.c.diff?r1=1.36&r2=1.37