Re: misc/49738 rc.subr leaks filedescriptors

[Robert Elz <kre%munnari.OZ.AU@localhost>]

The following reply was made to PR misc/49738; it has been noted by GNATS.

From: Robert Elz <kre%munnari.OZ.AU@localhost>
To: christos%zoulas.com@localhost (Christos Zoulas)
Cc: gnats-bugs%NetBSD.org@localhost, misc-bug-people%netbsd.org@localhost, gnats-admin%netbsd.org@localhost,
        netbsd-bugs%netbsd.org@localhost, mlelstv%serpens.de@localhost
Subject: Re: misc/49738 rc.subr leaks filedescriptors
Date: Fri, 21 Oct 2016 06:25:35 +0700

     Date:        Thu, 20 Oct 2016 11:02:12 -0400
     From:        christos%zoulas.com@localhost (Christos Zoulas)
     Message-ID:  <20161020150212.8EFD617FDAB%rebar.astron.com@localhost>
 
   | How about:
   | http://www.netbsd.org/~christos/cloexec.diff and adding:
   | 
   | cloexec +7 +8 in /etc/rc?
 
 It would want to include fd 9 as well ... but at the expense of adding
 more non-standard baggage to NetBSD's sh, and having the rc system not
 workable on any other sh, then that would work.
 
 I believe that mostly the rc system correctly closed the fd's it opens
 (which is the standard way to handle this) - but that this fails when the
 rc.d/script has its oeb start function (as for example, lvm does) which
 knows nothing about needing to close rc maintained (they are not from rc.subr
 the Subject of this PR is incorrect) file descriptors.
 
 An alternative would be to add a new function to the rc system for scripts
 to call, to close any fd's that the rc system has opened, and then have
 the relevant scripts call that function - it could be made reasonably general
 to allow for other similar future issues, perhaps "rc_exec_cleanup" or
 something.
 
 kre