22 May 2007
privbind - allow an unprivileged application to bind with reserved ports.
privbind -u user [ -g group] [ -n num] [ -l path] command [ arguments ... ]
Normally in Linux, only a superuser process can bind an Internet domain socket with a reserved port (port numbers less than 1024). Accordingly, server processes are typically run with superuser privileges, which can be dropped after binding the reserved port.
privbind can execute an application as an unprivileged user with just one extra privilege: it can bind to reserved ports.
privbind is useful in several situations. It can be used when the application is not trusted enough; It can be used when the server is written in a language without the setuid(2) feature (e.g., Java(TM)); It can also be used to run applications which don’t manipulate their own user id and need to be able to bind to a reserved port without needing any other root privileges.
|-u||The -u option is mandatory, and specifies under which user to run the given command. The user can be specified using either a username or a numeric user id. It should be an unprivileged (non-root) user.|
|-g||Specifies the group to switch to when running the given command. If this option is missing, then the given user’s default group is used.|
|-n||privbind’s default behaviour is to allow the application to call bind(2) with reserved ports an unlimited number of times. In order to do that (see "HOW IT WORKS" below), the privbind helper process needs to wait for the application to exit before it terminates.
The -n num option tells privbind that it can assume that only num binds need to be given elevated privileges. After this number of bind(2) calls have been executed, privbind’s helper process will exit, leaving behind only the unprivileged application running.
|-l||Mostly for internal use during build. Gives the explicit path to the LD_PRELOAD library.|
|-h||Shows a short help screen, and exits.|
Using technical jargon, privbind execs command as its main process, running itself in the background (as a child of the application’s process). The practical upshot of this, in layman’s terms, is that the user never sees privbind’s exit status. When running privbind, the process will exit whenever, and with whatever exit status, command does.
The above point should be particularly noted when using privbind to run daemons.
privbind has no SUID parts, and runs within the confines of a single process. This serves to minimize the security implications of using it. It is strongly advised that privbind not be made SUID, as this would allow any user that can run it to run any process as any other (non-root) user. At the moment privbind detects such a situation and warns about it, but will continue with the execution.
HOW IT WORKS
In a nutshell, privbind works by starting two processes. One drops privileges and runs (exec(2)) the command, the other remains as root. Privbind makes sure to keep a unix domain socket connecting the two processes.
Privbind uses LD_PRELOAD to intercept every call to bind(2) made by the program. Calls that can be completed non-privileged are done so. Calls that require root privileges are forwarded to the root process, that carry them out on the program’s behalf.
A more detailed explanation is available in the README file.
privbind currently uses "SOCK_SEQPACKET" for communication between privileged and non-privileged processes. This socket type is only implemented on Linux kernel 2.6.4 and later, which makes privbind none portable to older Linux kernels and many other non-Linux platforms.
The version of privbind described by this manual page is 1.0 (June 12, 2007)
Copyright (C) 2006-2007, Shachar Shemesh plus others. See the AUTHORS file.
privbind was written by Shachar Shemesh, with contributions from Amos Shapira and Nadav Har’El.
privbind is free software, released under the GNU General Public License (GPL). See the COPYING file for more information and the exact license terms.
The latest version of this software can be found in
Java is a registered trademark of Sun Microsystems.