freebsd-dev/usr.sbin/jail/jail.8

376 lines
13 KiB
Groff
Raw Normal View History

.\"
.\"Copyright (c) 2000 Robert N. M. Watson
.\"All rights reserved.
.\"
.\"Redistribution and use in source and binary forms, with or without
.\"modification, are permitted provided that the following conditions
.\"are met:
.\"1. Redistributions of source code must retain the above copyright
.\" notice, this list of conditions and the following disclaimer.
.\"2. Redistributions in binary form must reproduce the above copyright
.\" notice, this list of conditions and the following disclaimer in the
.\" documentation and/or other materials provided with the distribution.
.\"
.\"THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
.\"ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
.\"IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
.\"ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
.\"FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
.\"DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
.\"OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
.\"HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
.\"LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
.\"OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
.\"SUCH DAMAGE.
.\"
.\"
.\"----------------------------------------------------------------------------
.\""THE BEER-WARE LICENSE" (Revision 42):
.\"<phk@FreeBSD.ORG> wrote this file. As long as you retain this notice you
.\"can do whatever you want with this stuff. If we meet some day, and you think
.\"this stuff is worth it, you can buy me a beer in return. Poul-Henning Kamp
.\"----------------------------------------------------------------------------
.\"
1999-08-28 01:35:59 +00:00
.\"$FreeBSD$
.\"
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Dd April 28, 1999
.Dt JAIL 8
.Os FreeBSD 4.0
.Sh NAME
.Nm jail
.Nd imprison process and its descendants
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Sh SYNOPSIS
.Nm jail
.Ar path
.Ar hostname
.Ar ip-number
.Ar command
.Ar ...
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Sh DESCRIPTION
The
.Nm
command imprisons a process and all future descendants.
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Pp
Please see the
.Xr jail 2
man page for further details.
1999-07-09 21:35:50 +00:00
.Sh EXAMPLES
.Ss Setting up a Jail Directory Tree
1999-07-09 21:35:50 +00:00
This shows how to setup a jail directory tree:
.Bd -literal
D=/here/is/the/jail
cd /usr/src
make hierarchy DESTDIR=$D
make obj
make depend
1999-07-09 21:35:50 +00:00
make all
make install DESTDIR=$D
cd etc
make distribution DESTDIR=$D NO_MAKEDEV=yes
1999-07-09 21:35:50 +00:00
cd $D/dev
sh MAKEDEV jail
cd $D
ln -sf dev/null kernel
.Ed
.Ss Setting Up a Jail
Do what was described in
.Sx Setting Up a Jail Directory Tree
to build the jail directory tree. For the sake of this example, we will
assume you built it in
.Pa /data/jail/192.168.11.100 ,
named for the jailed IP address. Substitute below as needed with your
own directory, IP address, and hostname.
.Pp
First, you will want to set up your real system's environment to be
.Dq jail-friendly.
For consistency, we will refer to the parent box as the
.Dq host environment,
and to the jailed virtual machine as the
.Dq jail environment.
Because jail is implemented using IP aliases, one of the first things to do
is to disable IP services on the host system that listen on all local
IP addresses for a service. This means changing
.Xr inetd 8
to only listen on the
appropriate IP address, and so forth. Add the following to
.Pa /etc/rc.conf
in the host environment:
.Bd -literal -offset indent
sendmail_enable="NO"
inetd_flags="-wW -a 192.168.11.23"
portmap_enable="NO"
syslogd_flags="-ss"
.Ed
.Pp
.Li 192.169.11.23
is the native IP address for the host system, in this example. Daemons that
run out of
.Xr inetd 8
can be easily set to use only the specified host IP address. Other daemons
will need to be manually configured--for some this is possible through
the
.Xr rc.conf 5
flags entries, for others it is not possible without munging
the per-application configuration files, or even recompiling. For those
applications that cannot specify the IP they run on, it is better to disable
them, if possible.
.Pp
A number of daemons ship with the base system that may have problems when
run from outside of a jail in a jail-centric environment. This includes
.Xr syslogd 8 ,
.Xr sendmail 8 ,
.Xr named 8 ,
and
.Xr portmap 8 .
While sendmail and named can be configured to listen only on a specific
IP using their configuration files, in most cases it is easier to simply
run the daemons in jails only, and not in the host environment. Syslogd
cannot be configured to bind only a single IP, but can be configured to
not bind a network port, using the ``-ss'' argument. Attempting to serve
NFS from the host environment may also cause confusion, and cannot be
easily reconfigured to use only specific IPs, as some NFS services are
hosted directly from the kernel. Any third party network software running
in the host environment should also be checked and configured so that it
does not bind all IP addresses, which would result in those services also
appearing to be offered by the jail environments.
.Pp
Once
these daemons have been disabled or fixed in the host environment, it is
best to reboot so that all daemons are in a known state, to reduce the
potential for confusion later (such as finding that when you send mail
to a jail, and its sendmail is down, the mail is delivered to the host,
etc.)
.Pp
Start any jails for the first time without configuring the network
interface so that you can clean it up a little and set up accounts. As
with any machine (virtual or not) you will need to set a root password, time
zone, etc. Before beginning, you may want to copy
.Xr sysinstall 8
into the tree so that you can use it to set things up easily. Do this using:
.Bd -literal -offset indent
# mkdir /data/jail/192.168.11.100/stand
# cp /stand/sysinstall /data/jail/192.168.11.100/stand
.Ed
.Pp
Now start the jail:
.Bd -literal -offset indent
# jail /data/jail/192.168.11.100 testhostname 192.168.11.100 /bin/sh
.Ed
.Pp
You will end up with a shell prompt, assuming no errors, within the jail. You
can now run
.Pa /stand/sysinstall
and do the post-install configuration to set various configuration options,
or perform these actions manually by editing rc.conf, etc.
.Pp
.Bl -bullet -offset indent -compact
.It
Create an empty /etc/fstab to quell startup warnings about missing fstab
.It
Disable the port mapper (rc.conf: portmap_enable="NO")
.It
Run
.Xr newaliases 1
to quell sendmail warnings.
.It
Disable interface configuration to quell startup warnings about ifconfig
(network_interfaces="")
.It
Configure /etc/resolv.conf
so that name resolution within the jail will work correctly
.It
Set a root password, probably different from the real host system
.It
Set the timezone
.It
Add accounts for users in the jail environment
.It
Install any packages that you think the environment requires
.El
.Pp
You may also want to perform any package-specific configuration (web servers,
SSH servers, etc), patch up /etc/syslog.conf so it logs as you'd like, etc.
.Pp
Exit from the shell, and the jail will be shut down.
.Ss Starting the Jail
You are now ready to restart the jail and bring up the environment with
all of its daemons and other programs. To do this, first bring up the
virtual host interface, and then start the jail's
.Pa /etc/rc
script from within the jail.
.Pp
NOTE: If you plan to allow untrusted users to have root access inside the
jail, you may wish to consider setting the jail.set_hostname_allowed to
0. Please see the management reasons why this is a good idea. If you
do decide to set this variable, it must be set before starting any jails,
and once each boot.
.Bd -literal -offset indent
# ifconfig ed0 inet alias 192.168.11.100 netmask 255.255.255.255
# mount -t procfs proc /data/jail/192.168.11.100/proc
# jail /data/jail/192.168.11.100 testhostname 192.168.11.100 \\
/bin/sh /etc/rc
.Ed
.Pp
A few warnings will be produced, because most
.Xr sysctl 8
configuration variables cannot be set from within the jail, as they are
global across all jails and the host environment.
However, it should all
work properly.
You should be able to see
.Xr inetd 8 ,
.Xr syslogd 8 ,
and other processes running within the jail using
.Xr ps 1 ,
with the
.Dq J
flag appearing beside jailed processes. You should also be able to
telnet to the hostname or IP address of the jailed environment, and log
in using the accounts you created previously.
.Ss Managing the jail
Normal machine shutdown commands, such as
.Xr halt 8 ,
.Xr reboot 8 ,
and
.Xr shutdown 8 ,
cannot be used successfully within the jail. To kill all processes in a
jail, you may log into the jail and, as root, use one of the following
commands, depending on what you want to accomplish:
.Pp
.Bl -bullet -offset indent -compact
.It
.Li kill -TERM -1
.It
.Li kill -KILL -1
.El
.Pp
This will send the
.Dq TERM
or
.Dq KILL
signals to all processes in the jail from within the jail. Depending on
the intended use of the jail, you may also want to run
.Pa /etc/rc.shutdown
from within the jail. Currently there is no way to insert new processes
into a jail, so you must first log into the jail before performing these
actions.
.Pp
To kill processes from outside the jail, you must individually identify the
PID of each process to be killed. The
.Pa /proc/ Ns Va pid Ns Pa /status
file contains, as its last field, the hostname of the jail in which the
process runs, or
.Dq -
to indicate that the process is not running within a jail. The
.Xr ps 1
command also shows a
.Dq J
flag for processes in a jail. However, the hostname for a jail may be, by
default, modified from within the jail, so the
.Pa /proc
status entry is unreliable by default. To disable the setting of the hostname
from within a jail, set the
.Dq Va jail.set_hostname_allowed
sysctl variable in the host environment to 0, which will affect all jails.
You can have this sysctl set each boot using
.Xr sysctl.conf 5 .
Just add the following line to sysctl.conf:
.Bd -literal -offset indent
jail.set_hostname_allowed=0
.Ed
.Pp
In a future version of FreeBSD, the mechanisms for managing jails will be
more refined.
.Ss Sysctl MIB Entries
Certain aspects of the jail containments environment may be modified from
the host environment using
.Xr sysctl 8
MIB variables.
Currently, these variables affect all jails on the system, although in
the future this functionality may be finer grained.
.Bl -tag -width XXX
.It jail.set_hostname_allowed
This MIB entry determines whether or not processes within a jail are
allowed to change their hostname via
.Xr hostname 1
or
.Xr sethostname 3 .
In the current jail implementation, the ability to set the hostname from
within the jail can impact management tools relying on the accuracy of jail
information in
.Pa /proc .
As such, this should be disabled in environments where privileged access to
jails is given out to untrusted parties.
.It jail.socket_unixiproute_only
The jail functionality binds an IPv4 address to each jail, and limits
access to other network addresses in the IPv4 space that may be available
in the host environment.
However, jail is not currently able to limit access to other network
protocol stacks that have not had jail functionality added to them.
As such, by default, processes within jails may only access protocols
in the following domains:
.Dv PF_LOCAL ,
.Dv PF_INET ,
and
.Dv PF_ROUTE ,
permitting them access to UNIX domain sockets,
IPv4 addresses, and routing sockets.
To enable access to other domains, this MIB variable may be set to
0.
.It jail.sysvipc_allowed
This MIB entry determines whether or not processes within a jail have access
to System V IPC primitives.
In the current jail implementation, System V primitives share a single
namespace across the host and jail environments, meaning that processes
within a jail would be able to communicate with (and potentially interfere
with) processes outside of the jail, and in other jails.
As such, this functionality is disabled by default, but can be enabled
by setting this MIB entry to 1.
.El
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Sh SEE ALSO
.Xr newaliases 1 ,
.Xr ps 1 ,
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Xr chroot 2 ,
.Xr jail 2 ,
.Xr procfs 5 ,
.Xr rc.conf 5 ,
.Xr sysctl.conf 5 ,
.Xr halt 8 ,
.Xr inetd 8 ,
.Xr named 8 ,
.Xr portmap 8 ,
.Xr reboot 8 ,
.Xr sendmail 8 ,
.Xr shutdown 8 ,
.Xr sysctl 8 ,
.Xr syslogd 8
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Sh HISTORY
The
.Fn jail
function call appeared in
.Fx 4.0 .
1999-12-21 11:25:10 +00:00
.Sh AUTHORS
The jail feature was written by
.An Poul-Henning Kamp
for R&D Associates
This Implements the mumbled about "Jail" feature. This is a seriously beefed up chroot kind of thing. The process is jailed along the same lines as a chroot does it, but with additional tough restrictions imposed on what the superuser can do. For all I know, it is safe to hand over the root bit inside a prison to the customer living in that prison, this is what it was developed for in fact: "real virtual servers". Each prison has an ip number associated with it, which all IP communications will be coerced to use and each prison has its own hostname. Needless to say, you need more RAM this way, but the advantage is that each customer can run their own particular version of apache and not stomp on the toes of their neighbors. It generally does what one would expect, but setting up a jail still takes a little knowledge. A few notes: I have no scripts for setting up a jail, don't ask me for them. The IP number should be an alias on one of the interfaces. mount a /proc in each jail, it will make ps more useable. /proc/<pid>/status tells the hostname of the prison for jailed processes. Quotas are only sensible if you have a mountpoint per prison. There are no privisions for stopping resource-hogging. Some "#ifdef INET" and similar may be missing (send patches!) If somebody wants to take it from here and develop it into more of a "virtual machine" they should be most welcome! Tools, comments, patches & documentation most welcome. Have fun... Sponsored by: http://www.rndassociates.com/ Run for almost a year by: http://www.servetheweb.com/
1999-04-28 11:38:52 +00:00
.Dq Li http://www.rndassociates.com/
who contributed it to FreeBSD.
.Pp
Robert Watson wrote the extended documentation, found a few bugs, added
a few new features, and cleaned up the userland jail environment.
.Sh BUGS
Jail currently lacks strong management functionality, such as the ability
to deliver signals to all processes in a jail, and to allow access to
specific jail information via
.Xr ps 1
as opposed to
.Xr procfs 5 .
Similarly, it might be a good idea to add an
address alias flag such that daemons listening on all IPs (INADDR_ANY)
will not bind on that address, which would facilitate building a safe
host environment such that host daemons do not impose on services offered
from within jails. Currently, the simplist answer is to minimize services
offered on the host, possibly limiting it to services offered from
.Xr inetd 8
which is easily configurable.