xref: /freebsd/share/man/man8/diskless.8 (revision 6af83ee0d2941d18880b6aaa2b4facd1d30c6106)
1.\" Copyright (c) 1994 Gordon W. Ross, Theo de Raadt
2.\" Updated by Luigi Rizzo, Robert Watson
3.\" All rights reserved.
4.\"
5.\" Redistribution and use in source and binary forms, with or without
6.\" modification, are permitted provided that the following conditions
7.\" are met:
8.\" 1. Redistributions of source code must retain the above copyright
9.\"    notice, this list of conditions and the following disclaimer.
10.\" 2. Redistributions in binary form must reproduce the above copyright
11.\"    notice, this list of conditions and the following disclaimer in the
12.\"    documentation and/or other materials provided with the distribution.
13.\" 3. The name of the author may not be used to endorse or promote products
14.\"    derived from this software without specific prior written permission.
15.\"
16.\" THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17.\" IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18.\" OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19.\" IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
20.\" INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
21.\" NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
22.\" DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
23.\" THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
24.\" (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
25.\" THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
26.\"
27.\" $FreeBSD$
28.\"
29.Dd September 28, 2004
30.Dt DISKLESS 8
31.Os
32.Sh NAME
33.Nm diskless
34.Nd booting a system over the network
35.Sh DESCRIPTION
36The ability to boot a machine over the network is useful for
37.Em diskless
38or
39.Em dataless
40machines, or as a temporary measure while repairing or
41re-installing file systems on a local disk.
42This file provides a general description of the interactions between
43a client and its server when a client is booting over the network.
44.Sh OPERATION
45When booting a system over the network, there are three
46phases of interaction between client and server:
47.Bl -enum
48.It
49The stage-1 bootstrap, typically PXE built into your Ethernet
50card, loads a second-stage boot program.
51.It
52The second-stage boot program, typically
53.Xr pxeboot 8 ,
54loads modules and
55the kernel, and boots the kernel.
56.It
57The kernel
58.Tn NFS
59mounts the root directory and continues from there.
60.El
61.Pp
62Each of these phases are described in further detail below.
63.Pp
64First, the stage-1 bootstrap loads the stage-2 boot program over
65the network.
66The stage-1 bootstrap typically uses
67.Tn BOOTP
68or
69.Tn DHCP
70to obtain the filename to load, then uses
71.Tn TFTP
72to load the file.
73This file is typically called
74.Pa pxeboot ,
75and should be copied from
76.Pa /boot/pxeboot
77into the
78.Tn TFTP
79directory on the server, which is typically
80.Pa /tftpdir .
81.Pp
82The stage-2 boot program then loads additional modules and the kernel.
83These files may not exist on the
84.Tn DHCP
85or
86.Tn BOOTP
87server.
88You can use the
89.Ic next-server
90option available in
91.Tn DHCP
92configurations to specify the server holding
93the second stage boot files and kernel.
94The stage-2 program uses
95.Tn NFS
96or
97.Tn TFTP
98to obtain these files.
99By default,
100.Tn NFS
101is used.
102If you are using
103.Xr pxeboot 8 ,
104you can install a version that uses
105.Tn TFTP
106by setting
107.Li LOADER_TFTP_SUPPORT=YES
108in your
109.Pa /etc/make.conf ,
110then recompiling and reinstalling
111.Xr pxeboot 8
112via the command listed below.
113It is often necessary to use
114.Tn TFTP
115here so you can place a custom kernel
116in
117.Pa /tftpdir/ .
118If you use
119.Tn NFS
120and do not have a custom root file system for the
121.Nm
122client, the stage-2 boot will load your server's kernel as the kernel for
123the
124.Nm
125machine, which may not be what you want to have happen.
126.Bd -literal -offset indent
127cd /usr/src/sys/boot/i386
128make clean; make; make install
129cp /boot/pxeboot /tftpdir/
130.Ed
131.Pp
132In phase 3, the kernel acquires IP networking configuration in one
133of two ways, and then proceeds to mount the root file system and start
134operation.
135If the phase 2 loader supports passing network configuration to the
136kernel using the kernel environment, then the kernel will configure
137the network interface using that information.
138Otherwise, it must use
139.Tn DHCP
140or
141.Tn BOOTP
142to acquire
143configuration information.
144The boot
145scripts recognize a
146.Nm
147startup and perform
148the actions found in
149.Pa /etc/rc.d/initdiskless ,
150.Pa /etc/rc.d/resolv ,
151.Pa /etc/rc.d/tmp ,
152and
153.Pa /etc/rc.d/var .
154On early 5.x releases, the functions of latter three scripts where
155implemented in
156.Pa /etc/rc.d/diskless .
157On older systems, the scripts are located in
158.Pa /etc/rc.diskless1
159and
160.Pa /etc/rc.diskless2 .
161.Sh CONFIGURATION
162In order to run a
163.Nm
164client, you need the following:
165.Bl -bullet
166.It
167An
168.Tn NFS
169server which exports a root and
170.Pa /usr
171partitions with appropriate permissions.
172The
173.Nm
174scripts work with read-only partitions, as long as root is exported with
175.Fl maproot Ns =0
176so that some system files can be accessed.
177As an example,
178.Pa /etc/exports
179can contain the following lines:
180.Bd -literal -offset indent
181<ROOT> -ro -maproot=0 -alldirs <list of diskless clients>
182/usr -ro -alldirs <list of diskless clients>
183.Ed
184.Pp
185where
186.Aq ROOT
187is the mount point on the server of the root partition.
188The script
189.Pa /usr/share/examples/diskless/clone_root
190can be used to create a shared read-only root partition,
191but in many cases you may decide to export
192(again as read-only) the root directory used by
193the server itself.
194.It
195A
196.Tn BOOTP
197or
198.Tn DHCP
199server.
200.Xr bootpd 8
201can be enabled by
202uncommenting the
203.Dq Li bootps
204line in
205.Pa /etc/inetd.conf .
206A sample
207.Pa /etc/bootptab
208can be the following:
209.Bd -literal -offset indent
210 .default:\\
211    hn:ht=1:vm=rfc1048:\\
212    :sm=255.255.255.0:\\
213    :sa=<SERVER>:\\
214    :gw=<GATEWAY>:\\
215    :rp="<SERVER>:<ROOT>":
216
217<CLIENT>:ha=0123456789ab:tc=.default
218.Ed
219.Pp
220where
221.Aq SERVER ,
222.Aq GATEWAY
223and
224.Aq ROOT
225have the obvious meanings.
226.It
227A properly initialized root partition.
228The script
229.Pa /usr/share/examples/diskless/clone_root
230can help in creating it, using the server's root partition
231as a reference.
232If you are just starting out, you should
233simply use the server's own root directory,
234.Pa / ,
235and not try to clone it.
236.Pp
237You often do not want to use the same
238.Pa rc.conf
239or
240.Pa rc.local
241files for the
242.Nm
243boot as you do on the server.
244The
245.Nm
246boot
247scripts provide a mechanism through which you can override various files
248in
249.Pa /etc
250(as well as other subdirectories of root).
251The scripts provide four
252overriding directories situated in
253.Pa /conf/base ,
254.Pa /conf/default ,
255.Pa /conf/<broadcast-ip> ,
256and
257.Pa /conf/<machine-ip> .
258You should always create
259.Pa /conf/base/etc ,
260which will entirely replace the server's
261.Pa /etc
262on the
263.Nm
264machine.
265You can clone the server's
266.Pa /etc
267here or you can create a special file which tells the
268.Nm
269boot scripts
270to remount the server's
271.Pa /etc
272onto
273.Pa /conf/base/etc .
274You do this by creating the file
275.Pa /conf/base/etc/diskless_remount
276containing the mount point to use as a basis of the
277.Nm
278machine's
279.Pa /etc .
280For example, the file might contain:
281.Pp
282.Dl 10.0.0.1:/etc
283.Pp
284Alternatively, if the server contains several independent roots, the file
285might contain:
286.Pp
287.Dl 10.0.0.1:/usr/diskless/4.7-RELEASE/etc
288.Pp
289This would work, but if you copied
290.Pa /usr/diskless/4.7-RELEASE
291to
292.Pa /usr/diskless/4.8-RELEASE
293and upgraded the installation, you would need to modify the
294.Pa diskless_remount
295files to reflect that move.
296To avoid that, paths in
297.Pa diskless_remount
298files beginning with
299.Pa /
300have the actual path of the client's root prepended to them so the file
301could instead contain:
302.Pp
303.Dl /etc
304.Pp
305The
306.Nm
307scripts create memory file systems to hold the overridden
308directories.
309Only a 2MB partition is created by default, which may not
310be sufficient for your purposes.
311To override this, you can create the
312file
313.Pa /conf/base/etc/md_size
314containing the size, in 512 byte sectors, of the memory disk to create
315for that directory.
316.Pp
317You then typically provide file-by-file overrides in the
318.Pa /conf/default/etc
319directory.
320At a minimum, you must provide overrides for
321.Pa /etc/fstab , /etc/rc.conf ,
322and
323.Pa /etc/rc.local
324via
325.Pa /conf/default/etc/fstab , /conf/default/etc/rc.conf ,
326and
327.Pa /conf/default/etc/rc.local .
328.Pp
329Overrides are hierarchical.
330You can supply network-specific defaults
331in the
332.Pa /conf/ Ns Ao Ar BROADCASTIP Ac Ns Pa /etc
333directory, where
334.Aq Ar BROADCASTIP
335represents the broadcast IP address of
336the
337.Nm
338system as given to it via
339.Tn BOOTP .
340The
341.Pa diskless_remount
342and
343.Pa md_size
344features work in any of these directories.
345The configuration feature works on directories other then
346.Pa /etc ,
347you simply create the directory you wish to replace or override in
348.Pa /conf/{base,default,<broadcast>,<ip>}/*
349and work it in the same way that you work
350.Pa /etc .
351.Pp
352Since you normally clone the server's
353.Pa /etc
354using the
355.Pa /conf/base/etc/diskless_remount ,
356you might wish to remove unneeded files from the memory file system.
357For example,
358if the server has a firewall but you do not, you might wish
359to remove
360.Pa /etc/ipfw.conf .
361You can do this by creating a
362.Pa /conf/base/ Ns Ao Ar DIRECTORY Ac Ns Pa .remove
363file.
364For example,
365.Pa /conf/base/etc.remove ,
366which contains a list of relative paths that the boot scripts should remove
367from the memory file systems.
368.Pp
369As a minimum, you normally need to have the following in
370.Pa /conf/default/etc/fstab
371.Bd -literal -offset indent
372<SERVER>:<ROOT> /     nfs    ro 0 0
373<SERVER>:/usr   /usr  nfs    ro 0 0
374proc            /proc procfs rw 0 0
375.Ed
376.Pp
377You also need to create a customized version of
378.Pa /conf/default/etc/rc.conf
379which should contain
380the startup options for the
381.Nm
382client, and
383.Pa /conf/default/etc/rc.local
384which could be empty but prevents the server's own
385.Pa /etc/rc.local
386from leaking onto the
387.Nm
388system.
389.Pp
390In
391.Pa rc.conf ,
392most likely
393you will not need to set
394.Va hostname
395and
396.Va ifconfig_*
397because these will be already set by the startup code.
398Finally, it might be convenient to use a
399.Ic case
400statement using
401.Li `hostname`
402as the switch variable to do machine-specific configuration
403in case a number of
404.Nm
405clients share the same configuration
406files.
407.It
408The kernel for the
409.Nm
410clients, which will be loaded using
411.Tn NFS
412or
413.Tn TFTP ,
414must include support for the NFS client:
415.Pp
416.D1 Cd "options NFSCLIENT"
417.D1 Cd "options NFS_ROOT"
418.Pp
419If you are using a boot mechanism that does not pass network configuration
420to the kernel using the kernel environment, you will also need to include
421the following options:
422.Pp
423.D1 Cd "options BOOTP"
424.D1 Cd "options BOOTP_NFSROOT"
425.D1 Cd "options BOOTP_COMPAT"
426.Pp
427.Em Note :
428the PXE environment does not require these options.
429.Pp
430The
431.Nm
432booting environment relies on memory-backed file systems to
433support temporary local storage in the event that the root file system
434is mounted read-only; as such, it is necessary to add the following
435to the device section of the kernel configuration:
436.Pp
437.D1 Cd "device md"
438.Pp
439If you use the firewall, remember to default to
440.Dq open ,
441or your kernel
442will not be able to send/receive the
443.Tn BOOTP
444packets.
445.El
446.Sh SECURITY ISSUES
447Be warned that using unencrypted
448.Tn NFS
449to mount root and user
450partitions may expose information such as
451encryption keys.
452.Sh SEE ALSO
453.Xr ethers 5 ,
454.Xr exports 5 ,
455.Xr bootpd 8 ,
456.Xr mountd 8 ,
457.Xr nfsd 8 ,
458.Xr pxeboot 8 ,
459.Xr reboot 8 ,
460.Xr tftpd 8
461.Pp
462.Pa ports/net/etherboot
463.Sh BUGS
464This manpage is probably incomplete.
465.Pp
466.Fx
467sometimes requires to write onto
468the root partition, so the startup scripts mount MFS
469file systems on some locations (e.g.\&
470.Pa /etc
471and
472.Pa /var ) ,
473while
474trying to preserve the original content.
475The process might not handle all cases.
476