| 1 |
=============================================== |
| 2 |
Mounting the root filesystem via NFS (nfsroot) |
| 3 |
=============================================== |
| 4 |
|
| 5 |
:Authors: |
| 6 |
Written 1996 by Gero Kuhlmann <[email protected]> |
| 7 |
|
| 8 |
Updated 1997 by Martin Mares <[email protected]> |
| 9 |
|
| 10 |
Updated 2006 by Nico Schottelius <[email protected]> |
| 11 |
|
| 12 |
Updated 2006 by Horms <[email protected]> |
| 13 |
|
| 14 |
Updated 2018 by Chris Novakovic <[email protected]> |
| 15 |
|
| 16 |
|
| 17 |
|
| 18 |
In order to use a diskless system, such as an X-terminal or printer server for |
| 19 |
example, it is necessary for the root filesystem to be present on a non-disk |
| 20 |
device. This may be an initramfs (see |
| 21 |
Documentation/filesystems/ramfs-rootfs-initramfs.rst), a ramdisk (see |
| 22 |
Documentation/admin-guide/initrd.rst) or a filesystem mounted via NFS. The |
| 23 |
following text describes on how to use NFS for the root filesystem. For the rest |
| 24 |
of this text 'client' means the diskless system, and 'server' means the NFS |
| 25 |
server. |
| 26 |
|
| 27 |
|
| 28 |
|
| 29 |
|
| 30 |
Enabling nfsroot capabilities |
| 31 |
============================= |
| 32 |
|
| 33 |
In order to use nfsroot, NFS client support needs to be selected as |
| 34 |
built-in during configuration. Once this has been selected, the nfsroot |
| 35 |
option will become available, which should also be selected. |
| 36 |
|
| 37 |
In the networking options, kernel level autoconfiguration can be selected, |
| 38 |
along with the types of autoconfiguration to support. Selecting all of |
| 39 |
DHCP, BOOTP and RARP is safe. |
| 40 |
|
| 41 |
|
| 42 |
|
| 43 |
|
| 44 |
Kernel command line |
| 45 |
=================== |
| 46 |
|
| 47 |
When the kernel has been loaded by a boot loader (see below) it needs to be |
| 48 |
told what root fs device to use. And in the case of nfsroot, where to find |
| 49 |
both the server and the name of the directory on the server to mount as root. |
| 50 |
This can be established using the following kernel command line parameters: |
| 51 |
|
| 52 |
|
| 53 |
root=/dev/nfs |
| 54 |
This is necessary to enable the pseudo-NFS-device. Note that it's not a |
| 55 |
real device but just a synonym to tell the kernel to use NFS instead of |
| 56 |
a real device. |
| 57 |
|
| 58 |
|
| 59 |
nfsroot=[<server-ip>:]<root-dir>[,<nfs-options>] |
| 60 |
If the `nfsroot' parameter is NOT given on the command line, |
| 61 |
the default ``"/tftpboot/%s"`` will be used. |
| 62 |
|
| 63 |
<server-ip> Specifies the IP address of the NFS server. |
| 64 |
The default address is determined by the ip parameter |
| 65 |
(see below). This parameter allows the use of different |
| 66 |
servers for IP autoconfiguration and NFS. |
| 67 |
|
| 68 |
<root-dir> Name of the directory on the server to mount as root. |
| 69 |
If there is a "%s" token in the string, it will be |
| 70 |
replaced by the ASCII-representation of the client's |
| 71 |
IP address. |
| 72 |
|
| 73 |
<nfs-options> Standard NFS options. All options are separated by commas. |
| 74 |
The following defaults are used:: |
| 75 |
|
| 76 |
port = as given by server portmap daemon |
| 77 |
rsize = 4096 |
| 78 |
wsize = 4096 |
| 79 |
timeo = 7 |
| 80 |
retrans = 3 |
| 81 |
acregmin = 3 |
| 82 |
acregmax = 60 |
| 83 |
acdirmin = 30 |
| 84 |
acdirmax = 60 |
| 85 |
flags = hard, nointr, noposix, cto, ac |
| 86 |
|
| 87 |
|
| 88 |
ip=<client-ip>:<server-ip>:<gw-ip>:<netmask>:<hostname>:<device>:<autoconf>:<dns0-ip>:<dns1-ip>:<ntp0-ip> |
| 89 |
This parameter tells the kernel how to configure IP addresses of devices |
| 90 |
and also how to set up the IP routing table. It was originally called |
| 91 |
nfsaddrs, but now the boot-time IP configuration works independently of |
| 92 |
NFS, so it was renamed to ip and the old name remained as an alias for |
| 93 |
compatibility reasons. |
| 94 |
|
| 95 |
If this parameter is missing from the kernel command line, all fields are |
| 96 |
assumed to be empty, and the defaults mentioned below apply. In general |
| 97 |
this means that the kernel tries to configure everything using |
| 98 |
autoconfiguration. |
| 99 |
|
| 100 |
The <autoconf> parameter can appear alone as the value to the ip |
| 101 |
parameter (without all the ':' characters before). If the value is |
| 102 |
"ip=off" or "ip=none", no autoconfiguration will take place, otherwise |
| 103 |
autoconfiguration will take place. The most common way to use this |
| 104 |
is "ip=dhcp". |
| 105 |
|
| 106 |
<client-ip> IP address of the client. |
| 107 |
Default: Determined using autoconfiguration. |
| 108 |
|
| 109 |
<server-ip> IP address of the NFS server. |
| 110 |
If RARP is used to determine |
| 111 |
the client address and this parameter is NOT empty only |
| 112 |
replies from the specified server are accepted. |
| 113 |
|
| 114 |
Only required for NFS root. That is autoconfiguration |
| 115 |
will not be triggered if it is missing and NFS root is not |
| 116 |
in operation. |
| 117 |
|
| 118 |
Value is exported to /proc/net/pnp with the prefix "bootserver " |
| 119 |
(see below). |
| 120 |
|
| 121 |
Default: Determined using autoconfiguration. |
| 122 |
The address of the autoconfiguration server is used. |
| 123 |
|
| 124 |
<gw-ip> IP address of a gateway if the server is on a different subnet. |
| 125 |
Default: Determined using autoconfiguration. |
| 126 |
|
| 127 |
<netmask> Netmask for local network interface. |
| 128 |
If unspecified the netmask is derived from the client IP address |
| 129 |
assuming classful addressing. |
| 130 |
|
| 131 |
Default: Determined using autoconfiguration. |
| 132 |
|
| 133 |
<hostname> Name of the client. |
| 134 |
If a '.' character is present, anything |
| 135 |
before the first '.' is used as the client's hostname, and anything |
| 136 |
after it is used as its NIS domain name. May be supplied by |
| 137 |
autoconfiguration, but its absence will not trigger autoconfiguration. |
| 138 |
If specified and DHCP is used, the user-provided hostname (and NIS |
| 139 |
domain name, if present) will be carried in the DHCP request; this |
| 140 |
may cause a DNS record to be created or updated for the client. |
| 141 |
|
| 142 |
Default: Client IP address is used in ASCII notation. |
| 143 |
|
| 144 |
<device> Name of network device to use. |
| 145 |
Default: If the host only has one device, it is used. |
| 146 |
Otherwise the device is determined using |
| 147 |
autoconfiguration. This is done by sending |
| 148 |
autoconfiguration requests out of all devices, |
| 149 |
and using the device that received the first reply. |
| 150 |
|
| 151 |
<autoconf> Method to use for autoconfiguration. |
| 152 |
In the case of options |
| 153 |
which specify multiple autoconfiguration protocols, |
| 154 |
requests are sent using all protocols, and the first one |
| 155 |
to reply is used. |
| 156 |
|
| 157 |
Only autoconfiguration protocols that have been compiled |
| 158 |
into the kernel will be used, regardless of the value of |
| 159 |
this option:: |
| 160 |
|
| 161 |
off or none: don't use autoconfiguration |
| 162 |
(do static IP assignment instead) |
| 163 |
on or any: use any protocol available in the kernel |
| 164 |
(default) |
| 165 |
dhcp: use DHCP |
| 166 |
bootp: use BOOTP |
| 167 |
rarp: use RARP |
| 168 |
both: use both BOOTP and RARP but not DHCP |
| 169 |
(old option kept for backwards compatibility) |
| 170 |
|
| 171 |
if dhcp is used, the client identifier can be used by following |
| 172 |
format "ip=dhcp,client-id-type,client-id-value" |
| 173 |
|
| 174 |
Default: any |
| 175 |
|
| 176 |
<dns0-ip> IP address of primary nameserver. |
| 177 |
Value is exported to /proc/net/pnp with the prefix "nameserver " |
| 178 |
(see below). |
| 179 |
|
| 180 |
Default: None if not using autoconfiguration; determined |
| 181 |
automatically if using autoconfiguration. |
| 182 |
|
| 183 |
<dns1-ip> IP address of secondary nameserver. |
| 184 |
See <dns0-ip>. |
| 185 |
|
| 186 |
<ntp0-ip> IP address of a Network Time Protocol (NTP) server. |
| 187 |
Value is exported to /proc/net/ipconfig/ntp_servers, but is |
| 188 |
otherwise unused (see below). |
| 189 |
|
| 190 |
Default: None if not using autoconfiguration; determined |
| 191 |
automatically if using autoconfiguration. |
| 192 |
|
| 193 |
After configuration (whether manual or automatic) is complete, two files |
| 194 |
are created in the following format; lines are omitted if their respective |
| 195 |
value is empty following configuration: |
| 196 |
|
| 197 |
- /proc/net/pnp: |
| 198 |
|
| 199 |
#PROTO: <DHCP|BOOTP|RARP|MANUAL> (depending on configuration method) |
| 200 |
domain <dns-domain> (if autoconfigured, the DNS domain) |
| 201 |
nameserver <dns0-ip> (primary name server IP) |
| 202 |
nameserver <dns1-ip> (secondary name server IP) |
| 203 |
nameserver <dns2-ip> (tertiary name server IP) |
| 204 |
bootserver <server-ip> (NFS server IP) |
| 205 |
|
| 206 |
- /proc/net/ipconfig/ntp_servers: |
| 207 |
|
| 208 |
<ntp0-ip> (NTP server IP) |
| 209 |
<ntp1-ip> (NTP server IP) |
| 210 |
<ntp2-ip> (NTP server IP) |
| 211 |
|
| 212 |
<dns-domain> and <dns2-ip> (in /proc/net/pnp) and <ntp1-ip> and <ntp2-ip> |
| 213 |
(in /proc/net/ipconfig/ntp_servers) are requested during autoconfiguration; |
| 214 |
they cannot be specified as part of the "ip=" kernel command line parameter. |
| 215 |
|
| 216 |
Because the "domain" and "nameserver" options are recognised by DNS |
| 217 |
resolvers, /etc/resolv.conf is often linked to /proc/net/pnp on systems |
| 218 |
that use an NFS root filesystem. |
| 219 |
|
| 220 |
Note that the kernel will not synchronise the system time with any NTP |
| 221 |
servers it discovers; this is the responsibility of a user space process |
| 222 |
(e.g. an initrd/initramfs script that passes the IP addresses listed in |
| 223 |
/proc/net/ipconfig/ntp_servers to an NTP client before mounting the real |
| 224 |
root filesystem if it is on NFS). |
| 225 |
|
| 226 |
|
| 227 |
nfsrootdebug |
| 228 |
This parameter enables debugging messages to appear in the kernel |
| 229 |
log at boot time so that administrators can verify that the correct |
| 230 |
NFS mount options, server address, and root path are passed to the |
| 231 |
NFS client. |
| 232 |
|
| 233 |
|
| 234 |
rdinit=<executable file> |
| 235 |
To specify which file contains the program that starts system |
| 236 |
initialization, administrators can use this command line parameter. |
| 237 |
The default value of this parameter is "/init". If the specified |
| 238 |
file exists and the kernel can execute it, root filesystem related |
| 239 |
kernel command line parameters, including 'nfsroot=', are ignored. |
| 240 |
|
| 241 |
A description of the process of mounting the root file system can be |
| 242 |
found in Documentation/driver-api/early-userspace/early_userspace_support.rst |
| 243 |
|
| 244 |
|
| 245 |
Boot Loader |
| 246 |
=========== |
| 247 |
|
| 248 |
To get the kernel into memory different approaches can be used. |
| 249 |
They depend on various facilities being available: |
| 250 |
|
| 251 |
|
| 252 |
- Booting from a floppy using syslinux |
| 253 |
|
| 254 |
When building kernels, an easy way to create a boot floppy that uses |
| 255 |
syslinux is to use the zdisk or bzdisk make targets which use zimage |
| 256 |
and bzimage images respectively. Both targets accept the |
| 257 |
FDARGS parameter which can be used to set the kernel command line. |
| 258 |
|
| 259 |
e.g:: |
| 260 |
|
| 261 |
make bzdisk FDARGS="root=/dev/nfs" |
| 262 |
|
| 263 |
Note that the user running this command will need to have |
| 264 |
access to the floppy drive device, /dev/fd0 |
| 265 |
|
| 266 |
For more information on syslinux, including how to create bootdisks |
| 267 |
for prebuilt kernels, see https://syslinux.zytor.com/ |
| 268 |
|
| 269 |
.. note:: |
| 270 |
Previously it was possible to write a kernel directly to |
| 271 |
a floppy using dd, configure the boot device using rdev, and |
| 272 |
boot using the resulting floppy. Linux no longer supports this |
| 273 |
method of booting. |
| 274 |
|
| 275 |
- Booting from a cdrom using isolinux |
| 276 |
|
| 277 |
When building kernels, an easy way to create a bootable cdrom that |
| 278 |
uses isolinux is to use the isoimage target which uses a bzimage |
| 279 |
image. Like zdisk and bzdisk, this target accepts the FDARGS |
| 280 |
parameter which can be used to set the kernel command line. |
| 281 |
|
| 282 |
e.g:: |
| 283 |
|
| 284 |
make isoimage FDARGS="root=/dev/nfs" |
| 285 |
|
| 286 |
The resulting iso image will be arch/<ARCH>/boot/image.iso |
| 287 |
This can be written to a cdrom using a variety of tools including |
| 288 |
cdrecord. |
| 289 |
|
| 290 |
e.g:: |
| 291 |
|
| 292 |
cdrecord dev=ATAPI:1,0,0 arch/x86/boot/image.iso |
| 293 |
|
| 294 |
For more information on isolinux, including how to create bootdisks |
| 295 |
for prebuilt kernels, see https://syslinux.zytor.com/ |
| 296 |
|
| 297 |
- Using LILO |
| 298 |
|
| 299 |
When using LILO all the necessary command line parameters may be |
| 300 |
specified using the 'append=' directive in the LILO configuration |
| 301 |
file. |
| 302 |
|
| 303 |
However, to use the 'root=' directive you also need to create |
| 304 |
a dummy root device, which may be removed after LILO is run. |
| 305 |
|
| 306 |
e.g:: |
| 307 |
|
| 308 |
mknod /dev/boot255 c 0 255 |
| 309 |
|
| 310 |
For information on configuring LILO, please refer to its documentation. |
| 311 |
|
| 312 |
- Using GRUB |
| 313 |
|
| 314 |
When using GRUB, kernel parameter are simply appended after the kernel |
| 315 |
specification: kernel <kernel> <parameters> |
| 316 |
|
| 317 |
- Using loadlin |
| 318 |
|
| 319 |
loadlin may be used to boot Linux from a DOS command prompt without |
| 320 |
requiring a local hard disk to mount as root. This has not been |
| 321 |
thoroughly tested by the authors of this document, but in general |
| 322 |
it should be possible configure the kernel command line similarly |
| 323 |
to the configuration of LILO. |
| 324 |
|
| 325 |
Please refer to the loadlin documentation for further information. |
| 326 |
|
| 327 |
- Using a boot ROM |
| 328 |
|
| 329 |
This is probably the most elegant way of booting a diskless client. |
| 330 |
With a boot ROM the kernel is loaded using the TFTP protocol. The |
| 331 |
authors of this document are not aware of any no commercial boot |
| 332 |
ROMs that support booting Linux over the network. However, there |
| 333 |
are two free implementations of a boot ROM, netboot-nfs and |
| 334 |
etherboot, both of which are available on sunsite.unc.edu, and both |
| 335 |
of which contain everything you need to boot a diskless Linux client. |
| 336 |
|
| 337 |
- Using pxelinux |
| 338 |
|
| 339 |
Pxelinux may be used to boot linux using the PXE boot loader |
| 340 |
which is present on many modern network cards. |
| 341 |
|
| 342 |
When using pxelinux, the kernel image is specified using |
| 343 |
"kernel <relative-path-below /tftpboot>". The nfsroot parameters |
| 344 |
are passed to the kernel by adding them to the "append" line. |
| 345 |
It is common to use serial console in conjunction with pxelinux, |
| 346 |
see Documentation/admin-guide/serial-console.rst for more information. |
| 347 |
|
| 348 |
For more information on isolinux, including how to create bootdisks |
| 349 |
for prebuilt kernels, see https://syslinux.zytor.com/ |
| 350 |
|
| 351 |
|
| 352 |
|
| 353 |
|
| 354 |
Credits |
| 355 |
======= |
| 356 |
|
| 357 |
The nfsroot code in the kernel and the RARP support have been written |
| 358 |
by Gero Kuhlmann <[email protected]>. |
| 359 |
|
| 360 |
The rest of the IP layer autoconfiguration code has been written |
| 361 |
by Martin Mares <[email protected]>. |
| 362 |
|
| 363 |
In order to write the initial version of nfsroot I would like to thank |
| 364 |
Jens-Uwe Mager <[email protected]> for his help. |