1NFS(5) Linux Programmer's Manual NFS(5)
2
3
4
6 nfs - nfs and nfs4 fstab format and options
7
9 /etc/fstab
10
12 The fstab file contains information about which filesystems to mount
13 where and with what options. For NFS mounts, it contains the server
14 name and exported server directory to mount from, the local directory
15 that is the mount point, and the NFS specific options that control the
16 way the filesystem is mounted.
17
18 Three different versions of the NFS protocol are supported by the Linux
19 NFS client: NFS version 2, NFS version 3, and NFS version 4. To mount
20 via NFS version 2, use the nfs file system type and specify nfsvers=2.
21 To mount via NFS version 3, use the nfs file system type and specify
22 nfsvers=3. Version 3 is the default protocol version for the nfs file
23 system type when nfsvers= is not specified on the mount command and
24 both client and server support it. To mount via NFS version 4, use the
25 nfs4 file system type. The nfsvers= keyword is not supported for the
26 nfs4 file system type.
27
28 These file system types share similar mount options; the differences
29 are listed below.
30
31 Here is an example from an /etc/fstab file for an NFSv3 mount over TCP.
32
33 server:/usr/local/pub /pub nfs rsize=32768,wsize=32768,timeo=14,intr
34
35 Here is an example for an NFSv4 mount over TCP using Kerberos 5 mutual
36 authentication.
37
38 server:/usr/local/pub /pub nfs4 proto=tcp,sec=krb5,hard,intr
39
40 Options for the nfs file system type
41 rsize=n The number of bytes NFS uses when reading files from an
42 NFS server. The rsize is negotiated between the server
43 and client to determine the largest block size that both
44 can support. The value specified by this option is the
45 maximum size that could be used; however, the actual
46 size used may be smaller. Note: Setting this size to a
47 value less than the largest supported block size will
48 adversely affect performance.
49
50 wsize=n The number of bytes NFS uses when writing files to an
51 NFS server. The wsize is negotiated between the server
52 and client to determine the largest block size that both
53 can support. The value specified by this option is the
54 maximum size that could be used; however, the actual
55 size used may be smaller. Note: Setting this size to a
56 value less than the largest supported block size will
57 adversely affect performance.
58
59 timeo=n The value in tenths of a second before sending the first
60 retransmission after an RPC timeout. The default value
61 is 7 tenths of a second. After the first timeout, the
62 timeout is doubled after each successive timeout until a
63 maximum timeout of 60 seconds is reached or the enough
64 retransmissions have occured to cause a major timeout.
65 Then, if the filesystem is hard mounted, each new time‐
66 out cascade restarts at twice the initial value of the
67 previous cascade, again doubling at each retransmission.
68 The maximum timeout is always 60 seconds. Better over‐
69 all performance may be achieved by increasing the time‐
70 out when mounting on a busy network, to a slow server,
71 or through several routers or gateways.
72
73 retrans=n The number of minor timeouts and retransmissions that
74 must occur before a major timeout occurs. The default
75 is 3 timeouts. When a major timeout occurs, the file
76 operation is either aborted or a "server not responding"
77 message is printed on the console.
78
79 acregmin=n The minimum time in seconds that attributes of a regular
80 file should be cached before requesting fresh informa‐
81 tion from a server. The default is 3 seconds.
82
83 acregmax=n The maximum time in seconds that attributes of a regular
84 file can be cached before requesting fresh information
85 from a server. The default is 60 seconds.
86
87 acdirmin=n The minimum time in seconds that attributes of a direc‐
88 tory should be cached before requesting fresh informa‐
89 tion from a server. The default is 30 seconds.
90
91 acdirmax=n The maximum time in seconds that attributes of a direc‐
92 tory can be cached before requesting fresh information
93 from a server. The default is 60 seconds.
94
95 actimeo=n Using actimeo sets all of acregmin, acregmax, acdirmin,
96 and acdirmax to the same value. There is no default
97 value.
98
99 retry=n The number of minutes to retry an NFS mount operation in
100 the foreground or background before giving up. The
101 default value for forground mounts is 2 minutes. The
102 default value for background mounts is 10000 minutes,
103 which is roughly one week.
104
105 namlen=n When an NFS server does not support version two of the
106 RPC mount protocol, this option can be used to specify
107 the maximum length of a filename that is supported on
108 the remote filesystem. This is used to support the
109 POSIX pathconf functions. The default is 255 charac‐
110 ters.
111
112 port=n The numeric value of the port to connect to the NFS
113 server on. If the port number is 0 (the default) then
114 query the remote host's portmapper for the port number
115 to use. If the remote host's NFS daemon is not regis‐
116 tered with its portmapper, the standard NFS port number
117 2049 is used instead.
118
119 mountport=n The numeric value of the mountd port.
120
121 mounthost=name The name of the host running mountd .
122
123 mountprog=n Use an alternate RPC program number to contact the mount
124 daemon on the remote host. This option is useful for
125 hosts that can run multiple NFS servers. The default
126 value is 100005 which is the standard RPC mount daemon
127 program number.
128
129 mountvers=n Use an alternate RPC version number to contact the mount
130 daemon on the remote host. This option is useful for
131 hosts that can run multiple NFS servers. The default
132 value depends on which kernel you are using.
133
134 nfsprog=n Use an alternate RPC program number to contact the NFS
135 daemon on the remote host. This option is useful for
136 hosts that can run multiple NFS servers. The default
137 value is 100003 which is the standard RPC NFS daemon
138 program number.
139
140 nfsvers=n Use an alternate RPC version number to contact the NFS
141 daemon on the remote host. This option is useful for
142 hosts that can run multiple NFS servers. The default
143 value depends on which kernel you are using.
144
145 vers=n vers is an alternative to nfsvers and is compatible with
146 many other operating systems.
147
148 nolock Disable NFS locking. Do not start lockd. This is appro‐
149 priate for mounting the root filesystem or /usr or /var.
150 These filesystems are typically either read-only or not
151 shared, and in those cases, remote locking is not
152 needed. This also needs to be used with some old NFS
153 servers that don't support locking.
154 Note that applications can still get locks on files, but
155 the locks only provide exclusion locally. Other clients
156 mounting the same filesystem will not be able to detect
157 the locks.
158
159 bg If the first NFS mount attempt times out, retry the
160 mount in the background. After a mount operation is
161 backgrounded, all subsequent mounts on the same NFS
162 server will be backgrounded immediately, without first
163 attempting the mount. A missing mount point is treated
164 as a timeout, to allow for nested NFS mounts.
165
166 fg If the first NFS mount attempt times out, retry the
167 mount in the foreground. This is the complement of the
168 bg option, and also the default behavior.
169
170 soft If an NFS file operation has a major timeout then report
171 an I/O error to the calling program. The default is to
172 continue retrying NFS file operations indefinitely.
173
174 hard If an NFS file operation has a major timeout then report
175 "server not responding" on the console and continue
176 retrying indefinitely. This is the default.
177
178 intr If an NFS file operation has a major timeout and it is
179 hard mounted, then allow signals to interupt the file
180 operation and cause it to return EINTR to the calling
181 program. The default is to not allow file operations to
182 be interrupted.
183
184 posix Mount the NFS filesystem using POSIX semantics. This
185 allows an NFS filesystem to properly support the POSIX
186 pathconf command by querying the mount server for the
187 maximum length of a filename. To do this, the remote
188 host must support version two of the RPC mount protocol.
189 Many NFS servers support only version one.
190
191 nocto Suppress the retrieval of new attributes when creating a
192 file.
193
194 noac Disable all forms of attribute caching entirely. This
195 extracts a significant performance penalty but it allows
196 two different NFS clients to get reasonable results when
197 both clients are actively writing to a common export on
198 the server.
199
200 noacl Disables Access Control List (ACL) processing.
201
202 sec=mode Set the security flavor for this mount to "mode". The
203 default setting is sec=sys, which uses local unix uids
204 and gids to authenticate NFS operations (AUTH_SYS).
205 Other currently supported settings are: sec=krb5, which
206 uses Kerberos V5 instead of local unix uids and gids to
207 authenticate users; sec=krb5i, which uses Kerberos V5
208 for user authentication and performs integrity checking
209 of NFS operations using secure checksums to prevent data
210 tampering; and sec=krb5p, which uses Kerberos V5 for
211 user authentication and integrity checking, and encrypts
212 NFS traffic to prevent traffic sniffing (this is the
213 most secure setting). Note that there is a performance
214 penalty when using integrity or privacy.
215
216 tcp Mount the NFS filesystem using the TCP protocol. This
217 is the default if it is supported by both client and
218 server. Many NFS servers only support UDP.
219
220 udp Mount the NFS filesystem using the UDP protocol.
221
222 nordirplus Disables NFSv3 READDIRPLUS RPCs. Use this option when
223 mounting servers that don't support or have broken READ‐
224 DIRPLUS implementations.
225
226 nosharecache As of kernel 2.6.18, it is no longer possible to mount
227 the same same filesystem with different mount options to
228 a new mountpoint. It was deemed unsafe to do so, since
229 cached data cannot be shared between the two mount‐
230 points. In consequence, files or directories that were
231 common to both mountpoint subtrees could often be seen
232 to be out of sync following an update.
233 This option allows administrators to select the
234 pre-2.6.18 behaviour, permitting the same filesystem to
235 be mounted with different mount options.
236 Beware: Use of this option is not recommended unless you
237 are certain that there are no hard links or subtrees of
238 this mountpoint that are mounted elsewhere.
239
240 All of the non-value options have corresponding nooption forms. For
241 example, nointr means don't allow file operations to be interrupted.
242
243 Options for the nfs4 file system type
244 rsize=n The number of bytes nfs4 uses when reading files from
245 the server. The rsize is negotiated between the server
246 and client to determine the largest block size that both
247 can support. The value specified by this option is the
248 maximum size that could be used; however, the actual
249 size used may be smaller. Note: Setting this size to a
250 value less than the largest supported block size will
251 adversely affect performance.
252
253 wsize=n The number of bytes nfs4 uses when writing files to the
254 server. The wsize is negotiated between the server and
255 client to determine the largest block size that both can
256 support. The value specified by this option is the max‐
257 imum size that could be used; however, the actual size
258 used may be smaller. Note: Setting this size to a value
259 less than the largest supported block size will
260 adversely affect performance.
261
262 timeo=n The value in tenths of a second before sending the first
263 retransmission after an RPC timeout. The default value
264 depends on whether proto=udp or proto=tcp is in effect
265 (see below). The default value for UDP is 7 tenths of a
266 second. The default value for TCP is 60 seconds. After
267 the first timeout, the timeout is doubled after each
268 successive timeout until a maximum timeout of 60 seconds
269 is reached or the enough retransmissions have occured to
270 cause a major timeout. Then, if the filesystem is hard
271 mounted, each new timeout cascade restarts at twice the
272 initial value of the previous cascade, again doubling at
273 each retransmission. The maximum timeout is always 60
274 seconds.
275
276 retrans=n The number of minor timeouts and retransmissions that
277 must occur before a major timeout occurs. The default
278 is 5 timeouts for proto=udp and 2 timeouts for
279 proto=tcp. When a major timeout occurs, the file opera‐
280 tion is either aborted or a "server not responding" mes‐
281 sage is printed on the console.
282
283 acregmin=n The minimum time in seconds that attributes of a regular
284 file should be cached before requesting fresh informa‐
285 tion from a server. The default is 3 seconds.
286
287 acregmax=n The maximum time in seconds that attributes of a regular
288 file can be cached before requesting fresh information
289 from a server. The default is 60 seconds.
290
291 acdirmin=n The minimum time in seconds that attributes of a direc‐
292 tory should be cached before requesting fresh informa‐
293 tion from a server. The default is 30 seconds.
294
295 acdirmax=n The maximum time in seconds that attributes of a direc‐
296 tory can be cached before requesting fresh information
297 from a server. The default is 60 seconds.
298
299 actimeo=n Using actimeo sets all of acregmin, acregmax, acdirmin,
300 and acdirmax to the same value. There is no default
301 value.
302
303 retry=n The number of minutes to retry an NFS mount operation in
304 the foreground or background before giving up. The
305 default value for forground mounts is 2 minutes. The
306 default value for background mounts is 10000 minutes,
307 which is roughly one week.
308
309 port=n The numeric value of the port to connect to the NFS
310 server on. If the port number is 0 (the default) then
311 query the remote host's portmapper for the port number
312 to use. If the remote host's NFS daemon is not regis‐
313 tered with its portmapper, the standard NFS port number
314 2049 is used instead.
315
316 proto=n Mount the NFS filesystem using a specific network proto‐
317 col instead of the default UDP protocol. Many NFS ver‐
318 sion 4 servers only support TCP. Valid protocol types
319 are udp and tcp.
320
321 clientaddr=n On a multi-homed client, this causes the client to use a
322 specific callback address when communicating with an NFS
323 version 4 server. This option is currently ignored.
324
325 sec=mode Same as sec=mode for the nfs filesystem type (see
326 above).
327
328 bg If an NFS mount attempt times out, retry the mount in
329 the background. After a mount operation is back‐
330 grounded, all subsequent mounts on the same NFS server
331 will be backgrounded immediately, without first attempt‐
332 ing the mount. A missing mount point is treated as a
333 timeout, to allow for nested NFS mounts.
334
335 fg If the first NFS mount attempt times out, retry the
336 mount in the foreground. This is the complement of the
337 bg option, and also the default behavior.
338
339 soft If an NFS file operation has a major timeout then report
340 an I/O error to the calling program. The default is to
341 continue retrying NFS file operations indefinitely.
342
343 hard If an NFS file operation has a major timeout then report
344 "server not responding" on the console and continue
345 retrying indefinitely. This is the default.
346
347 intr If an NFS file operation has a major timeout and it is
348 hard mounted, then allow signals to interupt the file
349 operation and cause it to return EINTR to the calling
350 program. The default is to not allow file operations to
351 be interrupted.
352
353 nocto Suppress the retrieval of new attributes when creating a
354 file.
355
356 noac Disable attribute caching, and force synchronous writes.
357 This extracts a server performance penalty but it allows
358 two different NFS clients to get reasonable good results
359 when both clients are actively writing to common
360 filesystem on the server.
361
362 nosharecache As of kernel 2.6.18, it is no longer possible to mount
363 the same same filesystem with different mount options to
364 a new mountpoint. It was deemed unsafe to do so, since
365 cached data cannot be shared between the two mount‐
366 points. In consequence, files or directories that were
367 common to both mountpoint subtrees could often be seen
368 to be out of sync following an update.
369 This option allows administrators to select the
370 pre-2.6.18 behaviour, permitting the same filesystem to
371 be mounted with different mount options.
372 Beware: Use of this option is not recommended unless you
373 are certain that there are no hard links or subtrees of
374 this mountpoint that are mounted elsewhere.
375
376 All of the non-value options have corresponding nooption forms. For
377 example, nointr means don't allow file operations to be interrupted.
378
380 /etc/fstab
381
383 fstab(5), mount(8), umount(8), exports(5)
384
386 "Rick Sladkey" <jrs@world.std.com>
387
389 Checking files on NFS filesystem referenced by file descriptors (i.e.
390 the fcntl and ioctl families of functions) may lead to inconsistent
391 result due to the lack of consistency check in kernel even if noac is
392 used.
393
394
395
396Linux 0.99 20 November 1993 NFS(5)