2 .TH NFS 5 "2 November 2007"
4 nfs \- fstab format and options for the
10 NFS is an Internet Standard protocol
11 created by Sun Microsystems in 1984. NFS was developed
12 to allow file sharing between systems residing
13 on a local area network.
14 The Linux NFS client supports three versions
16 NFS version 2 [RFC1094],
17 NFS version 3 [RFC1813],
18 and NFS version 4 [RFC3530].
22 command attaches a file system to the system's
23 name space hierarchy at a given mount point.
28 should assemble a system's file name hierarchy
29 from various independent file systems
30 (including file systems exported by NFS servers).
33 file describes a single file system, its mount point,
34 and a set of default mount options for that mount point.
36 For NFS file system mounts, a line in the
38 file specifies the server name,
39 the path name of the exported server directory to mount,
40 the local directory that is the mount point,
41 the type of file system that is being mounted,
42 and a list of mount options that control
43 the way the filesystem is mounted and
44 how the NFS client behaves when accessing
45 files on this mount point.
46 The fifth and sixth fields on each line are not used
47 by NFS, thus conventionally each contain the digit zero. For example:
51 .TA 2.5i +0.75i +0.75i +1.0i
52 server:path /mountpoint fstype option,option,... 0 0
55 The server's hostname and export pathname
56 are separated by a colon, while
57 the mount options are separated by commas. The remaining fields
58 are separated by blanks or tabs.
60 The server's hostname can be an unqualified hostname,
61 a fully qualified domain name,
62 a dotted quad IPv4 address, or
63 an IPv6 address enclosed in square brackets.
64 Link-local and site-local IPv6 addresses must be accompanied by an
68 for details on specifying raw IPv6 addresses.
72 field contains "nfs", for whatever version of the protocol.
75 allow several mount options, which are described below.
79 for a description of generic mount options
80 available for all file systems. If you do not need to
81 specify any mount options, use the generic option
86 .SS "Options supported by all versions"
87 These options are valid to use with any NFS version.
90 Determines the recovery behavior of the NFS client
91 after an NFS request times out.
92 If neither option is specified (or if the
94 option is specified), NFS requests are retried indefinitely.
97 option is specified, then the NFS client fails an NFS request
100 retransmissions have been sent,
101 causing the NFS client to return an error
102 to the calling application.
105 A so-called "soft" timeout can cause
106 silent data corruption in certain cases. As such, use the
108 option only when client responsiveness
109 is more important than data integrity.
110 Using NFS over TCP or increasing the value of the
112 option may mitigate some of the risks of using the
117 The time (in tenths of a second) the NFS client waits for a
118 response before it retries an NFS request. If this
119 option is not specified, requests are retried every
120 60 seconds for NFS over TCP.
121 The NFS client does not perform any kind of timeout backoff
124 However, for NFS over UDP, the client uses an adaptive
125 algorithm to estimate an appropriate timeout value for frequently used
126 request types (such as READ and WRITE requests), but uses the
128 setting for infrequently used request types (such as FSINFO requests).
131 option is not specified,
132 infrequently used request types are retried after 1.1 seconds.
133 After each retransmission, the NFS client doubles the timeout for
135 up to a maximum timeout length of 60 seconds.
138 The number of times the NFS client retries a request before
139 it attempts further recovery action. If the
141 option is not specified, the NFS client tries each request
144 The NFS client generates a "server not responding" message
147 retries, then attempts further recovery (depending on whether the
149 mount option is in effect).
152 The maximum number of bytes in each network READ request
153 that the NFS client can receive when reading data from a file
155 The actual data payload size of each NFS READ request is equal to
158 setting. The largest read payload supported by the Linux NFS client
159 is 1,048,576 bytes (one megabyte).
163 value is a positive integral multiple of 1024.
166 values lower than 1024 are replaced with 4096; values larger than
167 1048576 are replaced with 1048576. If a specified value is within the supported
168 range but not a multiple of 1024, it is rounded down to the nearest
173 value is not specified, or if the specified
175 value is larger than the maximum that either client or server can support,
176 the client and server negotiate the largest
178 value that they can both support.
182 mount option as specified on the
184 command line appears in the
186 file. However, the effective
188 value negotiated by the client and server is reported in the
193 The maximum number of bytes per network WRITE request
194 that the NFS client can send when writing data to a file
195 on an NFS server. The actual data payload size of each
196 NFS WRITE request is equal to
199 setting. The largest write payload supported by the Linux NFS client
200 is 1,048,576 bytes (one megabyte).
206 value is a positive integral multiple of 1024.
209 values lower than 1024 are replaced with 4096; values larger than
210 1048576 are replaced with 1048576. If a specified value is within the supported
211 range but not a multiple of 1024, it is rounded down to the nearest
216 value is not specified, or if the specified
218 value is larger than the maximum that either client or server can support,
219 the client and server negotiate the largest
221 value that they can both support.
225 mount option as specified on the
227 command line appears in the
229 file. However, the effective
231 value negotiated by the client and server is reported in the
236 Selects whether the client may cache file attributes. If neither
237 option is specified (or if
239 is specified), the client caches file
242 To improve performance, NFS clients cache file
243 attributes. Every few seconds, an NFS client checks the server's version of each
244 file's attributes for updates. Changes that occur on the server in
245 those small intervals remain undetected until the client checks the
248 option prevents clients from caching file
249 attributes so that applications can more quickly detect file changes
252 In addition to preventing the client from caching file attributes,
255 option forces application writes to become synchronous so
256 that local changes to a file become visible on the server
257 immediately. That way, other clients can quickly detect recent
258 writes when they check the file's attributes.
262 option provides greater cache coherence among NFS clients
263 accessing the same files,
264 but it extracts a significant performance penalty.
265 As such, judicious use of file locking is encouraged instead.
266 The DATA AND METADATA COHERENCE section contains a detailed discussion
270 The minimum time (in seconds) that the NFS client caches
271 attributes of a regular file before it requests
272 fresh attribute information from a server.
273 If this option is not specified, the NFS client uses
277 The maximum time (in seconds) that the NFS client caches
278 attributes of a regular file before it requests
279 fresh attribute information from a server.
280 If this option is not specified, the NFS client uses
284 The minimum time (in seconds) that the NFS client caches
285 attributes of a directory before it requests
286 fresh attribute information from a server.
287 If this option is not specified, the NFS client uses
291 The maximum time (in seconds) that the NFS client caches
292 attributes of a directory before it requests
293 fresh attribute information from a server.
294 If this option is not specified, the NFS client uses
307 If this option is not specified, the NFS client uses
308 the defaults for each of these options listed above.
313 command behaves if an attempt to mount an export fails.
318 to exit with an error status if any part of the mount request
319 times out or fails outright.
320 This is called a "foreground" mount,
321 and is the default behavior if neither the
325 mount option is specified.
329 option is specified, a timeout or failure causes the
331 command to fork a child which continues to attempt
333 The parent immediately returns with a zero exit code.
334 This is known as a "background" mount.
336 If the local mount point directory is missing, the
338 command acts as if the mount request timed out.
339 This permits nested NFS mounts specified in
341 to proceed in any order during system initialization,
342 even if some NFS servers are not yet available.
343 Alternatively these issues can be addressed
344 using an automounter (refer to
349 The number of minutes that the
351 command retries an NFS mount operation
352 in the foreground or background before giving up.
353 If this option is not specified, the default value for foreground mounts
354 is 2 minutes, and the default value for background mounts is 10000 minutes
355 (80 minutes shy of one week).
356 If a value of zero is specified, the
358 command exits immediately after the first failure.
361 The RPCGSS security flavor to use for accessing files on this mount point.
364 option is not specified, or if
366 is specified, the NFS client uses the AUTH_SYS security flavor
367 for all NFS requests on this mount point.
368 Valid security flavors are
381 Refer to the SECURITY CONSIDERATIONS section for details.
383 .BR sharecache " / " nosharecache
384 Determines how the client's data cache and attribute cache are shared
385 when mounting the same export more than once concurrently. Using the
386 same cache reduces memory requirements on the client and presents
387 identical file contents to applications when the same remote file is
388 accessed via different mount points.
390 If neither option is specified, or if the
393 specified, then a single cache is used for all mount points that
394 access the same export. If the
397 then that mount point gets a unique cache. Note that when data and
398 attribute caches are shared, the mount options from the first mount
399 point take effect for subsequent concurrent mounts of the same export.
401 As of kernel 2.6.18, the behavior specified by
403 is legacy caching behavior. This
404 is considered a data risk since multiple cached copies
405 of the same file on the same client can become out of sync
406 following a local update of one of the copies.
408 .BR resvport " / " noresvport
409 Specifies whether the NFS client should use a privileged source port
410 when communicating with an NFS server for this mount point.
411 If this option is not specified, or the
413 option is specified, the NFS client uses a privileged source port.
416 option is specified, the NFS client uses a non-privileged source port.
417 This option is supported in kernels 2.6.28 and later.
419 Using non-privileged source ports helps increase the maximum number of
420 NFS mount points allowed on a client, but NFS servers must be configured
421 to allow clients to connect via non-privileged source ports.
423 Refer to the SECURITY CONSIDERATIONS section for important details.
425 .BI lookupcache= mode
426 Specifies how the kernel manages its cache of directory entries
427 for a given mount point.
435 This option is supported in kernels 2.6.28 and later.
437 The Linux NFS client caches the result of all NFS LOOKUP requests.
438 If the requested directory entry exists on the server,
439 the result is referred to as
441 If the requested directory entry does not exist on the server,
442 the result is referred to as
445 If this option is not specified, or if
447 is specified, the client assumes both types of directory cache entries
448 are valid until their parent directory's cached attributes expire.
451 .BR pos " or " positive
452 is specified, the client assumes positive entries are valid
453 until their parent directory's cached attributes expire, but
454 always revalidates negative entires before an application
460 the client revalidates both types of directory cache entries
461 before an application can use them.
462 This permits quick detection of files that were created or removed
463 by other clients, but can impact application and server performance.
465 The DATA AND METADATA COHERENCE section contains a
466 detailed discussion of these trade-offs.
467 .SS "Options for versions 2 and 3 only"
468 Use these options, along with the options in the above subsection,
469 for NFSv2/v3 only. They will be ignored for newer versions.
472 The transport protocol name and protocol family the NFS client uses
473 to transmit requests to the NFS server for this mount point.
474 If an NFS server has both an IPv4 and an IPv6 address, using a specific
475 netid will force the use of IPv4 or IPv6 networking to communicate
478 If support for TI-RPC is built into the
482 is a valid netid listed in
484 The value "rdma" may also be specified.
487 command does not have TI-RPC support, then
489 is one of "tcp," "udp," or "rdma," and only IPv4 may be used.
491 Each transport protocol uses different default
496 Refer to the description of these two mount options for details.
498 In addition to controlling how the NFS client transmits requests to
499 the server, this mount option also controls how the
501 command communicates with the server's rpcbind and mountd services.
502 Specifying a netid that uses TCP forces all traffic from the
504 command and the NFS client to use TCP.
505 Specifying a netid that uses UDP forces all traffic types to use UDP.
509 mount option is not specified, the
511 command discovers which protocols the server supports
512 and chooses an appropriate transport for each service.
513 Refer to the TRANSPORT METHODS section for more details.
518 option is an alternative to specifying
520 It is included for compatibility with other operating systems.
525 option is an alternative to specifying
527 It is included for compatibility with other operating systems.
532 option is an alternative to specifying
536 The numeric value of the server's NFS service port.
537 If the server's NFS service is not available on the specified port,
538 the mount request fails.
540 If this option is not specified, or if the specified port value is 0,
541 then the NFS client uses the NFS service port number
542 advertised by the server's rpcbind service.
543 The mount request fails if the server's rpcbind service is not available,
544 the server's NFS service is not registered with its rpcbind service,
545 or the server's NFS service is not available on the advertised port.
548 The numeric value of the server's mountd port.
549 If the server's mountd service is not available on the specified port,
550 the mount request fails.
552 If this option is not specified,
553 or if the specified port value is 0, then the
555 command uses the mountd service port number
556 advertised by the server's rpcbind service.
557 The mount request fails if the server's rpcbind service is not available,
558 the server's mountd service is not registered with its rpcbind service,
559 or the server's mountd service is not available on the advertised port.
561 This option can be used when mounting an NFS server
562 through a firewall that blocks the rpcbind protocol.
564 .BI mountproto= netid
565 The transport protocol name and protocol family the NFS client uses
566 to transmit requests to the NFS server's mountd service when performing
567 this mount request, and when later unmounting this mount point.
569 If support for TI-RPC is built into the
573 is a valid netid listed in
577 is one of "tcp" or "udp," and only IPv4 may be used.
579 This option can be used when mounting an NFS server
580 through a firewall that blocks a particular transport.
581 When used in combination with the
583 option, different transports for mountd requests and NFS requests
585 If the server's mountd service is not available via the specified
586 transport, the mount request fails.
588 Refer to the TRANSPORT METHODS section for more on how the
590 mount option interacts with the
595 The hostname of the host running mountd.
596 If this option is not specified, the
598 command assumes that the mountd service runs
599 on the same host as the NFS service.
602 The RPC version number used to contact the server's mountd.
603 If this option is not specified, the client uses a version number
604 appropriate to the requested NFS version.
605 This option is useful when multiple NFS services
606 are running on the same remote server host.
609 The maximum length of a pathname component on this mount.
610 If this option is not specified, the maximum length is negotiated
611 with the server. In most cases, this maximum length is 255 characters.
613 Some early versions of NFS did not support this negotiation.
614 Using this option ensures that
616 reports the proper maximum component length to applications
620 The NFS protocol version number used to contact the server's NFS service.
621 If the server does not support the requested version, the mount request fails.
622 If this option is not specified, the client negociate a suitable version with
623 the server, trying version 4 first, version 3 second, and version 2 last.
626 This option is an alternative to the
629 It is included for compatibility with other operating systems.
631 .BR lock " / " nolock
632 Selects whether to use the NLM sideband protocol to lock files on the server.
633 If neither option is specified (or if
635 is specified), NLM locking is used for this mount point.
638 option, applications can lock files,
639 but such locks provide exclusion only against other applications
640 running on the same client.
641 Remote applications are not affected by these locks.
643 NLM locking must be disabled with the
645 option when using NFS to mount
649 contains files used by the NLM implementation on Linux.
652 option is also required when mounting exports on NFS servers
653 that do not support the NLM protocol.
655 .BR intr " / " nointr
656 Selects whether to allow signals to interrupt file operations
657 on this mount point. If neither option
661 signals do not interrupt NFS file operations. If
663 is specified, system calls return EINTR if an in-progress NFS operation is interrupted by
668 option is preferred to using the
670 option because it is significantly less likely to result in data corruption.
673 .BR intr " / " nointr
674 mount option is deprecated after kernel 2.6.25.
675 Only SIGKILL can interrupt a pending NFS operation on these kernels,
676 and if specified, this mount option is ignored to provide backwards
677 compatibility with older kernels.
680 Selects whether to use close-to-open cache coherence semantics.
681 If neither option is specified (or if
683 is specified), the client uses close-to-open
684 cache coherence semantics. If the
686 option is specified, the client uses a non-standard heuristic to determine when
687 files on the server have changed.
691 option may improve performance for read-only mounts,
692 but should be used only if the data on the server changes only occasionally.
693 The DATA AND METADATA COHERENCE section discusses the behavior
694 of this option in more detail.
697 Selects whether to use the NFSACL sideband protocol on this mount point.
698 The NFSACL sideband protocol is a proprietary protocol
699 implemented in Solaris that manages Access Control Lists. NFSACL was never
700 made a standard part of the NFS protocol specification.
707 the NFS client negotiates with the server
708 to see if the NFSACL protocol is supported,
709 and uses it if the server supports it.
710 Disabling the NFSACL sideband protocol may be necessary
711 if the negotiation causes problems on the client or server.
712 Refer to the SECURITY CONSIDERATIONS section for more details.
714 .BR rdirplus " / " nordirplus
715 Selects whether to use NFS version 3 READDIRPLUS requests.
716 If this option is not specified, the NFS client uses READDIRPLUS requests
717 on NFS version 3 mounts to read small directories.
718 Some applications perform better if the client uses only READDIR requests
720 .SS "Options for version 4 only"
721 Use these options, along with the options in the first subsection above,
722 for NFSv4 only. They will be ignored with older versions.
725 The transport protocol name and protocol family the NFS client uses
726 to transmit requests to the NFS server for this mount point.
727 If an NFS server has both an IPv4 and an IPv6 address, using a specific
728 netid will force the use of IPv4 or IPv6 networking to communicate
731 If support for TI-RPC is built into the
735 is a valid netid listed in
739 is one of "tcp" or "udp," and only IPv4 may be used.
741 All NFS version 4 servers are required to support TCP,
742 so if this mount option is not specified, the NFS version 4 client
743 uses the TCP protocol.
744 Refer to the TRANSPORT METHODS section for more details.
747 The numeric value of the server's NFS service port.
748 If the server's NFS service is not available on the specified port,
749 the mount request fails.
751 If this mount option is not specified,
752 the NFS client uses the standard NFS port number of 2049
753 without first checking the server's rpcbind service.
754 This allows an NFS version 4 client to contact an NFS version 4
755 server through a firewall that may block rpcbind requests.
757 If the specified port value is 0,
758 then the NFS client uses the NFS service port number
759 advertised by the server's rpcbind service.
760 The mount request fails if the server's rpcbind service is not available,
761 the server's NFS service is not registered with its rpcbind service,
762 or the server's NFS service is not available on the advertised port.
764 .BR intr " / " nointr
765 Selects whether to allow signals to interrupt file operations
766 on this mount point. If neither option is specified (or if
768 is specified), system calls return EINTR if an in-progress NFS operation
769 is interrupted by a signal. If
771 is specified, signals do not
772 interrupt NFS operations.
776 option is preferred to using the
778 option because it is significantly less likely to result in data corruption.
781 .BR intr " / " nointr
782 mount option is deprecated after kernel 2.6.25.
783 Only SIGKILL can interrupt a pending NFS operation on these kernels,
784 and if specified, this mount option is ignored to provide backwards
785 compatibility with older kernels.
788 Selects whether to use close-to-open cache coherence semantics
789 for NFS directories on this mount point.
795 the default is to use close-to-open cache coherence
796 semantics for directories.
798 File data caching behavior is not affected by this option.
799 The DATA AND METADATA COHERENCE section discusses
800 the behavior of this option in more detail.
802 .BI clientaddr= n.n.n.n
803 Specifies a single IPv4 address (in dotted-quad form),
804 or a non-link-local IPv6 address,
805 that the NFS client advertises to allow servers
806 to perform NFS version 4 callback requests against
807 files on this mount point. If the server is unable to
808 establish callback connections to clients, performance
809 may degrade, or accesses to files may temporarily hang.
811 If this option is not specified, the
813 command attempts to discover an appropriate callback address automatically.
814 The automatic discovery process is not perfect, however.
815 In the presence of multiple client network interfaces,
816 special routing policies,
817 or atypical network topologies,
818 the exact address to use for callbacks may be nontrivial to determine.
819 .SH nfs4 FILE SYSTEM TYPE
822 file system type is an old syntax for specifying NFSv4 usage. It can still
823 be used with all NFSv4-specific and common options, excepted the
826 .SH MOUNT CONFIGURATION FILE
827 If the mount command is configured to do so, all of the mount options
828 described in the previous section can also be configured in the
829 .I /etc/nfsmount.conf
834 To mount an export using NFS version 2,
837 file system type and specify the
840 To mount using NFS version 3,
843 file system type and specify the
846 To mount using NFS version 4,
849 file system type, with the
855 The following example from an
857 file causes the mount command to negotiate
858 reasonable defaults for NFS behavior.
861 .TA 2.5i +0.7i +0.7i +.7i
862 server:/export /mnt nfs defaults 0 0
865 Here is an example from an /etc/fstab file for an NFS version 2 mount over UDP.
868 .TA 2.5i +0.7i +0.7i +.7i
869 server:/export /mnt nfs nfsvers=2,proto=udp 0 0
872 Try this example to mount using NFS version 4 over TCP
873 with Kerberos 5 mutual authentication.
876 .TA 2.5i +0.7i +0.7i +.7i
877 server:/export /mnt nfs4 sec=krb5 0 0
880 This example can be used to mount /usr over NFS.
883 .TA 2.5i +0.7i +0.7i +.7i
884 server:/export /usr nfs ro,nolock,nocto,actimeo=3600 0 0
887 This example shows how to mount an NFS server
888 using a raw IPv6 link-local address.
891 .TA 2.5i +0.7i +0.7i +.7i
892 [fe80::215:c5ff:fb3e:e2b1%eth0]:/export /mnt nfs defaults 0 0
894 .SH "TRANSPORT METHODS"
895 NFS clients send requests to NFS servers via
896 Remote Procedure Calls, or
898 The RPC client discovers remote service endpoints automatically,
899 handles per-request authentication,
900 adjusts request parameters for different byte endianness on client and server,
901 and retransmits requests that may have been lost by the network or server.
902 RPC requests and replies flow over a network transport.
906 command, NFS client, and NFS server
907 can automatically negotiate proper transport
908 and data transfer size settings for a mount point.
909 In some cases, however, it pays to specify
910 these settings explicitly using mount options.
912 Traditionally, NFS clients used the UDP transport exclusively for
913 transmitting requests to servers. Though its implementation is
914 simple, NFS over UDP has many limitations that prevent smooth
915 operation and good performance in some common deployment
916 environments. Even an insignificant packet loss rate results in the
917 loss of whole NFS requests; as such, retransmit timeouts are usually
918 in the subsecond range to allow clients to recover quickly from
919 dropped requests, but this can result in extraneous network traffic
922 However, UDP can be quite effective in specialized settings where
923 the networks MTU is large relative to NFSs data transfer size (such
924 as network environments that enable jumbo Ethernet frames). In such
925 environments, trimming the
929 settings so that each
930 NFS read or write request fits in just a few network frames (or even
931 in a single frame) is advised. This reduces the probability that
932 the loss of a single MTU-sized network frame results in the loss of
933 an entire large read or write request.
935 TCP is the default transport protocol used for all modern NFS
936 implementations. It performs well in almost every conceivable
937 network environment and provides excellent guarantees against data
938 corruption caused by network unreliability. TCP is often a
939 requirement for mounting a server through a network firewall.
941 Under normal circumstances, networks drop packets much more
942 frequently than NFS servers drop requests. As such, an aggressive
943 retransmit timeout setting for NFS over TCP is unnecessary. Typical
944 timeout settings for NFS over TCP are between one and ten minutes.
945 After the client exhausts its retransmits (the value of the
947 mount option), it assumes a network partition has occurred,
948 and attempts to reconnect to the server on a fresh socket. Since
949 TCP itself makes network data transfer reliable,
953 can safely be allowed to default to the largest values supported by
954 both client and server, independent of the network's MTU size.
955 .SS "Using the mountproto mount option"
956 This section applies only to NFS version 2 and version 3 mounts
957 since NFS version 4 does not use a separate protocol for mount
960 The Linux NFS client can use a different transport for
961 contacting an NFS server's rpcbind service, its mountd service,
962 its Network Lock Manager (NLM) service, and its NFS service.
963 The exact transports employed by the Linux NFS client for
964 each mount point depends on the settings of the transport
965 mount options, which include
968 .BR udp ", and " tcp .
970 The client sends Network Status Manager (NSM) notifications
971 via UDP no matter what transport options are specified, but
972 listens for server NSM notifications on both UDP and TCP.
973 The NFS Access Control List (NFSACL) protocol shares the same
974 transport as the main NFS service.
976 If no transport options are specified, the Linux NFS client
977 uses UDP to contact the server's mountd service, and TCP to
978 contact its NLM and NFS services by default.
980 If the server does not support these transports for these services, the
982 command attempts to discover what the server supports, and then retries
983 the mount request once using the discovered transports.
984 If the server does not advertise any transport supported by the client
985 or is misconfigured, the mount request fails.
988 option is in effect, the mount command backgrounds itself and continues
989 to attempt the specified mount request.
997 option is specified but the
999 option is not, the specified transport is used to contact
1000 both the server's mountd service and for the NLM and NFS services.
1004 option is specified but none of the
1005 .BR proto ", " udp " or " tcp
1006 options are specified, then the specified transport is used for the
1007 initial mountd request, but the mount command attempts to discover
1008 what the server supports for the NFS protocol, preferring TCP if
1009 both transports are supported.
1012 .BR mountproto " and " proto
1014 .BR udp " or " tcp )
1015 options are specified, then the transport specified by the
1017 option is used for the initial mountd request, and the transport
1021 .BR udp " or " tcp " options)"
1022 is used for NFS, no matter what order these options appear.
1023 No automatic service discovery is performed if these options are
1027 .BR proto ", " udp ", " tcp ", "
1030 options are specified more than once on the same mount command line,
1031 then the value of the rightmost instance of each of these options
1033 .SH "DATA AND METADATA COHERENCE"
1034 Some modern cluster file systems provide
1035 perfect cache coherence among their clients.
1036 Perfect cache coherence among disparate NFS clients
1037 is expensive to achieve, especially on wide area networks.
1038 As such, NFS settles for weaker cache coherence that
1039 satisfies the requirements of most file sharing types. Normally,
1040 file sharing is completely sequential:
1041 first client A opens a file, writes something to it, then closes it;
1042 then client B opens the same file, and reads the changes.
1044 .SS "Close-to-open cache consistency"
1045 When an application opens a file stored on an NFS server,
1046 the NFS client checks that it still exists on the server
1047 and is permitted to the opener by sending a GETATTR or ACCESS request.
1048 When the application closes the file,
1049 the NFS client writes back any pending changes
1050 to the file so that the next opener can view the changes.
1051 This also gives the NFS client an opportunity to report
1052 any server write errors to the application
1053 via the return code from
1055 The behavior of checking at open time and flushing at close time
1056 is referred to as close-to-open cache consistency.
1057 .SS "Weak cache consistency"
1058 There are still opportunities for a client's data cache
1059 to contain stale data.
1060 The NFS version 3 protocol introduced "weak cache consistency"
1061 (also known as WCC) which provides a way of efficiently checking
1062 a file's attributes before and after a single request.
1063 This allows a client to help identify changes
1064 that could have been made by other clients.
1066 When a client is using many concurrent operations
1067 that update the same file at the same time
1068 (for example, during asynchronous write behind),
1069 it is still difficult to tell whether it was
1070 that client's updates or some other client's updates
1071 that altered the file.
1072 .SS "Attribute caching"
1075 mount option to achieve attribute cache coherence
1076 among multiple clients.
1077 Almost every file system operation checks
1078 file attribute information.
1079 The client keeps this information cached
1080 for a period of time to reduce network and server load.
1083 is in effect, a client's file attribute cache is disabled,
1084 so each operation that needs to check a file's attributes
1085 is forced to go back to the server.
1086 This permits a client to see changes to a file very quickly,
1087 at the cost of many extra network operations.
1089 Be careful not to confuse the
1091 option with "no data caching."
1094 mount option prevents the client from caching file metadata,
1095 but there are still races that may result in data cache incoherence
1096 between client and server.
1098 The NFS protocol is not designed to support
1099 true cluster file system cache coherence
1100 without some type of application serialization.
1101 If absolute cache coherence among clients is required,
1102 applications should use file locking. Alternatively, applications
1103 can also open their files with the O_DIRECT flag
1104 to disable data caching entirely.
1105 .SS "Directory entry caching"
1106 The Linux NFS client caches the result of all NFS LOOKUP requests.
1107 If the requested directory entry exists on the server,
1108 the result is referred to as a
1109 .IR positive " lookup result.
1110 If the requested directory entry does not exist on the server
1111 (that is, the server returned ENOENT),
1112 the result is referred to as
1113 .IR negative " lookup result.
1115 To detect when directory entries have been added or removed
1117 the Linux NFS client watches a directory's mtime.
1118 If the client detects a change in a directory's mtime,
1119 the client drops all cached LOOKUP results for that directory.
1120 Since the directory's mtime is a cached attribute, it may
1121 take some time before a client notices it has changed.
1122 See the descriptions of the
1123 .BR acdirmin ", " acdirmax ", and " noac
1124 mount options for more information about
1125 how long a directory's mtime is cached.
1127 Caching directory entries improves the performance of applications that
1128 do not share files with applications on other clients.
1129 Using cached information about directories can interfere
1130 with applications that run concurrently on multiple clients and
1131 need to detect the creation or removal of files quickly, however.
1134 mount option allows some tuning of directory entry caching behavior.
1136 Before kernel release 2.6.28,
1137 the Linux NFS client tracked only positive lookup results.
1138 This permitted applications to detect new directory entries
1139 created by other clients quickly while still providing some of the
1140 performance benefits of caching.
1141 If an application depends on the previous lookup caching behavior
1142 of the Linux NFS client, you can use
1143 .BR lookupcache=positive .
1145 If the client ignores its cache and validates every application
1146 lookup request with the server,
1147 that client can immediately detect when a new directory
1148 entry has been either created or removed by another client.
1149 You can specify this behavior using
1150 .BR lookupcache=none .
1151 The extra NFS requests needed if the client does not
1152 cache directory entries can exact a performance penalty.
1153 Disabling lookup caching
1154 should result in less of a performance penalty than using
1156 and has no effect on how the NFS client caches the attributes of files.
1158 .SS "The sync mount option"
1159 The NFS client treats the
1161 mount option differently than some other file systems
1164 for a description of the generic
1173 is specified (or if the
1175 option is specified),
1176 the NFS client delays sending application
1177 writes to the server
1178 until any of these events occur:
1180 Memory pressure forces reclamation of system memory resources.
1182 An application flushes file data explicitly with
1188 An application closes a file with
1191 The file is locked/unlocked via
1194 In other words, under normal circumstances,
1195 data written by an application may not immediately appear
1196 on the server that hosts the file.
1200 option is specified on a mount point,
1201 any system call that writes data to files on that mount point
1202 causes that data to be flushed to the server
1203 before the system call returns control to user space.
1204 This provides greater data cache coherence among clients,
1205 but at a significant performance cost.
1207 Applications can use the O_SYNC open flag to force application
1208 writes to individual files to go to the server immediately without
1212 .SS "Using file locks with NFS"
1213 The Network Lock Manager protocol is a separate sideband protocol
1214 used to manage file locks in NFS version 2 and version 3.
1215 To support lock recovery after a client or server reboot,
1216 a second sideband protocol --
1217 known as the Network Status Manager protocol --
1220 file locking is supported directly in the main NFS protocol,
1221 and the NLM and NSM sideband protocols are not used.
1223 In most cases, NLM and NSM services are started automatically,
1224 and no extra configuration is required.
1225 Configure all NFS clients with fully-qualified domain names
1226 to ensure that NFS servers can find clients to notify them of server reboots.
1228 NLM supports advisory file locks only.
1229 To lock NFS files, use
1231 with the F_GETLK and F_SETLK commands.
1232 The NFS client converts file locks obtained via
1236 When mounting servers that do not support the NLM protocol,
1237 or when mounting an NFS server through a firewall
1238 that blocks the NLM service port,
1241 mount option. NLM locking must be disabled with the
1243 option when using NFS to mount
1247 contains files used by the NLM implementation on Linux.
1251 option may also be advised to improve the performance
1252 of a proprietary application which runs on a single client
1253 and uses file locks extensively.
1254 .SS "NFS version 4 caching features"
1255 The data and metadata caching behavior of NFS version 4
1256 clients is similar to that of earlier versions.
1257 However, NFS version 4 adds two features that improve
1259 .I change attributes
1261 .IR "file delegation" .
1265 is a new part of NFS file and directory metadata
1266 which tracks data changes.
1267 It replaces the use of a file's modification
1268 and change time stamps
1269 as a way for clients to validate the content
1271 Change attributes are independent of the time stamp
1272 resolution on either the server or client, however.
1276 is a contract between an NFS version 4 client
1277 and server that allows the client to treat a file temporarily
1278 as if no other client is accessing it.
1279 The server promises to notify the client (via a callback request) if another client
1280 attempts to access that file.
1281 Once a file has been delegated to a client, the client can
1282 cache that file's data and metadata aggressively without
1283 contacting the server.
1285 File delegations come in two flavors:
1291 delegation means that the server notifies the client
1292 about any other clients that want to write to the file.
1295 delegation means that the client gets notified about
1296 either read or write accessors.
1298 Servers grant file delegations when a file is opened,
1299 and can recall delegations at any time when another
1300 client wants access to the file that conflicts with
1301 any delegations already granted.
1302 Delegations on directories are not supported.
1304 In order to support delegation callback, the server
1305 checks the network return path to the client during
1306 the client's initial contact with the server.
1307 If contact with the client cannot be established,
1308 the server simply does not grant any delegations to
1310 .SH "SECURITY CONSIDERATIONS"
1311 NFS servers control access to file data,
1312 but they depend on their RPC implementation
1313 to provide authentication of NFS requests.
1314 Traditional NFS access control mimics
1315 the standard mode bit access control provided in local file systems.
1316 Traditional RPC authentication uses a number
1317 to represent each user
1318 (usually the user's own uid),
1319 a number to represent the user's group (the user's gid),
1320 and a set of up to 16 auxiliary group numbers
1321 to represent other groups of which the user may be a member.
1323 Typically, file data and user ID values appear unencrypted
1324 (i.e. "in the clear") on the network.
1325 Moreover, NFS versions 2 and 3 use
1326 separate sideband protocols for mounting,
1327 locking and unlocking files,
1328 and reporting system status of clients and servers.
1329 These auxiliary protocols use no authentication.
1331 In addition to combining these sideband protocols with the main NFS protocol,
1332 NFS version 4 introduces more advanced forms of access control,
1333 authentication, and in-transit data protection.
1334 The NFS version 4 specification mandates NFSv4 ACLs,
1335 RPCGSS authentication, and RPCGSS security flavors
1336 that provide per-RPC integrity checking and encryption.
1337 Because NFS version 4 combines the
1338 function of the sideband protocols into the main NFS protocol,
1339 the new security features apply to all NFS version 4 operations
1340 including mounting, file locking, and so on.
1341 RPCGSS authentication can also be used with NFS versions 2 and 3,
1342 but does not protect their sideband protocols.
1346 mount option specifies the RPCGSS security mode
1347 that is in effect on a given NFS mount point.
1350 provides cryptographic proof of a user's identity in each RPC request.
1351 This provides strong verification of the identity of users
1352 accessing data on the server.
1353 Note that additional configuration besides adding this mount option
1354 is required in order to enable Kerberos security.
1357 man page for details.
1359 Two additional flavors of Kerberos security are supported:
1365 security flavor provides a cryptographically strong guarantee
1366 that the data in each RPC request has not been tampered with.
1369 security flavor encrypts every RPC request
1370 to prevent data exposure during network transit; however,
1371 expect some performance impact
1372 when using integrity checking or encryption.
1373 Similar support for other forms of cryptographic security (such as lipkey and SPKM3)
1376 The NFS version 4 protocol allows
1377 clients and servers to negotiate among multiple security flavors
1378 during mount processing.
1379 However, Linux does not yet implement such negotiation.
1380 The Linux client specifies a single security flavor at mount time
1381 which remains in effect for the lifetime of the mount.
1382 If the server does not support this flavor,
1383 the initial mount request is rejected by the server.
1384 .SS "Using non-privileged source ports"
1385 NFS clients usually communicate with NFS servers via network sockets.
1386 Each end of a socket is assigned a port value, which is simply a number
1387 between 1 and 65535 that distinguishes socket endpoints at the same
1389 A socket is uniquely defined by a tuple that includes the transport
1390 protocol (TCP or UDP) and the port values and IP addresses of both
1393 The NFS client can choose any source port value for its sockets,
1394 but usually chooses a
1397 A privileged port is a port value less than 1024.
1398 Only a process with root privileges may create a socket
1399 with a privileged source port.
1401 The exact range of privileged source ports that can be chosen is
1402 set by a pair of sysctls to avoid choosing a well-known port, such as
1403 the port used by ssh.
1404 This means the number of source ports available for the NFS client,
1405 and therefore the number of socket connections that can be used
1407 is practically limited to only a few hundred.
1409 As described above, the traditional default NFS authentication scheme,
1410 known as AUTH_SYS, relies on sending local UID and GID numbers to identify
1411 users making NFS requests.
1412 An NFS server assumes that if a connection comes from a privileged port,
1413 the UID and GID numbers in the NFS requests on this connection have been
1414 verified by the client's kernel or some other local authority.
1415 This is an easy system to spoof, but on a trusted physical network between
1416 trusted hosts, it is entirely adequate.
1418 Roughly speaking, one socket is used for each NFS mount point.
1419 If a client could use non-privileged source ports as well,
1420 the number of sockets allowed,
1421 and thus the maximum number of concurrent mount points,
1422 would be much larger.
1424 Using non-privileged source ports may compromise server security somewhat,
1425 since any user on AUTH_SYS mount points can now pretend to be any other
1426 when making NFS requests.
1427 Thus NFS servers do not support this by default.
1428 They explicitly allow it usually via an export option.
1430 To retain good security while allowing as many mount points as possible,
1431 it is best to allow non-privileged client connections only if the server
1432 and client both require strong authentication, such as Kerberos.
1433 .SS "Mounting through a firewall"
1434 A firewall may reside between an NFS client and server,
1435 or the client or server may block some of its own ports via IP
1437 It is still possible to mount an NFS server through a firewall,
1440 command's automatic service endpoint discovery mechanisms may not work; this
1441 requires you to provide specific endpoint details via NFS mount options.
1443 NFS servers normally run a portmapper or rpcbind daemon to advertise
1444 their service endpoints to clients. Clients use the rpcbind daemon to determine:
1446 What network port each RPC-based service is using
1448 What transport protocols each RPC-based service supports
1450 The rpcbind daemon uses a well-known port number (111) to help clients find a service endpoint.
1451 Although NFS often uses a standard port number (2049),
1452 auxiliary services such as the NLM service can choose
1453 any unused port number at random.
1455 Common firewall configurations block the well-known rpcbind port.
1456 In the absense of an rpcbind service,
1457 the server administrator fixes the port number
1458 of NFS-related services so that the firewall
1459 can allow access to specific NFS service ports.
1460 Client administrators then specify the port number
1461 for the mountd service via the
1466 It may also be necessary to enforce the use of TCP or UDP
1467 if the firewall blocks one of those transports.
1468 .SS "NFS Access Control Lists"
1469 Solaris allows NFS version 3 clients direct access
1470 to POSIX Access Control Lists stored in its local file systems.
1471 This proprietary sideband protocol, known as NFSACL,
1472 provides richer access control than mode bits.
1473 Linux implements this protocol
1474 for compatibility with the Solaris NFS implementation.
1475 The NFSACL protocol never became a standard part
1476 of the NFS version 3 specification, however.
1478 The NFS version 4 specification mandates a new version
1479 of Access Control Lists that are semantically richer than POSIX ACLs.
1480 NFS version 4 ACLs are not fully compatible with POSIX ACLs; as such,
1481 some translation between the two is required
1482 in an environment that mixes POSIX ACLs and NFS version 4.
1490 option is not fully supported.
1491 Generic options, such as
1493 can be modified using the
1496 but NFS-specific options are not all supported.
1497 The underlying transport or NFS version
1498 cannot be changed by a remount, for example.
1499 Performing a remount on an NFS file system mounted with the
1501 option may have unintended consequences.
1504 option is a mixture of a generic option,
1506 and an NFS-specific option
1509 Before 2.4.7, the Linux NFS client did not support NFS over TCP.
1511 Before 2.4.20, the Linux NFS client used a heuristic
1512 to determine whether cached file data was still valid
1513 rather than using the standard close-to-open cache coherency method
1516 Starting with 2.4.22, the Linux NFS client employs
1517 a Van Jacobsen-based RTT estimator to determine retransmit
1518 timeout values when using NFS over UDP.
1520 Before 2.6.0, the Linux NFS client did not support NFS version 4.
1522 Before 2.6.8, the Linux NFS client used only synchronous reads and writes
1524 .BR rsize " and " wsize
1525 settings were smaller than the system's page size.
1527 The Linux NFS client does not yet support
1528 certain optional features of the NFS version 4 protocol,
1529 such as security negotiation, server referrals, and named attributes.
1544 .BR rpc.svcgssd (8),
1547 RFC 768 for the UDP specification.
1549 RFC 793 for the TCP specification.
1551 RFC 1094 for the NFS version 2 specification.
1553 RFC 1813 for the NFS version 3 specification.
1555 RFC 1832 for the XDR specification.
1557 RFC 1833 for the RPC bind specification.
1559 RFC 2203 for the RPCSEC GSS API protocol specification.
1561 RFC 3530 for the NFS version 4 specification.