2 .TH NFS 5 "2 November 2007"
4 nfs \- fstab format and options for the
12 NFS is an Internet Standard protocol
13 created by Sun Microsystems in 1984. NFS was developed
14 to allow file sharing between systems residing
15 on a local area network.
16 The Linux NFS client supports three versions
18 NFS version 2 [RFC1094],
19 NFS version 3 [RFC1813],
20 and NFS version 4 [RFC3530].
24 command attaches a file system to the system's
25 name space hierarchy at a given mount point.
30 should assemble a system's file name hierarchy
31 from various independent file systems
32 (including file systems exported by NFS servers).
35 file describes a single file system, its mount point,
36 and a set of default mount options for that mount point.
38 For NFS file system mounts, a line in the
40 file specifies the server name,
41 the path name of the exported server directory to mount,
42 the local directory that is the mount point,
43 the type of file system that is being mounted,
44 and a list of mount options that control
45 the way the filesystem is mounted and
46 how the NFS client behaves when accessing
47 files on this mount point.
48 The fifth and sixth fields on each line are not used
49 by NFS, thus conventionally each contain the digit zero. For example:
53 .TA 2.5i +0.75i +0.75i +1.0i
54 server:path /mountpoint fstype option,option,... 0 0
57 The server's hostname and export pathname
58 are separated by a colon, while
59 the mount options are separated by commas. The remaining fields
60 are separated by blanks or tabs.
61 The server's hostname can be an unqualified hostname,
62 a fully qualified domain name,
63 or a dotted quad IPv4 address.
66 field contains either "nfs" (for version 2 or version 3 NFS mounts)
67 or "nfs4" (for NFS version 4 mounts).
72 file system types share similar mount options,
73 which are described below.
77 for a description of generic mount options
78 available for all file systems. If you do not need to
79 specify any mount options, use the generic option
84 .SS "Valid options for either the nfs or nfs4 file system type"
85 These options are valid to use when mounting either
90 They imply the same behavior
91 and have the same default for both file system types.
94 Determines the recovery behavior of the NFS client
95 after an NFS request times out.
96 If neither option is specified (or if the
98 option is specified), NFS requests are retried indefinitely.
101 option is specified, then the NFS client fails an NFS request
104 retransmissions have been sent,
105 causing the NFS client to return an error
106 to the calling application.
109 A so-called "soft" timeout can cause
110 silent data corruption in certain cases. As such, use the
112 option only when client responsiveness
113 is more important than data integrity.
114 Using NFS over TCP or increasing the value of the
116 option may mitigate some of the risks of using the
121 The time (in tenths of a second) the NFS client waits for a
122 response before it retries an NFS request. If this
123 option is not specified, requests are retried every
124 60 seconds for NFS over TCP.
125 The NFS client does not perform any kind of timeout backoff
128 However, for NFS over UDP, the client uses an adaptive
129 algorithm to estimate an appropriate timeout value for frequently used
130 request types (such as READ and WRITE requests), but uses the
132 setting for infrequently used request types (such as FSINFO requests).
135 option is not specified,
136 infrequently used request types are retried after 1.1 seconds.
137 After each retransmission, the NFS client doubles the timeout for
139 up to a maximum timeout length of 60 seconds.
142 The number of times the NFS client retries a request before
143 it attempts further recovery action. If the
145 option is not specified, the NFS client tries each request
148 The NFS client generates a "server not responding" message
151 retries, then attempts further recovery (depending on whether the
153 mount option is in effect).
156 The maximum number of bytes in each network READ request
157 that the NFS client can receive when reading data from a file
159 The actual data payload size of each NFS READ request is equal to
162 setting. The largest read payload supported by the Linux NFS client
163 is 1,048,576 bytes (one megabyte).
167 value is a positive integral multiple of 1024.
170 values lower than 1024 are replaced with 4096; values larger than
171 1048576 are replaced with 1048576. If a specified value is within the supported
172 range but not a multiple of 1024, it is rounded down to the nearest
177 value is not specified, or if the specified
179 value is larger than the maximum that either client or server can support,
180 the client and server negotiate the largest
182 value that they can both support.
186 mount option as specified on the
188 command line appears in the
190 file. However, the effective
192 value negotiated by the client and server is reported in the
197 The maximum number of bytes per network WRITE request
198 that the NFS client can send when writing data to a file
199 on an NFS server. The actual data payload size of each
200 NFS WRITE request is equal to
203 setting. The largest write payload supported by the Linux NFS client
204 is 1,048,576 bytes (one megabyte).
210 value is a positive integral multiple of 1024.
213 values lower than 1024 are replaced with 4096; values larger than
214 1048576 are replaced with 1048576. If a specified value is within the supported
215 range but not a multiple of 1024, it is rounded down to the nearest
220 value is not specified, or if the specified
222 value is larger than the maximum that either client or server can support,
223 the client and server negotiate the largest
225 value that they can both support.
229 mount option as specified on the
231 command line appears in the
233 file. However, the effective
235 value negotiated by the client and server is reported in the
240 Selects whether the client may cache file attributes. If neither
241 option is specified (or if
243 is specified), the client caches file
246 To improve performance, NFS clients cache file
247 attributes. Every few seconds, an NFS client checks the server's version of each
248 file's attributes for updates. Changes that occur on the server in
249 those small intervals remain undetected until the client checks the
252 option prevents clients from caching file
253 attributes so that applications can more quickly detect file changes
256 In addition to preventing the client from caching file attributes,
259 option forces application writes to become synchronous so
260 that local changes to a file become visible on the server
261 immediately. That way, other clients can quickly detect recent
262 writes when they check the file's attributes.
266 option provides greater cache coherence among NFS clients
267 accessing the same files,
268 but it extracts a significant performance penalty.
269 As such, judicious use of file locking is encouraged instead.
270 The DATA AND METADATA COHERENCE section contains a detailed discussion
274 The minimum time (in seconds) that the NFS client caches
275 attributes of a regular file before it requests
276 fresh attribute information from a server.
277 If this option is not specified, the NFS client uses
281 The maximum time (in seconds) that the NFS client caches
282 attributes of a regular file before it requests
283 fresh attribute information from a server.
284 If this option is not specified, the NFS client uses
288 The minimum time (in seconds) that the NFS client caches
289 attributes of a directory before it requests
290 fresh attribute information from a server.
291 If this option is not specified, the NFS client uses
295 The maximum time (in seconds) that the NFS client caches
296 attributes of a directory before it requests
297 fresh attribute information from a server.
298 If this option is not specified, the NFS client uses
311 If this option is not specified, the NFS client uses
312 the defaults for each of these options listed above.
317 command behaves if an attempt to mount an export fails.
322 to exit with an error status if any part of the mount request
323 times out or fails outright.
324 This is called a "foreground" mount,
325 and is the default behavior if neither the
329 mount option is specified.
333 option is specified, a timeout or failure causes the
335 command to fork a child which continues to attempt
337 The parent immediately returns with a zero exit code.
338 This is known as a "background" mount.
340 If the local mount point directory is missing, the
342 command acts as if the mount request timed out.
343 This permits nested NFS mounts specified in
345 to proceed in any order during system initialization,
346 even if some NFS servers are not yet available.
347 Alternatively these issues can be addressed
348 using an automounter (refer to
353 The number of minutes that the
355 command retries an NFS mount operation
356 in the foreground or background before giving up.
357 If this option is not specified, the default value for foreground mounts
358 is 2 minutes, and the default value for background mounts is 10000 minutes (80 minutes shy of one week).
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 "Valid options for the nfs file system type"
468 Use these options, along with the options in the above subsection,
474 The transport the NFS client uses
475 to transmit requests to the NFS server for this mount point.
481 Each transport uses different default
485 settings; refer to the description of these two mount options for details.
487 In addition to controlling how the NFS client transmits requests to
488 the server, this mount option also controls how the
490 command communicates with the server's rpcbind and mountd services.
493 forces all traffic from the
495 command and the NFS client to use TCP.
498 forces all traffic types to use UDP.
502 mount option is not specified, the
504 command discovers which protocols the server supports
505 and chooses an appropriate transport for each service.
506 Refer to the TRANSPORT METHODS section for more details.
511 option is an alternative to specifying
513 It is included for compatibility with other operating systems.
518 option is an alternative to specifying
520 It is included for compatibility with other operating systems.
523 The numeric value of the server's NFS service port.
524 If the server's NFS service is not available on the specified port,
525 the mount request fails.
527 If this option is not specified, or if the specified port value is 0,
528 then the NFS client uses the NFS service port number
529 advertised by the server's rpcbind service.
530 The mount request fails if the server's rpcbind service is not available,
531 the server's NFS service is not registered with its rpcbind service,
532 or the server's NFS service is not available on the advertised port.
535 The numeric value of the server's mountd port.
536 If the server's mountd service is not available on the specified port,
537 the mount request fails.
539 If this option is not specified,
540 or if the specified port value is 0, then the
542 command uses the mountd service port number
543 advertised by the server's rpcbind service.
544 The mount request fails if the server's rpcbind service is not available,
545 the server's mountd service is not registered with its rpcbind service,
546 or the server's mountd service is not available on the advertised port.
548 This option can be used when mounting an NFS server
549 through a firewall that blocks the rpcbind protocol.
551 .BI mountproto= transport
552 The transport the NFS client uses
553 to transmit requests to the NFS server's mountd service when performing
554 this mount request, and when later unmounting this mount point.
561 This option can be used when mounting an NFS server
562 through a firewall that blocks a particular transport.
563 When used in combination with the
565 option, different transports for mountd requests and NFS requests
567 If the server's mountd service is not available via the specified
568 transport, the mount request fails.
569 Refer to the TRANSPORT METHODS section for more on how the
571 mount option interacts with the
576 The hostname of the host running mountd.
577 If this option is not specified, the
579 command assumes that the mountd service runs
580 on the same host as the NFS service.
583 The RPC version number used to contact the server's mountd.
584 If this option is not specified, the client uses a version number
585 appropriate to the requested NFS version.
586 This option is useful when multiple NFS services
587 are running on the same remote server host.
590 The maximum length of a pathname component on this mount.
591 If this option is not specified, the maximum length is negotiated
592 with the server. In most cases, this maximum length is 255 characters.
594 Some early versions of NFS did not support this negotiation.
595 Using this option ensures that
597 reports the proper maximum component length to applications
601 The NFS protocol version number used to contact the server's NFS service.
602 The Linux client supports version 2 and version 3 of the NFS protocol
603 when using the file system type
605 If the server does not support the requested version,
606 the mount request fails.
607 If this option is not specified, the client attempts to use version 3,
608 but negotiates the NFS version with the server if version 3 support
612 This option is an alternative to the
615 It is included for compatibility with other operating systems.
617 .BR lock " / " nolock
618 Selects whether to use the NLM sideband protocol to lock files on the server.
619 If neither option is specified (or if
621 is specified), NLM locking is used for this mount point.
624 option, applications can lock files,
625 but such locks provide exclusion only against other applications
626 running on the same client.
627 Remote applications are not affected by these locks.
629 NLM locking must be disabled with the
631 option when using NFS to mount
635 contains files used by the NLM implementation on Linux.
638 option is also required when mounting exports on NFS servers
639 that do not support the NLM protocol.
641 .BR intr " / " nointr
642 Selects whether to allow signals to interrupt file operations
643 on this mount point. If neither option
647 signals do not interrupt NFS file operations. If
649 is specified, system calls return EINTR if an in-progress NFS operation is interrupted by
654 option is preferred to using the
656 option because it is significantly less likely to result in data corruption.
659 .BR intr " / " nointr
660 mount option is deprecated after kernel 2.6.25.
661 Only SIGKILL can interrupt a pending NFS operation on these kernels,
662 and if specified, this mount option is ignored to provide backwards
663 compatibility with older kernels.
666 Selects whether to use close-to-open cache coherence semantics.
667 If neither option is specified (or if
669 is specified), the client uses close-to-open
670 cache coherence semantics. If the
672 option is specified, the client uses a non-standard heuristic to determine when
673 files on the server have changed.
677 option may improve performance for read-only mounts,
678 but should be used only if the data on the server changes only occasionally.
679 The DATA AND METADATA COHERENCE section discusses the behavior
680 of this option in more detail.
683 Selects whether to use the NFSACL sideband protocol on this mount point.
684 The NFSACL sideband protocol is a proprietary protocol
685 implemented in Solaris that manages Access Control Lists. NFSACL was never
686 made a standard part of the NFS protocol specification.
693 the NFS client negotiates with the server
694 to see if the NFSACL protocol is supported,
695 and uses it if the server supports it.
696 Disabling the NFSACL sideband protocol may be necessary
697 if the negotiation causes problems on the client or server.
698 Refer to the SECURITY CONSIDERATIONS section for more details.
700 .BR rdirplus " / " nordirplus
701 Selects whether to use NFS version 3 READDIRPLUS requests.
702 If this option is not specified, the NFS client uses READDIRPLUS requests
703 on NFS version 3 mounts to read small directories.
704 Some applications perform better if the client uses only READDIR requests
706 .SS "Valid options for the nfs4 file system type"
707 Use these options, along with the options in the first subsection above,
713 The transport the NFS client uses
714 to transmit requests to the NFS server for this mount point.
720 All NFS version 4 servers are required to support TCP,
721 so if this mount option is not specified, the NFS version 4 client
722 uses the TCP transport protocol.
723 Refer to the TRANSPORT METHODS section for more details.
726 The numeric value of the server's NFS service port.
727 If the server's NFS service is not available on the specified port,
728 the mount request fails.
730 If this mount option is not specified,
731 the NFS client uses the standard NFS port number of 2049
732 without first checking the server's rpcbind service.
733 This allows an NFS version 4 client to contact an NFS version 4
734 server through a firewall that may block rpcbind requests.
736 If the specified port value is 0,
737 then the NFS client uses the NFS service port number
738 advertised by the server's rpcbind service.
739 The mount request fails if the server's rpcbind service is not available,
740 the server's NFS service is not registered with its rpcbind service,
741 or the server's NFS service is not available on the advertised port.
743 .BR intr " / " nointr
744 Selects whether to allow signals to interrupt file operations
745 on this mount point. If neither option is specified (or if
747 is specified), system calls return EINTR if an in-progress NFS operation
748 is interrupted by a signal. If
750 is specified, signals do not
751 interrupt NFS operations.
755 option is preferred to using the
757 option because it is significantly less likely to result in data corruption.
760 .BR intr " / " nointr
761 mount option is deprecated after kernel 2.6.25.
762 Only SIGKILL can interrupt a pending NFS operation on these kernels,
763 and if specified, this mount option is ignored to provide backwards
764 compatibility with older kernels.
767 Selects whether to use close-to-open cache coherence semantics
768 for NFS directories on this mount point.
774 the default is to use close-to-open cache coherence
775 semantics for directories.
777 File data caching behavior is not affected by this option.
778 The DATA AND METADATA COHERENCE section discusses
779 the behavior of this option in more detail.
781 .BI clientaddr= n.n.n.n
782 Specifies a single IPv4 address (in dotted-quad form)
783 that the NFS client advertises to allow servers
784 to perform NFS version 4 callback requests against
785 files on this mount point. If the server is unable to
786 establish callback connections to clients, performance
787 may degrade, or accesses to files may temporarily hang.
789 If this option is not specified, the
791 command attempts to discover an appropriate callback address automatically.
792 The automatic discovery process is not perfect, however.
793 In the presence of multiple client network interfaces,
794 special routing policies,
795 or atypical network topologies,
796 the exact address to use for callbacks may be nontrivial to determine.
798 To mount an export using NFS version 2,
801 file system type and specify the
804 To mount using NFS version 3,
807 file system type and specify the
810 To mount using NFS version 4,
816 mount option is not supported for the
820 The following example from an
822 file causes the mount command to negotiate
823 reasonable defaults for NFS behavior.
826 .TA 2.5i +0.7i +0.7i +.7i
827 server:/export /mnt nfs defaults 0 0
830 Here is an example from an /etc/fstab file for an NFS version 2 mount over UDP.
833 .TA 2.5i +0.7i +0.7i +.7i
834 server:/export /mnt nfs nfsvers=2,proto=udp 0 0
837 Try this example to mount using NFS version 4 over TCP
838 with Kerberos 5 mutual authentication.
841 .TA 2.5i +0.7i +0.7i +.7i
842 server:/export /mnt nfs4 sec=krb5 0 0
845 This example can be used to mount /usr over NFS.
848 .TA 2.5i +0.7i +0.7i +.7i
849 server:/export /usr nfs ro,nolock,nocto,actimeo=3600 0 0
851 .SH "TRANSPORT METHODS"
852 NFS clients send requests to NFS servers via
853 Remote Procedure Calls, or
855 The RPC client discovers remote service endpoints automatically,
856 handles per-request authentication,
857 adjusts request parameters for different byte endianness on client and server,
858 and retransmits requests that may have been lost by the network or server.
859 RPC requests and replies flow over a network transport.
863 command, NFS client, and NFS server
864 can automatically negotiate proper transport
865 and data transfer size settings for a mount point.
866 In some cases, however, it pays to specify
867 these settings explicitly using mount options.
869 Traditionally, NFS clients used the UDP transport exclusively for
870 transmitting requests to servers. Though its implementation is
871 simple, NFS over UDP has many limitations that prevent smooth
872 operation and good performance in some common deployment
873 environments. Even an insignificant packet loss rate results in the
874 loss of whole NFS requests; as such, retransmit timeouts are usually
875 in the subsecond range to allow clients to recover quickly from
876 dropped requests, but this can result in extraneous network traffic
879 However, UDP can be quite effective in specialized settings where
880 the networks MTU is large relative to NFSs data transfer size (such
881 as network environments that enable jumbo Ethernet frames). In such
882 environments, trimming the
886 settings so that each
887 NFS read or write request fits in just a few network frames (or even
888 in a single frame) is advised. This reduces the probability that
889 the loss of a single MTU-sized network frame results in the loss of
890 an entire large read or write request.
892 TCP is the default transport protocol used for all modern NFS
893 implementations. It performs well in almost every conceivable
894 network environment and provides excellent guarantees against data
895 corruption caused by network unreliability. TCP is often a
896 requirement for mounting a server through a network firewall.
898 Under normal circumstances, networks drop packets much more
899 frequently than NFS servers drop requests. As such, an aggressive
900 retransmit timeout setting for NFS over TCP is unnecessary. Typical
901 timeout settings for NFS over TCP are between one and ten minutes.
902 After the client exhausts its retransmits (the value of the
904 mount option), it assumes a network partition has occurred,
905 and attempts to reconnect to the server on a fresh socket. Since
906 TCP itself makes network data transfer reliable,
910 can safely be allowed to default to the largest values supported by
911 both client and server, independent of the network's MTU size.
912 .SS "Using the mountproto mount option"
913 This section applies only to NFS version 2 and version 3 mounts
914 since NFS version 4 does not use a separate protocol for mount
917 The Linux NFS client can use a different transport for
918 contacting an NFS server's rpcbind service, its mountd service,
919 its Network Lock Manager (NLM) service, and its NFS service.
920 The exact transports employed by the Linux NFS client for
921 each mount point depends on the settings of the transport
922 mount options, which include
925 .BR udp ", and " tcp .
927 The client sends Network Status Manager (NSM) notifications
928 via UDP no matter what transport options are specified, but
929 listens for server NSM notifications on both UDP and TCP.
930 The NFS Access Control List (NFSACL) protocol shares the same
931 transport as the main NFS service.
933 If no transport options are specified, the Linux NFS client
934 uses UDP to contact the server's mountd service, and TCP to
935 contact its NLM and NFS services by default.
937 If the server does not support these transports for these services, the
939 command attempts to discover what the server supports, and then retries
940 the mount request once using the discovered transports.
941 If the server does not advertise any transport supported by the client
942 or is misconfigured, the mount request fails.
945 option is in effect, the mount command backgrounds itself and continues
946 to attempt the specified mount request.
954 option is specified but the
956 option is not, the specified transport is used to contact
957 both the server's mountd service and for the NLM and NFS services.
961 option is specified but none of the
962 .BR proto ", " udp " or " tcp
963 options are specified, then the specified transport is used for the
964 initial mountd request, but the mount command attempts to discover
965 what the server supports for the NFS protocol, preferring TCP if
966 both transports are supported.
969 .BR mountproto " and " proto
972 options are specified, then the transport specified by the
974 option is used for the initial mountd request, and the transport
978 .BR udp " or " tcp " options)"
979 is used for NFS, no matter what order these options appear.
980 No automatic service discovery is performed if these options are
984 .BR proto ", " udp ", " tcp ", "
987 options are specified more than once on the same mount command line,
988 then the value of the rightmost instance of each of these options
990 .SH "DATA AND METADATA COHERENCE"
991 Some modern cluster file systems provide
992 perfect cache coherence among their clients.
993 Perfect cache coherence among disparate NFS clients
994 is expensive to achieve, especially on wide area networks.
995 As such, NFS settles for weaker cache coherence that
996 satisfies the requirements of most file sharing types. Normally,
997 file sharing is completely sequential:
998 first client A opens a file, writes something to it, then closes it;
999 then client B opens the same file, and reads the changes.
1001 .SS "Close-to-open cache consistency"
1002 When an application opens a file stored on an NFS server,
1003 the NFS client checks that it still exists on the server
1004 and is permitted to the opener by sending a GETATTR or ACCESS request.
1005 When the application closes the file,
1006 the NFS client writes back any pending changes
1007 to the file so that the next opener can view the changes.
1008 This also gives the NFS client an opportunity to report
1009 any server write errors to the application
1010 via the return code from
1012 The behavior of checking at open time and flushing at close time
1013 is referred to as close-to-open cache consistency.
1014 .SS "Weak cache consistency"
1015 There are still opportunities for a client's data cache
1016 to contain stale data.
1017 The NFS version 3 protocol introduced "weak cache consistency"
1018 (also known as WCC) which provides a way of efficiently checking
1019 a file's attributes before and after a single request.
1020 This allows a client to help identify changes
1021 that could have been made by other clients.
1023 When a client is using many concurrent operations
1024 that update the same file at the same time
1025 (for example, during asynchronous write behind),
1026 it is still difficult to tell whether it was
1027 that client's updates or some other client's updates
1028 that altered the file.
1029 .SS "Attribute caching"
1032 mount option to achieve attribute cache coherence
1033 among multiple clients.
1034 Almost every file system operation checks
1035 file attribute information.
1036 The client keeps this information cached
1037 for a period of time to reduce network and server load.
1040 is in effect, a client's file attribute cache is disabled,
1041 so each operation that needs to check a file's attributes
1042 is forced to go back to the server.
1043 This permits a client to see changes to a file very quickly,
1044 at the cost of many extra network operations.
1046 Be careful not to confuse the
1048 option with "no data caching."
1051 mount option prevents the client from caching file metadata,
1052 but there are still races that may result in data cache incoherence
1053 between client and server.
1055 The NFS protocol is not designed to support
1056 true cluster file system cache coherence
1057 without some type of application serialization.
1058 If absolute cache coherence among clients is required,
1059 applications should use file locking. Alternatively, applications
1060 can also open their files with the O_DIRECT flag
1061 to disable data caching entirely.
1062 .SS "Directory entry caching"
1063 The Linux NFS client caches the result of all NFS LOOKUP requests.
1064 If the requested directory entry exists on the server,
1065 the result is referred to as a
1066 .IR positive " lookup result.
1067 If the requested directory entry does not exist on the server
1068 (that is, the server returned ENOENT),
1069 the result is referred to as
1070 .IR negative " lookup result.
1072 To detect when directory entries have been added or removed
1074 the Linux NFS client watches a directory's mtime.
1075 If the client detects a change in a directory's mtime,
1076 the client drops all cached LOOKUP results for that directory.
1077 Since the directory's mtime is a cached attribute, it may
1078 take some time before a client notices it has changed.
1079 See the descriptions of the
1080 .BR acdirmin ", " acdirmax ", and " noac
1081 mount options for more information about
1082 how long a directory's mtime is cached.
1084 Caching directory entries improves the performance of applications that
1085 do not share files with applications on other clients.
1086 Using cached information about directories can interfere
1087 with applications that run concurrently on multiple clients and
1088 need to detect the creation or removal of files quickly, however.
1091 mount option allows some tuning of directory entry caching behavior.
1093 Before kernel release 2.6.28,
1094 the Linux NFS client tracked only positive lookup results.
1095 This permitted applications to detect new directory entries
1096 created by other clients quickly while still providing some of the
1097 performance benefits of caching.
1098 If an application depends on the previous lookup caching behavior
1099 of the Linux NFS client, you can use
1100 .BR lookupcache=positive .
1102 If the client ignores its cache and validates every application
1103 lookup request with the server,
1104 that client can immediately detect when a new directory
1105 entry has been either created or removed by another client.
1106 You can specify this behavior using
1107 .BR lookupcache=none .
1108 The extra NFS requests needed if the client does not
1109 cache directory entries can exact a performance penalty.
1110 Disabling lookup caching
1111 should result in less of a performance penalty than using
1113 and has no effect on how the NFS client caches the attributes of files.
1115 .SS "The sync mount option"
1116 The NFS client treats the
1118 mount option differently than some other file systems
1121 for a description of the generic
1130 is specified (or if the
1132 option is specified),
1133 the NFS client delays sending application
1134 writes to the server
1135 until any of these events occur:
1137 Memory pressure forces reclamation of system memory resources.
1139 An application flushes file data explicitly with
1145 An application closes a file with
1148 The file is locked/unlocked via
1151 In other words, under normal circumstances,
1152 data written by an application may not immediately appear
1153 on the server that hosts the file.
1157 option is specified on a mount point,
1158 any system call that writes data to files on that mount point
1159 causes that data to be flushed to the server
1160 before the system call returns control to user space.
1161 This provides greater data cache coherence among clients,
1162 but at a significant performance cost.
1164 Applications can use the O_SYNC open flag to force application
1165 writes to individual files to go to the server immediately without
1169 .SS "Using file locks with NFS"
1170 The Network Lock Manager protocol is a separate sideband protocol
1171 used to manage file locks in NFS version 2 and version 3.
1172 To support lock recovery after a client or server reboot,
1173 a second sideband protocol --
1174 known as the Network Status Manager protocol --
1177 file locking is supported directly in the main NFS protocol,
1178 and the NLM and NSM sideband protocols are not used.
1180 In most cases, NLM and NSM services are started automatically,
1181 and no extra configuration is required.
1182 Configure all NFS clients with fully-qualified domain names
1183 to ensure that NFS servers can find clients to notify them of server reboots.
1185 NLM supports advisory file locks only.
1186 To lock NFS files, use
1188 with the F_GETLK and F_SETLK commands.
1189 The NFS client converts file locks obtained via
1193 When mounting servers that do not support the NLM protocol,
1194 or when mounting an NFS server through a firewall
1195 that blocks the NLM service port,
1198 mount option. NLM locking must be disabled with the
1200 option when using NFS to mount
1204 contains files used by the NLM implementation on Linux.
1208 option may also be advised to improve the performance
1209 of a proprietary application which runs on a single client
1210 and uses file locks extensively.
1211 .SS "NFS version 4 caching features"
1212 The data and metadata caching behavior of NFS version 4
1213 clients is similar to that of earlier versions.
1214 However, NFS version 4 adds two features that improve
1216 .I change attributes
1218 .IR "file delegation" .
1222 is a new part of NFS file and directory metadata
1223 which tracks data changes.
1224 It replaces the use of a file's modification
1225 and change time stamps
1226 as a way for clients to validate the content
1228 Change attributes are independent of the time stamp
1229 resolution on either the server or client, however.
1233 is a contract between an NFS version 4 client
1234 and server that allows the client to treat a file temporarily
1235 as if no other client is accessing it.
1236 The server promises to notify the client (via a callback request) if another client
1237 attempts to access that file.
1238 Once a file has been delegated to a client, the client can
1239 cache that file's data and metadata aggressively without
1240 contacting the server.
1242 File delegations come in two flavors:
1248 delegation means that the server notifies the client
1249 about any other clients that want to write to the file.
1252 delegation means that the client gets notified about
1253 either read or write accessors.
1255 Servers grant file delegations when a file is opened,
1256 and can recall delegations at any time when another
1257 client wants access to the file that conflicts with
1258 any delegations already granted.
1259 Delegations on directories are not supported.
1261 In order to support delegation callback, the server
1262 checks the network return path to the client during
1263 the client's initial contact with the server.
1264 If contact with the client cannot be established,
1265 the server simply does not grant any delegations to
1267 .SH "SECURITY CONSIDERATIONS"
1268 NFS servers control access to file data,
1269 but they depend on their RPC implementation
1270 to provide authentication of NFS requests.
1271 Traditional NFS access control mimics
1272 the standard mode bit access control provided in local file systems.
1273 Traditional RPC authentication uses a number
1274 to represent each user
1275 (usually the user's own uid),
1276 a number to represent the user's group (the user's gid),
1277 and a set of up to 16 auxiliary group numbers
1278 to represent other groups of which the user may be a member.
1280 Typically, file data and user ID values appear unencrypted
1281 (i.e. "in the clear") on the network.
1282 Moreover, NFS versions 2 and 3 use
1283 separate sideband protocols for mounting,
1284 locking and unlocking files,
1285 and reporting system status of clients and servers.
1286 These auxiliary protocols use no authentication.
1288 In addition to combining these sideband protocols with the main NFS protocol,
1289 NFS version 4 introduces more advanced forms of access control,
1290 authentication, and in-transit data protection.
1291 The NFS version 4 specification mandates NFSv4 ACLs,
1292 RPCGSS authentication, and RPCGSS security flavors
1293 that provide per-RPC integrity checking and encryption.
1294 Because NFS version 4 combines the
1295 function of the sideband protocols into the main NFS protocol,
1296 the new security features apply to all NFS version 4 operations
1297 including mounting, file locking, and so on.
1298 RPCGSS authentication can also be used with NFS versions 2 and 3,
1299 but does not protect their sideband protocols.
1303 mount option specifies the RPCGSS security mode
1304 that is in effect on a given NFS mount point.
1307 provides cryptographic proof of a user's identity in each RPC request.
1308 This provides strong verification of the identity of users
1309 accessing data on the server.
1310 Note that additional configuration besides adding this mount option
1311 is required in order to enable Kerberos security.
1314 man page for details.
1316 Two additional flavors of Kerberos security are supported:
1322 security flavor provides a cryptographically strong guarantee
1323 that the data in each RPC request has not been tampered with.
1326 security flavor encrypts every RPC request
1327 to prevent data exposure during network transit; however,
1328 expect some performance impact
1329 when using integrity checking or encryption.
1330 Similar support for other forms of cryptographic security (such as lipkey and SPKM3)
1333 The NFS version 4 protocol allows
1334 clients and servers to negotiate among multiple security flavors
1335 during mount processing.
1336 However, Linux does not yet implement such negotiation.
1337 The Linux client specifies a single security flavor at mount time
1338 which remains in effect for the lifetime of the mount.
1339 If the server does not support this flavor,
1340 the initial mount request is rejected by the server.
1341 .SS "Using non-privileged source ports"
1342 NFS clients usually communicate with NFS servers via network sockets.
1343 Each end of a socket is assigned a port value, which is simply a number
1344 between 1 and 65535 that distinguishes socket endpoints at the same
1346 A socket is uniquely defined by a tuple that includes the transport
1347 protocol (TCP or UDP) and the port values and IP addresses of both
1350 The NFS client can choose any source port value for its sockets,
1351 but usually chooses a
1354 A privileged port is a port value less than 1024.
1355 Only a process with root privileges may create a socket
1356 with a privileged source port.
1358 The exact range of privileged source ports that can be chosen is
1359 set by a pair of sysctls to avoid choosing a well-known port, such as
1360 the port used by ssh.
1361 This means the number of source ports available for the NFS client,
1362 and therefore the number of socket connections that can be used
1364 is practically limited to only a few hundred.
1366 As described above, the traditional default NFS authentication scheme,
1367 known as AUTH_SYS, relies on sending local UID and GID numbers to identify
1368 users making NFS requests.
1369 An NFS server assumes that if a connection comes from a privileged port,
1370 the UID and GID numbers in the NFS requests on this connection have been
1371 verified by the client's kernel or some other local authority.
1372 This is an easy system to spoof, but on a trusted physical network between
1373 trusted hosts, it is entirely adequate.
1375 Roughly speaking, one socket is used for each NFS mount point.
1376 If a client could use non-privileged source ports as well,
1377 the number of sockets allowed,
1378 and thus the maximum number of concurrent mount points,
1379 would be much larger.
1381 Using non-privileged source ports may compromise server security somewhat,
1382 since any user on AUTH_SYS mount points can now pretend to be any other
1383 when making NFS requests.
1384 Thus NFS servers do not support this by default.
1385 They explicitly allow it usually via an export option.
1387 To retain good security while allowing as many mount points as possible,
1388 it is best to allow non-privileged client connections only if the server
1389 and client both require strong authentication, such as Kerberos.
1390 .SS "Mounting through a firewall"
1391 A firewall may reside between an NFS client and server,
1392 or the client or server may block some of its own ports via IP
1394 It is still possible to mount an NFS server through a firewall,
1397 command's automatic service endpoint discovery mechanisms may not work; this
1398 requires you to provide specific endpoint details via NFS mount options.
1400 NFS servers normally run a portmapper or rpcbind daemon to advertise
1401 their service endpoints to clients. Clients use the rpcbind daemon to determine:
1403 What network port each RPC-based service is using
1405 What transport protocols each RPC-based service supports
1407 The rpcbind daemon uses a well-known port number (111) to help clients find a service endpoint.
1408 Although NFS often uses a standard port number (2049),
1409 auxiliary services such as the NLM service can choose
1410 any unused port number at random.
1412 Common firewall configurations block the well-known rpcbind port.
1413 In the absense of an rpcbind service,
1414 the server administrator fixes the port number
1415 of NFS-related services so that the firewall
1416 can allow access to specific NFS service ports.
1417 Client administrators then specify the port number
1418 for the mountd service via the
1423 It may also be necessary to enforce the use of TCP or UDP
1424 if the firewall blocks one of those transports.
1425 .SS "NFS Access Control Lists"
1426 Solaris allows NFS version 3 clients direct access
1427 to POSIX Access Control Lists stored in its local file systems.
1428 This proprietary sideband protocol, known as NFSACL,
1429 provides richer access control than mode bits.
1430 Linux implements this protocol
1431 for compatibility with the Solaris NFS implementation.
1432 The NFSACL protocol never became a standard part
1433 of the NFS version 3 specification, however.
1435 The NFS version 4 specification mandates a new version
1436 of Access Control Lists that are semantically richer than POSIX ACLs.
1437 NFS version 4 ACLs are not fully compatible with POSIX ACLs; as such,
1438 some translation between the two is required
1439 in an environment that mixes POSIX ACLs and NFS version 4.
1447 option is not fully supported.
1448 Generic options, such as
1450 can be modified using the
1453 but NFS-specific options are not all supported.
1454 The underlying transport or NFS version
1455 cannot be changed by a remount, for example.
1456 Performing a remount on an NFS file system mounted with the
1458 option may have unintended consequences.
1461 option is a mixture of a generic option,
1463 and an NFS-specific option
1466 Before 2.4.7, the Linux NFS client did not support NFS over TCP.
1468 Before 2.4.20, the Linux NFS client used a heuristic
1469 to determine whether cached file data was still valid
1470 rather than using the standard close-to-open cache coherency method
1473 Starting with 2.4.22, the Linux NFS client employs
1474 a Van Jacobsen-based RTT estimator to determine retransmit
1475 timeout values when using NFS over UDP.
1477 Before 2.6.0, the Linux NFS client did not support NFS version 4.
1479 Before 2.6.8, the Linux NFS client used only synchronous reads and writes
1481 .BR rsize " and " wsize
1482 settings were smaller than the system's page size.
1484 The Linux NFS client does not yet support
1485 certain optional features of the NFS version 4 protocol,
1486 such as security negotiation, server referrals, and named attributes.
1499 .BR rpc.svcgssd (8),
1502 RFC 768 for the UDP specification.
1504 RFC 793 for the TCP specification.
1506 RFC 1094 for the NFS version 2 specification.
1508 RFC 1813 for the NFS version 3 specification.
1510 RFC 1832 for the XDR specification.
1512 RFC 1833 for the RPC bind specification.
1514 RFC 2203 for the RPCSEC GSS API protocol specification.
1516 RFC 3530 for the NFS version 4 specification.