Skip to main content

NFS fstab options man page

NFS(5)   Linux Programmer’s Manual NFS(5)

NAME
       nfs - nfs and nfs4 fstab format and options

SYNOPSIS
       /etc/fstab

DESCRIPTION
       The  fstab  file contains information about which filesystems to mount
       where and with what options.  For NFS mounts, it contains  the server
       name  and  exported server directory to mount from, the local directory
       that is the mount point, and the NFS specific options that control  the
       way the filesystem is mounted.

       Three different versions of the NFS protocol are supported by the Linux
       NFS client: NFS version 2, NFS version 3, and NFS version 4.  To mount
       via  NFS version 2, use the nfs file system type and specify nfsvers=2.
       To mount via NFS version 3, use the nfs file system  type  and  specify
       nfsvers=3.   Version 3 is the default protocol version for the nfs file
       system type when nfsvers= is not specified on the  mount command.   To
       mount  via  NFS version 4, use the nfs4 file system type.  The nfsvers=
       keyword is not supported for the nfs4 file system type.

       These file system types share similar mount  options;  the  differences
       are listed below.

       Here is an example from an /etc/fstab file for an NFSv2 mount over UDP.

       server:/usr/local/pub /pub   nfs    rsize=32768,wsize=32768,timeo=14,intr

       Here is an example for an NFSv4 mount over TCP using Kerberos 5 mutual
       authentication.

       server:/usr/local/pub /pub   nfs4   proto=tcp,sec=krb5,hard,intr

   Options for the nfs file system type
       rsize=n      The  number of bytes NFS uses when reading files from an
     NFS server.  The rsize is negotiated between the server
     and client to determine the largest block size that both
     can support.  The value specified by this option is  the
     maximum  size  that  could  be used; however, the actual
     size used may be smaller. Note: Setting this size to  a
     value  less  than the largest supported block size will
     adversely affect performance.

       wsize=n      The number of bytes NFS uses when writing files to  an
     NFS  server.  The wsize is negotiated between the server
     and client to determine the largest block size that both
     can  support.  The value specified by this option is the
     maximum size that could be  used; however,  the actual
     size  used may be smaller.  Note: Setting this size to a
     value less than the largest supported  block  size  will
     adversely affect performance.

       timeo=n      The value in tenths of a second before sending the first
     retransmission after an RPC timeout.  The default value
     is  7  tenths of a second.  After the first timeout, the
     timeout is doubled after each successive timeout until a
     maximum  timeout of 60 seconds is reached or the enough
     retransmissions have occured to cause a  major  timeout.
     Then,  if the filesystem is hard mounted, each new time-
     out cascade restarts at twice the initial value  of  the
     previous cascade, again doubling at each retransmission.
     The maximum timeout is always 60 seconds. Better over-
     all  performance may be achieved by increasing the time-
     out when mounting on a busy network, to a slow  server,
     or through several routers or gateways.

       retrans=n      The  number  of  minor timeouts and retransmissions that
     must occur before a major timeout occurs.  The  default
     is  3  timeouts. When a major timeout occurs, the file
     operation is either aborted or a "server not responding"
     message is printed on the console.

       acregmin=n     The minimum time in seconds that attributes of a regular
     file should be cached before requesting  fresh  informa-
     tion from a server.  The default is 3 seconds.

       acregmax=n     The maximum time in seconds that attributes of a regular
     file can be cached before requesting  fresh  information
     from a server.  The default is 60 seconds.

       acdirmin=n     The  minimum time in seconds that attributes of a direc-
     tory should be cached before requesting  fresh  informa-
     tion from a server.  The default is 30 seconds.

       acdirmax=n     The  maximum time in seconds that attributes of a direc-
     tory can be cached before requesting  fresh  information
     from a server.  The default is 60 seconds.

       actimeo=n      Using  actimeo sets all of acregmin, acregmax, acdirmin,
     and acdirmax to the same value. There  is  no  default
     value.

       retry=n      The number of minutes to retry an NFS mount operation in
     the foreground or background  before  giving  up.   The
     default  value  for  forground mounts is 2 minutes.  The
     default value for background mounts  is  10000  minutes,
     which is roughly one week.

       namlen=n      When  an NFS server does not support version two of the
     RPC mount protocol, this option can be used  to  specify
     the  maximum  length  of a filename that is supported on
     the remote filesystem.  This  is used  to  support  the
     POSIX  pathconf  functions.   The default is 255 charac-
     ters.

       port=n      The numeric value of the port  to connect  to  the  NFS
     server  on.   If the port number is 0 (the default) then
     query the remote host’s portmapper for the  port number
     to  use. If the remote host’s NFS daemon is not regis-
     tered with its portmapper, the standard NFS port number
     2049 is used instead.


       proto=n      Mount the NFS filesystem using a specific network
     protocol instead of  the default TCP  protocol.
     Valid protocol types are udp and tcp.

       mountport=n    The  numeric  value  of  the  mountd port.  moun-
     thost=name The name of the host running mountd  .

       mountprog=n    Use  an  alternate  RPC program number to contact
     the mount daemon on the remote host.  This option
     is  useful  for  hosts  that can run multiple NFS
     servers. The default value is  100005  which  is
     the standard RPC mount daemon program number.

       mountvers=n    Use  an  alternate  RPC version number to contact
     the mount daemon on the remote host.  This option
     is  useful  for  hosts  that can run multiple NFS
     servers. The default value depends on which ker-
     nel you are using.

       nfsprog=n      Use  an  alternate  RPC program number to contact
     the NFS daemon on the remote host.   This option
     is  useful  for  hosts  that can run multiple NFS
     servers. The default value is  100003  which  is
     the standard RPC NFS daemon program number.

       nfsvers=n      Use  an  alternate  RPC version number to contact
     the NFS daemon on the remote host.   This option
     is  useful  for  hosts  that can run multiple NFS
     servers. The default value depends on which ker-
     nel you are using.

       vers=n      vers is an alternative to nfsvers and is compati-
     ble with many other operating systems.

       nolock      Disable NFS locking. Do not  start  lockd.   This
     has  to  be  used with some old NFS servers that
     don’t support locking.

       bg      If the first NFS mount attempt times  out,  retry
     the mount in the background.  After a mount oper-
     ation is backgrounded, all subsequent  mounts  on
     the  same NFS server will be backgrounded immedi-
     ately, without first  attempting the  mount.   A
     missing  mount  point is treated as a timeout, to
     allow for nested NFS mounts.

       fg      If the first NFS mount attempt times  out,  retry
     the mount in the foreground.  This is the comple-
     ment of the  bg  option, and  also  the default
     behavior.

       soft      If an NFS file operation has a major timeout then
     report an I/O error to the calling program.   The
     default  is  to continue retrying NFS file opera-
     tions indefinitely.

       hard      If an NFS file operation has a major timeout then
     report "server not responding" on the console and
     continue retrying  indefinitely.  This is  the
     default.

       intr      If  an NFS file operation has a major timeout and
     it  is  hard  mounted,  then  allow  signals   to
     interupt the  file  operation  and  cause  it to
     return EINTR to the calling program.  The default
     is  to  not  allow  file operations to be inter-
     rupted.

       posix      Mount the NFS filesystem using  POSIX  semantics.
     This allows an NFS filesystem to properly support
     the POSIX pathconf command by querying the  mount
     server  for the maximum length of a filename.  To
     do this, the remote host must support version two
     of the RPC mount protocol.  Many NFS servers sup-
     port only version one.

       nocto      Suppress the retrieval  of  new  attributes  when
     creating a file.

       noac      Disable  all forms of attribute caching entirely.
     This extracts a significant  performance penalty
     but  it  allows  two different NFS clients to get
     reasonable results when both clients are actively
     writing to a common export on the server.

       noacl      Disables Access Control List (ACL) processing.

       sec=mode      Set the security flavor for this mount to "mode".
     The default setting is sec=sys, which uses  local
     unix uids and gids to authenticate NFS operations
     (AUTH_SYS).  Other currently  supported  settings
     are:  sec=krb5, which uses Kerberos V5 instead of
     local unix uids and gids to  authenticate users;
     sec=krb5i,   which  uses Kerberos  V5  for  user
     authentication and performs integrity checking of
     NFS  operations using secure checksums to prevent
     data tampering; and sec=krb5p,  which  uses  Ker-
     beros  V5 for  user authentication and integrity
     checking, and encrypts  NFS  traffic  to prevent
     traffic  sniffing (this is the most secure set-
     ting).  Note that there is a performance penalty
     when using integrity or privacy.

       tcp      Mount  the NFS filesystem using the TCP protocol.
     This is the default protocol.

       udp      Mount the NFS filesystem using the  UDP  protocol
     instead of the default TCP protocol.

       nordirplus     Disables NFSv3 READDIRPLUS RPCs. Use this options
     when mounting servers that don’t support or  have
     broken READDIRPLUS implementations.

       nosharecache   As  of kernel 2.6.18, it is no longer possible to
     mount the same  same  filesystem with  different
     mount options to a new mountpoint.  It was deemed
     unsafe to do so, since  cached  data  cannot  be
     shared  between  the  two mountpoints. In conse-
     quence, files or directories that were common  to
     both  mountpoint subtrees could often be seen to
     be out of sync following an update.
     This option allows administrators to  select  the
     pre-2.6.18   behaviour, permitting   the   same
     filesystem to be mounted with  different  mount
     options.
     Beware:  Use  of this  option is not recommended
     unless you are certain that  there  are  no  hard
     links  or subtrees  of  this mountpoint that are
     mounted elsewhere.

       lookupcache=type
     This option dictates how directories  and  files
     should  be  cached when they are accessed -- i.e.
     "looked up" -- on the server.  A lookup can  be
     either  positive (directory/file was  found) or
     negative (directory/file was  not  found);  both
     types of lookups can be cached.

     By  default,  both  positive and negative lookups
     are cached ( lookupcache=all ). lookupcache=pos
     prevents negative  lookups  from being cached,
     while lookupcache=none prevents all lookups  from
     being cached.

     Note:  lookupcache=none can adversely affect per-
     formance, but may be necessary  if  shared  files
     created or deleted on the server need to be imme-
     diately visible to any  applications  running  on
     NFS clients.

       All  of the non-value options have corresponding nooption forms.
       For example, nointr means don’t allow  file  operations to  be
       interrupted.

   Options for the nfs4 file system type
       rsize=n      The  number of bytes nfs4 uses when reading files
     from the server. The rsize is negotiated between
     the  server  and client to determine the largest
     block size that  both  can  support.   The  value
     specified by this option is the maximum size that
     could be used; however, the actual size used  may
     be  smaller.   Note: Setting this size to a value
     less than the largest supported block  size  will
     adversely affect performance.

       wsize=n      The  number of bytes nfs4 uses when writing files
     to the server.  The wsize is  negotiated between
     the  server  and client to determine the largest
     block size that  both  can  support.   The  value
     specified by this option is the maximum size that
     could be used; however, the actual size used  may
     be  smaller.   Note: Setting this size to a value
     less than the largest supported block  size  will
     adversely affect performance.

       timeo=n      The  value  in  tenths of a second before sending
     the first retransmission after  an  RPC  timeout.
     The default value depends on whether proto=udp or
     proto=tcp is in effect (see below).  The default
     value  for  UDP  is  7  tenths  of a second.  The
     default value for TCP is 60 seconds.   After  the
     first  timeout, the timeout is doubled after each
     successive timeout until a maximum timeout of  60
     seconds  is reached or the enough retransmissions
     have occured to cause a major timeout.  Then,  if
     the  filesystem is hard mounted, each new timeout
     cascade restarts at twice the  initial  value  of
     the  previous  cascade,  again  doubling at each
     retransmission.  The maximum timeout is always 60
     seconds.

       retrans=n      The  number of minor timeouts and retransmissions
     that must occur before a major  timeout occurs.
     The  default  is 5  timeouts for proto=udp and 2
     timeouts for proto=tcp. When  a major timeout
     occurs, the file operation is either aborted or a
     "server not responding" message is printed on the
     console.

       acregmin=n     The  minimum time in seconds that attributes of a
     regular file should be cached  before  requesting
     fresh  information from a server. The default is
     3 seconds.

       acregmax=n     The maximum time in seconds that attributes of  a
     regular  file  can  be  cached  before requesting
     fresh information from a server. The default  is
     60 seconds.

       acdirmin=n     The  minimum time in seconds that attributes of a
     directory should be  cached  before  requesting
     fresh  information from a server. The default is
     30 seconds.

       acdirmax=n     The maximum time in seconds that attributes of  a
     directory can  be cached before requesting fresh
     information from a server.   The default is  60
     seconds.

       actimeo=n      Using  actimeo  sets  all of acregmin, acregmax,
     acdirmin, and acdirmax to the same value.  There
     is no default value.

       retry=n      The number of minutes to retry an NFS mount oper-
     ation in the foreground or background before giv-
     ing  up. The default value for forground mounts
     is 2 minutes.  The default value for  background
     mounts  is  10000 minutes,  which is roughly one
     week.

       port=n      The numeric value of the port to connect to  the
     NFS  server  on. If  the  port number is 0 (the
     default) then query the remote host’s  portmapper
     for the port number to use.  If the remote host’s
     NFS daemon is not registered with its portmapper,
     the   standard  NFS  port number 2049  is  used
     instead.

       proto=n      Mount the NFS filesystem using a specific network
     protocol instead of  the  default TCP protocol.
     Valid protocol types are udp and tcp.   Many  NFS
     version  4 servers only support the TCP protocol.

       clientaddr=n   On a multi-homed client, this causes  the client
     to  use a specific callback address when communi-
     cating with an NFS version 4 server.  This option
     is currently ignored.

       sec=mode      Same as sec=mode for the nfs filesystem type (see
     above).

       bg      If an NFS mount  attempt times  out,  retry  the
     mount in the background. After a mount operation
     is backgrounded, all  subsequent mounts on  the
     same NFS server will be backgrounded immediately,
     without first attempting the  mount.   A missing
     mount point is treated as a timeout, to allow for
     nested NFS mounts.

       fg      If the first NFS mount attempt times  out,  retry
     the mount in the foreground.  This is the comple-
     ment of the  bg  option, and  also  the default
     behavior.

       soft      If an NFS file operation has a major timeout then
     report an I/O error to the calling program.   The
     default  is  to continue retrying NFS file opera-
     tions indefinitely.

       hard      If an NFS file operation has a major timeout then
     report "server not responding" on the console and
     continue retrying  indefinitely.  This is  the
     default.

       intr      If  an NFS file operation has a major timeout and
     it  is  hard  mounted,  then  allow  signals   to
     interupt the  file  operation  and  cause  it to
     return EINTR to the calling program.  The default
     is  to  not  allow  file operations to be inter-
     rupted.

       nocto      Suppress the retrieval  of  new  attributes  when
     creating a file.

       noac      Disable  attribute caching, and force synchronous
     writes. This  extracts a  server   performance
     penalty  but  it allows two different NFS clients
     to get reasonable good results when both clients
     are  actively writing to common filesystem on the
     server.

       nosharecache   As of kernel 2.6.18, it is no longer possible  to
     mount  the  same same  filesystem with different
     mount options to a new mountpoint.  It was deemed
     unsafe  to  do  so,  since  cached data cannot be
     shared between the  two  mountpoints.  In conse-
     quence,  files or directories that were common to
     both mountpoint subtrees could often be  seen  to
     be out of sync following an update.
     This  option  allows administrators to select the
     pre-2.6.18   behaviour, permitting   the   same
     filesystem  to  be  mounted  with different mount
     options.
     Beware: Use of this  option  is  not  recommended
     unless  you  are certain that there are no hard
     links or subtrees of  this  mountpoint  that  are
     mounted elsewhere.

       fsc      Enable the use of persistent caching to the local
     disk using the FS-Cache facility for  the  given
     mount point.

       lookupcache=type
     This  option  dictates  how directories and files
     should be cached when they are accessed  --  i.e.
     "looked  up"  -- on  the server. A lookup can be
     either positive  (directory/file was  found)  or
     negative (directory/file was  not  found); both
     types of lookups can be cached.

     By default, both positive and  negative lookups
     are  cached ( lookupcache=all ). lookupcache=pos
     prevents negative  lookups  from being cached,
     while  lookupcache=none prevents all lookups from
     being cached.

     Note: lookupcache=none can adversely affect  per-
     formance, but  may  be necessary if shared files
     created or deleted on the server need to be imme-
     diately  visible to  any applications running on
     NFS clients.

       All of the non-value options have corresponding nooption forms.
       For  example,  nointr  means  don’t  allow file operations to be
       interrupted.

       All of the non-value options have corresponding nooption forms.
       For  example,  nointr  means  don’t  allow file operations to be
       interrupted.

FILES
       /etc/fstab

SEE ALSO
       fstab(5), mount(8), umount(8), exports(5)

AUTHOR
       "Rick Sladkey" <jrs@world.std.com>

BUGS
       The posix option is implemented but is currently ignored by  the
       Linux kernel.

       Checking files on NFS filesystem referenced by file descriptors
       (i.e. the fcntl and ioctl families of  functions)  may  lead  to
       inconsistent result due to the lack of consistency check in ker-
       nel even if noac is used.

Linux 0.99       20 November 1993 NFS(5)

Comments

Popular posts from this blog