exportfs, srvfs – network file server plumbing


exportfs [ options ]

srvfs [ -dR ] [ -p perm ] [ -P patternfile ] [ -e exportprog ] name path


Exportfs is a user level file server that allows Plan 9 compute servers, rather than file servers, to export portions of a name space across networks. The service is started either by the cpu(1) command or by a network listener process. An initial protocol establishes a root directory for the exported name space. The connection to exportfs is then mounted, typically on /mnt/term. Exportfs then acts as a relay file server: operations in the imported file tree are executed on the remote server and the results returned. This gives the appearance of exporting a name space from a remote machine into a local file tree.

The options are:

-A address

Use the network address to announce aan(8) connections, if requested by the initial protocol.  


Authenticate the user with the p9any protocol before running the regular exportfs session; used when exportfs is invoked to handle an incoming network connection. Exportfs creates a new name space for each connection, using /lib/namespace by default (see namespace(6)).  

-B address

Dial address, authenticate as a p9any client, and then serve that network connection. Requires setting the root of the name space with -r or -s. The remote system should run import -B to handle the call. See import(4) for an example.  

-d -f dbgfile

Log all 9P traffic to dbgfile (default /tmp/exportdb).  

-e ’enc auth

Set the encryption and authentication algorithms to use for encrypting the wire traffic (see ssl(3)). The defaults are rc4_256 and sha1.  

-m msize

Set the maximum message size that exportfs should offer to send (see version(5)); this helps tunneled 9P connections to avoid unnecessary fragmentation.  

-N nsfile

Serve the name space described by nsfile.  


Disallow mounts by user none.  

-P patternfile

Restrict the set of exported files. Patternfile contains one regular expression per line, to be matched against path names relative to the current working directory and starting with /. For a file to be exported, all lines with a prefix + must match and all those with prefix - must not match.  


Make the served name space read only.  

-r root

Bypass the initial protocol, serving the name space rooted at root.  

-S service

bypass the initial protocol, serving the result of mounting service. A separate mount is used for each attach(5) message, to correctly handle servers in which each mount corresponds to a different client e.g.,( rio(4)).  


equivalent to -r /; kept for compatibility.  

The cpu command uses exportfs to serve device files in the terminal. The import(4) command calls exportfs on a remote machine, permitting users to access arbitrary pieces of name space on other systems.

Because the kernel disallows reads and writes on mounted pipes (as might be found in /srv), exportfs calls itself (with appropriate -m and -S options) to simulate reads and writes on such files.

Srvfs invokes exportprog (default /bin/exportfs) to create a mountable file system from a name space and posts it at /srv/name, which is created with mode perm (default 0600). The name space is the directory tree rooted at path. The -d, -P, and -R options, if present, are relayed to exportprog.


To export the archive of one user for one month, except for secrets,


cd /n/dump
echo '+ ^/(2003(/10..(/usr(/glenda/?)?)?)?)?' > /tmp/pattern
echo '- \.(aes|pgp)$' >> /tmp/pattern
exportfs -P /tmp/pattern

Use srvfs to enable mounting of an FTP file system (see ftpfs(4)) in several windows, or to publish a /proc (see proc(3)) with a broken process so a remote person may debug the program:


srvfs ftp /n/ftp
srvfs broke /mnt/term/proc

Use srvfs to obtain a copy of a service to be manipulated directly by a user program like nfsserver(8):


srvfs nfs.boot /srv/boot
aux/nfsserver -f /srv/nfs.boot

Use srvfs to spy on all accesses to a particular subtree:


srvfs -d spy /
tail -f /tmp/exportdb &
mount /srv/spy /n/spy
cd /n/spy; ls




dial(2), import(4), aan(8), listen(8)