man(1) Manual page archive

     U9FS(4)                                                   U9FS(4)

          u9fs - serve 9P from Unix

          u9fs [ -Dnz ] [ -a authtype ] [ -A autharg ] [ -l logfile ]
          [ -m msize ] [ -u onlyuser ] fsroot

          U9fs is not a Plan 9 program.  Instead it is a program that
          serves Unix files to Plan 9 machines using the 9P protocol
          (see intro(5)). It is typically invoked on a Unix machine by
          inetd with its standard input and output connected to a net-
          work connection, typically TCP on an Ethernet.  It typically
          runs as user root and multiplexes access to multiple Plan 9
          clients over the single wire.  It assumes Plan 9 uids match
          Unix login names, and changes to the corresponding Unix
          effective uid when processing requests.  Characters in file
          and directory names unacceptable to Plan 9 are translated
          into a three-character sequence: `\' followed by two hex-
          adecimal digits.  U9fs serves both 9P1 (the 9P protocol as
          used by the second and third editions of Plan 9) and 9P2000.

          The options are:

          -D          Write very chatty debugging output to the log
                      file (see -l option below).

          -n          Signals that u9fs is not being invoked with a
                      network connection on standard input and output,
                      and thus should not try to determine the remote
                      address of the connection.  This is useful when
                      u9fs is not invoked from inetd (see examples

          -z          Truncate the log file on startup.  This is use-
                      ful mainly when debugging with -D.

          -a authtype Sets the authentication method to be used.
                      Authtype should be rhosts, none, or p9any.  The
                      default is rhosts, which uses the ruserok
                      library call to authenticate users by entries in
                      /etc/hosts.equiv or $HOME/.rhosts.  This default
                      is discouraged for all but the most controlled
                      networks.  Specifying none turns off authentica-
                      tion altogether.  This is useful when u9fs is
                      not invoked from inetd (see examples below, or
                      srvssh in srv(4)). Specifying p9any uses the
                      fourth edition Plan 9 authentication mechanisms.
                      The file /etc/u9fs.key, or autharg if specified

     U9FS(4)                                                   U9FS(4)

                      (see the -A option), is consulted for the
                      authentication data and should be suitably pro-
                      tected.  This file must contain exactly three
                      lines: secret (plaintext password), u9fs-user
                      (user id), and plan9-auth.dom (authentication

                      Finally, factotum must be taught a key of the

                      key proto=p9sk1 dom=plan9-auth.dom user=u9fs-user !password=secret

          -A autharg  Used to specify an argument to the authentica-
                      tion method.  See the authentication descrip-
                      tions above.

          -l logfile  Specifies the file which should contain debug-
                      ging output and other messages.  The out-of-
                      the-box compile-time default is /tmp/u9fs.log.

          -m msize    Set msize for 9P2000 (see open(5)).

          -u user     Treat all attaches as coming from user. This is
                      useful in some cases when running without inetd;
                      see the examples.

          If fsroot is specified, u9fs will serve only that tree; oth-
          wise, it will serve the entire Unix file system.

          Plan 9 calls 9P file service 9fs with TCP port number 564.
          Set up this way on a machine called, say, kremvax, u9fs may
          be connected to the name space of a Plan 9 process by

               9fs kremvax

          For more information on this procedure, see srv(4) and

          By default, u9fs serves the entire file system of the Unix
          machine.  It forbids access to devices because the program
          is single-threaded and may block unpredictably.  Using the
          attach specifier device connects to a file system identical
          to the usual system except it only permits device access
          (and may block unpredictably):

               srv tcp!kremvax!9fs
               mount -c /srv/tcp!kremvax!9fs /n/kremvax device

          (The 9fs command does not accept an attach specifier.)  Even
          so, device access may produce unpredictable results if the
          block size of the device is greater than 8192, the maximum

     U9FS(4)                                                   U9FS(4)

          data size of a 9P message.

          The source to u9fs is in the Plan 9 directory
          /sys/src/cmd/unix/u9fs.  To install u9fs on a Unix system
          with an ANSI C compiler, copy the source to a directory on
          that system and run make.  Then install the binary in
          /usr/etc/u9fs.  Add this line to inetd.conf:

               9fs     stream  tcp     nowait  root   /usr/etc/u9fs   u9fs

          and this to services:

               9fs     564/tcp       9fs  # Plan 9 fs

          Due to a bug in their IP software, some systems will not
          accept the service name 9fs, thinking it a service number
          because of the initial digit.  If so, run the service as
          u9fs or 564.

          On systems where listeners cannot be started, execnet(4) is
          useful for running u9fs via other network mechanisms; the
          script srvssh in srv(4) provides this for the ssh protocol.


          Problems are reported to the log file specified with the -l
          option (default /tmp/u9fs.log).  The -D flag enables chatty

          bind(1), execnet(4), srv(4), ip(3), nfsserver(8)

          The implementation of devices is unsatisfactory.

          Semantics like remove-on-close or the atomicity of wstat are
          hard to provide exactly.