p The use of .Xr tftp 1 does not require an account or password on the remote system. Due to the lack of authentication information, .Nm will allow only publicly readable files to be accessed. Filenames beginning in ``\|.\|.\|/'' or containing ``/\|.\|.\|/'' are not allowed. Files may be written to only if they already exist and are publicly writable.
p Note that this extends the concept of .Qq public to include all users on all hosts that can be reached through the network; this may not be appropriate on all systems, and its implications should be considered before enabling tftp service. The server should have the user ID with the lowest possible privilege.
p Access to files may be restricted by invoking .Nm with a list of directories by including up to 20 pathnames as server program arguments in
a /etc/inetd.conf . In this case access is restricted to files whose names are prefixed by the one of the given directories. The given directories are also treated as a search path for relative filename requests.
p The options are: l -tag -width "directory" t Fl d Enable verbose debugging messages to .Xr syslogd 8 . t Fl g Ar group Change gid to that of .Ar group on startup. If this isn't specified, the gid is set to that of the .Ar user specified with .Fl u . t Fl l Logs all requests using .Xr syslog 3 . t Fl n Suppresses negative acknowledgement of requests for nonexistent relative filenames. t Fl p Ar path separator All occurances of the single character .Ar path separator in the requested filename are replaced with .Sq / . t Fl s Ar directory .Nm will .Xr chroot 2 to .Ar directory on startup. This is recommended for security reasons (so that files other than those in the
a /tftpboot directory aren't accessible). If the remote host passes the directory name as part of the file name to transfer, you may have to create a symbolic link from .Sq tftpboot to .Sq . under
a /tftpboot . t Fl u Ar user Change uid to that of .Ar user on startup. If .Fl u isn't given, .Ar user defaults to .Dq nobody . If .Fl g isn't also given, change the gid to that of .Ar user as well. .El .Sh SEE ALSO .Xr tftp 1 , .Xr inetd 8 .Rs .%R RFC .%N 1350 .%D July 1992 .%T "The TFTP Protocol (Revision 2)" .Re .Rs .%R RFC .%N 2347 .%D May 1998 .%T "TFTP Option Extension" .Re .Rs .%R RFC .%N 2348 .%D May 1998 .%T "TFTP Blocksize Option" .Re .Rs .%R RFC .%N 2349 .%D May 1998 .%T "TFTP Timeout Interval and Transfer Size Options" .Re .Sh HISTORY The .Nm command appeared in x 4.2 .
p The .Fl s flag appeared in .Nx 1.0 .
p The .Fl g and .Fl u flags appeared in .Nx 1.4 .
p IPv6 support was implemented by WIDE/KAME project in 1999.
p TFTP options were implemented by Wasabi Systems, Inc., in 2003, and first appeared in .Nx 2.0 . .Sh BUGS Files larger than 33,553,919 octets (65535 blocks, last one <512 octets) cannot be correctly transferred without client and server supporting blocksize negotiation (RFCs 2347 and 2348). As a kludge, .Nm accepts a sequence of block numbers which wrap to zero after 65535.
p Many tftp clients will not transfer files over 16,776,703 octets (32767 blocks), as they incorrectly count the block number using a signed rather than unsigned 16-bit integer. .Sh SECURITY CONSIDERATIONS You are .Em strongly advised to set up .Nm using the .Fl s flag in conjunction with the name of the directory that contains the files that .Nm will serve to remote hosts (e.g.,
a /tftpboot ) . This ensures that only the files that should be served to remote hosts can be accessed by them.
p Because there is no user-login or validation within the .Tn TFTP protocol, the remote site will probably have some sort of file-access restrictions in place. The exact methods are specific to each site and therefore difficult to document here.