All files accessible to netplan are stored in the /var/lib/plan/netplan.dir directory. netplan will not access any files that are not in this directory or in subdirectories of this directory. It will also refuse to access softlinks and files with multiple hard links. This prevents users from linking normally inaccessible files to netplan.dir and accessing them through netplan . Finally, files beginning with a dot are rejected to prevent access to netplan-acl and other future configuration files.
/etc/plan/ may also contain a file netplan-acl that controls which user can access which file. If the file is missing, no restrictions are imposed unless netplan is started with the -a option, in which case no access to any file is granted. The syntax for netplan-acl file is a sequence of rules like this:
name | owner | * : [permit | deny] [read] [write] [delete] [netmask n.n.n.n] [[user | group | host] data [data...]]
name is the file the rule applies to; an asterisk (*) applies to all files. The special name owner applies to the file by the name of current user, containing that user's ``own'' appointments.
Permit is the default. If none of read, write, or delete are specified, all three are the default. The netmask applies to the client's IP address. The default netmask is 255.255.255.255. data is one or more user names or numerical UIDs, group names or numerical GIDs, or host names or numerical n.n.n.n IP addresses for user, group, and host clauses, respectively. In user clauses, values of the form user@host are also accepted. Symbolic names will be looked up on the server host (where netplan is running) and will be converted to numerical values while the ACL file is read. This assumes that all hosts agree on symbolic and numeric user, group, and host IDs; the plan/netplan protocol always uses numerical IDs and assumes that they correspond to the same symbolic names on both hosts. If no user, group, or host keyword and data list is specified, the meaning is ``any''.
Trailing n=0 IP address components are not assumed to denote nets; use the netmask specifier for subnet masking. All whitespace is ignored.
Pound signs (#) introduce comments that extend to the end of the line.
*: permit read *: permit write host daphne vacation: permit write user 207 thomas: deny user * thomas: permit read write delete user 207 208 announce: permit write netmask 255.255.255.0 host 10.0.1.0
first permits reading all files to everybody, and restricts write access to users on host daphne. The third line permits write access to the file vacation to user 207, in addition to the read access permitted in lines 1-2. Next, read and write access for file thomas is granted to users 207 and 208 only. Finally, the file announce can be written by all users on hosts whose network address begins with 10.0.1. Trailing ".0" parts need not be entered. The netmask basically specifies which bits of the client address are compared; all addresses are binary-ANDed with the netmask before comparison.
When opening a file, the rules are scanned sequentially. All rules whose file part (before the colon) matches the opened file, set or clear permission flags for reading and writing, based on the identity of the plan client as given by his user ID, group IDs, and IP address. The final settings of these flags determine the permissions of the file for the client. The final permission setting is the AND result of the permissions derived for the host/netmask, and user/group part, respectively.
netplan tries to verify the identity of the client user with the IDENTD (RFC 1413) protocol. If the identification succeeds, the client username is mapped to UID and GIDs per the local passwd and group files on the server host. If RFC 1413 identification is unsuccessful, netplan trusts the (numerical) identity provided by the client.
If the -a option is given on the invocation of netplan, RFC 1413 identification becomes mandatory, and a failed identification will map the client to the NOBODY UID and GID.
Note that although the ACL syntax was roughly patterned after TIS fwtk firewall configuration files, the code and interpretation is rather different.
netplan trusts IP addresses to determine host (network) access restrictions. If IP addresses cannot be trusted, host access restrictions become meaningless.
Without RFC 1413 authetication, netplan trusts whatever user id and group id the client provides. If netplan is used through the regular plan front-end application, the access list file specifications are honored, but any half-witted programmer can fake his identity using telnet or a hacked version of plan (the sources are, after all, freely available) to circumvent the access restrictions.
If RFC 1413 authentication is mandatory (-a flag), netplan still trusts whatever the remote identd provides. If you cannot trust root on the remote host, you cannot trust the identd result. (And if you cannot trust IP addresses, you effectively cannot trust the remote root any more.)
In this version of netplan, no challenge-response encryption is used to guarantee secure transactions. This may or may not change in future versions. In this version, access lists provide only a moderate protection.
The location for /etc/plan/netplan-acl is specific to Debian GNU/Linux. For compliance with FSSTND/FHS, it has been moved there from its traditional /var/lib/plan/netplan.dir/.netplan-acl location. The program still accesses that file via a symlink located at the traditional location.
Please send all complaints, comments, bug fixes, and porting experiences to me. Always include your plan version as reported by "plan -v" in your mail. To be added to the mailing list, send mail to email@example.com with the line "subscribe plan" (without the quotes) in the message body (not the subject).
See http://www.bitrot.de/plan.html for new releases.