mirror of
https://gitlab.com/fabinfra/fabaccess/bffh.git
synced 2024-12-22 11:43:49 +01:00
250 lines
12 KiB
Plaintext
250 lines
12 KiB
Plaintext
{- Main configuration file for bffh
|
|
- ================================
|
|
-
|
|
- In this configuration file you configure almost all parts of how bffh operates, but most importantly:
|
|
- * Machines
|
|
- * Initiators and Actors
|
|
- * Which Initiators and Actors relate to which machine(s)
|
|
- * Roles and the permissions granted by them
|
|
-}
|
|
|
|
-- The config is in the configuration format/language dhall. You can find more information about dhall over at
|
|
-- https://dhall-lang.org
|
|
|
|
-- (Our) Dhall is somewhat similar to JSON and YAML in that it expects a top-level object containing the
|
|
-- configuration values
|
|
{
|
|
-- Configure the addresses and ports bffh listens on
|
|
listens = [
|
|
-- BFFH binds a port for every listen object in this array.
|
|
-- Each listen object is of the format { address = <STRING>, port = <INTEGER> }
|
|
-- If you don't specify a port bffh will use the default of `59661`
|
|
-- 'address' can be a IP address or a hostname
|
|
-- If bffh can not bind a port for the specified combination if will log an error but *continue with the remaining ports*
|
|
{ address = "127.0.0.1", port = 59661 },
|
|
{ address = "::1", port = 59661 },
|
|
{ address = "steak.fritz.box", port = 59661 }
|
|
],
|
|
|
|
-- Configure TLS. BFFH requires a PEM-encoded certificate and the associated key as two separate files
|
|
certfile = "examples/self-signed-cert.pem",
|
|
keyfile = "examples/self-signed-key.pem",
|
|
|
|
-- BFFH right now requires a running MQTT broker.
|
|
mqtt_url = "tcp://localhost:1883",
|
|
|
|
-- Path to the database file for bffh. bffh will in fact create two files; ${db_path} and ${db_path}.lock.
|
|
-- BFFH will *not* create any directories so ensure that the directory exists and the user running bffh has write
|
|
-- access into them.
|
|
db_path = "/tmp/bffh",
|
|
|
|
-- Audit log path. Bffh will log state changes into this file, one per line.
|
|
-- Audit log entries are for now JSON:
|
|
-- {"timestamp":1641497361,"machine":"Testmachine","state":{"state":{"InUse":{"uid":"Testuser","subuid":null,"realm":null}}}}
|
|
auditlog_path = "/tmp/bffh.audit",
|
|
|
|
-- In dhall you can also easily import definitions from other files, e.g. you could write
|
|
-- roles = ./roles.dhall
|
|
roles = {
|
|
-- Role definitions
|
|
-- A role definition is of the form
|
|
-- rolename = {
|
|
-- parents = [<list of role names to inherit from>],
|
|
-- permissions = [<list of perm rules>],
|
|
-- }
|
|
--
|
|
-- Role names are case sensitive, so RoleName != rolename.
|
|
--
|
|
-- If you want either parents or permissions to be empty its best to completely skip it:
|
|
testrole = {
|
|
permissions = [ "lab.some.admin" ]
|
|
},
|
|
somerole = {
|
|
parents = ["testparent"],
|
|
-- "Permissions" are formatted as Perm Rules, so you can use the wildcards '*' and '+'
|
|
permissions = [ "lab.test.*" ]
|
|
},
|
|
-- Roles can inherit from each other. In that case a member of e.g. 'somerole' that inherits from
|
|
-- 'testparent' will have all the permissions of 'somerole' AND 'testparent' assigned to them.
|
|
-- Right now permissions are stricly additive so you can't take a permission away in a child role that a parent
|
|
-- role grants.
|
|
testparent = {
|
|
permissions = [
|
|
"lab.some.write",
|
|
"lab.some.read",
|
|
"lab.some.disclose"
|
|
]
|
|
}
|
|
},
|
|
|
|
-- Configure machines
|
|
-- "Machines" (which in future will be more appropiately named "resources") are the main thing bffh is concerned
|
|
-- with.
|
|
-- You can define an almost limitless amount of machines (well 2^64 - 1, so 18_446_744_073_709_551_615 to be precise)
|
|
-- Each of these machines can then have several "actors" and "initiators" assigned
|
|
machines = {
|
|
Testmachine = {
|
|
-- A machine comes with two "names". The id above ("Testmachine") and the "name" ("MachineA").
|
|
-- The id is what you'll use in the config format and is strictly limited to alphanumeric characters and '_'
|
|
-- and must begin with a letter. Most importantly you CAN NOT use '-' or spaces in an identifier
|
|
-- (dhall makes this technically possible but you can break things in subtle ways)
|
|
|
|
-- REQUIRED. The "name" of a machine is what will be presented to humans. It can contain all unicode
|
|
-- including spaces and nonprintable characters.
|
|
-- A name SHOULD be short but unique.
|
|
name = "MachineA",
|
|
|
|
-- OPTIONAL. A description can be assigned to machines. It will also only be shown to humans. Thus it is
|
|
-- once again limited only to unicode. If you want to provide your users with important additional
|
|
-- information other than the name this is the place to do it.
|
|
description = "A test machine",
|
|
|
|
-- OPTIONAL. If you have a wiki going into more detail how to use a certain machine or what to keep in
|
|
-- mind when using it you can provide a URL here that will be presented to users.
|
|
wiki = "https://wiki.example.org/machineA",
|
|
|
|
-- OPTIONAL. You can assign categories to machines to allow clients to group/filter machines by them.
|
|
category = "Testcategory",
|
|
|
|
-- REQUIRED.
|
|
-- Each machine MUST have *all* Permission levels assigned to it.
|
|
-- Permissions aren't PermRules as used in the 'roles' definitions but must be precise without wildcards.
|
|
-- Permission levels aren't additive, so a user having 'manage' permission does not automatically get
|
|
-- 'read' or 'write' permission.
|
|
|
|
-- (Note, disclose is not fully implemented at the moment)
|
|
-- Users lacking 'disclose' will not be informed about this machine in any way and it will be hidden from
|
|
-- them in the client. Usually the best idea is to assign 'read' and 'disclose' to the same permission.
|
|
disclose = "lab.test.read",
|
|
|
|
-- Users lacking 'read' will be shown a machine including name, description, category and wiki but not
|
|
-- it's current state. The current user is not disclosed.
|
|
read = "lab.test.read",
|
|
|
|
-- The 'write' permission allows to 'use' the machine.
|
|
write = "lab.test.write",
|
|
|
|
-- Manage represents the 'superuser' permission. Users with this permission can force set any state and
|
|
-- read out the current user
|
|
manage = "lab.test.admin"
|
|
},
|
|
Another = {
|
|
wiki = "test_another",
|
|
category = "test",
|
|
disclose = "lab.test.read",
|
|
manage = "lab.test.admin",
|
|
name = "Another",
|
|
read = "lab.test.read",
|
|
write = "lab.test.write"
|
|
},
|
|
Yetmore = {
|
|
description = "Yet more test machines",
|
|
disclose = "lab.test.read",
|
|
manage = "lab.test.admin",
|
|
name = "Yetmore",
|
|
read = "lab.test.read",
|
|
write = "lab.test.write"
|
|
}
|
|
},
|
|
|
|
-- Actor configuration. Actors are how bffh affects change in the real world by e.g. switching a power socket
|
|
-- using a shelly
|
|
actors = {
|
|
-- Actors similarly to machines have an 'id'. This id (here "Shelly1234") is limited to Alphanumeric ASCII
|
|
-- and must begin with a letter.
|
|
Shelly1234 = {
|
|
-- Actors are modular pieces of code that are loaded as required. The "Shelly" module will send
|
|
-- activation signals to a shelly switched power socket over MQTT
|
|
module = "Shelly",
|
|
-- Actors can have arbitrary parameters passed to them, varying by actor module.
|
|
params = {
|
|
-- For Shelly you can configure the MQTT topic segment it uses. Shellies listen to a specific topic
|
|
-- containing their name (which is usually of the form "shelly_<id>" but can be changed).
|
|
-- If you do not configure a topic here the actor will use it's 'id' (in this case "Shelly1234").
|
|
topic = "Topic1234"
|
|
}
|
|
},
|
|
|
|
Bash = {
|
|
-- The "Process" module runs a given script or command on state change.
|
|
-- bffh invoces the given cmd as `$ ${cmd} ${args} ${id} ${state}` so e.g. as
|
|
-- `$ ./examples/actor.sh your ad could be here Bash inuse`
|
|
module = "Process",
|
|
params = {
|
|
-- which is configured by the (required) 'cmd' parameter. Paths are relative to PWD of bffh. Systemd
|
|
-- and similar process managers may change this PWD so it's usually the most future-proof to use
|
|
-- absolute paths.
|
|
cmd = "./examples/actor.sh",
|
|
-- You can pass static args in here, these will be passed to every invocation of the command by this actor.
|
|
-- args passed here are split by whitespace, so these here will be passed as 5 separate arguments
|
|
args = "your ad could be here"
|
|
}
|
|
},
|
|
|
|
DoorControl1 = {
|
|
-- This actor calls the actor.py script in examples/
|
|
-- It gets passed it's own name, so you can have several actors
|
|
-- from the same script.
|
|
-- If you need to pass more arguments to the command you can use the `args` key in
|
|
-- `params` as is done with the actor `Bash`
|
|
module = "Process",
|
|
-- the `args` are passed in front of all other parameters so they are best suited to
|
|
-- optional parameters like e.g. the verboseness
|
|
params = { cmd = "./examples/actor.py", args = "-vvv" }
|
|
},
|
|
DoorControl2 = {
|
|
module = "Process",
|
|
params = { cmd = "./examples/actor.py" }
|
|
},
|
|
DoorControl3 = {
|
|
-- This is an example for how it looks like if an actor is misconfigured.
|
|
-- the actor.py doesn't know anything about DoorControl3 and, if this actor is enabled,
|
|
-- will return with an error showing up in the server logs.
|
|
module = "Process",
|
|
params = { cmd = "./examples/actor.py" }
|
|
},
|
|
|
|
Bash2 = { module = "Process", params = { cmd = "./examples/actor.sh" , args = "this is a different one" }},
|
|
FailBash = { module = "Process", params = { cmd = "./examples/fail-actor.sh" }}
|
|
},
|
|
|
|
-- Linkng up machines to actors
|
|
-- Actors need to be connected to machines to be useful. A machine can be connected to multiple actors, but one
|
|
-- actor can only be connected to one machine.
|
|
actor_connections = [
|
|
{ machine = "Testmachine", actor = "Shelly1234" },
|
|
{ machine = "Another", actor = "Bash" },
|
|
{ machine = "Yetmore", actor = "Bash2" },
|
|
{ machine = "Yetmore", actor = "FailBash"}
|
|
],
|
|
|
|
-- Initiators are configured almost the same way as Actors, refer to actor documentation for more details
|
|
-- The below '{=}' is what you need if you want to define *no* initiators at all and only use the API with apps
|
|
-- to let people use machines.
|
|
initiators = {=},
|
|
-- The "Dummy" initiator will try to use and return a machine as the given user every few seconds. It's good to
|
|
-- test your system but will spam your log so is disabled by default.
|
|
--initiators = { Initiator = { module = "Dummy", params = { uid = "Testuser" } } },
|
|
|
|
-- Linking up machines to initiators. Similar to actors a machine can have several initiators assigned but an
|
|
-- initiator can only be assigned to one machine.
|
|
-- The below is once again how you have to define *no* initiators.
|
|
init_connections = [] : List { machine : Text, initiator : Text },
|
|
--init_connections = [{ machine = "Testmachine", initiator = "Initiator" }]
|
|
|
|
instanceurl = "https://example.com",
|
|
spacename = "examplespace",
|
|
|
|
oidc = Some {
|
|
-- the configuration endpoint is sent to the client to discover
|
|
-- authentication and token exchange endpoints
|
|
configuration_endpoint = "https://keycloak.thejonny.de/realms/fabaccess/.well-known/openid-configuration",
|
|
-- the userinfo endpoint is queried to check an access token and to discover information about the user.
|
|
userinfo_endpoint = "https://keycloak.thejonny.de/realms/fabaccess/protocol/openid-connect/userinfo",
|
|
-- identifying attribute. defaults to "sub"
|
|
userid_attribute = "username",
|
|
-- TODO: autodiscover the endpoints from the issuer.
|
|
-- issuer = "https://keycloak.thejonny.de/realms/fabaccess/"
|
|
}
|
|
}
|