This module is a plugin for the Seneca framework. It provides a permissions system for actions. It's also a good example of how you can compose actions by layering plugins onto each other.
This plugin works by wrapping existing actions with a permission checking action. If the permission test passes, the parent action can proceed. If not, a permission error is generated.
The possible permission checks are:
- allow: simple yes/no
- act: allow only specific actions to pass
- entity: allow only specific actions on entities
- own: allow on specific actions on entities that are 'owned' by given users
This plugin also understands when it is used in a web server context, and will add a permission specification to the req.seneca object if it exists.
A full example, in the context of the seneca data editor, is provided in the seneca examples repository. TODO!
If you're using this module, feel free to contact me on twitter if you have any questions! :) @rjrodger
Current Version: 0.4.0
Tested on: Node 0.10.36, Seneca 0.6.1
You use this plugin mainly by adding a perm object to user entities:
var userent = seneca.make('sys','user')
userent.find({email:'[email protected]'}, function(err,alice){
alice.perm = {
act:[
{foo:'bar', perm$:true},
{qaz:'lol', perm$:false},
]
}
alice.save$()
})
This permission specification allows the user alice to execute foo:bar actions, but not qaz:lol ones. The perm$ metadata property specifies the permissions for each action pattern in the list.
Here's another example, this time for entities:
var userent = seneca.make('sys','user')
userent.find({email:'[email protected]'}, function(err,alice){
alice.perm = {
entity:[
{base:'shop', name:'cart', perm$:'r'},
{base:'qaz', perm$:'crudq'},
]
}
alice.save$()
})
In this case, alice can only read ('r') from -/shop/cart entities, but can do anything to -/qaz/- entities. There are more details on the seneca data entity model here.
Of course, this example code does not actually work, as all of the setup and configuration is missing. Here is some working code that provides a minimal example by setting up some of the values manually.
This code applies a permissions test to the echo plugin. The echo plugin has one action role:echo, that just gives you back the same arguments you put it. The example code adds a permission check to this action, and shows you how to trigger it manually, using the perm$ argument.
var seneca = require('seneca')()
seneca.use('echo')
seneca.use( 'perm', {act:[
{role:'echo'},
]})
seneca.act({role:'perm', cmd:'init'})
seneca.act({role:'echo', foo:'bar', perm$:{allow:true}},function(err,out){
console.log('foo='+out.foo)
})
seneca.act({role:'echo', foo:'bar', perm$:{allow:false}},function(err,out){
console.log(err)
})
The perm$ metadata argument is recognised by the perm plugin and contains the permission specification to test against. Normally, you won't set this manually, but generate it using the role:perm, cmd:makeperm action. In a web context, this plugin also does this for you, based on the perm property of the current user.
Here's an outline of how you would use it in a connect app, using the user and auth plugins to provide user accounts:
var seneca = require('seneca')()
var connect = require('connect')
seneca.use('user')
seneca.use('auth')
seneca.use('perm',{
act:[{role:'echo'}]
})
seneca.use('echo')
seneca.act({role:'perm', cmd:'init'})
var app = connect()
app.use( seneca.service() )
app.listen(3000)
The perm plugin does not wrap other actions immediately when it is registered. Rather, you call the role:perm, cmd:init action when you're ready. First you need to add any other plugins and actions that you want to apply permissions checking to.
npm install seneca
npm install seneca-perm
This plugin has two elements. First, the options define the set of actions that permission checks will apply to. Second, permission checks only occur if there is a perm$ metadata argument, containing a permissions specification.
- act: array of action pins (needed for both allow and act checks)
- entity: array of entity type specifications
- own: array of entity type specifications
These specify the actions to wrap. For example:
seneca.use( 'perm', {act:[
{role:'echo'}
]})
This wraps any action with a role:echo argument, which means that it will have a permission check applied.
You need to specify explicitly the actions to which you wish to apply permission checks.
As a convenience, you can apply permission checks to entities by simply specifying their zone, base and name (all optional):
seneca.use( 'perm', {entity:[
{base:'sys'},
{base:'foo',name:'bar'}
]})
The above code specifies that actions on any entity of type -/sys/- or -/foo/bar will have an permission check applied.
The entity option saves you from having to specify a permission check for all the entity actions, otherwise you would have to do this:
seneca.use( 'perm', {act:[
{role:'entity',cmd:'save',base:'sys'},
{role:'entity',cmd:'load',base:'sys'},
{role:'entity',cmd:'list',base:'sys'},
...
]})
The allowed entity operations (create, read, update, delete, query) are specified in the perm$ metadata argument to the entity actions (see below).
The own option works in much the same way as the entity option, except that the user must be the owner of the entity in question. Entities should have an owner property containing the identifier of the -/sys/user entity for this to work.
seneca.use( 'perm', {own:[
{base:'foo'}
]})
To trigger a permissions check, an action must contain a perm$ metadata argument. This is an object that contains one or more of the following properties:
- allow: boolean, true if action is permitted.
- act: an action router object; only matching actions can be executed
- entity: an entity type router object, matching the entity action, and specifying the operations permitted
- own: an entity type router object, matching the entity action, and specifying the operations permitted, if also owner of the entity
You do not normally construct the perm$ object directly, but instead use the role:perm, cmd:makeperm action to create one from a literal definition (you can store this in the -/sys/user entity, for example). If you store the definition in a perm property on -/sys/user, and use the perm plugin in a web context, then this is done for you automatically.
To store as a literal, use this structure in the perm property:
{allow:true|false}
This will converted to the perm$ metadata argument, by the role:perm, cmd:makeperm action:
{allow:true|false}
In general, this check is not particularly useful for individual users, serving rather to provide a global block on certain actions. You would do this by adding an allow property to any perm$ you generate.
To store as a literal, use this structure in the perm property:
{act:[
{ role:'...', cmd:'...', perm$:true },
{ role:'...', perm$:false },
{ foo:'bar', perm$:true },
...
]}
The act sub property is an array of action pins. Each pin specifies the argument properties to match against. The perm$ value indicates if the action is allowed or not.
This will converted to the perm$ metadata argument:
{act:router}
where router is a seneca router TODO! (the same thing that routes action arguments to plugin functions). The router matches a given set of action arguments to the permission specification.
You could construct it manually, like so:
var router = seneca.util.router()
router.add( {role:'...',cmd:'...'}, true )
router.add( {role:'...'}, false )
router.add( {foo:'bar'}, true )
Note that as with all routers, the action arguments are matched in alphabetical order.
To store as a literal, use this structure in the perm property:
{act:[
{ base:'sys', perm$:'' },
{ base:'foo', perm$:'rq' },
{ base:'foo': name:'bar', perm$:'crudq' },
...
]}
This specification is similar to the act specification above, except that the entity type is matched against. The permission value encodes the allowed operations. There are:
- create (r): create new entities
- read (r): load an entity by identifier
- update (r): modify an entity by identifier
- delete (r): remove an entity by identifier
- query (r): perform queries to list multiple entites
The query (q) permission also allows you to perform queries for the read, update and delete operations, if you have permission for those too.
In the example specification above, the user has no permissions on -/sys/- entities, can only read and query -/foo/- entities, and can perform any operations on -/foo/bar entities.
The perm$ metadata argument form is:
{entity:router}
where the router is constructed in the same way as the act permission, except using the entity zone, base and name. The data values in the router are the crudq operation specifications.
To store as a literal, use this structure in the perm property:
{act:[
{ base:'sys', perm$:'' },
{ base:'foo', perm$:'rq' },
{ base:'foo': name:'bar', perm$:'crudq' },
...
]}
This is the same as the entity permission. However, the permissions only apply if the entity has an owner property that matches the identifier of the user executing the action.
The perm plugin handles all the set up for you when used in a web context. See the test/perm.app.js example code.
The perm$ metadata argument form is:
{own:{
entity:router
owner:'...'
}}
Where the entity property is a router on the entity zone, base and name, has data values of the form 'crudq', and the owner is the identifier of the user.
An access control procedure runs a set of ACLs against a given pair of entity
and action
An ACL is composed of:
- a list of roles which are required for this ACL to authorize
- a set of actions (save, update, get, list)
- on a given entity (the type as well as specific attributes values)
- a control type (one of required|requisite|sufficient) that determine what happens should the ACL fail or succeed:
required
— The service result must be successful for authentication to continue. If the test fails at this point, the user is not notified until the results of all service tests that reference that interface are complete.requisite
— The service result must be successful for authentication to continue. However, if a test fails at this point, the user is notified immediately with a message reflecting the first failed required or requisite service test.sufficient
— The service result is ignored if it fails. However, if the result of a service flagged sufficient is successful and no previous services flagged required have failed, then no other results are required and the user is authenticated to the service.
IMPORTANT: The order in which
required
ACLs are called is not critical. Only thesufficient
andrequisite
control flags cause order to become important.
Examples:
si.use( '..', {
accessControls: [{
name: 'access to foobar entities',
roles: ['foobar'],
entities: [{
zone: undefined,
base: undefined,
name: 'foobar'
}],
control: 'required',
actions: 'crudq',
conditions: []
},{
name: 'access to foobar EMEA entities',
roles: ['EMEA'],
entities: [{
zone: undefined,
base: undefined,
name: 'foobar'
}],
control: 'required',
actions: 'crud',
conditions: [{
attributes: {
'region': 'EMEA'
}
}
]
},{
name: 'access to foobar EMEA entities',
roles: ['private_items'],
entities: [{
zone: undefined,
base: undefined,
name: 'item'
}],
control: 'required',
actions: 'r',
conditions: [{
attributes: {
'status': 'private'
}
}
]
}]
})
Out of the box, this plugin exports a web filter that looks at the logged-in user and runs the ACLs against the user's perm.roles
profile attribute:
user = {
email: '[email protected]',
perm: {
roles: ['foobar', 'private_items']
}
}
You can manually invoke the ACLs by setting the perm$
attribute in the arguments:
var publicAccess = si.delegate({perm$:{roles:[]}})
var pf1 = publicAccess.make('item',{number: 1, status: 'public'})
var privateAccess = si.delegate({perm$:{roles:['private_items']}})
var pf2 = privateAccess.make('item',{number: 2, status: 'private'})
In some cases, you want to run access controls against the current logged in user. For this, you can reference the current user in an ACL:
si.use( '..', {
accessControls: [{
name: 'todos: owner only',
roles: ['my_todos'],
entities: [{
zone: undefined,
base: undefined,
name: 'todo'
}],
control: 'required',
actions: 'crud',
conditions: [{
attributes: {
'owner': '{user.id}'
}
}
]
}]
})
The above will allow users to only create, read, update or delete 'todo' objects where they are the owner.
mocha test/*.test.js --seneca.log.print
If you need to trace what seneca-perm is doing you can active DEBUG before running your application:
DEBUG=seneca-perm:* node yourapp.js