This changeset removes webauthn from the Notify.gov admin app. We are not using webauthn at all in our implementation and will be looking at an entirely different authentication system in the near future.
Signed-off-by: Carlo Costino <carlo.costino@gsa.gov>
the js `fetch` function will follow redirects blindly and return you the
final 200 response. when there's an error, we don't want to go anywhere,
and we want to use the flask `flash` functionality to pop up an error
page (the likely reason for seeing this is using a yubikey that isn't
associated with your user). using `flash` and then
`window.location.reload()` handles this fine.
However, when the user does log in succesfully we need to properly log
them in - this includes:
* checking their account isn't over the max login count
* resetting failed login count to 0 if not
* setting a new session id in the database (so other browser windows are
logged out)
* checking if they need to revalidate their email access (every 90 days)
* clearing old user out of the cache
This code all happens in the ajax function rather than being in a
separate redirect, so that you can't just navigate to the login flow. I
wasn't able to unit test that function due how it uses the session and
other flask globals, so moved the auth into its own function so it's
easy to stub out all that CBOR nonsense.
TODO: We still need to pass any `next` URLs through the chain from login
page all the way through the javascript AJAX calls and redirects to the
log_in_user function
This naming was introduced in 2016 without explanation [1]. I find it
confusing because:
- It's reminiscent of "_app", which is a Python convention indicating
the variable is internal, so maybe avoid using it.
- It suggests there's some other "app" fixture I should be using (there
isn't, though).
The Python style guide describes using an underscore suffix to avoid
clashes with inbuilt names [1], which is sort of applicable if we need
to import the "app" module [2]. However, we can also avoid clashes by
choosing a different name, without the strange underscore.
[1]: 3b1d521c10
[2]: 78824f54fd/tests/app/main/views/test_forgot_password.py (L5)
This links up the `get_webauthn_credentials_for_user` and
`create_webauthn_credential_for_user` methods of the user api client to
notifications-api.
To send data to the API we need strings to be unicode, so we call
decode('utf-8') on base64 objects.
Co-authored-by: Leo Hemsted <leo.hemsted@digital.cabinet-office.gov.uk>
This passes existing credentials in the server response, to allow
the browser to prevent re-registering the same key for the same
user. Registering the same key multiple times doesn't seem to be
an issue technically; the user has likely got their keys mixed up.
- Chrome says "you don't need to register it again".
- Safari exits with an InvalidStateError.
- Firefox exits with a DOMException.
This adds Yubico's FIDO2 library and two APIs for working with the
"navigator.credentials.create()" function in JavaScript. The GET
API uses the library to generate options for the "create()" function,
and the POST API decodes and verifies the resulting credential. While
the options and response are dict-like, CBOR is necessary to encode
some of the byte-level values, which can't be represented in JSON.
Much of the code here is based on the Yubico library example [1][2].
Implementation notes:
- There are definitely better ways to alert the user about failure, but
window.alert() will do for the time being. Using location.reload() is
also a bit jarring if the page scrolls, but not a major issue.
- Ideally we would use window.fetch() to do AJAX calls, but we don't
have a polyfill for this, and we use $.ajax() elsewhere [3]. We need
to do a few weird tricks [6] to stop jQuery trashing the data.
- The FIDO2 server doesn't serve web requests; it's just a "server" in
the sense of WebAuthn terminology. It lives in its own module, since it
needs to be initialised with the app / config.
- $.ajax returns a promise-like object. Although we've used ".fail()"
elsewhere [3], I couldn't find a stub object that supports it, so I've
gone for ".catch()", and used a Promise stub object in tests.
- WebAuthn only works over HTTPS, but there's an exception for "localhost"
[4]. However, the library is a bit too strict [5], so we have to disable
origin verification to avoid needing HTTPS for dev work.
[1]: c42d9628a4/examples/server/server.py
[2]: c42d9628a4/examples/server/static/register.html
[3]: 91453d3639/app/assets/javascripts/updateContent.js (L33)
[4]: https://stackoverflow.com/questions/55971593/navigator-credentials-is-null-on-local-server
[5]: c42d9628a4/fido2/rpid.py (L69)
[6]: https://stackoverflow.com/questions/12394622/does-jquery-ajax-or-load-allow-for-responsetype-arraybuffer
This adds a new platform admin settings row, leading a page which
shows any existing keys and allows a new one to be registered. Until
the APIs for this are implemented, the user API client just returns
some stubbed data for manual testing.
This also includes a basic JavaScript module to do the main work of
registering a new authenticator, to be implemented in the next commits.
Some more minor notes:
- Setting the headings in the mapping_table is necessary to get the
horizontal rule along the top (to match the design).
- Setting caption to False in the mapping_table is necessary to stop
an extra margin appearing at the top.
We were using user fixtures in a lot of parameterized tests, but this is
no longer allowed in Pytest 5. To avoid having to split up the parametrized
tests (which would make the test files a lot longer and slightly more
difficult to read) this commit creates functions which return various types
of user json so that we can use these as the test parameters instead.
Before, the delete decorator would delete the keys from Redis and then
we made the request to api to change the data. However, it is possible
that the cache could get re-populated in between these two things
happening, and so would cache outdated data.
This changes the order to send the api request first. We then always
delete the specified keys from Redis. Changing the order of the code in
the decorator changes the order in which the cache keys get deleted, so
the tests have been updated.
Users can only be archived by Platform Admin from the user page
(/users/<user_id>). This removes them from all services and orgs and
updates their details.
The data flow of other bits of our application looks like this:
```
API (returns JSON)
⬇
API client (returns a built in type, usually `dict`)
⬇
Model (returns an instance, eg of type `Service`)
⬇
View (returns HTML)
```
The user API client was architected weirdly, in that it returned a model
directly, like this:
```
API (returns JSON)
⬇
API client (returns a model, of type `User`, `InvitedUser`, etc)
⬇
View (returns HTML)
```
This mixing of different layers of the application is bad because it
makes it hard to write model code that doesn’t have circular
dependencies. As our application gets more complicated we will be
relying more on models to manage this complexity, so we should make it
easy, not hard to write them.
It also means that most of our mocking was of the User model, not just
the underlying JSON. So it would have been easy to introduce subtle bugs
to the user model, because it wasn’t being comprehensively tested. A lot
of the changed lines of code in this commit mean changing the tests to
mock only the JSON, which means that the model layer gets implicitly
tested.
For those reasons this commit changes the user API client to return
JSON, not an instance of `User` or other models.
The api endpoint to get all template folders also returns the users who
can see each folder.
We need to clear the template-folder cache when adding a user to a service so
that we are not using out of date data about who can see each folder.
Added a folder permissions form to the page to invite users to services.
This only shows if the service has 'edit_folder_permissions' enabled,
and all folder checkboxes are checked by default. This change means that
InviteApiClient.create_invite now sends folder_permissions through to
notifications_api (so invites get created with folder permissions).
Started passing the folder_permissions through to notifications-api when
accepting an invite. This changes UserApiClient.add_user_to_service to
send folder_permissions to notifications_api so that new users get folder
permissions when they are added to the service.
The endpoint for adding a user to a service in api will now deal with
both user permissions and a user's folder permissions, so this changes
the format of the data we pass through.
The endpoint for setting permissions in api will now be used for both
user permissions and a user's folder permissions, so this changes the
format of the data we pass through.
when clients are defined in app/__init__.py, it increases the chance of
cyclical imports. By moving module level client singletons out to a
separate extensions file, we stop cyclical imports, but keep the same
code flow - the clients are still initialised in `create_app` in
`__init__.py`.
The redis client in particular is no longer separate - previously redis
was set up on the `NotifyAdminAPIClient` base class, but now there's one
singleton in `app.extensions`. This was done so that we can access redis
from outside of the existing clients.
This commit is the first step to disentangling the models from the API
clients. With the models in the same folder as the API clients it makes
it hard to import the API clients within the model without getting a
circular import.
After this commit the user API clients still has this problem, but at
least the service API client doesn’t.
Pytest is deprecating the direct calling of fixtures. One fixture that
we call directly quite a lot is `fake_uuid`. Since it just returns the
value of `sample_uuid()` we can either call that instead (where we need
a fixed value) or generate a new UUID each time (where a fixed value is
not needed).
we were seeing isort produce different outputs locally and in docker -
this was due to it having different opinions about whether the tests
module (ie all our unit tests) is a first party (local) or third party
(pip installed) import. It's a first party import, so by defining this
in the setup.cfg isort settings, we can force it to be consistent
between environments.
Note: I don't know why it was different in the first place though
Redis is giving us a big performance boost (it’s roughly halved the
median request time on the admin app).
Once we’re confident that it’s working properly[1] we can eke out a bit
more performance from it by keeping the caches alive for longer. As
far as I can tell we’re still using Redis in a very low-volume way[2],
so increasing the number of things we’re storing shouldn’t start taxing
our Redis server at all. But reducing the number of times we have to
hit the API to refresh the cache _should_ result in some performance
increase.
---
1. ie we’re not seeing instances of stale caches not being invalidated
2. We have 2.5G of available space in Redis. Here is our current usage:
```
used_memory:7728960
used_memory_human:7.37M
used_memory_rss:7728960
used_memory_peak:16563776
used_memory_peak_human:15.79M
used_memory_lua:37888
```
Accepting an invite changes:
- the `user_to_service` list of users returned by `GET /service/<id>`
- the `services` list return by `GET /user/<id>`
The latter change is causing the functional tests to fail.
In the same way, and for the same reasons that we’re caching the service
object.
Here’s a sample of the data returned by the API – so we should make sure
that any changes to this data invalidate the cache.
If we ever change a user’s phone number (for example) directly in the
database, then we will need to invalidate this cache manually.
```python
{
'data':{
'organisations':[
'4c707b81-4c6d-4d33-9376-17f0de6e0405'
],
'logged_in_at':'2018-04-10T11:41:03.781990Z',
'id':'2c45486e-177e-40b8-997d-5f4f81a461ca',
'email_address':'test@example.gov.uk',
'platform_admin':False,
'password_changed_at':'2018-01-01 10:10:10.100000',
'permissions':{
'42a9d4f2-1444-4e22-9133-52d9e406213f':[
'manage_api_keys',
'send_letters',
'manage_users',
'manage_templates',
'view_activity',
'send_texts',
'send_emails',
'manage_settings'
],
'a928eef8-0f25-41ca-b480-0447f29b2c20':[
'manage_users',
'manage_templates',
'manage_settings',
'send_texts',
'send_emails',
'send_letters',
'manage_api_keys',
'view_activity'
],
},
'state':'active',
'mobile_number':'07700900123',
'failed_login_count':0,
'name':'Example',
'services':[
'6078a8c0-52f5-4c4f-b724-d7d1ff2d3884',
'6afe3c1c-7fda-4d8d-aa8d-769c4bdf7803',
],
'current_session_id':'fea2ade1-db0a-4c90-93e7-c64a877ce83e',
'auth_type':'sms_auth'
}
}
```
in the db, we have several rows for single permissions - we separate
`send_messages` into `send_texts`, `send_emails` and `send_letters`,
and also `manage_service` into `manage_users` and `manage_settings`.
But on the front end we don't do anything with this distinction. It's
unhelpful for us to have to think about permissions as groups of things
when we can never split them up at all. So we should combine them. This
commit makes sure:
* when user models are read (from JSON direct from the API), we
should transform them from db permissions into roles.
* when permissions are persisted (editing permissions, and creating
invites), we should send db permissions to the API.
All other interaction with permissions (should just be the endpoint
decorator and checks in html templates generally) should use admin
roles.
Done using isort[1], with the following command:
```
isort -rc ./app ./tests
```
Adds linting to the `run_tests.sh` script to stop badly-sorted imports
getting re-introduced.
Chosen style is ‘Vertical Hanging Indent’ with trailing commas, because
I think it gives the cleanest diffs, eg:
```
from third_party import (
lib1,
lib2,
lib3,
lib4,
)
```
1. https://pypi.python.org/pypi/isort
One of the things we need to know for a service to go live is whether
they have at least two users with the ‘manage service’ permission.
So this commit adds a method to the client to count how many users have
a given permission. We can do logic on this count later. But having the
counting done in the client feels like a cleaner separation of concerns.
Meant some refactoring of the way `service_id` is extracted from the
request, in order to make it easier to mock.
