feat: Implement querying openedx-authz for publish permissions

[rodmgwgu]

Description

Related bug: openedx/openedx-authz#139

Adds validating publish permissions based on the new openedx-authz model.

Hook implementation mostly copied from frontend-app-admin-console

Context:

The new openedx-authz authorization model adds the possibility to have fine-grained permissions for different roles in the system. Currently, the MVP involves implementing these new roles over the Content Libraries only.

The new roles are:

• Library Admin
• Library Author
• Library Contributor
• Library User

The specific permissions for these roles can be found here.

The enforcement mechanism for MFEs as implemented in this PR is described here.

This change mainly concerns the new "Library Contributor" role, which should have permission to edit library content, but not to publish it. This was not taken into account on the previous permission model, because before, anyone with write permissions could publish. This change makes it possible to disable the Publish button for "Library Contributors" by using the new enforcement mechanism.

Screenshots:

Before changes, as a "Library Contributor" user, the Publish button was shown, but the action failed due to lack of permissions:

After changes, the Publish button is hidden:

Supporting information

Project: RBAC AuthZ

Related bug: openedx/openedx-authz#139

The enforcement mechanism for MFEs as implemented in this PR is described here.

Testing instructions

1. Create a new non-staff user via tutor with tutor dev do createuser nonstaff nonstaff@example.com
2. Log in as an admin user to Studio
3. Create a new Content Library (non-legacy Library)
4. Add the "nonstaff" user to the Library team, as a "Library Contributor"

5. In an incognito window, log in to studio as the "nonstaff" user
6. Go to the direct url for the library details page
7. You should be able to see the Library details page, but the Publish button should not be shown

Other information

I implemented the validateUserPermissions hook and it's related code as an independent module in src/authz, with the idea of externalizing it as a library in the near future, perhaps as a part of frontend-base?

Because of that, I'm not using the helpers in src/testUtils.tsx on the hook tests.

Concern: I'm mocking the validateUserPermissions API call (used by the validateUserPermissions hook) on the initializeMocks util, so it doesn't interfere with several tests that are checking against axios mocks without validating the request url. I'm not sure if this is the best practice or if there is a better way?

Best Practices Checklist

We're trying to move away from some deprecated patterns in this codebase. Please
check if your PR meets these recommendations before asking for a review:

• Any new files are using TypeScript (.ts, .tsx).
• Avoid propTypes and defaultProps in any new or modified code.
• Tests should use the helpers in src/testUtils.tsx (specifically initializeMocks)
• Do not add new fields to the Redux state/store. Use React Context to share state among multiple components.
• Use React Query to load data from REST APIs. See any apiHooks.ts in this repo for examples.
• All new i18n messages in messages.ts files have a description for translators to use.
• Avoid using ../ in import paths. To import from parent folders, use @src, e.g. import { initializeMocks } from '@src/testUtils'; instead of from '../../../../testUtils'

[openedx-webhooks]

Thanks for the pull request, @rodmgwgu!

This repository is currently maintained by @bradenmacdonald.

Once you've gone through the following steps feel free to tag them in a comment and let them know that your changes are ready for engineering review.

🔘 Get product approval

If you haven't already, check this list to see if your contribution needs to go through the product review process.

• If it does, you'll need to submit a product proposal for your contribution, and have it reviewed by the Product Working Group.
  • This process (including the steps you'll need to take) is documented here.
• If it doesn't, simply proceed with the next step.

🔘 Provide context

To help your reviewers and other members of the community understand the purpose and larger context of your changes, feel free to add as much of the following information to the PR description as you can:

• Dependencies

  This PR must be merged before / after / at the same time as ...

• Blockers

  This PR is waiting for OEP-1234 to be accepted.

• Timeline information

  This PR must be merged by XX date because ...

• Partner information

  This is for a course on edx.org.

• Supporting documentation
• Relevant Open edX discussion forum threads

🔘 Get a green build

If one or more checks are failing, continue working on your changes until this is no longer the case and your build turns green.

---

Where can I find more information?

If you'd like to get more details on all aspects of the review process for open source pull requests (OSPRs), check out the following resources:

• Overview of Review Process for Community Contributions
• Pull Request Status Guide
• Making changes to your pull request

When can I expect my changes to be merged?

Our goal is to get community contributions seen and reviewed as efficiently as possible.

However, the amount of time that it takes to review and merge a PR can vary significantly based on factors such as:

• The size and impact of the changes that it introduces
• The need for product review
• Maintenance status of the parent repository

💡 As a result it may take up to several weeks or months to complete a review and merge your PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 94.87%. Comparing base (ef36156) to head (6d8852c).
⚠️ Report is 8 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2685      +/-   ##
==========================================
+ Coverage   94.85%   94.87%   +0.01%     
==========================================
  Files        1232     1237       +5     
  Lines       27899    27973      +74     
  Branches     6316     6155     -161     
==========================================
+ Hits        26464    26538      +74     
- Misses       1364     1377      +13     
+ Partials       71       58      -13     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[MaferMazu]

Thanks @rodmgwgu, I tested in a Ulmo env, and it works as expected! ✨

The code looks good to me, but I would prefer someone with more frontend skills to help with a review as well.

Before

After

[MaferMazu]

Something we could improve is being explicit about the action we are requesting instead of using userPermissions[0].

[rodmgwgu]

According to the ADR, the API guarantees that the order of the response will match the requested permissions, that's why I'm not trying to match it explicitly.

[MaferMazu]

I didn't know about the order 🤔

The suggestion was more about readability and clarity regarding the permission I am requesting, and I still think it is important. Since that index depends on the order of the elements in LIBRARY_PERMISSIONS, if that list grows, I don't think it will be clear enough to use only the indexes.

[bradenmacdonald]

+1. Also, hard-coding 0 here means that this code would become wrong if someone else changed the order of the LIBRARY_PERMISSIONS constant. Which could definitely happen, and then there would be a security hole.

[MaferMazu]

I am thinking about this issue: openedx/openedx-authz#144. I haven't refined it yet, but I would probably need to add more params to the request to see other permissions.

[bradenmacdonald]

I would suggest an API like this (no change to the REST API or the internal arrays, just implement some helper logic in the hook to support this):

const possiblePermissions = {
  canPublish: CONTENT_LIBRARY_PERMISSIONS.PUBLISH_LIBRARY_CONTENT,
];

const {
  isLoading: isLoadingUserPermissions,
  data: userPermissions,
} = useScopedUserPermissions(possiblePermissions, { scope: libraryId });

// API is useScopedUserPermissions(actions object, extra fields to mix in);

const canPublish = userPermissions?.canPublish;
// or
const canPublish = userPermissions?.canPublish.allowed; // (this is more verbose, and requiring these creates security bugs whenever users forget to include `.allowed`, but if you know there will likely be other fields besides .allowed in the future, it's better to be more verbose now)

[rodmgwgu]

Thanks for the suggestions, I changed it to the following format:

// Assign an arbitrary key to each permission request to retrieve results later:
const { isLoading, data } = useUserPermissions({
    canRead: {
        action: "content_libraries.view_library",
        scope: "lib:OpenedX:CSPROB"
    }
});

// Retrieve the boolean result using the specified key:
if (data?.canRead) { ... }

This way the permissions query is easier to read and understand, and using results is simpler and less error prone.

What do you think about this approach? @bradenmacdonald @MaferMazu

[MaferMazu]

Thanks @rodmgwgu, I think it's more verbose/secure in that way.

[bradenmacdonald]

I implemented the validateUserPermissions hook and it's related code as an independent module in src/authz, with the idea of externalizing it as a library in the near future, perhaps as a part of frontend-base?

Because of that, I'm not using the helpers in src/testUtils.tsx on the hook tests.

Sounds good. I think it would be great to have this in frontend-base.

Concern: I'm mocking the validateUserPermissions API call (used by the validateUserPermissions hook) on the initializeMocks util, so it doesn't interfere with several tests that are checking against axios mocks without validating the request url. I'm not sure if this is the best practice or if there is a better way?

Your approach seems fine, but we shouldn't have tests that break just because another axios call/mock was introduced. Were there many such cases? Usually axios mocks only mock one specific endpoint, and the asserts should be made on that same endpoint-specific mock.

[bradenmacdonald]

This is minor/optional feedback about the name of this hook:

To me, "validate user permissions" sounds like an action, like it would throw an exception if the user doesn't have some permissions. But this is just fetching some data, not making an action.

I think "useScopedPermissions" or just "useUserPermissions" or something like that would better reflect that this is just getting the user permissions, but you still have to validate/check that they're allowed or not yourself.

[rodmgwgu]

Thanks, changed it to useUserPermissions

[bradenmacdonald]

I would suggest an API like this (no change to the REST API or the internal arrays, just implement some helper logic in the hook to support this):

const possiblePermissions = {
  canPublish: CONTENT_LIBRARY_PERMISSIONS.PUBLISH_LIBRARY_CONTENT,
];

const {
  isLoading: isLoadingUserPermissions,
  data: userPermissions,
} = useScopedUserPermissions(possiblePermissions, { scope: libraryId });

// API is useScopedUserPermissions(actions object, extra fields to mix in);

const canPublish = userPermissions?.canPublish;
// or
const canPublish = userPermissions?.canPublish.allowed; // (this is more verbose, and requiring these creates security bugs whenever users forget to include `.allowed`, but if you know there will likely be other fields besides .allowed in the future, it's better to be more verbose now)

[rodmgwgu]

Concern: I'm mocking the validateUserPermissions API call (used by the validateUserPermissions hook) on the initializeMocks util, so it doesn't interfere with several tests that are checking against axios mocks without validating the request url. I'm not sure if this is the best practice or if there is a better way?

Your approach seems fine, but we shouldn't have tests that break just because another axios call/mock was introduced. Were there many such cases? Usually axios mocks only mock one specific endpoint, and the asserts should be made on that same endpoint-specific mock.

@bradenmacdonald About the tests that break, I'm seeing about 34 test files that potentially have the issue, basically they do calls like:

expect(axiosMock.history.post.length).toBe(1);
expect(axiosMock.history.post[0].data).toBe(...);

Now, because the useUserPermissions hook is calling an extra endpoint, axiosMock.history.post.length becomes 2, and axiosMock.history.post[0] may be different than what we are expecting.

I started fixing those tests by waiting for a specific history item with a specific url, but after seeing the amount of files affected I think that should be fixed separately.

An example here: https://github.com/WGU-Open-edX/frontend-app-authoring/blob/master/src/library-authoring/components/ComponentCard.test.tsx#L86

[bradenmacdonald]

I started fixing those tests by waiting for a specific history item with a specific url, but after seeing the amount of files affected I think that should be fixed separately.

makes sense, yep. Would you mind opening an issue about it with the examples you provided?

[rodmgwgu]

I started fixing those tests by waiting for a specific history item with a specific url, but after seeing the amount of files affected I think that should be fixed separately.

makes sense, yep. Would you mind opening an issue about it with the examples you provided?

Sure, here is the issue: #2723

Thanks!

[MaferMazu]

Thanks for addressing the comments @rodmgwgu.

@bradenmacdonald, could you please help review this again?

[bradenmacdonald]

@bradenmacdonald, could you please help review this again?

I am not able to test this. The instructions are perhaps missing some setup details? I have the latest master/main version of edx-platform, tutor, and tutor-mfe, but when my Django superuser tests this PR, the publish button is hidden. I checked the requests and it's saying I'm not authorized, even though I'm a superuser:

POST http://local.openedx.io:8000/api/authz/v1/permissions/validate/me 200
[
    {
        "action": "content_libraries.publish_library_content",
        "scope": "lib:BradenX:LC1",
        "allowed": false
    }
]

In addition, the admin console MFE is not running at http://localhost:2025/admin-console/ ("connection refused"). Shouldn't it be running by default with the latest version of tutor and tutor-mfe?

[rodmgwgu]

@bradenmacdonald, could you please help review this again?

I am not able to test this. The instructions are perhaps missing some setup details? I have the latest master/main version of edx-platform, tutor, and tutor-mfe, but when my Django superuser tests this PR, the publish button is hidden. I checked the requests and it's saying I'm not authorized, even though I'm a superuser:

POST http://local.openedx.io:8000/api/authz/v1/permissions/validate/me 200
[
    {
        "action": "content_libraries.publish_library_content",
        "scope": "lib:BradenX:LC1",
        "allowed": false
    }
]

This can happen if the authz policies are not loaded, or the library existed before and the migrations were not run.

The authz policies are loaded with ./manage.py lms load_policies, however I think those should be automatically run on init, do you know if that's the case @MaferMazu?

There is also a migration in the openedx-authz module that fills any preexisting permissions in the new system, however for admin users it should always be true anyways.

In addition, the admin console MFE is not running at http://localhost:2025/admin-console/ ("connection refused"). Shouldn't it be running by default with the latest version of tutor and tutor-mfe?

This should be running by default, at least in my local setup it is.

[bradenmacdonald]

This can happen if the authz policies are not loaded, or the library existed before and the migrations were not run.

The authz policies are loaded with ./manage.py lms load_policies, however I think those should be automatically run on init, do you know if that's the case @MaferMazu?

I ran tutor dev launch -I and ./manage.py lms load_policies but I'm still seeing the same issue.

In addition, the admin console MFE is not running at http://localhost:2025/admin-console/ ("connection refused"). Shouldn't it be running by default with the latest version of tutor and tutor-mfe?

This should be running by default, at least in my local setup it is.

OK, I figured it out - my version of tutor-mfe was not installed properly (not using pip install -e; it reverted to the pypi version 20 instead of main from git). So now I see the MFE is working but when I click on the "Library Team" button, the MFE just gives me a 403 error. I assume this is the same problem as above.

[rodmgwgu]

This can happen if the authz policies are not loaded, or the library existed before and the migrations were not run.
The authz policies are loaded with ./manage.py lms load_policies, however I think those should be automatically run on init, do you know if that's the case @MaferMazu?

I ran tutor dev launch -I and ./manage.py lms load_policies but I'm still seeing the same issue.

@bradenmacdonald I just tried clearing and rebuilding my local enviroment to try to reproduce the issue, but it's working correctly for me.

Here are the steps I followed:

1. Stop the containers with tutor dev stop
2. Clear the tutor folder (delete the folder in the fs)
3. Clear the containers and volumes from docker
4. Pull latest changes from tutor and re-install tutor main pip install -e './tutor[full]'
5. Pull latest changes from edx-platform and mount it
6. Pull this branch from frontend-app-authoring and mount it
7. Disable indigo tutor plugins disable indigo
8. tutor dev launch -I
9. tutor dev restart
10. Create an admin user with tutor tutor dev do createuser --staff --superuser admin admin@example.com
11. Login with that user to studio
12. Create a new Library
13. Go into the Library, the Publish button is visible and the API is responding as expected

[bradenmacdonald]

@rodmgwgu I suspect the problem is with existing devstacks, not new ones. I tried creating a new library and a new admin user, but I'm still seeing the same issue - no publish button and a 403 error when I try to access the "Library Team" page. I tried with both my existing superuser and the new admin user.

There is also a migration in the openedx-authz module that fills any preexisting permissions in the new system, however for admin users it should always be true anyways.

How can I check if the migration worked? Is there a page in the django admin that shows the permissions models?

This can happen if the authz policies are not loaded, or the library existed before and the migrations were not run.

The authz policies are loaded with ./manage.py lms load_policies, however I think those should be automatically run on init

Here's the output of running lms load_policies. It seems wrong that it mentions only the newly-created library, and only one user, when I have many libraries and users.

[ip None] enforcer.py:25 - Performing deferred casbin enforcer initialisation
model.py:112 - Model:
model.py:115 - r.r: sub, act, scope
model.py:115 - p.p: sub, act, scope, eft
model.py:115 - e.e: some(where (p_eft == allow)) && !some(where (p_eft == deny))
model.py:115 - m.m: (g(r_sub, p_sub, r_scope) || g(r_sub, p_sub, "*")) && keyMatch(r_scope, p_scope) && (r_act == p_act || g2(p_act, r_act))
model.py:115 - g.g: _, _, _
model.py:115 - g.g2: _, _
policy.py:73 - Policy:
policy.py:79 - p : sub, act, scope, eft : [['role^library_admin', 'act^view_library', 'lib^*', 'allow'], ['role^library_admin', 'act^manage_library_tags', 'lib^*', 'allow'], ['role^library_admin', 'act^delete_library', 'lib^*', 'allow'], ['role^library_admin', 'act^edit_library_content', 'lib^*', 'allow'], ['role^library_admin', 'act^publish_library_content', 'lib^*', 'allow'], ['role^library_admin', 'act^reuse_library_content', 'lib^*', 'allow'], ['role^library_admin', 'act^view_library_team', 'lib^*', 'allow'], ['role^library_admin', 'act^manage_library_team', 'lib^*', 'allow'], ['role^library_admin', 'act^create_library_collection', 'lib^*', 'allow'], ['role^library_admin', 'act^edit_library_collection', 'lib^*', 'allow'], ['role^library_admin', 'act^delete_library_collection', 'lib^*', 'allow'], ['role^library_author', 'act^view_library', 'lib^*', 'allow'], ['role^library_author', 'act^manage_library_tags', 'lib^*', 'allow'], ['role^library_author', 'act^edit_library_content', 'lib^*', 'allow'], ['role^library_author', 'act^publish_library_content', 'lib^*', 'allow'], ['role^library_author', 'act^reuse_library_content', 'lib^*', 'allow'], ['role^library_author', 'act^view_library_team', 'lib^*', 'allow'], ['role^library_author', 'act^create_library_collection', 'lib^*', 'allow'], ['role^library_author', 'act^edit_library_collection', 'lib^*', 'allow'], ['role^library_author', 'act^delete_library_collection', 'lib^*', 'allow'], ['role^library_contributor', 'act^view_library', 'lib^*', 'allow'], ['role^library_contributor', 'act^manage_library_tags', 'lib^*', 'allow'], ['role^library_contributor', 'act^edit_library_content', 'lib^*', 'allow'], ['role^library_contributor', 'act^reuse_library_content', 'lib^*', 'allow'], ['role^library_contributor', 'act^view_library_team', 'lib^*', 'allow'], ['role^library_contributor', 'act^create_library_collection', 'lib^*', 'allow'], ['role^library_contributor', 'act^edit_library_collection', 'lib^*', 'allow'], ['role^library_contributor', 'act^delete_library_collection', 'lib^*', 'allow'], ['role^library_user', 'act^view_library', 'lib^*', 'allow'], ['role^library_user', 'act^reuse_library_content', 'lib^*', 'allow'], ['role^library_user', 'act^view_library_team', 'lib^*', 'allow']]
policy.py:79 - g : _, _, _ : [['user^braden', 'role^library_admin', 'lib^lib:OpenCraftX:NEWTEST']]
policy.py:79 - g2 : _, _ : [['act^manage_library_tags', 'act^edit_library_content'], ['act^delete_library', 'act^edit_library_content'], ['act^publish_library_content', 'act^edit_library_content'], ['act^edit_library_content', 'act^view_library'], ['act^reuse_library_content', 'act^view_library'], ['act^publish_library_content', 'act^view_library'], ['act^manage_library_team', 'act^view_library_team'], ['act^delete_library_collection', 'act^edit_library_collection'], ['act^create_library_collection', 'act^edit_library_collection'], ['act^edit_library_collection', 'act^view_library']]
assertion.py:48 - Role links for: g
assertion.py:48 - Role links for: g2
role_manager.py:218 - act^manage_library_tags < act^edit_library_content, act^edit_library_content < act^view_library, act^delete_library < act^edit_library_content, act^publish_library_content < (act^edit_library_content, act^view_library), act^reuse_library_content < act^view_library, act^manage_library_team < act^view_library_team, act^delete_library_collection < act^edit_library_collection, act^edit_library_collection < act^view_library, act^create_library_collection < act^edit_library_collection
utils.py:30 - Loaded 31 policies from source enforcer.
utils.py:34 - Target enforcer has 31 existing policies before migration.
utils.py:44 - Policy ['role^library_admin', 'act^view_library', 'lib^*', 'allow'] already exists in target, skipping.
... many similar lines ("already exists") ...
utils.py:44 - Policy ['role^library_user', 'act^view_library_team', 'lib^*', 'allow'] already exists in target, skipping.
utils.py:57 - Grouping policy g2, ['act^manage_library_tags', 'act^edit_library_content'] already exists in target, skipping.
... many similar lines ("already exists") ...
utils.py:57 - Grouping policy g2, ['act^edit_library_collection', 'act^view_library'] already exists in target, skipping.
utils.py:67 - Skipping g3 policies: 'g3' not found in source enforcer.
utils.py:67 - Skipping g4 policies: 'g4' not found in source enforcer.
utils.py:67 - Skipping g5 policies: 'g5' not found in source enforcer.
utils.py:67 - Skipping g6 policies: 'g6' not found in source enforcer.
utils.py:68 - Successfully loaded policies from <casbin.model.model.Model object at 0xffff6ab81c10> into the database.

[MaferMazu]

The authz policies are loaded with ./manage.py lms load_policies, however, I think those should be automatically run on init, do you know if that's the case @MaferMazu?

Yes, it should work automatically when you init the env.

How can I check if the migration worked? Is there a page in the Django admin that shows the permissions models?

You could check the Casbin rules in the Django admin. Example: https://ulmo.demo.edly.io/admin/casbin_adapter/casbinrule/ (credentials: https://docs.google.com/spreadsheets/d/1-gEXI4IiAn5mdROGDMxNcFHAaZCRu3mxVLNDmTdJFI8/edit?gid=1326754537#gid=1326754537)

In that you can verify all the policies are loaded (41 policies at least).

To check specific permissions, you could use: https://ulmo.demo.edly.io/api-docs/#/authz/authz_v1_permissions_validate_me_create (using your local domain), and for example:

[
  {
    "action": "content_libraries.view_library_team",
    "scope": "lib:BradenX:LC1"
  }
]

And regarding this, the staff and superusers will always return true.

[MaferMazu]

@bradenmacdonald, based on your logs, the policy load is working. But I notice they're using an old version (e.g., the permission doesn't have the content_libraries prefix). So I think the problem you are having is that you had installed an old version of openedx-authz. Can you install the latest version that is in edx-platform (0.20.0)?

It will be:
pip install openedx-authz==0.20.0
And then you would need to run the load policy
./manage.py lms load_policies --clear-existing
That will remove the current policies and role assignments that use the new authz system and update the policies.
Note: The load_policies doesn't migrate old roles to new ones; it only sets the rules we use to evaluate permissions.
Then:
./manage.py lms migrate openedx_authz
That last one should transform your current roles into the new ones.