While in theory users should be unable to get a valid session in order
to access these endpoints in the first place, this provides an extra
layer of hardering to prevent a deactivated user with a session from
accessing data via the old-style JSON API.
The security model for deactivated users (and users in deactivated
realms) being unable to access the service is intended to work via two
mechanisms:
* All active user sessions are deleted, and all login code paths
(where a user could get a new session) check whether the user (or
realm) is inactive before authorizing the request, preventing the
user from accessing the website and AJAX endpoints.
* All API code paths (which don't require a session) check whether the
user (and realm) are active.
However, this security model was not implemented correctly. In
particular, the check for whether a user has an active account in the
login process was done inside the login form's validators, which meant
that authentication mechanisms that did not use the login form
(e.g. Google and REMOTE_USER auth) could succeed in granting a session
even with an inactive account. The Zulip homepage would still fail to
load because the code for / includes an API call to Tornado authorized
by the user's token that would fail, but this mechanism could allow an
inactive user to access realm data or users to access data in a
deactivated realm.
This fixes the issue by adding explicit checks for inactive users and
inactive realms in all authentication backends (even those that were
already protected by the login form validator).
Mirror dummy users are already inactive, so we can remove the explicit
code around mirror dummy users.
The following commits add a complete set of tests for Zulip's inactive
user and realm security model.
In a deactivated realm, webhooks would still successfully send
messages, since there was no check for whether the realm was active in
api_key_only_webhook_view.
Long ago, there was work on an experimental integration model where
every user in a realm would have administrative control over all bots,
with the goal of simplifying the process of setting up communally
administered bots for smaller teams. While that new model was never
fully implemented (and thus never setup as an option), an error in
that original implementation meant that the data on all bots in a
realm, including their API keys, was sent to the browsers of users via
the `realm_bots` variable in `page_params`. The data wasn't displayed
in the UI for non-admin users, but was available via e.g. the
javascript console.
This commit updates this behavior to only send sensitive bot data like
API keys to the owner of the bot (and realm admins).
We may in the future implement a model simplifying communally
administered integrations, but if we do that, those bots should be
limited in their capabilities (e.g. only able to send webhook
messages).
This bug has been present since Zulip was released as open source.
The old code for this lookup was unnecessarily complicated because we
were working around Guardian, where the `is_realm_admin` check was
extremely expensive.
Previously we relied on having two matching list of fields for the
get_active_user_dicts_in_realm, one in the actual code and the other
in the caching system. By unifying these lists to have a single
source, we eliminate a class of caching bugs we might otherwise
regularly introduce.
This results in a substantial performance improvement for all of
Zulip's backend templates.
Changes in templates:
- Change `block.super` to `super()`.
- Remove `load` tag because Jinja2 doesn't support it.
- Use `minified_js()|safe` instead of `{% minified_js %}`.
- Use `compressed_css()|safe` instead of `{% compressed_css %}`.
- `forloop.first` -> `loop.first`.
- Use `{{ csrf_input }}` instead of `{% csrf_token %}`.
- Use `{# ... #}` instead of `{% comment %}`.
- Use `url()` instead of `{% url %}`.
- Use `_()` instead of `{% trans %}` because in Jinja `trans` is a block tag.
- Use `{% trans %}` instead of `{% blocktrans %}`.
- Use `{% raw %}` instead of `{% verbatim %}`.
Changes in tools:
- Check for `trans` block in `check-templates` instead of `blocktrans`
Changes in backend:
- Create custom `render_to_response` function which takes `request` objects
instead of `RequestContext` object. There are two reasons to do this:
1. `RequestContext` is not compatible with Jinja2
2. `RequestContext` in `render_to_response` is deprecated.
- Add Jinja2 related support files in zproject/jinja2 directory. It
includes a custom backend and a template renderer, compressors for js
and css and Jinja2 environment handler.
- Enable `slugify` and `pluralize` filters in Jinja2 environment.
Fixes#620.
To avoid the potential for introducing regressions here, we carefully
pass a default to REQ or not based on how the existing webhook's
parsing code worked. In the longer term, we'll want to make the
behavior consistent.
This fixes an exception where client_id was never set in an error code
path. It shouldn't be needed, but I think this makes the code clearer
and this will help in debugging the actual problem.
Related to #753.
The error message for a test file that doesn't import properly was
previously pretty difficult to understand and it wasn't clear how to
debug the issue.
This commit adds the capability to keep track and remove uploaded
files. Unclaimed attachments are files that have been uploaded to the
server but are not referred in any messages. A management command to
remove old unclaimed files after a week is also included.
Tests for getting the file referred in messages are also included.
Since we don't have a stable way to get the Dropbox preview failure
image (and it was sorta a weird setup anyway), it seems best to just
remove the condition.
Previously we needed to use a specified password when activating a
formerly mirror dummy user, in order for that user to be able to
(re)set their password and login. Now that we have our own password
reset form, this is no longer required.
Previously, if a user had only authenticated via Google auth, they
would be unable to reset their password in order to set one (which is
needed to setup the mobile apps, for example).
Several recently merged webhooks were incorrectly not checking that
the actual webhook result didn't return an error. While they would
usually still fail in most cases when checking whether the message
came back correctly, this hid the root cause errors and thus made it
much harder to debug.
We were incorrectly applying the rate limiting rules to webhooks even
if rate limiting was disabled (as in the test suite), causing test
failures when the total number of webhook tests in Zulip got too high.
This integration relies on the Teamcity "tcWebHooks" plugin which is
available at
https://netwolfuk.wordpress.com/category/teamcity/tcplugins/tcwebhooks/
It posts build fail and success notifications to a stream specified in
the webhook URL.
It uses the name of the build configuration as the topic.
For personal builds, it tries to map the Teamcity username to a Zulip
username, and sends a private message to that person.
As documented in https://github.com/zulip/zulip/issues/441, Guardian
has quite poor performance, and in fact almost 50% of the time spent
running the Zulip backend test suite on my laptop was inside Guardian.
As part of this migration, we also clean up the old API_SUPER_USERS
variable used to mark EMAIL_GATEWAY_BOT as an API super user; now that
permission is managed entirely via the database.
When rebasing past this commit, developers will need to do a
`manage.py migrate` in order to apply the migration changes before the
server will run again.
We can't yet remove Guardian from INSTALLED_APPS, requirements.txt,
etc. in this release, because otherwise the reverse migration won't
work.
Fixes#441.
When uploaded avatar image is not a valid image file, PIL raises
IOError. Catch the IOError raised by PIL and raise JsonableError.
This will return a response with status code 400.
S3Test is now only the S3-specific test (which isn't even run), so we
can now invest in making FileUploadTest have good coverage of the
(local) file upload code paths.
Previously, the UserProfile objects were created in the order
generated by a Set, which meant tests would randomly start failing if
the code that runs before this part of populate_db changed (and thus
caused the Set object used to pass users into bulk_create_users to
have a different order when enumerated).
This fixes the issue in two ways -- one by sorting the users inside
bulk_create_users, and second by attaching subscriptions to users
based on a deterministic ordering.
The restarted Tornado processes seemed to escape the process group and
thus continue running after run-dev.py finished.
While we're at it, we don't need to dump/reload event queues in the
test suite either.
The previous version of sanitize_name dropped all unicode characters
and mangled filenames with multiple `.`s in the extension, leading to
confusing URLs for files uploaded to Zulip.
Fixes#321.
[tweaked significantly by tabbott]
It's always been the case that in production, Tornado dumps all the
event queues when shut down so that they can be reloaded by the
replacement Tornado process. This never worked in development because
the codepath for auto-reload didn't go through either a signal or
sys.exit (it re-execs the process instead).
This meant that we didn't have a mechanism for testing the event queue
dump/load functionality in the development environment. We fix this
by adding such dumping/loading. However, this breaks the automatic
reloading of open browser windows on a server restart, so we add that
back in by adjusting the special `restart` events to pass a special
`immediate` flag when used in development.
This also has the benefit of removing the "Bad event queue" errors one
would get on every file save induced restart on the Python console.
This is a no-op right now, but we'll want the new structure for the
next commit, and splitting this out makes it a lot easier to read what
is actually changed in the next commit.
This change drops the memory used for Python processes run by Zulip in
development from about 1GB to 300MB on my laptop.
On the front of safety, http://pika.readthedocs.org/en/latest/faq.html
explains "Pika does not have any notion of threading in the code. If
you want to use Pika with threading, make sure you have a Pika
connection per thread, created in that thread. It is not safe to share
one Pika connection across threads.". Since this code only connects
to rabbitmq inside the individual threads, I believe this should be
safe.
Progress towards #32.
Apparently, our event queue garbage collection logic never actually
disconnected any existing handler objects.
We fix this by disconnecting the handler inside cleanup(), adding a
special check to avoid creating a pointless timeout object.
The new Tornado handler tracking logic properly handled requests that
threw an exception or followed the RespondAsynchronously code path,
but did not properly de-allocated the handler in the syncronous case.
An easy reproducer for this is to load a new Zulip browser window;
that will leak 2 handler objects for the 2 synchronous requests made
from Django to Tornado as part of initial state fetching.
This line appears to have been lost in rebasing from the original
implementation of 1396eb7022faec4c2d91553800a35781a96dd5bd; so the
previous fix actually only addressed the issue in a rare exception
case.
The recent Tornado memory leak fix
(1396eb7022) didn't use the correct
variable name for the current handler ID, causing this cleanup code to
fail in the event that a view raised an exception.
Replaced calls to ifilterfalse by list comprehensions because
ifilterfalse is not part of python 3. Also changed some lists to sets
for faster lookup.
Refer to #256.
In 2ea0daab19, handlers were moved to
being tracked via the handlers_by_id dict, but nothing cleared this
dict, resulting in every handler object being leaked. Since a Tornado
process uses a different handler object for every request, this
resulted in a significant memory leak. We fix this by clearing the
handlers_by_id dict in the two code paths that would result in a
Tornado handler being de-allocated: the exception codepath and the
handler disconnect codepath.
Fixes#463.
Add a function email_allowed_for_realm that checks whether a user with
given email is allowed to join a given realm (either because the email
has the right domain, or because the realm is open), and use it
whenever deciding whether to allow adding a user to a realm.
This commit is not intended to change any behavior, except in one case
where the Zulip realm's domain was not being converted to lowercase.
Tim Abbott
Umair Khan
Preston Hansen
Tomasz Kolek
rahuldeve
Ashish Kumar
Antek Grzanka
Vladislav Manchev
David Payne
Eklavya Sharma
Anindya Chakravarti
Varshit
Ryan Moore
Tim Abbott
Reid Barton