This can happen for various reasons including rate limiting and middleware bugs. This should resolve the warning we're seeing in the logs
```
RequestTracker.get_data failed : NoMethodError : undefined method `[]' for nil:NilClass
```
Our 'page_view_crawler' / 'page_view_anon' metrics are based purely on the User Agent sent by clients. This means that 'badly behaved' bots which are imitating real user agents are counted towards 'anon' page views.
This commit introduces a new method of tracking visitors. When an initial HTML request is made, we assume it is a 'non-browser' request (i.e. a bot). Then, once the JS application has booted, we notify the server to count it as a 'browser' request. This reliance on a JavaScript-capable browser matches up more closely to dedicated analytics systems like Google Analytics.
Existing data collection and graphs are unchanged. Data collected via the new technique is available in a new 'experimental' report.
- Run the CSP-nonce-related middlewares on the generated response
- Fix the readonly mode checking to avoid empty strings being passed (the `check_readonly_mode` before_action will not execute in the case of these re-dispatched exceptions)
- Move the BlockRequestsMiddleware cookie-setting to the middleware, so that it is included even for unusual HTML responses like these exceptions
The strict-dynamic CSP directive is supported in all our target browsers, and makes for a much simpler configuration. Instead of allowlisting paths, we use a per-request nonce to authorize `<script>` tags, and then those scripts are allowed to load additional scripts (or add additional inline scripts) without restriction.
This becomes especially useful when admins want to add external scripts like Google Tag Manager, or advertising scripts, which then go on to load a ton of other scripts.
All script tags introduced via themes will automatically have the nonce attribute applied, so it should be zero-effort for theme developers. Plugins *may* need some changes if they are inserting their own script tags.
This commit introduces a strict-dynamic-based CSP behind an experimental `content_security_policy_strict_dynamic` site setting.
This will make it easier to analyze rate limiting in reverse-proxy logs. To make this possible without a database lookup, we add the username to the encrypted `_t` cookie data.
I took the wrong approach here, need to rethink.
* Revert "FIX: Use Guardian.basic_user instead of new (anon) (#24705)"
This reverts commit 9057272ee2.
* Revert "DEV: Remove unnecessary method_missing from GuardianUser (#24735)"
This reverts commit a5d4bf6dd2.
* Revert "DEV: Improve Guardian devex (#24706)"
This reverts commit 77b6a038ba.
* Revert "FIX: Introduce Guardian::BasicUser for oneboxing checks (#24681)"
This reverts commit de983796e1.
c.f. de983796e1
There will soon be additional login_required checks
for Guardian, and the intent of many checks by automated
systems is better fulfilled by using BasicUser, which
simulates a logged in TL0 forum user, rather than an
anon user.
In some cases the use of anon still makes sense (e.g.
anonymous_cache), and in that case the more explicit
`Guardian.anon_user` is used
In the past we would build the stack of Omniauth providers at boot, which meant that plugins had to register any authenticators in the root of their plugin.rb (i.e. not in an `after_initialize` block). This could be frustrating because many features are not available that early in boot (e.g. Zeitwerk autoloading).
Now that we build the omniauth strategy stack 'just in time', it is safe for plugins to register their auth methods in an `after_initialize` block. This commit relaxes the old restrictions so that plugin authors have the option to move things around.
Previously, we would build the stack of omniauth authenticators once on boot. That meant that all strategies had to be included, even if they were disabled. We then used the `before_request_phase` to ensure disabled strategies could not be used. This works well, but it means that omniauth is often doing unnecessary work running logic in disabled strategies.
This commit refactors things so that we build the stack of strategies on each request. That means we only need to include the enabled strategies in the stack - disabled strategies are totally ignored. Building the stack on-demand like this does add some overhead to auth requests, but on the majority of sites that will be significantly outweighed by the fact we're now skipping logic for disabled authenticators.
As well as the slight performance improvement, this new approach means that:
- Broken (i.e. exception-raising) strategies cannot cause issues on a site if they're disabled
- `other_phase` of disabled strategies will never appear in the backtrace of other authentication errors
Why this change?
This is a follow up to e8f7b62752.
Tracking of GC stats didn't really belong in the `MethodProfiler` class
so we want to extract that concern into its own class.
As part of this PR, the `track_gc_stat_per_request` site setting has
also been renamed to `instrument_gc_stat_per_request`.
Under some situations, we would inadvertently return a public (unauthenticated) result to an authenticated API request. This commit adds the `Api-Key` header to our anonymous cache bypass logic.
This patch introduces a cookies rotator as indicated in the Rails
upgrade guide. This allows to migrate from the old SHA1 digest to the
new SHA256 digest.
Adds stats for API and user API requests similar to regular page views.
This comes with a new report to visualize API requests per day like the
consolidated page views one.
This will allow consumers (e.g. the discourse-prometheus plugin) to separate topic-timings and message-bus requests. It also fixes the is_background boolean for subfolder sites.
Our discourse_public_exceptions middleware is designed to catch bubbled exceptions from lower in the stack, and then use `ApplicationController.rescue_with_handler` to render an appropriate error response.
When the request itself is invalid, we had an escape-hatch to skip re-dispatching the request to ApplicationController. However, it was possible to work around this by 'layering' the errors. For example, if you made a request which resulted in a 404, but **also** had some other invalidity, the escape hatch would not be triggered.
This commit ensures that these kind of 'layered' errors are properly handled, without logging warnings. It also adds detection for invalid JSON bodies and badly-formed multipart requests.
The user-facing behavior is unchanged. This commit simply prevents warnings being logged for invalid requests.
Currently, Discourse rate limits all incoming requests by the IP address they
originate from regardless of the user making the request. This can be
frustrating if there are multiple users using Discourse simultaneously while
sharing the same IP address (e.g. employees in an office).
This commit implements a new feature to make Discourse apply rate limits by
user id rather than IP address for users at or higher than the configured trust
level (1 is the default).
For example, let's say a Discourse instance is configured to allow 200 requests
per minute per IP address, and we have 10 users at trust level 4 using
Discourse simultaneously from the same IP address. Before this feature, the 10
users could only make a total of 200 requests per minute before they got rate
limited. But with the new feature, each user is allowed to make 200 requests
per minute because the rate limits are applied on user id rather than the IP
address.
The minimum trust level for applying user-id-based rate limits can be
configured by the `skip_per_ip_rate_limit_trust_level` global setting. The
default is 1, but it can be changed by either adding the
`DISCOURSE_SKIP_PER_IP_RATE_LIMIT_TRUST_LEVEL` environment variable with the
desired value to your `app.yml`, or changing the setting's value in the
`discourse.conf` file.
Requests made with API keys are still rate limited by IP address and the
relevant global settings that control API keys rate limits.
Before this commit, Discourse's auth cookie (`_t`) was simply a 32 characters
string that Discourse used to lookup the current user from the database and the
cookie contained no additional information about the user. However, we had to
change the cookie content in this commit so we could identify the user from the
cookie without making a database query before the rate limits logic and avoid
introducing a bottleneck on busy sites.
Besides the 32 characters auth token, the cookie now includes the user id,
trust level and the cookie's generation date, and we encrypt/sign the cookie to
prevent tampering.
Internal ticket number: t54739.
Allow admins to configure exceptions to our Rails rate limiter.
Configuration happens in the environment variables, and work with both
IPs and CIDR blocks.
Example:
```
env:
DISCOURSE_MAX_REQS_PER_IP_EXCEPTIONS: >-
14.15.16.32/27
216.148.1.2
```
Before this change, calling `StyleSheet::Manager.stylesheet_details`
for the first time resulted in multiple queries to the database. This is
because the code was modelled in a way where each `Theme` was loaded
from the database one at a time.
This PR restructures the code such that it allows us to load all the
theme records in a single query. It also allows us to eager load the
required associations upfront. In order to achieve this, I removed the
support of loading multiple themes per request. It was initially added
to support user selectable theme components but the feature was never
completed and abandoned because it wasn't a feature that we thought was
worth building.
Over the years we accrued many spelling mistakes in the code base.
This PR attempts to fix spelling mistakes and typos in all areas of the code that are extremely safe to change
- comments
- test descriptions
- other low risk areas
For sites with login_required set to true, counting anonymous pageviews is
confusing. Requests to /login and other pages would make it look like
anonymous users have access to site's content.
Fixes `Rack::Lint::LintError: a header value must be a String, but the value of 'Retry-After' is a Integer`. (see: 14a236b4f0/lib/rack/lint.rb (L676))
I found it when I got flooded by those warning a while back in a test-related accident 😉 (ember CLI tests were hitting a local rails server at a fast rate)
The regular expression to detect private IP addresses did not always detect them successfully.
Changed to use ruby's in-built IPAddr.new(ip_address).private? method instead
which does the same thing but covers all cases.
It returned a 429 error code with a 'Retry-After' header if a
RateLimiter::LimitExceeded was raised and unhandled, but the header was
missing if the request was limited in the 'RequestTracker' middleware.
`/srv/status` routes should not be cached at all. Also, we want to
decouple the route from Redis which `AnonymouseCache` relies on. The
`/srv/status` should continue to return a success response even if Redis
is down.
Previous to this change our anonymous rate limits acted as a throttle.
New implementation means we now also consider rate limited requests towards
the limit.
This means that if an anonymous user is hammering the server it will not be
able to get any requests through until it subsides with traffic.
This reverts commit e3de45359f.
We need to improve out strategy by adding a cache breaker with this change ... some assets on CDNs and clients may have incorrect CORS headers which can cause stuff to break.
David Taylor
Michael Brown
Jarek Radosz
Martin Brennan
Alan Guo Xiang Tan
OsamaSayegh
Daniel Waterworth
Loïc Guitaut
Penar Musaraj
Bianca Nenciu
Vinoth Kannan
Rafael dos Santos Silva
Dan Ungureanu
Josh Soref
Martin Brennan
Vinoth Kannan
Tobias Eigen
Sam