When there are many synchronous requests waiting on a
`_PerHostRatelimiter`, each request will be started recursively just
after the previous request has completed. Under the right conditions,
this leads to stack exhaustion.
A common way for requests to become synchronous is when the remote
client disconnects early, because the homeserver is overloaded and slow
to respond.
Avoid stack exhaustion under these conditions by deferring subsequent
requests until the next reactor tick.
Fixes#14480.
Signed-off-by: Sean Quah <seanq@matrix.org>
PKCE can protect against certain attacks and is enabled by default. Support
can be controlled manually by setting the pkce_method of each oidc_providers
entry to 'auto' (default), 'always', or 'never'.
This is required by Twitter OAuth 2.0 support.
Then adapts calling code to retry when needed so it doesn't 500
to clients.
Signed-off-by: Mathieu Velten <mathieuv@matrix.org>
Co-authored-by: Sean Quah <8349537+squahtx@users.noreply.github.com>
An empty cache does not mean the entity has no changed, if
it is earlier than the earliest known stream position return that
the entity *has* changed since the cache cannot accurately
answer that query.
The internal methods of the StreamChangeCache were inconsistently
treating the earliest known stream position as valid. It is now treated as
invalid, meaning the cache cannot determine if an entity at the earliest
known stream position has changed or not.
Avoid an n+1 query problem and fetch the bundled aggregations for
m.annotation relations in a single query instead of a query per event.
This applies similar logic for as was previously done for edits in
8b309adb43 (#11660) and threads
in b65acead42 (#11752).
Remove type hints from comments which have been added
as Python type hints. This helps avoid drift between comments
and reality, as well as removing redundant information.
Also adds some missing type hints which were simple to fill in.
This code was factored out to a method, but also left in-place.
Calling this twice in a row makes no sense: the first call will reduce
the size appropriately, but the loop will immediately exit since the
cache size was already reduced.
1. `federation_client.timestamp_to_event(...)` now handles all `destination` looping and uses our generic `_try_destination_list(...)` helper.
2. Consistently handling `NotRetryingDestination` and `FederationDeniedError` across `get_pdu` , backfill, and the generic `_try_destination_list` which is used for many places we use this pattern.
3. `get_pdu(...)` now returns `PulledPduInfo` so we know which `destination` we ended up pulling the PDU from
Fix https://github.com/matrix-org/synapse/issues/13856
Fix https://github.com/matrix-org/synapse/issues/13865
> Discovered while trying to make Synapse fast enough for [this MSC2716 test for importing many batches](https://github.com/matrix-org/complement/pull/214#discussion_r741678240). As an example, disabling the `have_seen_event` cache saves 10 seconds for each `/messages` request in that MSC2716 Complement test because we're not making as many federation requests for `/state` (speeding up `have_seen_event` itself is related to https://github.com/matrix-org/synapse/issues/13625)
>
> But this will also make `/messages` faster in general so we can include it in the [faster `/messages` milestone](https://github.com/matrix-org/synapse/milestone/11).
>
> *-- https://github.com/matrix-org/synapse/issues/13856*
### The problem
`_invalidate_caches_for_event` doesn't run in monolith mode which means we never even tried to clear the `have_seen_event` and other caches. And even in worker mode, it only runs on the workers, not the master (AFAICT).
Additionally there was bug with the key being wrong so `_invalidate_caches_for_event` never invalidates the `have_seen_event` cache even when it does run.
Because we were using the `@cachedList` wrong, it was putting items in the cache under keys like `((room_id, event_id),)` with a `set` in a `set` (ex. `(('!TnCIJPKzdQdUlIyXdQ:test', '$Iu0eqEBN7qcyF1S9B3oNB3I91v2o5YOgRNPwi_78s-k'),)`) and we we're trying to invalidate with just `(room_id, event_id)` which did nothing.
This speeds things up by ~2x.
The vast majority of the time is now spent in `LruCache` moving things around the linked lists.
We do this via two things:
1. Don't create a deferred per-key during bulk set operations in `DeferredCache`. Instead, only create them if a subsequent caller asks for the key.
2. Add a bulk lookup API to `DeferredCache` rather than use a loop.
Still maintains local in memory lookup optimisation, but does any external
lookup as part of the deferred that prevents duplicate lookups for the same
event at once. This makes the assumption that fetching from an external
cache is a non-zero load operation.
Some experimental prep work to enable external event caching based on #9379 & #12955. Doesn't actually move the cache at all, just lays the groundwork for async implemented caches.
Signed off by Nick @ Beeper (@Fizzadar)
* Cast to postgres types when handling postgres db
* Remove unused method
* Easy annotations
* Annotate create_room
* Use `ParamSpec` to annotate looping_call
* Annotate `default_config`
* Track `now` as a float
`time_ms` returns an int like the proper Synapse `Clock`
* Introduce a `Timer` dataclass
* Introduce a Looper type
* Suppress checking of a mock
* tests.utils is typed
* Changelog
* Whoops, import ParamSpec from typing_extensions
* ditch the psycopg2 casts
This simplifies the access token verification logic by removing the `rights`
parameter which was only ever used for the unsubscribe link in email
notifications. The latter has been moved under the `/_synapse` namespace,
since it is not a standard API.
This also makes the email verification link more secure, by embedding the
app_id and pushkey in the macaroon and verifying it. This prevents the user
from tampering the query parameters of that unsubscribe link.
Macaroon generation is refactored:
- Centralised all macaroon generation and verification logic to the
`MacaroonGenerator`
- Moved to `synapse.utils`
- Changed the constructor to require only a `Clock`, hostname, and a secret key
(instead of a full `Homeserver`).
- Added tests for all methods.
By always returning all requested values from the function
wrapped by cachedList. Otherwise implicit None values get
added into the cache, which are unexpected.
Over time we've begun to use newer versions of mypy, typeshed, stub
packages---and of course we've improved our own annotations. This makes
some type ignore comments no longer necessary. I have removed them.
There was one exception: a module that imports `select.epoll`. The
ignore is redundant on Linux, but I've kept it ignored for those of us
who work on the source tree using not-Linux. (#11771)
I'm more interested in the config line which enforces this. I want
unused ignores to be reported, because I think it's useful feedback when
annotating to know when you've fixed a problem you had to previously
ignore.
* Installing extras before typechecking
Lacking an easy way to install all extras generically, let's bite the bullet and
make install the hand-maintained `all` extra before typechecking.
Now that https://github.com/matrix-org/backend-meta/pull/6 is merged to
the release/v1 branch.
This will mainly be useful when dealing with module callbacks, which are
all typed as returning `Awaitable`s instead of coroutines or
`Deferred`s.
Signed-off-by: Sean Quah <seanq@element.io>
Just after a task acquires a contended `Linearizer` lock, it sleeps.
If the task is cancelled during this sleep, we need to release the lock.
Signed-off-by: Sean Quah <seanq@element.io>
Refactor and convert `Linearizer` to async. This makes a `Linearizer`
cancellation bug easier to fix.
Also refactor to use an async context manager, which eliminates an
unlikely footgun where code that doesn't immediately use the context
manager could forget to release the lock.
Signed-off-by: Sean Quah <seanq@element.io>
These decorators mostly support cancellation already. Add cancellation
tests and fix use of finished logging contexts by delaying cancellation,
as suggested by @erikjohnston.
Signed-off-by: Sean Quah <seanq@element.io>
`delay_cancellation` behaves like `stop_cancellation`, except it
delays `CancelledError`s until the original `Deferred` resolves.
This is handy for unifying cleanup paths and ensuring that uncancelled
coroutines don't use finished logcontexts.
Signed-off-by: Sean Quah <seanq@element.io>
* `@cached` can now take an `uncached_args` which is an iterable of names to not use in the cache key.
* Requires `@cached`, @cachedList` and `@lru_cache` to use keyword arguments for clarity.
* Asserts that keyword-only arguments in cached functions are not accepted. (I tested this briefly and I don't believe this works properly.)
* Fix incorrect argument in test case
* Add copyright header
* Docstring and __all__
* Exclude dev depenencies
* Use changelog from #12088
* Include version in error messages
This will hopefully distinguish between the version of the source code
and the version of the distribution package that is installed.
* Linter script is your friend
* Pull runtime dep checks into their own module
* Reimplement `check_requirements` using `importlib`
I've tried to make this clearer. We start by working out which of
Synapse's requirements we need to be installed here and now. I was
surprised that there wasn't an easier way to see which packages were
installed by a given extra.
I've pulled out the error messages into functions that deal with "is
this for an extra or not". And I've rearranged the loop over two
different sets of requirements into one loop with a "must be instaled"
flag.
I hope you agree that this is clearer.
* Test cases
* Fix 'Unhandled error in Deferred'
Fixes a CRITICAL "Unhandled error in Deferred" log message which happened when
a function wrapped with `@cachedList` failed
* Minor optimisation to cachedListDescriptor
we can avoid re-using `missing`, which saves looking up entries in
`deferreds_map`, and means we don't need to copy it.
* Improve type annotation on CachedListDescriptor
* fix incorrect unwrapFirstError import
this was being imported from the wrong place
* Refactor `concurrently_execute` to use `yieldable_gather_results`
* Improve exception handling in `yieldable_gather_results`
Try to avoid swallowing so many stack traces.
* mark unwrapFirstError deprecated
* changelog
Part of the Tchap Synapse mainlining.
This allows modules to implement extra logic to figure out whether a given 3PID can be added to the local homeserver. In the Tchap use case, this will allow a Synapse module to interface with the custom endpoint /internal_info.
This adds some opentracing annotations to ResponseCache, to make it easier to see what's going on; in particular, it adds a link back to the initial trace which is actually doing the work of generating the response.
When all entries in an `LruCache` have a size of 0 according to the
provided `size_callback`, and `drop_from_cache` is called on a cache
node, the node would be unlinked from the LRU linked list but remain in
the cache dictionary. An assertion would be later be tripped due to the
inconsistency.
Avoid unintentionally calling `__len__` and use a strict `is None`
check instead when unwrapping the weak reference.
