Weakness in auth chain indexing allows DoS from remote room members
through disk fill and high CPU usage.
A remote Matrix user with malicious intent, sharing a room with Synapse
instances before 1.104.1, can dispatch specially crafted events to
exploit a weakness in how the auth chain cover index is calculated. This
can induce high CPU consumption and accumulate excessive data in the
database of such instances, resulting in a denial of service.
Servers in private federations, or those that do not federate, are not
affected.
Resurrecting https://github.com/matrix-org/synapse/pull/13918.
This should reduce IOPs incurred by joining to the events table to
lookup stream ordering, which happens in many receipt handling code
paths. Like the previous PR I believe sufficient time has passed between
the original migration in DB schema 72 and now to merge this as-is. It's
highly unlikely that both the migration is still ongoing AND (active)
users still have any receipts prior to that date.
In the unlikely event there is a receipt without a populated
`event_stream_ordering` synapse will behave just as it does now when
receipts exist for events that don't (yet): for push action calculation
the receipts are just ignored.
I've removed the validation on event IDs as this is already covered
here:
59ceabcb97/synapse/handlers/receipts.py (L189-L192)
PR #16942 removed an invalid optimisation that avoided pulling out state
for non-gappy syncs. This causes a large increase in DB usage. c.f.
#16941 for why that optimisation was wrong.
However, we can still optimise in the simple case where the events in
the timeline are a linear chain without any branching/merging of the
DAG.
cc. @richvdh
Before we were pulling out *all* read receipts for a user for every
event we pushed. Instead let's only pull out the relevant receipts.
This also pulled out the event rows for each receipt, causing load on
the events table.
This PR fixes a very, very niche edge-case, but I've got some more work
coming which will otherwise make the problem worse.
The bug happens when the syncing user leaves a room, and has a sync
filter which includes "left" rooms, but sets the timeline limit to 0. In
that case, the state returned in the `state` section is calculated
incorrectly.
The fix is to pass a token corresponding to the point that the user
leaves the room through to `compute_state_delta`.
Requests may require a User-Agent header, and the change in #16972
accidentally removed it, resulting in requests getting rejected causing
login to fail.
Fixes https://github.com/element-hq/synapse/issues/16680, as well as a
related bug, where servers which we had *never* successfully sent an
event to would not be retried.
In order to fix the case of pending to-device messages, we hook into the
existing `wake_destinations_needing_catchup` process, by extending it to
look for destinations that have pending to-device messages. The
federation transmission loop then attempts to send the pending to-device
messages as normal.
When a lot of locks are waiting for a single lock, notifying all locks
independently with `call_later` on each release is really costly and
incurs some kind of async contention, where the CPU is spinning a lot
for not much.
The included test is taking around 30s before the change, and 0.5s
after.
It was found following failing tests with
https://github.com/element-hq/synapse/pull/16827.
This PR aims to fix#16895, caused by a regression in #7 and not fixed
by #16903. The PR #16903 only fixes a starvation issue, where the CPU
isn't released. There is a second issue, where the execution is blocked.
This theory is supported by the flame graphs provided in #16895 and the
fact that I see the CPU usage reducing and far below the limit.
Since the changes in #7, the method `check_state_independent_auth_rules`
is called with the additional parameter `batched_auth_events`:
6fa13b4f92/synapse/handlers/federation_event.py (L1741-L1743)
It makes the execution enter this if clause, introduced with #151956fa13b4f92/synapse/event_auth.py (L178-L189)
There are two issues in the above code snippet.
First, there is the blocking issue. I'm not entirely sure if this is a
deadlock, starvation, or something different. In the beginning, I
thought the copy operation was responsible. It wasn't. Then I
investigated the nested `store.get_events` inside the function `update`.
This was also not causing the blocking issue. Only when I replaced the
set difference operation (`-` ) with a list comprehension, the blocking
was resolved. Creating and comparing sets with a very large amount of
events seems to be problematic.
This is how the flamegraph looks now while persisting outliers. As you
can see, the execution no longer locks up in the above function.
Second, the copying here doesn't serve any purpose, because only a
shallow copy is created. This means the same objects from the original
dict are referenced. This fails the intention of protecting these
objects from mutation. The review of the original PR
https://github.com/matrix-org/synapse/pull/15195 had an extensive
discussion about this matter.
Various approaches to copying the auth_events were attempted:
1) Implementing a deepcopy caused issues due to
builtins.EventInternalMetadata not being pickleable.
2) Creating a dict with new objects akin to a deepcopy.
3) Creating a dict with new objects containing only necessary
attributes.
Concluding, there is no easy way to create an actual copy of the
objects. Opting for a deepcopy can significantly strain memory and CPU
resources, making it an inefficient choice. I don't see why the copy is
necessary in the first place. Therefore I'm proposing to remove it
altogether.
After these changes, I was able to successfully join these rooms,
without the main worker locking up:
- #synapse:matrix.org
- #element-android:matrix.org
- #element-web:matrix.org
- #ecips:matrix.org
- #ipfs-chatter:ipfs.io
- #python:matrix.org
- #matrix:matrix.org
