Commit Graph

243 Commits (58c657322a18cadf8036819ef87fce9889d65fb0)

Author SHA1 Message Date
Richard van der Hoff 77ebef9d43
Merge pull request #3118 from matrix-org/rav/reject_prev_events
Reject events which have lots of prev_events
2018-04-23 17:51:38 +01:00
Richard van der Hoff dc875d2712
Merge pull request #3106 from NotAFile/py3-six-itervalues-1
Use six.itervalues in some places
2018-04-20 15:43:52 +01:00
Richard van der Hoff 0c280d4d99 Reinstate linearizer for federation_server.on_context_state_request 2018-04-20 11:10:04 +01:00
Richard van der Hoff 1f4b498b73 Add some comments 2018-04-18 00:15:36 +01:00
Adrian Tschira 36c59ce669 Use six.itervalues in some places
There's more where that came from

Signed-off-by: Adrian Tschira <nota@notafile.com>
2018-04-15 20:39:43 +02:00
Richard van der Hoff b78395b7fe Refactor ResponseCache usage
Adds a `.wrap` method to ResponseCache which wraps up the boilerplate of a
(get, set) pair, and then use it throughout the codebase.

This will be largely non-functional, but does include the following functional
changes:

* federation_server.on_context_state_request: drops use of _server_linearizer
  which looked redundant and could cause incorrect cache misses by yielding
  between the get and the set.
* RoomListHandler.get_remote_public_room_list(): fixes logcontext leaks
* the wrap function includes some logging. I'm hoping this won't be too noisy
  on production.
2018-04-12 13:02:15 +01:00
Richard van der Hoff b3384232a0 Add metrics for ResponseCache 2018-04-10 23:14:47 +01:00
Erik Johnston 56e709857c
Merge pull request #2979 from matrix-org/erikj/no_handlers
Don't build handlers on workers unnecessarily
2018-03-13 13:46:38 +00:00
Erik Johnston 9a2d9b4789
Merge pull request #2977 from matrix-org/erikj/replication_move_props
Move property setting from ReplicationLayer to base classes
2018-03-13 11:45:25 +00:00
Erik Johnston f43b6d6d9b Fix docstring types 2018-03-13 11:29:35 +00:00
Erik Johnston 265b993b8a Split replication layer into two 2018-03-13 10:55:47 +00:00
Erik Johnston e05bf34117 Move property setting from ReplicationLayer to FederationBase 2018-03-13 10:51:30 +00:00
Erik Johnston c3f79c9da5 Split out edu/query registration to a separate class 2018-03-13 10:24:27 +00:00
Richard van der Hoff 3079f80d4a Factor out `event_from_pdu_json`
turns out we have two copies of this, and neither needs to be an instance
method
2017-12-30 18:40:19 +00:00
Richard van der Hoff 65abc90fb6 federation_server: clean up imports 2017-12-30 18:40:19 +00:00
Erik Johnston 2a7e9faeec Do logcontexts outside ResponseCache 2017-10-25 15:21:08 +01:00
Richard van der Hoff 582bd19ee9 Fix 500 error when we get an error handling a PDU
FederationServer doesn't have a send_failure (and nor does its subclass,
ReplicationLayer), so this was failing.

I'm not really sure what the idea behind send_failure is, given (a) we don't do
anything at the other end with it except log it, and (b) we also send back the
failure via the transaction response. I suspect there's a whole lot of dead
code around it, but for now I'm just removing the broken bit.
2017-10-17 20:52:40 +01:00
Richard van der Hoff 8dd0c85ac5 Merge pull request #2529 from matrix-org/rav/fix_transaction_failure_handling
log pdu_failures from incoming transactions
2017-10-11 17:29:14 +01:00
Richard van der Hoff b75d443caf log pdu_failures from incoming transactions
... even if we have no EDUs.

This appears to have been introduced in
476899295f.
2017-10-11 14:36:13 +01:00
Richard van der Hoff 6a6cc27aee fed server: process PDUs for different rooms in parallel
With luck, this will give a real-time improvement when there are many rooms and
the server ends up calling out to fetch missing events.
2017-10-09 18:30:31 +01:00
Richard van der Hoff 4c7c4d4061 Fed server: use a linearizer for ongoing transactions
We don't want to process the same transaction multiple times concurrently, so
use a linearizer.
2017-10-09 18:30:10 +01:00
Richard van der Hoff ba5b9b80a5 fed server: refactor on_incoming_transaction
Move as much as possible to after the have_responded check, and reduce the
number of times we iterate over the pdu list.
2017-10-09 18:10:53 +01:00
Richard van der Hoff a6e3222fe5 Fed server: Move origin-check code to _handle_received_pdu
The response-building code expects there to be an entry in the `results` list
for each entry in the pdu_list, so the early `continue` was messing this
up. That doesn't really matter, because all that the federation client does is
log any errors, but it's pretty poor form.
2017-10-09 17:53:32 +01:00
Richard van der Hoff de042b3b88 Do some logging when one-time-keys get claimed
might help us figure out if https://github.com/vector-im/riot-web/issues/3868
has happened.
2017-05-09 19:04:56 +01:00
Richard van der Hoff 30bcbf775a Accept join events from all servers
Make sure that we accept join events from any server, rather than just the
origin server, to make the federation join dance work correctly.

(Fixes #1893).
2017-04-03 15:58:07 +01:00
Richard van der Hoff 29235901b8 Move FederationServer._handle_new_pdu to FederationHandler
Unfortunately this significantly increases the size of the already-rather-big
FederationHandler, but the code fits more naturally here, and it paves the way
for the tighter integration that I need between handling incoming PDUs and
doing the join dance.

Other than renaming the existing `FederationHandler.on_receive_pdu` to
`_process_received_pdu` to make way for it, this just consists of the move, and
replacing `self.handler` with `self` and `self` with `self.replication_layer`.
2017-03-09 16:20:13 +00:00
Richard van der Hoff e8b1721290 Move sig check out of _handle_new_pdu
When we receive PDUs via `get_missing_events`, we have already checked their
sigs, so there is no need to do it again.
2017-03-09 15:50:44 +00:00
Richard van der Hoff 3406333a58 Factor _get_missing_events_for_pdu out of _handle_new_pdu
This should be functionally identical: it just seeks to improve readability by
reducing indentation.
2017-03-09 15:50:44 +00:00
Erik Johnston c974116f19 Implement device key caching over federation 2017-01-26 16:07:24 +00:00
Erik Johnston f7085ac84f Name linearizer's for better logs 2017-01-09 17:17:10 +00:00
Matthew 4304e7e593 do the discard check in the right place to avoid grabbing dependent events 2017-01-07 03:44:18 +00:00
Matthew e10c527930 Discard PDUs from invalid origins due to #1753 in 0.18.[56] 2017-01-07 02:13:14 +00:00
Matthew Hodgson 468749c9fc fix comment 2017-01-05 12:00:11 +00:00
Matthew Hodgson eedf400d05 limit total timeout for get_missing_events to 10s 2017-01-05 11:58:15 +00:00
Mark Haines 62ce3034f3 s/aquire/acquire/g 2016-12-30 20:04:44 +00:00
Mark Haines 0aff09f6c9 Add more useful logging when we block fetching events 2016-12-30 20:00:44 +00:00
Mark Haines d4a35ada28 Send device messages over federation 2016-09-06 18:16:20 +01:00
Erik Johnston 2854ee2a52 Only pull out IDs from DB for /state_ids/ request 2016-09-02 10:53:36 +01:00
Erik Johnston b4e2290d89 Merge branch 'develop' of github.com:matrix-org/synapse into erikj/state_ids_api 2016-08-04 14:04:35 +01:00
Erik Johnston edb33eb163 Rename fields to _ids 2016-08-03 17:19:15 +01:00
Erik Johnston bcc9cda8ca Fix copy + paste fails 2016-08-03 17:17:26 +01:00
Erik Johnston e3a720217a Add /state_ids federation API
The new API only returns the event_ids for the state, as most
requesters will already have the vast majority of the events already.
2016-08-03 14:47:37 +01:00
Mark Haines 921f17f938 Merge branch 'develop' into rav/refactor_device_query 2016-08-03 11:12:47 +01:00
Richard van der Hoff 1efee2f52b E2E keys: Make federation query share code with client query
Refactor the e2e query handler to separate out the local query, and then make
the federation handler use it.
2016-08-02 18:12:00 +01:00
Erik Johnston c9154b970c Don't double wrap 200 2016-08-02 16:45:53 +01:00
Erik Johnston b3d5c4ad9d Fix response cache 2016-08-02 16:42:21 +01:00
Erik Johnston 248e6770ca Cache federation state responses 2016-07-21 10:30:12 +01:00
Erik Johnston 8f4a9bbc16 Linearize some federation endpoints based on (origin, room_id) 2016-06-17 16:43:45 +01:00
Erik Johnston 2884712ca7 Only re-sign our own events 2016-06-17 14:47:33 +01:00
Erik Johnston d41a1a91d3 Linearize fetching of gaps on incoming events
This potentially stops the server from doing multiple requests for the
same data.
2016-06-15 15:16:14 +01:00
Erik Johnston 17aab5827a Add some logging for when servers ask for missing events 2016-06-08 11:55:31 +01:00
Mark Haines 9c272da05f Add an openidish mechanism for proving to third parties that you own a given user_id 2016-05-05 13:42:44 +01:00
Mark Haines 5244c0b48e Remove unused backfilled parameter from persist_event 2016-03-21 18:06:08 +00:00
Erik Johnston 9adf0e92bc Catch exceptions from EDU handling 2016-03-18 15:12:50 +00:00
Erik Johnston 3c5f25507b Yield on EDU handling 2016-03-18 13:55:16 +00:00
Daniel Wagner-Hall 577951b032 Allow third_party_signed to be specified on /join 2016-02-23 15:11:25 +00:00
Erik Johnston 2c1fbea531 Fix up logcontexts 2016-02-08 14:26:45 +00:00
Matthew Hodgson 6c28ac260c copyrights 2016-01-07 04:26:29 +00:00
Daniel Wagner-Hall 2cebe53545 Exchange 3pid invites for m.room.member invites 2015-11-05 16:43:19 +00:00
Daniel Wagner-Hall 137fafce4e Allow rejecting invites
This is done by using the same /leave flow as you would use if you had
already accepted the invite and wanted to leave.
2015-10-20 11:58:58 +01:00
Daniel Wagner-Hall 0c38e8637f Remove unnecessary class-wrapping 2015-10-13 18:00:38 +01:00
Daniel Wagner-Hall 5b3e9713dd Implement third party identifier invites 2015-10-01 17:49:52 +01:00
Mark Haines 0cceb2ac92 Add a few strategic new lines to break up the on_query_client_keys and on_claim_client_keys methods in federation_server.py 2015-08-13 17:27:46 +01:00
Mark Haines 62c010283d Add federation support for end-to-end key requests 2015-07-23 16:03:38 +01:00
Erik Johnston 476899295f Change the way we do logging contexts so that they survive divergences 2015-05-08 16:32:18 +01:00
Erik Johnston 23c639ff32 Split a storage function in two so that we don't have to do extra work. 2015-05-01 10:17:19 +01:00
Paul "LeoNerd" Evans 2e4f0b2bd7 Replace the @metrics.counted annotations in federation with specifically-written counters and distributions 2015-03-12 16:24:51 +00:00
Paul "LeoNerd" Evans 9470412316 Initial attempt at sprinkling some @metrics.counted decorations around the federation code 2015-03-12 16:24:50 +00:00
Erik Johnston 9708f49abf Docs 2015-03-05 16:35:16 +00:00
Erik Johnston ae702d161a Handle if get_missing_pdu returns 400 or not all events. 2015-03-05 16:08:02 +00:00
Erik Johnston 0a036944bd Merge branch 'develop' of github.com:matrix-org/synapse into batched_get_pdu 2015-03-02 13:53:30 +00:00
Erik Johnston 29481690c5 If we're yielding don't add errback 2015-03-02 11:50:43 +00:00
Erik Johnston 23d9bd1d74 Process transactions serially.
Since the events received in a transaction are ordered, later events
might depend on earlier events and so we shouldn't blindly process them
in parellel.
2015-03-02 11:39:57 +00:00
Erik Johnston db215b7e00 Implement and use new batched get missing pdu 2015-02-23 13:58:02 +00:00
Erik Johnston 0ac2a79faa Initial stab at implementing a batched get_missing_pdus request 2015-02-19 17:24:14 +00:00
Erik Johnston 676e8ee78a Remove debug raise 2015-02-17 15:22:45 +00:00
Erik Johnston c02da58a9d Merge branch 'develop' of github.com:matrix-org/synapse into failures 2015-02-17 15:15:07 +00:00
Erik Johnston 659ead082f Format the response of transaction request in a nicer way 2015-02-17 15:11:44 +00:00
Erik Johnston c82e26ad4b Actually respond with JSON to incoming transaction 2015-02-17 13:24:13 +00:00
Erik Johnston 02bfa889de Handle recieving failures in transactions 2015-02-17 13:13:14 +00:00
Erik Johnston 72a4de2ce6 Use consumeErrors=True on all DeferredLists.
This is so that the DeferredLists actually consume the error instead of
propogating down the non-existent errback chain. This should reduce the
number of unhandled errors we are seeing.
2015-02-17 10:07:01 +00:00
Erik Johnston 91fc5eef1d Mark old events as outliers.
This is to fix the issue where if a remote server sends an event
that references a really "old" event, then the local server will pull
that in and send to all clients.

We decide if an event is old if its depth is less than the minimum depth
of the room.
2015-02-16 14:27:40 +00:00
Erik Johnston 6138584651 Don't return anything from _handle_new_pdu, since we ignore the return value anyway 2015-02-16 14:08:02 +00:00
Erik Johnston 963256638d Correctly handle all the places that can throw exceptions 2015-02-12 18:17:11 +00:00
Erik Johnston 697ab75a34 Sign auth_chains when returned by /state/ requests 2015-02-10 15:46:24 +00:00
Erik Johnston 8dae5c8108 Remove unused imports 2015-02-03 15:01:12 +00:00
Erik Johnston 0dd3aea319 Keep around the old (buggy) version of the prune_event function so that we can use it to check signatures for events on old servers 2015-02-03 14:58:30 +00:00
Erik Johnston 776ac820f9 Briefly doc structure of query_auth API. 2015-01-30 15:58:28 +00:00
Erik Johnston a70a801184 Fix bug where we superfluously asked for current state. Change API of /query_auth/ so that we don't duplicate events in the response. 2015-01-30 13:34:01 +00:00
Erik Johnston c1d860870b Fix regression where we no longer correctly handled the case of gaps in our event graph 2015-01-30 10:48:47 +00:00
Erik Johnston 78015948a7 Initial implementation of auth conflict resolution 2015-01-29 16:52:33 +00:00
Erik Johnston c92d64a6c3 Make it the responsibility of the replication layer to check signature and hashes. 2015-01-26 14:33:11 +00:00
Erik Johnston 7b88619241 Split up replication_layer module into client, server and transaction queue 2015-01-26 10:45:24 +00:00