Squashed commit of the following:
commit e5e2d793119733ecbcf9b85f966e018ab0318741
Author: Neil Alexander <neilalexander@users.noreply.github.com>
Date: Wed Jan 13 17:28:12 2021 +0000
Add dendrite_roomserver_processroomevent_duration_millis to prometheus
* Add RoomInfo cache, remove RoomServerRoomNID cache, ensure caches are thread-safe
* Don't panic if the roomInfo isn't known yet
* LRU package is already threadsafe
* Use RoomInfo cache to find room version if possible in Events()
* Adding comments about RoomInfoCache safety
* Hit the database far less to find room NIDs for event NIDs
* Close the rows
* Fix SQLite selectRoomNIDsForEventNIDsSQL
* Give same treatment to room version lookups
* Update GMSL
* Add MSC2836EventRelationships to fedsender
* Call MSC2836EventRelationships in reqCtx
* auth remote servers
* Extract room ID and servers from previous events; refactor a bit
* initial cut of federated threading
* Use the right client/fed struct in the response
* Add QueryAuthChain for use with MSC2836
* Add auth chain to federated response
* Fix pointers
* under CI: more logging and enable mscs, nil fix
* Handle direction: up
* Actually send message events to the roomserver..
* Add children and children_hash to unsigned, with tests
* Add logic for exploring threads and tracking children; missing storage functions
* Implement storage functions for children
* Add fetchUnknownEvent
* Do federated hits for include_children if we have unexplored children
* Use /ev_rel rather than /event as the former includes child metadata
* Remove cross-room threading impl
* Enable MSC2836 in the p2p demo
* Namespace mscs db
* Enable msc2836 for ygg
Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>
* Don't recalculate event IDs so often
* Revert invite change
* Make sure we're using the right NIDs
* Update gomatrixserverlib
* Update to NewEventFromTrustedJSONWithEventID
* Fix go.mod
* Update gomatrixserverlib to matrix-org/gomatrixserverlib#243
* Use BulkSelectEventID
* Add mscs/hooks package, begin work for msc2836
* Flesh out hooks and add SQL schema
* Begin implementing core msc2836 logic
* Add test harness
* Linting
* Implement visibility checks; stub out APIs for tests
* Flesh out testing
* Flesh out walkThread a bit
* Persist the origin_server_ts as well
* Edges table instead of relationships
* Add nodes table for event metadata
* LEFT JOIN to extract origin_server_ts for children
* Add graph walking structs
* Implement walking algorithm
* Add more graph walking tests
* Add auto_join for local rooms
* Fix create table syntax on postgres
* Add relationship_room_id|servers to the unsigned section of events
* Persist the parent room_id/servers in edge metadata
Other events cannot assert the true room_id/servers for the
parent event, only make claims to them, hence why this is
edge metadata.
* guts to pass through room_id/servers
* Refactor msc2836 to allow handling from federation
* Add JoinedVia to PerformJoin responses
* Fix tests; review comments
The previous implementation was only checking if room history was
"shared", which it wasn't for rooms where a user was invited, or world
readable rooms.
This implementation leverages the IsServerAllowed method, which already
implements the complete verification algorithm.
Signed-off-by: `Mayeul Cantan <oss+matrix@mayeul.net>`
Co-authored-by: Kegsay <kegan@matrix.org>
* Add basic storage methods
* Add internal api handler
* Add check for forgotten room
* Add /rooms/{roomID}/forget endpoint
* Add missing rsAPI method
* Remove unused parameters
* Add passing tests
Signed-off-by: Till Faelligen <tfaelligen@gmail.com>
* Add missing file
* Add postgres migration
* Add sqlite migration
* Use Forgetter to forget room
* Remove empty line
* Update HTTP status codes
It looks like the spec calls for these to be 400, rather than 403: https://matrix.org/docs/spec/client_server/r0.6.1#post-matrix-client-r0-rooms-roomid-forget
Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>
* Add configuration for max_message_bytes for sarama
* Log all errors when sending multiple messages
Signed-off-by: Till Faelligen <tfaelligen@gmail.com>
* Add missing config
* - Better comments on what MaxMessageBytes is used for
- Also sets the size the consumer may use
* Set RewritesState once
* Check if any new state provided
* Obey rewritesState
* Don't nuke everything the sync API knows when purging state
* Fix panic from duplicate insert
* Consistency
* Use HasState
* Remove nolint
* Clean up joined rooms on state rewrite
* Add resolve-state helper
* Tweaks
* Refactor forward extremities, again
* Tweaks
* Minor optimisation
* Make path a bit clearer
* Only process state/membership if forward extremities have changed
* Usage comments in resolve-state
* Fix sqite locking bugs present on sytest
Comments do the explaining.
* Fix deadlock in sqlite mode
Caused by starting a writer whilst within a writer
* Only complain about invalid state deltas for non-overwrite events
* Do not re-process outlier unnecessarily
* Add KindOld
* Don't process latest events/memberships for old events
* Allow federationsender to ignore duplicate key entries when LatestEventIDs is duplicated by RS output events
* Signal to downstream components if an event has become a forward extremity
* Don't exclude from sync
* Soft-fail checks on KindNew
* Don't run the latest events updater at all for KindOld
* Don't make federation sender change after all
* Kind in federation sender join
* Don't send isForwardExtremity
* Fix syncapi
* Update comments
* Fix SendEventWithState
* Update sytest-whitelist
* Generate old output events
* Sync API consumes old room events
* Update comments
* Start Kafka connection for each component that needs one
* Fix roomserver unit tests
* Rename to naffkaInstance (@Kegsay review comment)
* Fix import cycle
* Capture errors
* Don't request only state key tuples needed for auth (we end up discarding room state this way)
* QueryStateAfterEvent returns all state when no tuples supplied
* Resolve state
* Comments
* Adjust backfill to send backward extremity with state before other backfilled events, include prev_events with no state amongst missing events
* Not finished refactor
* Fix test
* Remove isInboundTxn
* Remove debug logging
We expect to have missing events as we walk back in the DAG over federation
as we didn't always create the room. When checking if the server is allowed
to see those events, just give up and stop rather than fail the request.
* Rename serverkeyapi to signingkeyserver
We use "api" for public facing stuff and "server" for internal stuff.
As the server key API is internal only, we call it 'signing key server',
which also clarifies the type of key (as opposed to TLS keys, E2E keys, etc)
* Convert docker/scripts to use signing-key-server
* Rename missed bits
* Deep forward extremity calculation
* Use updater txn
* Update error
* Update error
* Create previous event references in StoreEvent
* Use latest events updater to row-lock prev events
* Fix unexpected fallthrough
* Fix deadlock
* Don't roll back
* Update comments in calculateLatest
* Don't include events that we can't find references for in the forward extremities
* Add another passing test
* Don't send rewrite events
* Remove final traces of rewrite events
* Remove test that is no longer needed
* Revert "Remove test that is no longer needed"
This reverts commit 9a45babff6.
* Update test to use KindOutlier
* Resolve state after event against current room state when determining latest state changes
* Update sytest-whitelist
* Update sytest-whitelist, blacklist
* Try to ask other servers in the room for missing events if the origin won't provide them
* Logging
* More logging
* Implement QueryMissingAuthPrevEvents
* Try to get missing auth events badly
* Use processEvent
* Logging
* Update QueryMissingAuthPrevEvents
* Try to find missing auth events
* Patchy fix for test
* Logging tweaks
* Send auth events as outliers
* Update check in QueryMissingAuthPrevEvents
* Error responses
* More return codes
* Don't return error on reject/soft-fail since it was ultimately handled
* More tweaks
* More error tweaks
* Sanity-check room version on RS event input
* Update gomatrixserverlib
* Reject make_join when no room members are left
* Revert some changes from wrong branch
* Distinguish between room not existing and room being abandoned on this server
* nolint
* Replace all usages of txn.Stmt with sqlutil.TxStmt
Signed-off-by: Sam Day <me@samcday.com>
* Fix sign off link in PR template.
Signed-off-by: Sam Day <me@samcday.com>
Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>
* WIP Event rejection
* Still send back errors for rejected events
Instead, discard them at the federationapi /send layer rather than
re-implementing checks at the clientapi/PerformJoin layer.
* Implement rejected events
Critically, rejected events CAN cause state resolution to happen
as it can merge forks in the DAG. This is fine, _provided_ we
do not add the rejected event when performing state resolution,
which is what this PR does. It also fixes the error handling
when NotAllowed happens, as we were checking too early and needlessly
handling NotAllowed in more than one place.
* Update test to match reality
* Modify InputRoomEvents to no longer return an error
Errors do not serialise across HTTP boundaries in polylith mode,
so instead set fields on the InputRoomEventsResponse. Add `Err()`
function to make the API shape basically the same.
* Remove redundant returns; linting
* Update blacklist
* SendEventWithState events as new
* Use cumulative state IDs for final event
* Error wrapping in calculateAndSetState
* Handle overwriting same event type and state key
* Hacky way to spot historical events
* Don't exclude from sync
* Don't generate output events when rewriting forward extremities
* Update output event check
* Historical output events
* Define output room event type
* Notify key changes on state
* Don't send our membership event twice
* Deduplicate state entries
* Tweaks
* Remove unnecessary nolint
* Fix current state upsert in sync API
* Send auth events as outliers, state events as rewrite
* Sync API don't consume state events
* Process events actually
* Improve outlier check
* Fix local room check
* Remove extra room check, it seems to break the whole damn world
* Fix federated join check
* Fix nil pointer exception
* Better comments on DeduplicateStateEntries
* Reflow forced federated joins
* Don't force federated join for possibly even local invites
* Comment SendEventWithState better
* Rewrite room state in sync API storage
* Add TODO
* Clean up all room data when receiving create event
* Don't generate output events for rewrites, but instead notify that state is rewritten on the final new event
* Rename to PurgeRoom
* Exclude backfilled messages from /sync
* Split out rewriting state from updating state from state res
Co-authored-by: Kegan Dougal <kegan@matrix.org>
* Remove QueryBulkStateContent from current state server
Expected fail due to db impl not existing
* Implement query bulk state content
* Fix up rejecting invites over federation
* Fix bulk content marshalling
* Move currentstateserver API to roomserver
Stub out DB functions for now, nothing uses the roomserver version yet.
* Allow it to startup
* Implement some current-state-server storage interface functions
* Add missing package
* Initial FIFOing of roomserver inputs
* Remove EventID response from api.InputRoomEventsResponse
* Don't send back event ID unnecessarily
* Fix ordering hopefully
* Reduce copies, use buffered task channel to reduce contention on other rooms
* Fix error handling
* Add Queryer and use embedded structs
* Add Inputer and factor out more RS API stuff
This neatly splits up the RS API based on the functionality it provides,
whilst providing a useful place for code sharing via the `helpers` package.
* Use federation sender for backfill and getting missing events
* Fix internal URL paths
* Update go.mod/go.sum for matrix-org/gomatrixserverlib#218
* Add missing server implementations in HTTP interface
- New package `perform` which contains all `Perform` functions
- New package `helpers` which contains helper functions used by both
perform and query/input functions.
- Perform invite/leave have no idea how to `WriteOutputEvents` and this
is now returned from `PerformInvite` or `PerformLeave` respectively.
Still to do:
- RSAPI is fed into the inviter/joiner/leaver - this introduces circular
logic so will need to be removed.
- Put query operations in a `query` package.
- Put input operations (and output) in an `input` package.
- Factor out helper functions as much as possible, possibly rejigging the
storage layer in the process.
* Initial work on roomserver NID caches
* Give caches to roomserver storage
* Populate caches
* Fix bugs
* Fix WASM build
* Don't hit cache twice in RoomNIDExcludingStubs
* Store reverse room ID-room NID mapping, consult caches when assigning NIDs
* Offset updates take place using TransactionWriter
* Refactor TransactionWriter in current state server
* Refactor TransactionWriter in federation sender
* Refactor TransactionWriter in key server
* Refactor TransactionWriter in media API
* Refactor TransactionWriter in server key API
* Refactor TransactionWriter in sync API
* Refactor TransactionWriter in user API
* Fix deadlocking Sync API tests
* Un-deadlock device database
* Fix appservice API
* Rename TransactionWriters to Writers
* Move writers up a layer in sync API
* Document sqlutil.Writer interface
* Add note to Writer documentation
* Per-room input mutex
* GetMembership should use transaction when assigning state key NID
* Actually use writer transactions rather than ignoring them
* Limit per-room mutexes to Postgres
* Flip the check in InputRoomEvents
* Updated TransactionWriters, moved locks in roomserver, various other tweaks
* Fix redaction deadlocks
* Fix lint issue
* Rename SQLiteTransactionWriter to ExclusiveTransactionWriter
* Fix us not sending transactions through in latest events updater
* Make PerformJoin send input membership event
* Invite input room events in separate goroutine
* Don't limit roomserver input events using request context
* Synchronous input room events
* Nope, that didn't work
* oops send state key to GetMembership
* Don't generate stripped state in client API more times than necessary, generate output events on receiving end of federated invite
* Commit membership updater changes
* Tweaks
* Initial pass at refactoring config (not finished)
* Don't forget current state and EDU servers
* More shifting around
* Update server key API tests
* Fix roomserver test
* Fix more tests
* Further tweaks
* Fix current state server test (sort of)
* Maybe fix appservices
* Fix client API test
* Include database connection string in database options
* Fix sync API build
* Update config test
* Fix unit tests
* Fix federation sender build
* Fix gobind build
* Set Listen address for all services in HTTP monolith mode
* Validate config, reinstate appservice derived in directory, tweaks
* Tweak federation API test
* Set MaxOpenConnections/MaxIdleConnections to previous values
* Update generate-config
* Modify /state/{eventType}/{stateKey} to return the event at the time the user left
Or live, depending on their current state. Hopefully fixes some sytests!
* Linting
* Set HasBeenInRoom
* Fix cases for world-readable history visibility
* Fix bug in finding the requested state event
Co-authored-by: Neil Alexander <neilalexander@users.noreply.github.com>
* Use TransactionWriter on other component SQLites
* Fix sync API tests
* Fix panic in media API
* Fix a couple of transactions
* Fix wrong query, add some logging output
* Add debug logging into StoreEvent
* Adjust InsertRoomNID
* Update logging
* Add a bit more logging to the fedsender
* bugfix: continue sending PDUs if ones are added whilst sending another PDU
Without this, the queue goes back to sleep on `<-oq.notifyPDUs` which won't
fire because `pendingPDUs` is already > 0. This should fix a flakey sytest.
* Break if no txn is sent
* WIP syncapi work
* More debugging
* Bump GMSL version to pull in working Event.Redact
* Remove logging
* Make redactions work on v3+
* Fix more tests
* Emit redacted_event from the roomserver when redactions are validated
- Consume them in the currentstateserver and act accordingly.
- Add integration test for the roomserver to check that injecting
`m.room.redaction` events result in `redacted_event` being emitted.
* Linting
* Ignore events that redact themselves
* Implement core redaction logic
- Add a new `redactions_table.go` which tracks the mapping of
the redaction event ID and the redacted event ID
- Mark redactions as 'validated' when we have both events.
- When redactions are validated, add `unsigned.redacted_because`
and modify the `eventJSON` accordingly.
Note: We currently do NOT redact the event content - it's gated
behind a feature flag - until we have tested redactions a bit more.
* Linting
* Use content_value instead of membership
* Fix build
* Replace publicroomsapi with a combination of clientapi/roomserver/currentstateserver
- All public rooms paths are now handled by clientapi
- Requests to (un)publish rooms are sent to the roomserver via `PerformPublish`
which are stored in a new `published_table.go`
- Requests for public rooms are handled in clientapi by:
* Fetch all room IDs which are published using `QueryPublishedRooms` on the roomserver.
* Apply pagination parameters to the slice.
* Do a `QueryBulkStateContent` request to the currentstateserver to pull out
required state event *content* (not entire events).
* Aggregate and return the chunk.
Mostly but not fully implemented (DB queries on currentstateserver are missing)
* Fix pq query
* Make postgres work
* Make sqlite work
* Fix tests
* Unbreak pagination tests
* Linting
* Return remote errors from FS.PerformJoin
Follows the same pattern as PerformJoin on roomserver (no error return).
Also return the right format for incompatible room version errors.
Makes a bunch of tests pass!
* Handle network errors better when returning remote HTTP errors
* Linting
* Fix tests
* Update whitelist, pass network errors through in API=1 mode
* Add PerformInvite and refactor how errors get handled
- Rename `JoinError` to `PerformError`
- Remove `error` from the API function signature entirely. This forces
errors to be bundled into `PerformError` which makes it easier for callers
to detect and handle errors. On network errors, HTTP clients will make a
`PerformError`.
* Unbreak everything; thanks Go!
* Send back JSONResponse according to the PerformError
* Update federation invite code too
* Pass join errors through internal API boundaries
Required for certain invite sytests. We will need to think of a
better way of handling this going forwards.
* Include m.room.avatar in stripped state; handle trailing slashes when GETing state events
* Update whitelist
* Update whitelist
* Minor perf/debugging improvements
- publicroomsapi: Don't call QueryEventsByID with no event IDs
- appservice: Consume only if there are 1 or more ASes
- roomserver: don't keep a copy of the request "for debugging" - we trace now
* fedsender: return early if we have no destinations
* Unbreak tests
This is a wrapper around whatever impl we have which then logs
the function name/request/response/error.
Also tweak when we log on kafka streams: only log on the producer
side not the consumer side: we've never had issues with comms and
having 1 message rather than N would be nice.
* s/QueryBackfill/PerformBackfill/g
* OutputEvent now includes AddStateEvents which contain the full event of extra state events
* Only include adds not the current event
* Get adding state right
* Remove clientapi producers which aren't actually producers
They are actually just convenience wrappers around the internal APIs
for roomserver/eduserver. Move their logic to their respective `api`
packages and call them directly.
* Remove TODO
* unbreak ygg
* Split out adding HTTP routes from making internal APIs for clarity
* Split out more components
* Split out more things
* Finish converting
* internal mux for internal routes
* Move Updater structs to shared and use it for postgres
* Add constructors for NewXXXUpdater and a useTxns flag
In sqlite, we set useTxns=false and comment why.
* Handle nil txn
* Handle nil in transaction
* Missed one
* Close the txn at the right time
* Don't close the transaction as we reuse it between calls
* Add missing routing for PerformDirectoryLookupRequest
* Tweak output
* Fix some bugs in devices
* Don't default to federated room joins in response to invite
* Update sytest-whitelist
* Update comments
* Return correct room ID from PerformJoin
* Fix appservice and EDU server API setup, update sytest-whitelist
* Update sytest-whitelist
* Separate muxes for public and internal APIs
* Update client-api-proxy and federation-api-proxy so they don't add /api to the path
* Tidy up
* Consistent HTTP setup
* Set up prefixes properly
* sytest: Make 'Inbound federation can backfill events' pass
This breaks 'Outbound federation can backfill events' because now
we are returning the right number of events, which the previous
test was relying on.
Previously, /messages was backfilling the membership event, causing
the test to pass. Now we are no longer backfilling the membership
event due to the change in this commit, causing the test to fail.
The test should instead be returning the membership event locally
from synacpis database, but it doesn't do it fast enough, resulting
in a no-op /sync response with a next_batch=s0_0 which will never
pick up the local membership event when it rolls in. The test
does attempt to retry, but doesn't take the new next_batch=s1_0
resulting in it missing from the /messages response.
* Linting
* Comment out updaters a bit, add overwrite flag to latest events
* Make sure we don't send fast-forwarded state changes over federation, start with empty set when overwriting
* Remove redundant check for overwrite
* WIP get_missing_events work
* More WIP get_missing_events work
* First working /get_missing_events implementation
Flakey currently due to racing between /sync and /send
* Final tweaks
* Remove log lines
* Linting
* go mod tidy
* Clamp min depth to 0
* sort events by depth because sytest makes me sad
Specifically I think it's
4172585c25/lib/SyTest/Federation/Client.pm (L265)
to blame here.
* Improve federation sender performance and behaviour, add backoff
* Tweaks
* Tweaks
* Tweaks
* Take copies of events before passing to destination queues
* Don't accidentally drop queued messages
* Don't take copies again
* Tidy up a bit
* Break out statistics (tracked component-wide), report success and failures from Perform actions
* Fix comment, use atomic add
* Improve logic a bit, don't block on wakeup, move idle check
* Don't retry sucessful invites, don't dispatch sendEvent, sendInvite etc
* Dedupe destinations, fix other bug hopefully
* Dispatch sends again
* Federation sender to ignore invites that are destined locally
* Loopback invite events
* Remodel a bit with channels
* Linter
* Only loopback invite event if we know the room
* We should tell other resident servers about the invite if we know about the room
* Correct invite signing
* Fix invite loopback
* Check HTTP response codes, push new invites to front of queue
* Review comments
* Add PerformJoin template
* Try roomserver perform join
* Send correct server name to FS API
* Pass through content, try to handle multiple server names
* Fix local server checks
* Don't refer to non-existent error
* Add directory lookups of aliases
* Remove unneeded parameters
* Don't repeat join events into the roomserver
* Unmarshal the content, that would help
* Check if the user is already in the room in the fedeationapi too
* Return incompatible room version error
* Use Membership, don't try more servers than needed
* Review comments, make FS API take list of servernames, dedupe them, break out of loop properly on success
* Tweaks
* Limit database connections (#564)
- Add new options to the config file database:
max_open_conns: 100
max_idle_conns: 2
conn_max_lifetime: -1
- Implement connection parameter setup on the *DB (database/sql) in internal/sqlutil/trace.go:Open()
- Propagate the values in the form of DbProperties interface via all the
Open() and NewDatabase() functions
Signed-off-by: Tomas Jirka <tomas.jirka@email.cz>
* Fix wasm builds
* Remove file accidentally added from working tree
Co-authored-by: Tomas Jirka <tomas.jirka@email.cz>
* Consolidation of roomserver APIs
* Comment out alias tests for now, they are broken
* Wire AS API into roomserver again
* Roomserver didn't take asAPI param before so return to that
* Prevent roomserver asking AS API for alias info
* Rename some files
* Remove alias_test, incoherent tests and unwanted appservice integration
* Remove FS API inject on syncapi component
* Make backfill work for shared history visibility
* fetch missing state on backfill to remember snapshots correctly
* Fix gmsl to not mux in auth events into room state
* Whoops
* Linting
* Define an input API for the federationsender
* Wiring for rooomserver input API and federation sender input API
* Whoops, commit common too
* Merge input API into query API
* Rename FederationSenderQueryAPI to FederationSenderInternalAPI
* Fix dendritejs
* Rename Input to Perform
* Fix a couple of inputs -> performs
* Remove needless storage interface, add comments
* Initial cut for backfilling
The syncserver now asks the roomserver via QueryBackfill (which already
existed to *handle* backfill requests) which then makes federation requests
via gomatrixserverlib.RequestBackfill.
Currently, tests fail on subsequent /messages requests because we don't know
which servers are in the room, because we are unable to get state snapshots
from a backfilled event because that code doesn't exist yet.
* WIP backfill, doesn't work
* Make initial backfill pass checks
* Persist backfilled events with state snapshots
* Remove debug lines
* Linting
* Review comments
* Update gomatixserverlib
* Try to build invite stripped state if not given to us
* SendInvite improvements
* Transpose invite_room_state into invite_state.events for sync API
* Remove syncapi debugging output
* Use RespInviteV2
* Update gomatrixserverlib
* Send the invite event as a normal roomserver event too, for incorporating into room (should this be done by the roomserver automatically for invite inputs?)
* Federation sender use invite_room_state, room server try to insert membership state
* Check supported room versions on the invite endpoint
* Prevent roomserver query API from trying to handle requests for stub rooms
* Adding a nolint
* Replace IsRoomStub with RoomNIDExcludingStubs, fix query API to use that instead
* Review comments
* Experimental LRU cache for room versions
* Don't accidentally try to type-assert nil
* Also reduce hits on query API
* Use hashicorp implementation which mutexes for us
* Define const for max cache entries
* Rename to be specifically immutable, panic if we try to mutate a cache entry
* Review comments
* Remove nil guards, give roomserver integration test a cache
* go mod tidy
