Files
Warp/server/internal/roomserver/roomserver.go
T
tajniak81andClaude Opus 5 d187ad0d17
CI / Test (ubuntu-latest) (push) Canceled after 0s
CI / Test (windows-latest) (push) Canceled after 0s
CI / Storage tests on PostgreSQL (push) Canceled after 0s
CI / Complement (federation off) (push) Canceled after 0s
CI / golangci-lint (push) Canceled after 0s
CI / Smoke test the binary (push) Canceled after 0s
Phase 4: sync
Implements PLAN.md Phase 4: the read model. /sync, /messages and /context,
the filters that narrow them, and the streams a response carries that are
not room events — typing, receipts, presence and to-device messages. With
this a client can be used against Warp, which is milestone M1.

The one architectural choice is that the syncapi has no copy of the rooms:
it reads them through the roomserver. Dendrite gives its syncapi its own
tables and feeds them from the output stream, which is right for a design
where the two can be separate processes, and Dendrite's could be. Warp is
one process by decision, and there the copy buys nothing and costs the
hardest class of bug a homeserver has — a read model that disagrees with
the room, where a client is shown a message that was rejected and the only
way to find out is to compare two tables. So the roomserver grew a read
model (stream.go, roomsync.go, visibility.go) rather than exporting its
store: reading an event means parsing it under its room's version, and a
second copy of the version registry, the parser and the client projection
would be a second place for them to disagree.

syncapi: stream tokens naming a position in each of five independent
streams, so a client that has seen receipt 7 but only event 4 can say so in
one opaque string. A token with fewer positions than this build writes is
accepted and the rest read as zero, which is what lets a token stored
before a stream existed go on working. The PDU position comes from
MAX(stream_ordering) and not from the counter row the other four use: the
counter is advanced inside the writing transaction, so a token taken from
it would name a position whose event no reader can see yet, and the event
would be skipped for good once that transaction committed.

Long polling goes through a notifier that wakes by room and by user rather
than globally — an event wakes the listeners of its room, a membership
event also wakes the user it names, so an invite reaches a client that is
not in the room yet. A wake-up is only ever a hint: the listener re-reads
the streams and decides for itself what changed, which is what makes it
safe for the notifier to hold no state about what each client has seen. The
listener registers before the read, so an event landing in between leaves
the channel signalled and the sync reads again rather than blocking on a
change it just missed. EventPersisted must not block, because it is called
with the room's write lock still held.

"Something happened" is not "a token moved". A stream can move without this
user being concerned by it, and returning on every such move would turn a
long poll back into a busy loop. The same trap caught presence twice over:
a client says it is still online by syncing, so ?set_presence= makes every
sync a presence write, and a stream that moved each time would wake every
peer, which would sync, which would move it again. SetPresence therefore
reports whether anything a peer would notice actually changed, and a write
that only refreshes the last-active timestamp leaves the stream alone.

Timeline windows are named by their two ends and a direction, Low exclusive
and High inclusive in both directions, rather than by "from" and "to" as
the wire format spells them. Each of those would otherwise be the low bound
in one direction and the high bound in the other, which is exactly the
off-by-one every pagination bug is made of; the handlers translate once and
nothing below them thinks about it again. A sync's timeline is read
backwards and reported forwards, because a client wants the newest events
of the window it missed — a forward read with a limit hands it the oldest
and then calls the rest a gap.

visibility: history visibility is now a span rather than a yes or no, which
is what Phase 3 deferred here. A user who has left a room is served the
room as it was when they left — their state reads, their members and their
timeline — instead of being refused, and is bounded above at their own
membership event so a kicked client cannot go on streaming. Their span is
found from their own member events, so a user who joined, left and rejoined
sees their latest span. ?at= on /members is honoured for the same reason.
What is left is that the span is computed from the room's current
history_visibility where the spec judges each event by the one in force
when it was sent; that is a state read per event and belongs with open
question 1.

Typing is held in memory and nowhere else: notifications expire in seconds,
are worthless across a restart, and arrive one per keystroke-burst per
room, so persisting them would be the busiest table in the database holding
the least valuable data in it. Receipts only ever advance, compared on the
position of the event acknowledged rather than on arrival, because two of a
user's clients read at different speeds and the slower one must not drag
the marker backwards. To-device queues are per device and drained by the
next sync rather than by the one that delivered them, so a client that
crashes in between gets the batch again — the right way round, since a lost
key share leaves a message permanently undecryptable and a duplicate is
ignored. A wildcard addressee is expanded when sent, not stored, because
"every device" means the ones that existed then.

Presence is off by default, which answers PLAN.md open question 3. It is
the noisiest thing in Matrix and the one genuinely fan-out notification
here; a private homeserver gains little from it. With it off the endpoints
refuse rather than disappearing, because a client that gets a 404 from
/presence concludes the server is too old to have it and stops asking.

storage: one migration per driver adding the receipt, to-device, presence
and filter tables with a counter row each, and a stream position on the
existing account data table rather than a parallel one — there is exactly
one current value per (user, room, type), and a second table would be a
second thing to keep in step. Account data is reported in an event list but
is not an event, so it gets its own response type: a room event with an
empty sender and no timestamp would make a client decide what that meant.

Deliberately missing, each named in docs/sync.md: notification counts (they
need push rules, Phase 8), device lists (the keyserver, Phase 7), anything
remote (Phase 5), threads (Phase 10). device_lists is sent as two empty
lists because a client reads a missing one as malformed; the notification
counts are left out entirely, because a zero there is a claim and an absent
field is not.

Tested through HTTP rather than against the builder, because what is worth
pinning down is the response a client sees — which section a room lands in,
whether a token handed back can be handed in again, whether paging
terminates and loses nothing. The storage tests run against both SQL
drivers; the notifier has its own tests for the concurrent case. A live
server was driven end to end, which is what caught the two bugs the unit
tests had missed: a backwards /messages page must be in reverse
chronological order, and `start` must name where the page began rather than
position zero when ?from= was omitted.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-10-16 11:55:27 +02:00

360 lines
13 KiB
Go

// Package roomserver owns rooms: it is the only thing in Warp that creates
// an event, and the only thing that decides whether one belongs in a room.
//
// Everything here runs through one pipeline — validate, authorise, resolve,
// persist, emit — because every event has to be judged the same way whether
// it came from a client, from `createRoom` or, from Phase 5, from another
// server. The rules themselves are not here: they are in internal/matrix,
// which is pure library code with no storage and no HTTP. This package is
// what connects those rules to a database and to a room's history.
//
// What the Client-Server API handlers get is a small set of operations —
// create a room, change a membership, send an event, redact one, read the
// state — each of which takes user IDs and content and returns an event ID
// or an error. Handlers do no authorisation of their own, and could not: a
// membership change is allowed or refused by the room's power levels and
// join rules, which only this package reads.
package roomserver
import (
"context"
"encoding/json"
"errors"
"fmt"
"log/slog"
"sync"
"time"
"github.com/dariuszjaniak/warp/server/internal/matrix/auth"
"github.com/dariuszjaniak/warp/server/internal/matrix/event"
"github.com/dariuszjaniak/warp/server/internal/matrix/id"
"github.com/dariuszjaniak/warp/server/internal/matrix/roomversion"
"github.com/dariuszjaniak/warp/server/internal/matrix/signing"
"github.com/dariuszjaniak/warp/server/internal/storage"
)
// Errors callers have to tell apart.
//
// The storage and auth sentinels are re-exported rather than translated, so
// that errors.Is works whichever package a caller reaches for.
var (
ErrNoSuchRoom = storage.ErrNoSuchRoom
ErrNoSuchEvent = storage.ErrNoSuchEvent
ErrNoSuchAlias = storage.ErrNoSuchAlias
ErrAliasExists = storage.ErrAliasExists
// ErrRejected means the authorisation rules refused the event. The
// wrapped message names the rule that said no, and is safe to show a
// client: it says why the room would not accept what they sent.
ErrRejected = auth.ErrRejected
// ErrForbidden is a refusal from a check that sits above the
// authorisation rules — not being in a room whose state you asked for,
// or deleting somebody else's alias.
ErrForbidden = errors.New("forbidden")
// ErrFederationDisabled is an operation that would need to talk to
// another server while federation is off.
ErrFederationDisabled = errors.New("federation is disabled")
// ErrInvalidContent is event content this server will not sign: not an
// object, too large, or missing something its type requires.
ErrInvalidContent = errors.New("invalid event content")
// ErrUnsupportedRoomVersion is a room version Warp cannot take part in.
ErrUnsupportedRoomVersion = roomversion.ErrUnsupportedVersion
// ErrUnknownRoomVersion is a room version that does not exist.
ErrUnknownRoomVersion = roomversion.ErrUnknownVersion
)
// Profiles supplies the display name and avatar that a membership event
// carries a copy of.
//
// Membership events duplicate the profile on purpose: a client showing a
// room's history has to render the name the sender had at the time, and the
// member event is the only place that is recorded. *userapi.API satisfies
// this as it stands.
type Profiles interface {
Account(ctx context.Context, localpart string) (storage.Account, error)
}
// Notification is one persisted event, as the emit stage reports it.
type Notification struct {
RoomID string
// Event is the event as it was persisted.
Event *event.Event
// StreamOrdering is the position the event was given in this server's
// event stream.
StreamOrdering int64
// StateChanged reports whether the room's current state moved.
StateChanged bool
}
// Notifier is told about every event this server persists, in the order
// they were persisted.
//
// It is the emit stage of the input pipeline and the seam the sync API
// plugs into: a client long-polling /sync has to be woken the moment an
// event lands, and the input pipeline is the only place that knows one
// has. *syncapi.Notifier satisfies it; nil skips the stage, which is what
// a build with no sync API does.
//
// An implementation must not block. It is called with the room's write
// lock still held, so anything slow here serialises the room's writers
// behind the slowest reader of it.
type Notifier interface {
EventPersisted(ctx context.Context, n Notification)
}
// API is the roomserver component.
type API struct {
serverName string
store storage.RoomStore
profiles Profiles
signingKey *signing.KeyPair
notifier Notifier
log *slog.Logger
now func() time.Time
// federationEnabled reports the live state of the federation switch.
// Several operations are refused outright while it is off, and the
// answer has to be read per request because an admin can flip it.
federationEnabled func() bool
// defaultRoomVersion is what createRoom uses when the client does not
// ask for one.
defaultRoomVersion roomversion.Version
// locks serialises the writers of each room. Two events racing into one
// room would each pick up the same forward extremities and both become
// one, forking the room for no reason — and then needing state
// resolution to put it back together. Taking a lock per room instead
// keeps every locally created event a child of the one before it.
locks sync.Map // room ID -> *sync.Mutex
}
// Options configure the component.
type Options struct {
// ServerName is the Matrix server name. It is the domain part of the
// user IDs this server may send events as, and of the room IDs and
// aliases it mints.
ServerName string
// Store is the persistence layer.
Store storage.RoomStore
// Profiles supplies the display name and avatar of a local user.
Profiles Profiles
// SigningKey signs every event this server creates.
SigningKey *signing.KeyPair
// DefaultRoomVersion overrides the registry's default for createRoom.
// Empty uses roomversion.Default.
DefaultRoomVersion string
// FederationEnabled reports the live federation switch. Nil means off,
// which is the safe default for a half-wired server.
FederationEnabled func() bool
// Notifier receives every persisted event. Nil skips the emit stage.
Notifier Notifier
Logger *slog.Logger
// Now is the clock, so that tests can stamp events deterministically.
Now func() time.Time
}
// New builds the component.
func New(opts Options) (*API, error) {
switch {
case opts.ServerName == "":
return nil, errors.New("roomserver: ServerName must not be empty")
case opts.Store == nil:
return nil, errors.New("roomserver: Store must not be nil")
case opts.SigningKey == nil:
// Without a key nothing here can produce an event, and finding that
// out on the first createRoom rather than at startup would mean a
// server that looks healthy and cannot be used.
return nil, errors.New("roomserver: SigningKey must not be nil")
}
version := roomversion.Default
if opts.DefaultRoomVersion != "" {
if _, err := roomversion.Get(roomversion.Version(opts.DefaultRoomVersion)); err != nil {
return nil, fmt.Errorf("roomserver: default room version %q: %w", opts.DefaultRoomVersion, err)
}
version = roomversion.Version(opts.DefaultRoomVersion)
}
if opts.FederationEnabled == nil {
opts.FederationEnabled = func() bool { return false }
}
if opts.Logger == nil {
opts.Logger = slog.Default()
}
if opts.Now == nil {
opts.Now = time.Now
}
return &API{
serverName: opts.ServerName,
store: opts.Store,
profiles: opts.Profiles,
signingKey: opts.SigningKey,
notifier: opts.Notifier,
log: opts.Logger,
now: opts.Now,
federationEnabled: opts.FederationEnabled,
defaultRoomVersion: version,
}, nil
}
// ServerName returns the server name this roomserver signs events as.
func (r *API) ServerName() string { return r.serverName }
// DefaultRoomVersion returns the version createRoom uses by default.
func (r *API) DefaultRoomVersion() string { return string(r.defaultRoomVersion) }
// nowMS is the current time in the unit events carry.
func (r *API) nowMS() int64 { return r.now().UnixMilli() }
// lock serialises the writers of one room; see API.locks.
func (r *API) lock(roomID string) *sync.Mutex {
actual, _ := r.locks.LoadOrStore(roomID, &sync.Mutex{})
mu, ok := actual.(*sync.Mutex)
if !ok {
// Unreachable: nothing else ever stores into this map.
panic("roomserver: room lock map holds something that is not a mutex")
}
return mu
}
// isLocal reports whether a user ID belongs to this server.
func (r *API) isLocal(userID string) bool {
serverName, err := id.ServerNameOf(userID)
return err == nil && serverName == r.serverName
}
// localpartOf returns the localpart of a user ID of this server.
func (r *API) localpartOf(userID string) (string, bool) {
localpart, serverName, err := id.SplitUserID(userID)
if err != nil || serverName != r.serverName {
return "", false
}
return localpart, true
}
// rules returns the room version rules a room's events are read under.
//
// A room whose version is no longer supported is a real possibility — a
// future build could drop one — and it has to fail as a room-level error
// rather than a parse error on every event in it.
func (r *API) rules(room storage.Room) (roomversion.Rules, error) {
rules, err := roomversion.Get(roomversion.Version(room.Version))
if err != nil {
return roomversion.Rules{}, fmt.Errorf("room %s is version %q: %w", room.ID, room.Version, err)
}
return rules, nil
}
// parseEvent turns a stored event back into an Event.
//
// A redacted event parses from its redacted form and keeps its original
// event ID, because the ID is the reference hash of the redacted form in
// the first place.
func parseEvent(stored storage.StoredEvent, rules roomversion.Rules) (*event.Event, error) {
e, err := event.Parse(stored.JSON, rules)
if err != nil {
return nil, fmt.Errorf("stored event %s is unreadable: %w", stored.EventID, err)
}
return e, nil
}
// parseEvents parses a batch, failing on the first unreadable one.
//
// An unreadable stored event is a corrupted database or a bug in this
// server's own encoder, never input: everything in the store was parsed and
// validated on the way in. It must not be skipped quietly, because the
// events being parsed here are a room's state and dropping one would
// silently widen what the room allows.
func parseEvents(stored []storage.StoredEvent, rules roomversion.Rules) ([]*event.Event, error) {
out := make([]*event.Event, 0, len(stored))
for _, s := range stored {
e, err := parseEvent(s, rules)
if err != nil {
return nil, err
}
out = append(out, e)
}
return out, nil
}
// stateEntries projects a state map onto the rows the store keeps.
func stateEntries(st auth.State) []storage.StateEntry {
out := make([]storage.StateEntry, 0, len(st))
for key, e := range st {
out = append(out, stateEntry(key.Type, key.StateKey, e))
}
return out
}
// stateEntry projects one piece of state onto a storage row, carrying the
// membership along for an m.room.member event so that the store can index
// it.
func stateEntry(eventType, stateKey string, e *event.Event) storage.StateEntry {
entry := storage.StateEntry{Type: eventType, StateKey: stateKey, EventID: e.EventID()}
if eventType == event.TypeMember {
// A member event whose content has no membership would have been
// refused on the way in, so an error here cannot happen; the empty
// string it leaves behind is also the correct "not a member".
entry.Membership, _ = e.Membership()
}
return entry
}
// eventProvider lets state resolution read a room's auth chains.
type eventProvider struct {
store storage.RoomStore
roomID string
rules roomversion.Rules
}
// EventsByID implements stateres.EventProvider.
//
// Events this server does not hold are left out rather than reported as an
// error, which is what the algorithm requires: a rejected or missing auth
// event takes no part in resolution.
func (p eventProvider) EventsByID(ctx context.Context, ids []string) ([]*event.Event, error) {
stored, err := p.store.EventsByID(ctx, p.roomID, ids)
if err != nil {
return nil, err
}
return parseEvents(stored, p.rules)
}
// validateContent checks event content before this server signs it.
//
// The size limit is the spec's limit on a whole event, applied to the
// content alone: the rest of an event is at most a few hundred bytes, so
// anything bigger than this cannot be made to fit and the client should hear
// about it before the event is built, hashed and signed.
func validateContent(content json.RawMessage) (json.RawMessage, error) {
if len(content) == 0 {
return json.RawMessage("{}"), nil
}
if len(content) > event.MaxEventSize {
return nil, fmt.Errorf("%w: content is %d bytes, the limit for a whole event is %d",
ErrInvalidContent, len(content), event.MaxEventSize)
}
var probe map[string]json.RawMessage
if err := json.Unmarshal(content, &probe); err != nil {
return nil, fmt.Errorf("%w: content must be a JSON object", ErrInvalidContent)
}
return content, nil
}