soju/user.go

870 lines
20 KiB
Go
Raw Normal View History

2020-03-16 10:18:41 +00:00
package soju
import (
"crypto/sha256"
"encoding/binary"
"encoding/hex"
"fmt"
2020-03-16 10:18:41 +00:00
"time"
"gopkg.in/irc.v3"
2020-03-16 10:18:41 +00:00
)
type event interface{}
type eventUpstreamMessage struct {
msg *irc.Message
uc *upstreamConn
}
type eventUpstreamConnectionError struct {
net *network
err error
}
type eventUpstreamConnected struct {
uc *upstreamConn
}
type eventUpstreamDisconnected struct {
uc *upstreamConn
}
type eventUpstreamError struct {
uc *upstreamConn
err error
}
type eventDownstreamMessage struct {
msg *irc.Message
dc *downstreamConn
}
type eventDownstreamConnected struct {
dc *downstreamConn
}
type eventDownstreamDisconnected struct {
dc *downstreamConn
}
type eventChannelDetach struct {
uc *upstreamConn
name string
}
type eventBroadcast struct {
msg *irc.Message
}
2020-08-03 16:45:13 +00:00
type eventStop struct{}
type deliveredClientMap map[string]string // client name -> msg ID
type deliveredStore struct {
m deliveredCasemapMap
}
func newDeliveredStore() deliveredStore {
return deliveredStore{deliveredCasemapMap{newCasemapMap(0)}}
}
func (ds deliveredStore) HasTarget(target string) bool {
return ds.m.Value(target) != nil
}
func (ds deliveredStore) LoadID(target, clientName string) string {
clients := ds.m.Value(target)
if clients == nil {
return ""
}
return clients[clientName]
}
func (ds deliveredStore) StoreID(target, clientName, msgID string) {
clients := ds.m.Value(target)
if clients == nil {
clients = make(deliveredClientMap)
ds.m.SetValue(target, clients)
}
clients[clientName] = msgID
}
func (ds deliveredStore) ForEachTarget(f func(target string)) {
for _, entry := range ds.m.innerMap {
f(entry.originalKey)
}
}
func (ds deliveredStore) ForEachClient(f func(clientName string)) {
clients := make(map[string]struct{})
for _, entry := range ds.m.innerMap {
delivered := entry.value.(deliveredClientMap)
for clientName := range delivered {
clients[clientName] = struct{}{}
}
}
for clientName := range clients {
f(clientName)
}
}
2020-03-16 10:18:41 +00:00
type network struct {
Network
user *user
2021-04-13 18:12:49 +00:00
logger Logger
stopped chan struct{}
2020-03-20 21:48:17 +00:00
conn *upstreamConn
channels channelCasemapMap
delivered deliveredStore
lastError error
casemap casemapping
2020-03-16 10:18:41 +00:00
}
func newNetwork(user *user, record *Network, channels []Channel) *network {
2021-04-13 18:12:49 +00:00
logger := &prefixLogger{user.logger, fmt.Sprintf("network %q: ", record.GetName())}
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
m := channelCasemapMap{newCasemapMap(0)}
for _, ch := range channels {
ch := ch
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
m.SetValue(ch.Name, &ch)
}
2020-03-16 10:18:41 +00:00
return &network{
Network: *record,
user: user,
2021-04-13 18:12:49 +00:00
logger: logger,
stopped: make(chan struct{}),
channels: m,
delivered: newDeliveredStore(),
casemap: casemapRFC1459,
2020-03-16 10:18:41 +00:00
}
}
func (net *network) forEachDownstream(f func(*downstreamConn)) {
net.user.forEachDownstream(func(dc *downstreamConn) {
if dc.network == nil && dc.caps["soju.im/bouncer-networks"] {
return
}
if dc.network != nil && dc.network != net {
return
}
f(dc)
})
}
2020-06-03 15:28:31 +00:00
func (net *network) isStopped() bool {
select {
case <-net.stopped:
return true
default:
return false
}
}
func userIdent(u *User) string {
// The ident is a string we will send to upstream servers in clear-text.
// For privacy reasons, make sure it doesn't expose any meaningful user
// metadata. We just use the base64-encoded hashed ID, so that people don't
// start relying on the string being an integer or following a pattern.
var b [64]byte
binary.LittleEndian.PutUint64(b[:], uint64(u.ID))
h := sha256.Sum256(b[:])
return hex.EncodeToString(h[:16])
}
2020-03-16 10:18:41 +00:00
func (net *network) run() {
2021-05-26 08:49:52 +00:00
if !net.Enabled {
return
}
2020-03-16 10:18:41 +00:00
var lastTry time.Time
for {
2020-06-03 15:28:31 +00:00
if net.isStopped() {
return
}
if dur := time.Now().Sub(lastTry); dur < retryConnectDelay {
delay := retryConnectDelay - dur
2021-04-13 18:12:49 +00:00
net.logger.Printf("waiting %v before trying to reconnect to %q", delay.Truncate(time.Second), net.Addr)
2020-03-16 10:18:41 +00:00
time.Sleep(delay)
}
lastTry = time.Now()
uc, err := connectToUpstream(net)
if err != nil {
2021-04-13 18:12:49 +00:00
net.logger.Printf("failed to connect to upstream server %q: %v", net.Addr, err)
net.user.events <- eventUpstreamConnectionError{net, fmt.Errorf("failed to connect: %v", err)}
2020-03-16 10:18:41 +00:00
continue
}
if net.user.srv.Identd != nil {
net.user.srv.Identd.Store(uc.RemoteAddr().String(), uc.LocalAddr().String(), userIdent(&net.user.User))
}
2020-03-16 10:18:41 +00:00
uc.register()
2020-04-01 10:14:36 +00:00
if err := uc.runUntilRegistered(); err != nil {
text := err.Error()
if regErr, ok := err.(registrationError); ok {
text = string(regErr)
}
uc.logger.Printf("failed to register: %v", text)
net.user.events <- eventUpstreamConnectionError{net, fmt.Errorf("failed to register: %v", text)}
2020-04-01 10:14:36 +00:00
uc.Close()
continue
}
2020-03-16 10:18:41 +00:00
2020-06-03 15:28:31 +00:00
// TODO: this is racy with net.stopped. If the network is stopped
// before the user goroutine receives eventUpstreamConnected, the
// connection won't be closed.
net.user.events <- eventUpstreamConnected{uc}
if err := uc.readMessages(net.user.events); err != nil {
2020-03-16 10:18:41 +00:00
uc.logger.Printf("failed to handle messages: %v", err)
net.user.events <- eventUpstreamError{uc, fmt.Errorf("failed to handle messages: %v", err)}
2020-03-16 10:18:41 +00:00
}
uc.Close()
net.user.events <- eventUpstreamDisconnected{uc}
if net.user.srv.Identd != nil {
net.user.srv.Identd.Delete(uc.RemoteAddr().String(), uc.LocalAddr().String())
}
2020-03-16 10:18:41 +00:00
}
}
func (net *network) stop() {
2020-06-03 15:28:31 +00:00
if !net.isStopped() {
close(net.stopped)
}
if net.conn != nil {
net.conn.Close()
}
}
func (net *network) detach(ch *Channel) {
if ch.Detached {
return
}
2021-04-13 18:12:49 +00:00
net.logger.Printf("detaching channel %q", ch.Name)
ch.Detached = true
if net.user.msgStore != nil {
nameCM := net.casemap(ch.Name)
lastID, err := net.user.msgStore.LastMsgID(net, nameCM, time.Now())
if err != nil {
2021-04-13 18:12:49 +00:00
net.logger.Printf("failed to get last message ID for channel %q: %v", ch.Name, err)
}
ch.DetachedInternalMsgID = lastID
}
if net.conn != nil {
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
uch := net.conn.channels.Value(ch.Name)
if uch != nil {
uch.updateAutoDetach(0)
}
}
net.forEachDownstream(func(dc *downstreamConn) {
dc.SendMessage(&irc.Message{
Prefix: dc.prefix(),
Command: "PART",
Params: []string{dc.marshalEntity(net, ch.Name), "Detach"},
})
})
}
func (net *network) attach(ch *Channel) {
if !ch.Detached {
return
}
2021-04-13 18:12:49 +00:00
net.logger.Printf("attaching channel %q", ch.Name)
detachedMsgID := ch.DetachedInternalMsgID
ch.Detached = false
ch.DetachedInternalMsgID = ""
var uch *upstreamChannel
if net.conn != nil {
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
uch = net.conn.channels.Value(ch.Name)
net.conn.updateChannelAutoDetach(ch.Name)
}
net.forEachDownstream(func(dc *downstreamConn) {
dc.SendMessage(&irc.Message{
Prefix: dc.prefix(),
Command: "JOIN",
Params: []string{dc.marshalEntity(net, ch.Name)},
})
if uch != nil {
forwardChannel(dc, uch)
}
if detachedMsgID != "" {
dc.sendTargetBacklog(net, ch.Name, detachedMsgID)
2021-04-13 15:49:37 +00:00
}
})
}
func (net *network) deleteChannel(name string) error {
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
ch := net.channels.Value(name)
if ch == nil {
return fmt.Errorf("unknown channel %q", name)
}
if net.conn != nil {
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
uch := net.conn.channels.Value(ch.Name)
if uch != nil {
uch.updateAutoDetach(0)
}
}
if err := net.user.srv.db.DeleteChannel(ch.ID); err != nil {
return err
}
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
net.channels.Delete(name)
return nil
}
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
func (net *network) updateCasemapping(newCasemap casemapping) {
net.casemap = newCasemap
net.channels.SetCasemapping(newCasemap)
net.delivered.m.SetCasemapping(newCasemap)
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
if net.conn != nil {
net.conn.channels.SetCasemapping(newCasemap)
for _, entry := range net.conn.channels.innerMap {
uch := entry.value.(*upstreamChannel)
uch.Members.SetCasemapping(newCasemap)
}
}
}
func (net *network) storeClientDeliveryReceipts(clientName string) {
if !net.user.hasPersistentMsgStore() {
return
}
var receipts []DeliveryReceipt
net.delivered.ForEachTarget(func(target string) {
msgID := net.delivered.LoadID(target, clientName)
if msgID == "" {
return
}
receipts = append(receipts, DeliveryReceipt{
Target: target,
InternalMsgID: msgID,
})
})
if err := net.user.srv.db.StoreClientDeliveryReceipts(net.ID, clientName, receipts); err != nil {
2021-04-13 18:12:49 +00:00
net.logger.Printf("failed to store delivery receipts for client %q: %v", clientName, err)
}
}
func (net *network) isHighlight(msg *irc.Message) bool {
if msg.Command != "PRIVMSG" && msg.Command != "NOTICE" {
return false
}
text := msg.Params[1]
nick := net.Nick
if net.conn != nil {
nick = net.conn.nick
}
// TODO: use case-mapping aware comparison here
return msg.Prefix.Name != nick && isHighlight(text, nick)
}
func (net *network) detachedMessageNeedsRelay(ch *Channel, msg *irc.Message) bool {
highlight := net.isHighlight(msg)
return ch.RelayDetached == FilterMessage || ((ch.RelayDetached == FilterHighlight || ch.RelayDetached == FilterDefault) && highlight)
}
2020-03-16 10:18:41 +00:00
type user struct {
User
srv *Server
logger Logger
2020-03-16 10:18:41 +00:00
events chan event
done chan struct{}
2020-03-16 10:18:41 +00:00
networks []*network
downstreamConns []*downstreamConn
msgStore messageStore
Add LIST support This commit adds support for downstream LIST messages from multiple concurrent downstreams to multiple concurrent upstreams, including support for multiple pending LIST requests from the same downstream. Because a unique RPL_LISTEND message must be sent to the requesting downstream, and that there might be multiple upstreams, each sending their own RPL_LISTEND, a cache of RPL_LISTEND replies of some sort is required to match RPL_LISTEND together in order to only send one back downstream. This commit adds a list of "pending LIST" structs, which each contain a map of all upstreams that yet need to send a RPL_LISTEND, and the corresponding LIST request associated with that response. This list of pending LISTs is sorted according to the order that the requesting downstreams sent the LIST messages in. Each pending set also stores the id of the requesting downstream, in order to only forward the replies to it and no other downstream. (This is important because LIST replies can typically amount to several thousands messages on large servers.) When a single downstream makes multiple LIST requests, only the first one will be immediately sent to the upstream servers. The next ones will be buffered until the first one is completed. Distinct downstreams can make concurrent LIST requests without any request buffering. Each RPL_LIST message is forwarded to the downstream of the first matching pending LIST struct. When an upstream sends an RPL_LISTEND message, the upstream is removed from the first matching pending LIST struct, but that message is not immediately forwarded downstream. If there are no remaining pending LIST requests in that struct is then empty, that means all upstreams have sent back all their RPL_LISTEND replies (which means they also sent all their RPL_LIST replies); so a unique RPL_LISTEND is sent to downstream and that pending LIST set is removed from the cache. Upstreams are removed from the pending LIST structs in two other cases: - when they are closed (to avoid stalling because of a disconnected upstream that will never reply to the LIST message): they are removed from all pending LIST structs - when they reply with an ERR_UNKNOWNCOMMAND or RPL_TRYAGAIN LIST reply, which is typically used when a user is not allowed to LIST because they just joined the server: they are removed from the first pending LIST struct, as if an RPL_LISTEND message was received
2020-03-26 01:40:30 +00:00
// LIST commands in progress
pendingLISTs []pendingLIST
Add LIST support This commit adds support for downstream LIST messages from multiple concurrent downstreams to multiple concurrent upstreams, including support for multiple pending LIST requests from the same downstream. Because a unique RPL_LISTEND message must be sent to the requesting downstream, and that there might be multiple upstreams, each sending their own RPL_LISTEND, a cache of RPL_LISTEND replies of some sort is required to match RPL_LISTEND together in order to only send one back downstream. This commit adds a list of "pending LIST" structs, which each contain a map of all upstreams that yet need to send a RPL_LISTEND, and the corresponding LIST request associated with that response. This list of pending LISTs is sorted according to the order that the requesting downstreams sent the LIST messages in. Each pending set also stores the id of the requesting downstream, in order to only forward the replies to it and no other downstream. (This is important because LIST replies can typically amount to several thousands messages on large servers.) When a single downstream makes multiple LIST requests, only the first one will be immediately sent to the upstream servers. The next ones will be buffered until the first one is completed. Distinct downstreams can make concurrent LIST requests without any request buffering. Each RPL_LIST message is forwarded to the downstream of the first matching pending LIST struct. When an upstream sends an RPL_LISTEND message, the upstream is removed from the first matching pending LIST struct, but that message is not immediately forwarded downstream. If there are no remaining pending LIST requests in that struct is then empty, that means all upstreams have sent back all their RPL_LISTEND replies (which means they also sent all their RPL_LIST replies); so a unique RPL_LISTEND is sent to downstream and that pending LIST set is removed from the cache. Upstreams are removed from the pending LIST structs in two other cases: - when they are closed (to avoid stalling because of a disconnected upstream that will never reply to the LIST message): they are removed from all pending LIST structs - when they reply with an ERR_UNKNOWNCOMMAND or RPL_TRYAGAIN LIST reply, which is typically used when a user is not allowed to LIST because they just joined the server: they are removed from the first pending LIST struct, as if an RPL_LISTEND message was received
2020-03-26 01:40:30 +00:00
}
type pendingLIST struct {
downstreamID uint64
// list of per-upstream LIST commands not yet sent or completed
pendingCommands map[int64]*irc.Message
2020-03-16 10:18:41 +00:00
}
func newUser(srv *Server, record *User) *user {
logger := &prefixLogger{srv.Logger, fmt.Sprintf("user %q: ", record.Username)}
var msgStore messageStore
if srv.LogPath != "" {
msgStore = newFSMessageStore(srv.LogPath, record.Username)
} else {
msgStore = newMemoryMessageStore()
}
2020-03-16 10:18:41 +00:00
return &user{
User: *record,
srv: srv,
logger: logger,
events: make(chan event, 64),
done: make(chan struct{}),
msgStore: msgStore,
2020-03-16 10:18:41 +00:00
}
}
func (u *user) forEachNetwork(f func(*network)) {
for _, network := range u.networks {
f(network)
}
}
func (u *user) forEachUpstream(f func(uc *upstreamConn)) {
for _, network := range u.networks {
if network.conn == nil {
2020-03-16 10:18:41 +00:00
continue
}
f(network.conn)
2020-03-16 10:18:41 +00:00
}
}
func (u *user) forEachDownstream(f func(dc *downstreamConn)) {
for _, dc := range u.downstreamConns {
f(dc)
}
}
func (u *user) getNetwork(name string) *network {
for _, network := range u.networks {
if network.Addr == name {
return network
}
if network.Name != "" && network.Name == name {
return network
}
2020-03-16 10:18:41 +00:00
}
return nil
}
func (u *user) getNetworkByID(id int64) *network {
for _, net := range u.networks {
if net.ID == id {
return net
}
}
return nil
}
2020-03-16 10:18:41 +00:00
func (u *user) run() {
defer func() {
if u.msgStore != nil {
if err := u.msgStore.Close(); err != nil {
u.logger.Printf("failed to close message store for user %q: %v", u.Username, err)
}
}
close(u.done)
}()
networks, err := u.srv.db.ListNetworks(u.ID)
2020-03-16 10:18:41 +00:00
if err != nil {
u.logger.Printf("failed to list networks for user %q: %v", u.Username, err)
2020-03-16 10:18:41 +00:00
return
}
for _, record := range networks {
record := record
channels, err := u.srv.db.ListChannels(record.ID)
if err != nil {
u.logger.Printf("failed to list channels for user %q, network %q: %v", u.Username, record.GetName(), err)
continue
}
network := newNetwork(u, &record, channels)
2020-03-16 10:18:41 +00:00
u.networks = append(u.networks, network)
if u.hasPersistentMsgStore() {
receipts, err := u.srv.db.ListDeliveryReceipts(record.ID)
if err != nil {
u.logger.Printf("failed to load delivery receipts for user %q, network %q: %v", u.Username, network.GetName(), err)
return
}
for _, rcpt := range receipts {
network.delivered.StoreID(rcpt.Target, rcpt.Client, rcpt.InternalMsgID)
}
}
2020-03-16 10:18:41 +00:00
go network.run()
}
for e := range u.events {
switch e := e.(type) {
case eventUpstreamConnected:
uc := e.uc
uc.network.conn = uc
uc.updateAway()
netIDStr := fmt.Sprintf("%v", uc.network.ID)
uc.forEachDownstream(func(dc *downstreamConn) {
dc.updateSupportedCaps()
if !dc.caps["soju.im/bouncer-networks"] {
sendServiceNOTICE(dc, fmt.Sprintf("connected to %s", uc.network.GetName()))
}
dc.updateNick()
dc.updateRealname()
})
u.forEachDownstream(func(dc *downstreamConn) {
if dc.caps["soju.im/bouncer-networks-notify"] {
dc.SendMessage(&irc.Message{
Prefix: dc.srv.prefix(),
Command: "BOUNCER",
Params: []string{"NETWORK", netIDStr, "state=connected"},
})
}
})
uc.network.lastError = nil
case eventUpstreamDisconnected:
u.handleUpstreamDisconnected(e.uc)
case eventUpstreamConnectionError:
net := e.net
stopped := false
select {
case <-net.stopped:
stopped = true
default:
}
if !stopped && (net.lastError == nil || net.lastError.Error() != e.err.Error()) {
net.forEachDownstream(func(dc *downstreamConn) {
sendServiceNOTICE(dc, fmt.Sprintf("failed connecting/registering to %s: %v", net.GetName(), e.err))
})
}
net.lastError = e.err
case eventUpstreamError:
uc := e.uc
uc.forEachDownstream(func(dc *downstreamConn) {
sendServiceNOTICE(dc, fmt.Sprintf("disconnected from %s: %v", uc.network.GetName(), e.err))
})
uc.network.lastError = e.err
case eventUpstreamMessage:
msg, uc := e.msg, e.uc
if uc.isClosed() {
uc.logger.Printf("ignoring message on closed connection: %v", msg)
break
}
if err := uc.handleMessage(msg); err != nil {
uc.logger.Printf("failed to handle message %q: %v", msg, err)
}
case eventChannelDetach:
uc, name := e.uc, e.name
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
c := uc.network.channels.Value(name)
if c == nil || c.Detached {
continue
}
uc.network.detach(c)
if err := uc.srv.db.StoreChannel(uc.network.ID, c); err != nil {
u.logger.Printf("failed to store updated detached channel %q: %v", c.Name, err)
}
case eventDownstreamConnected:
dc := e.dc
if err := dc.welcome(); err != nil {
dc.logger.Printf("failed to handle new registered connection: %v", err)
break
}
u.downstreamConns = append(u.downstreamConns, dc)
dc.forEachNetwork(func(network *network) {
if network.lastError != nil {
sendServiceNOTICE(dc, fmt.Sprintf("disconnected from %s: %v", network.GetName(), network.lastError))
}
})
u.forEachUpstream(func(uc *upstreamConn) {
uc.updateAway()
})
case eventDownstreamDisconnected:
dc := e.dc
for i := range u.downstreamConns {
if u.downstreamConns[i] == dc {
u.downstreamConns = append(u.downstreamConns[:i], u.downstreamConns[i+1:]...)
break
}
}
dc.forEachNetwork(func(net *network) {
net.storeClientDeliveryReceipts(dc.clientName)
})
u.forEachUpstream(func(uc *upstreamConn) {
uc.updateAway()
})
case eventDownstreamMessage:
msg, dc := e.msg, e.dc
if dc.isClosed() {
dc.logger.Printf("ignoring message on closed connection: %v", msg)
break
}
err := dc.handleMessage(msg)
if ircErr, ok := err.(ircError); ok {
ircErr.Message.Prefix = dc.srv.prefix()
dc.SendMessage(ircErr.Message)
} else if err != nil {
dc.logger.Printf("failed to handle message %q: %v", msg, err)
dc.Close()
}
case eventBroadcast:
msg := e.msg
u.forEachDownstream(func(dc *downstreamConn) {
dc.SendMessage(msg)
})
2020-08-03 16:45:13 +00:00
case eventStop:
u.forEachDownstream(func(dc *downstreamConn) {
dc.Close()
})
for _, n := range u.networks {
n.stop()
n.delivered.ForEachClient(func(clientName string) {
n.storeClientDeliveryReceipts(clientName)
})
2020-08-03 16:45:13 +00:00
}
return
default:
panic(fmt.Sprintf("received unknown event type: %T", e))
}
}
2020-03-16 10:18:41 +00:00
}
func (u *user) handleUpstreamDisconnected(uc *upstreamConn) {
uc.network.conn = nil
uc.endPendingLISTs(true)
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
for _, entry := range uc.channels.innerMap {
uch := entry.value.(*upstreamChannel)
uch.updateAutoDetach(0)
}
netIDStr := fmt.Sprintf("%v", uc.network.ID)
uc.forEachDownstream(func(dc *downstreamConn) {
dc.updateSupportedCaps()
})
u.forEachDownstream(func(dc *downstreamConn) {
if dc.caps["soju.im/bouncer-networks-notify"] {
dc.SendMessage(&irc.Message{
Prefix: dc.srv.prefix(),
Command: "BOUNCER",
Params: []string{"NETWORK", netIDStr, "state=disconnected"},
})
}
})
if uc.network.lastError == nil {
uc.forEachDownstream(func(dc *downstreamConn) {
if !dc.caps["soju.im/bouncer-networks"] {
sendServiceNOTICE(dc, fmt.Sprintf("disconnected from %s", uc.network.GetName()))
}
})
}
}
func (u *user) addNetwork(network *network) {
u.networks = append(u.networks, network)
go network.run()
}
func (u *user) removeNetwork(network *network) {
network.stop()
u.forEachDownstream(func(dc *downstreamConn) {
if dc.network != nil && dc.network == network {
dc.Close()
}
})
for i, net := range u.networks {
if net == network {
u.networks = append(u.networks[:i], u.networks[i+1:]...)
return
}
}
panic("tried to remove a non-existing network")
}
func (u *user) checkNetwork(record *Network) error {
for _, net := range u.networks {
if net.GetName() == record.GetName() && net.ID != record.ID {
return fmt.Errorf("a network with the name %q already exists", record.GetName())
}
}
return nil
}
func (u *user) createNetwork(record *Network) (*network, error) {
if record.ID != 0 {
panic("tried creating an already-existing network")
}
if err := u.checkNetwork(record); err != nil {
return nil, err
}
network := newNetwork(u, record, nil)
err := u.srv.db.StoreNetwork(u.ID, &network.Network)
2020-03-16 10:18:41 +00:00
if err != nil {
return nil, err
}
u.addNetwork(network)
idStr := fmt.Sprintf("%v", network.ID)
attrs := getNetworkAttrs(network)
u.forEachDownstream(func(dc *downstreamConn) {
if dc.caps["soju.im/bouncer-networks-notify"] {
dc.SendMessage(&irc.Message{
Prefix: dc.srv.prefix(),
Command: "BOUNCER",
Params: []string{"NETWORK", idStr, attrs.String()},
})
}
})
2020-03-16 10:18:41 +00:00
return network, nil
}
func (u *user) updateNetwork(record *Network) (*network, error) {
if record.ID == 0 {
panic("tried updating a new network")
}
if err := u.checkNetwork(record); err != nil {
return nil, err
}
network := u.getNetworkByID(record.ID)
if network == nil {
panic("tried updating a non-existing network")
}
if err := u.srv.db.StoreNetwork(u.ID, record); err != nil {
return nil, err
}
// Most network changes require us to re-connect to the upstream server
Implement casemapping TL;DR: supports for casemapping, now logs are saved in casemapped/canonical/tolower form (eg. in the #channel directory instead of #Channel... or something) == What is casemapping? == see <https://modern.ircdocs.horse/#casemapping-parameter> == Casemapping and multi-upstream == Since each upstream does not necessarily use the same casemapping, and since casemappings cannot coexist [0], 1. soju must also update the database accordingly to upstreams' casemapping, otherwise it will end up inconsistent, 2. soju must "normalize" entity names and expose only one casemapping that is a subset of all supported casemappings (here, ascii). [0] On some upstreams, "emersion[m]" and "emersion{m}" refer to the same user (upstreams that advertise rfc1459 for example), while on others (upstreams that advertise ascii) they don't. Once upstream's casemapping is known (default to rfc1459), entity names in map keys are made into casemapped form, for upstreamConn, upstreamChannel and network. downstreamConn advertises "CASEMAPPING=ascii", and always casemap map keys with ascii. Some functions require the caller to casemap their argument (to avoid needless calls to casemapping functions). == Message forwarding and casemapping == downstream message handling (joins and parts basically): When relaying entity names from downstreams to upstreams, soju uses the upstream casemapping, in order to not get in the way of the user. This does not brings any issue, as long as soju replies with the ascii casemapping in mind (solves point 1.). marshalEntity/marshalUserPrefix: When relaying entity names from upstreams with non-ascii casemappings, soju *partially* casemap them: it only change the case of characters which are not ascii letters. ASCII case is thus kept intact, while special symbols like []{} are the same every time soju sends them to downstreams (solves point 2.). == Casemapping changes == Casemapping changes are not fully supported by this patch and will result in loss of history. This is a limitation of the protocol and should be solved by the RENAME spec.
2021-03-16 09:00:34 +00:00
channels := make([]Channel, 0, network.channels.Len())
for _, entry := range network.channels.innerMap {
ch := entry.value.(*Channel)
channels = append(channels, *ch)
}
updatedNetwork := newNetwork(u, record, channels)
// If we're currently connected, disconnect and perform the necessary
// bookkeeping
if network.conn != nil {
network.stop()
// Note: this will set network.conn to nil
u.handleUpstreamDisconnected(network.conn)
}
// Patch downstream connections to use our fresh updated network
u.forEachDownstream(func(dc *downstreamConn) {
if dc.network != nil && dc.network == network {
dc.network = updatedNetwork
}
})
// We need to remove the network after patching downstream connections,
// otherwise they'll get closed
u.removeNetwork(network)
// This will re-connect to the upstream server
u.addNetwork(updatedNetwork)
// TODO: only broadcast attributes that have changed
idStr := fmt.Sprintf("%v", updatedNetwork.ID)
attrs := getNetworkAttrs(updatedNetwork)
u.forEachDownstream(func(dc *downstreamConn) {
if dc.caps["soju.im/bouncer-networks-notify"] {
dc.SendMessage(&irc.Message{
Prefix: dc.srv.prefix(),
Command: "BOUNCER",
Params: []string{"NETWORK", idStr, attrs.String()},
})
}
})
return updatedNetwork, nil
}
func (u *user) deleteNetwork(id int64) error {
network := u.getNetworkByID(id)
if network == nil {
panic("tried deleting a non-existing network")
}
if err := u.srv.db.DeleteNetwork(network.ID); err != nil {
return err
}
u.removeNetwork(network)
idStr := fmt.Sprintf("%v", network.ID)
u.forEachDownstream(func(dc *downstreamConn) {
if dc.caps["soju.im/bouncer-networks-notify"] {
dc.SendMessage(&irc.Message{
Prefix: dc.srv.prefix(),
Command: "BOUNCER",
Params: []string{"NETWORK", idStr, "*"},
})
}
})
return nil
}
func (u *user) updatePassword(hashed string) error {
u.User.Password = hashed
return u.srv.db.StoreUser(&u.User)
}
2020-08-03 16:45:13 +00:00
func (u *user) stop() {
u.events <- eventStop{}
<-u.done
2020-08-03 16:45:13 +00:00
}
func (u *user) hasPersistentMsgStore() bool {
if u.msgStore == nil {
return false
}
_, isMem := u.msgStore.(*memoryMessageStore)
return !isMem
}